7.1.1. Current practice

Pharmacotherapy is most frequently used to facilitate withdrawal from alcohol in people who are dependent; many fewer individuals receive medication for relapse prevention such as acamprosate, disulfiram or naltrexone. Indeed, some people may be reluctant to take medication and traditionally many residential rehabilitation units have not been prepared to accept or support people taking such medication, although this is slowly changing. A US survey revealed that only about 9% of people needing treatment for alcohol dependence received medication for relapse prevention; prescriptions of disulfiram declined by 3% between 2003 and 2007, while prescriptions for naltrexone rose by 3% and for acamprosate by 10% (Mark et al., 2009). The level of prescribing is likely to be similar or even lower in the UK. One estimate from data on prescriptions shows in 2008 that there were almost 135,000 prescriptions for acamprosate or disulfiram from primary care or NHS settings, with the majority (62%) for acamprosate (The NHS Information Centre, Lifestyles Statistics, 2009). In NHS hospitals, the use of disulfiram has increased with slightly more (54%) prescriptions issued than for acamprosate. There are regional variations with London issuing 104 prescriptions per 100,000 population and the North East issuing 417 per 100,000. Some doctors can be reluctant to prescribe pharmacological interventions such as acamprosate, naltrexone and disulfiram, due to lack of knowledge or familiarity (Mark et al., 2003). Barriers to prescribing naltrexone in the US have been described as including a ‘lack of awareness, a lack of evidence of efficacy in practice, side effects, time for patient management, a reluctance to take medications, medication addiction concerns, AA philosophy, and price' (Mark et al., 2003). Nevertheless, there are a variety of medications with proven effectiveness and others with emerging efficacy that deserve due consideration as part of any individual treatment package.

For relapse prevention, both acamprosate and disulfiram are licensed for relapse prevention in the UK, much of Europe, Australasia and North America. Naltrexone is used in the UK but licensed elsewhere (for example, in the US).

In this guideline, some pharmacotherapies described do not have a UK licence for the indication discussed. It is important to realise that in this area of medicine the absence of a licence can mean that one has not been applied for, rather than that the pharmacotherapy is not safe or appropriate. The terms ‘unlicensed’ and ‘off-label’ should not necessarily be taken to automatically imply disapproval, nor incorrect or improper use. There is no contraindication to prescribing a drug off-licence provided there is a body of evidence that supports its efficacy and safety (Healy & Nutt, 1998; Royal College of Psychiatrists, 2007), and often evidence of safety may come from its use in other disorders where a licence may have been granted. In particular, many drugs will not have a licence for use in young people, children or in older people, but this does not mean they necessarily lack efficacy or are unsafe. Nevertheless, when prescribing in these populations, due care must be taken in terms of dosage and monitoring of side effects, as well as potential interactions with other medications or physical morbidity.

7.1.2. The effects of alcohol on brain chemistry and how this relates to medication

As described in Chapter 2, alcohol affects many of the brain's chemical systems. The pharmacology of most of the medications commonly used, such as benzodiazepines for alcohol withdrawal and disulfiram, acamprosate and naltrexone for relapse prevention, is well characterised and provides a potential neurobiological rationale for their effectiveness. Understanding more about how alcohol interacts with the brain has revealed many potential targets of interest, for example to reduce drinking or craving. In many cases, medication already exists with the desired pharmacology but is used for another indication, for example baclofen for muscle spasm. Most new medication is being developed to prevent relapse rather than for use in alcohol withdrawal, or to improve cognition or prevent toxicity.

7.1.3. Brain chemistry and medication for relapse prevention

7.1.4. Brain chemistry and medication for alcohol withdrawal

A significant number of alcohol's effects on the brain involve interacting with the inhibitory GABA system. In addition to the GABA-B system described above, there is a GABA-A or GABA-benzodiazepine system that plays several important roles in mediating effects of alcohol on the brain.

The GABA-A receptor is made of different subunits on which there are various binding sites for benzodiazepines, barbiturates, neurosteroids and some anaesthetics, as well as for GABA. Alcohol interacts with the GABA-benzodiazepine receptor and increases its inhibitory activity, resulting in reduced anxiety and sedation, and can contribute to ataxia, slurred speech and respiratory depression. Thus alcohol has a similar effect to benzodiazepines such as diazepam. Alcohol is often used for its anxiolytic or sedative effects rather than pleasurable effects and anxiety and sleep disorders are associated with vulnerability to alcohol misuse.

Tolerance is the need to drink more alcohol to obtain the same or desired effect, and it develops in those drinking more heavily and regularly. A reduced sensitivity of the GABA system to alcohol underlies tolerance. It is thought that changes in the subunit profile of the GABA-A receptor complex are involved (Krystal et al., 2006). In alcohol withdrawal, benzodiazepines such as chlordiazepoxide (Librium) or diazepam (Valium) will boost this reduced GABA-ergic function to increase the inhibitory activity in the brain. This is important to control symptoms such as anxiety and tremor, and to reduce the risk of complications such as seizures and delirium tremens.

In addition to boosting the inhibitory GABA system, alcohol antagonises the excitatory neurotransmitter system, glutamate and particularly the NMDA receptor. To overcome this blockade, the number of NMDA receptors increase in response to continued drinking. This increase has been associated with memory impairment in animal models and may therefore underlie amnesia or blackouts, which can be experienced by people who drink heavily (Krystal et al., 2003). In alcohol withdrawal, therefore, the increased glutamatergic activity significantly contributes to the associated symptoms and risks such as tremor and seizures. Anticonvulsants, which reduce glutamatergic activity as well as increasing GABA-ergic activity, can therefore be used to treat alcohol withdrawal. In addition to this GABA-glutamate activity, anticonvulsants will also inhibit voltage-activated sodium channels and, consequently, further excitatory activity.

Another consequence of increased glutamatergic and calcium channel activity is cell death. Therefore a potential advantage of antagonising this increased activity in withdrawal is neuroprotection or preventing cell death. In animal models, acamprosate has been shown to reduce increased glutamatergic activity in withdrawal but robust clinical evidence is lacking. Whether it occurs with anticonvulsants has not been systematically studied.

7.2.1. Aim of review

The focus of this chapter is the effectiveness and cost-effectiveness of pharmacological interventions to prevent relapse or reduce alcohol consumption. The use of drugs alone or in combination with a range of other psychosocial interventions were considered. The drugs considered for inclusion in the review are set out in Table 89.

Table 89

Pharmacology of medications for the treatment of alcohol misuse.

The review aimed to evaluate all available pharmacological interventions for relapse prevention. This was conducted for adults and, where evidence was available, separately for special populations such as children and young people or older people. The GDG decided to conduct a meta-analysis only on the drugs that were licensed for alcohol use in the UK or drugs that are in common usage with a large amount of clinical evidence of efficacy. From these criteria, the drugs identified for review were acamprosate, naltrexone and disulfiram. For naltrexone and disulfiram, only the oral delivery preparations of these drugs was considered for meta-analysis due to the lack of available evidence and the uncommon usage of the extended-release and subcutaneous implantation preparations of these drugs. These drugs are evaluated in the first instance for both adults and for special populations (children and young people). The narrative review of the available literature for the use of pharmacological interventions for special populations can be found in Section 7.12. For other pharmacological interventions that are not licensed for used in the UK for the treatment of alcohol misuse, meta-analyses were not conducted. The reasons for this and the narrative synthesis of the evidence can be found in Section 7.14.

Literature evaluating pharmacological interventions for less severely dependent and non-dependent drinkers is limited and a meta-analysis could not be conducted. A narrative synthesis of the available literature is provided in Section 7.16. Similarly, trials where the participant sample included a very high prevalence of comorbid mental health disorders were excluded from the meta-analysis and are reviewed separately in Section 7.17. See Chapter 3 for a further discussion of the review methods used in this chapter. Lastly, a review of the long-term management of WKS can be found in Section 7.18.

7.2.2. Review questions

The review question that the GDG addressed, and from which the literature searches were developed, is:

1. For people with alcohol dependence or harmful alcohol use, what pharmacological interventions are more clinically and cost-effective?

In addition:

1. What are the impacts of severity and comorbidities on outcomes?
2. When should pharmacological treatments be initiated and for what duration should they be prescribed?

7.4.1. Studies considered for review⁴³

The review team conducted a systematic search for RCTs that assessed the benefits and disadvantages of acamprosate for relapse prevention. The clinical review protocol for this section can be found in Section 7.3. Study characteristics are summarised in Table 91. For the related health economic evidence see Section 7.10.

Table 91

Summary of study characteristics for acamprosate versus placebo.

There were a total of 19 trials (including one study still awaiting translation) comparing acamprosate with placebo. These were typically large, high-quality studies, of which ten were sponsored by the drug company. A number of psychosocial interventions were used in addition to the trial medication, in line with the drug licensing agreement, which included alcohol counselling, medication management and relapse prevention as well as high-intensity alcohol treatment programmes. Data on participants lapsing to alcohol consumption was acquired from the authors of two meta-analyses (Mann, 2004; Rosner et al., 2008), who had access to unpublished data and therefore allowed for the development of a more complete dataset. Both the PAILLE1995 and PELC1997 studies were three-armed trials where two different doses of acamprosate (1.3 g and 2 g) were compared with placebo. To avoid the double counting of the control data, only data for the groups taking 2 g of acamprosate were used, because this is the dose recommended by the BNF. Reasons for exclusion of studies from this review included not providing an acceptable diagnosis of alcohol dependence, not being an RCT, having fewer than ten participants per group, not being double blind and not reporting any relevant outcomes. Further information about both included and excluded studies can be found in Appendix 16e.

The populations within these trials were typically presenting with moderate to severe alcohol dependence, either indicated through alcohol consumption or dependency scale shown at baseline. These studies were mainly conducted in Europe, with only one (CHICK2000A) conducted in the UK. Acamprosate was started after the participant completed assisted withdrawal (if required) in all trials except one, GUAL2001, when it was started during assisted withdrawal.

7.4.2. Evidence summary

Evidence from the important outcomes and overall quality of evidence are presented in Table 92. The full evidence profiles and associated forest plots can be found in Appendix 18d and Appendix 17d, respectively.

Table 92

Evidence summary table for trials of acamprosate versus placebo.

There was a significant but small effect of acamprosate in promoting abstinence in participants when compared with placebo (RR = 0.83; 95% CI = 0.77 to 0.88). The effect was most pronounced at 6 months, but remained significant up to 12 months. In the one trial that continued up to 2 years (WHITWORTH1996) this small effect continued for up to 12 months after the termination of treatment. The number of individuals relapsing to heavy drinking was also significantly less in the acamprosate group. This effect was also small (RR = 0.90; 95% CI = 0.81 to 0.99) but suggests participants were more likely to stay in treatment if randomised to acamprosate instead of placebo. However, more participants left the trials due to adverse events in the acamprosate group, although this was not statistically significant.

The quality of the evidence for acamprosate is high, therefore further research is unlikely to have an important impact on confidence in the estimate of the effect. An evidence summary of the results of the meta-analyses can be seen in Table 92.

7.5.1. Studies considered

The review team conducted a systematic review of RCTs that assessed the beneficial or detrimental effects of naltrexone for relapse prevention. See Section 7.2 for the aim of the review and the review questions. The clinical review protocol for this section can be found in Section 7.3. See Table 93 for a summary of the study characteristics of the included studies.

Table 93

Summary of study characteristics for naltrexone.

A total of 27 trials compared oral naltrexone with placebo and four trials compared naltrexone with acamprosate. In addition, there were two studies comparing naltrexone with naltrexone plus sertraline and one trial comparing naltrexone with topiramate. The majority of the trials were large, high-quality studies with five trials sponsored by drug companies. Twenty-six of the trials (LATT2002 being the exception) included one of a number of different psychosocial interventions in addition to either naltrexone or placebo, including alcohol counselling, coping skills or relapse prevention as well as high-intensity alcohol treatment programmes. Unpublished data on individuals relapsing to heavy drinking was acquired from the authors of a metaanalysis (Rosner et al., 2008), who had access to unpublished data. Reasons for exclusion of studies from this review included not providing an acceptable diagnosis of alcohol dependence, not being an RCT, having fewer than ten participants per group, not being double blind and not reporting any relevant outcomes. Further information about both included and excluded studies can be found in Appendix 16e.

One additional study including naltrexone by Petrakis and colleagues (2005), although a high-quality trial, was excluded because the whole participant sample was comorbid with a range of Axis I disorders, with many participants having multiple coexisting disorders. This was unusual when compared with the included trials, where comorbidity was usually grounds for exclusion. This study is described more fully in the comorbidity in Section 7.17.

The participant population included in these trials ranged from mild to severe dependence based on baseline alcohol consumption and dependency scale scores. (This is in contrast to the studies included in the acamprosate review where participants generally presented with more severe dependence.) The majority of these trials were conducted in North America, and recruitment was most commonly through advertisements or referrals. If assisted withdrawal was required, then naltrexone was started after this was completed in these trials.

7.5.2. Evidence summary

Evidence on the important outcomes and overall quality of evidence are presented in Table 94. The full evidence profiles and associated forest plots can be found in Appendix 18d and Appendix 17d, respectively.

Table 94

Evidence summary table for trials of naltrexone.

The comparison of oral naltrexone versus placebo showed a small but significant effect favouring naltrexone on rates of relapse to heavy drinking (RR = 0.83; 95% CI, 0.75 to 0.91). The mean DDD within the trial duration was less in the naltrexone group when compared with placebo, with a small but significant effect (SMD = −0.28; 95% CI, −0.44 to −0.11). A significant but small effect favouring naltrexone was also found on days of heavy drinking during the trial (SMD = −0.43; 95% CI, −0.82 to −0.03). Although overall discontinuation rates favoured naltrexone over placebo, there was no significant difference between the two groups. However, participants were significantly more likely to leave treatment due to adverse events in the naltrexone group, with significantly fewer adverse events reported in the placebo group.

When comparing oral naltrexone with acamprosate, the four trials reviewed showed no significant difference in discontinuation for any reason or due to adverse events between the two interventions. On critical outcomes, there were no significant differences between naltrexone and acamprosate except for number of individuals returning to any drinking (RR = 0.71; 95% CI, 0.57 to 0.88) and DDD (SMD = −0.76; 95% CI, −1.09 to −0.44). However, these findings were based on only one study (RUBIO2001), which found that participants in the naltrexone group were significantly less likely to return to any drinking and consumed significantly less DDD during the trial period. When comparing naltrexone with topiramate, the analysis showed no significant differences between the groups on any outcomes except number of participants continuously abstinent and weeks until first relapse, both outcomes favouring naltrexone. The analysis of naltrexone versus naltrexone plus sertraline showed no significant differences between the groups on any outcomes. However, discontinuation rates were less in the combination group.

The quality of the evidence reviewed for oral naltrexone versus placebo was high, therefore further research is unlikely to have an important impact on confidence in the estimate of the effect. The quality of the evidence for naltrexone versus acamprosate was also high. However, the quality for the evidence for the naltrexone plus sertraline combination intervention versus naltrexone alone and for naltrexone versus topiramate was moderate, therefore further research is likely to have an important impact on confidence in the estimate of these effects.

7.6.1. Studies considered

The review team conducted a systematic review of RCTs that assessed the beneficial or detrimental effects of acamprosate plus naltrexone for relapse prevention. See Section 7.2 for the aim of the review and the review questions. The clinical review protocol for this section can be found in Section 7.3. See Table 95 for a summary of the study characteristics of included studies.

Table 95

Summary of study characteristics for naltrexone + acamprosate.

There were two trials comparing the combination of acamprosate and naltrexone with placebo, acamprosate alone and naltrexone alone. Both were large multiple-armed trials designed specifically to test the effects of the drugs in isolation and together. The KIEFER2003 trial included a population of people with severe dependence recruited from inpatient facilities; their mean pre-admission consumption of alcohol was 223 units per week. Each participant received a relapse prevention intervention in addition to pharmacological therapy. The ANTON2006 study included a less severe population of people with dependence who were recruited through advertisements or clinical referrals; their mean pre-admission consumption of alcohol was 97 units per week. In addition to being randomised to one of four pharmacological interventions, participants were also randomised to a cognitive-behavioural intervention with medication management or medication management alone. Reasons for exclusion of studies from this review included not providing an acceptable diagnosis of alcohol dependence, not being an RCT, having fewer than ten participants per group, not being double blind and not reporting any relevant outcomes. Further information about both included and excluded studies can be found in Appendix 16e.

7.6.2. Evidence summary

Evidence from the important outcomes and overall quality of evidence are presented in Table 96. The full evidence profiles and associated forest plots can be found in Appendix 18d and Appendix 17d, respectively.

Table 96

Evidence summary table for trials of acamprosate + naltrexone.

There was no significant difference between the combination of acamprosate and naltrexone and either drug alone on reducing the likelihood of returning to heavy drinking at 3 months (combination versus acamprosate: RR = 0.93; 95% CI, 0.74 to 1.17; combination versus naltrexone: RR = 1.03 [0.90 to 1.17]), and the one trial continuing up to 12 months showed no effect. In addition, there were no significant differences on any other outcomes between the combination group and either drug. The combined drug group was also equivalent to the placebo group on discontinuation rates and PDA. Relapse rates at 6 months were significantly different with a moderate effect in favour of the combined intervention group (RR = 0.44; 95% CI, 0.28 to 0.69); however, there was no difference between the groups on relapse rates at 3 months or 12 months.

The quality of the evidence was high, therefore further research is unlikely to have an important impact on confidence in the estimate of the effect. It was also noted that there was significant heterogeneity between comparisons of the KIEFER2003 and the ANTON2006 studies, which is very likely to be due to the differences in the populations, baseline drinking level and from where they were recruited (inpatient facility versus advertisement or referral).

7.7.1. Studies considered

The review team conducted a systematic review of RCTs that assessed the beneficial or detrimental effects of disulfiram for relapse prevention. See Section 7.2 for the aim of the review and the review questions. The clinical review protocol for this section can be found in Section 7.3. See Table 97 for the characteristics of the included studies. Unlike the reviews of acamprosate and naltrexone, there was much less high-quality evidence available on the efficacy and effectiveness of disulfiram, and for this reason the GDG decided to use open-label trials in the meta-analysis of disulfiram.

Table 97

Summary of study characteristics for oral disulfiram.

The reason for this was that due to the disulfiram–ethanol reaction, a number of the studies had to be open-label for ethical reasons so that participants were aware that they were taking a substance that can cause potentially dangerous side effects when taken with alcohol. This also contributes to the psychological effect of disulfiram, where the fear of the chemical reaction is believed to be as important as the pharmacological effects of the drug in determining the efficacy of the intervention. The FULLER1979 and FULLER1986 trials adapted their trials for this purpose and randomised participants to either the full dose of disulfiram (250 mg per day) or to 1 mg of disulfiram with a placebo agent that has been judged to have no clinical effect.

Due to the age of some of the trials, inclusion criteria for diagnosis was also relaxed to include papers that did not explicitly mention the diagnosis tool used to determine eligibility to the trial. The Petrakis and colleagues (2005) trial was also excluded from the meta-analysis as many participants had a range of Axis I disorders.

There were a total of three trials comparing oral disulfiram with placebo (FULLER1979; FULLER1986; CHICK1992), one trial comparing oral disulfiram with acamprosate (LAAKSONEN2008), two trials comparing oral disulfiram with naltrexone (DESOUSA2004; LAAKSONEN2008) and one trial comparing oral disulfiram with topiramate (DESOUSA2008). In addition, there was one trial comparing disulfiram and counselling with counselling alone (GERREIN1973).

The severity of dependence of the participants included in these trials was not reported for the older trials; however, in the more recent studies, dependency indicated through baseline consumption and dependency scales suggested that these participants were of moderate to severe dependency. The trials varied in terms of the country in which they were conducted, with CHICK1992 being the only trial conducted in the UK. Three studies were conducted in the US (FULLER1979; FULLER1986; GERREIN1973), two were conducted in India (DESOUSA2004; DESOUSA2008) and one was conducted in Finland (LAAKSONEN2008).

7.7.2. Evidence summary

Evidence on the important outcomes and the overall quality of evidence are presented in Table 98. The full evidence profiles and associated forest plots can be found in Appendix 18d and Appendix 17d, respectively.

Table 98

Evidence summary table for trials of oral disulfiram.

Oral disulfiram was not significantly different from placebo in preventing participants lapsing to alcohol consumption (RR = 1.05; 95% CI, 0.96 to 1.15). There was also no difference in rates of discontinuation between the two groups. However, LAAKSONEN2008 showed that, in comparison with acamprosate, disulfiram was significantly more likely to increase the time until participants first drank any alcohol (SMD = −0.84; 95% CI, −1.28 to −0.40) and drank heavily (SMD = −1.17; 95% CI, −1.66 to −0.68), and also decreased the amount of alcohol consumed and the number of drinking days. In comparison with naltrexone, disulfiram was also significantly more likely to increase the time to first heavy drinking day and the number of abstinent days. Participants in the naltrexone group were significantly more likely to return to any drinking (RR = 0.18; 95% CI, 0.08 to 0.42) or relapse to heavy drinking (RR = 0.28; 95% CI, 0.13 to 0.59) when compared with the oral disulfiram group, although this was based on two open-label studies (DESOUSA2004; LAAKSONEN2008).

The comparison of disulfiram and topiramate also showed a significant difference in the number of participants relapsing to heavy drinking (RR = 0.23; 95% CI, 0.09 to 0.55), time to first drink and time to first relapse in favour of disulfiram, but this was based on just one open-label study (DESOUSA2008). It may be that the psychological effects of knowing they were taking disulfiram may have contributed significantly to the results. The comparison of disulfiram with counselling versus counselling alone showed no significant differences between the groups on numbers of participants returning to drinking (RR = 0.86; 95% CI, 0.55 to 1.34).

The quality of the evidence was moderate, therefore further research is likely to have an important impact on confidence in the estimate of the effect. The main reason for the lower quality of the evidence was that the studies reviewed were generally not conducted in a double-blind trial.

Acamprosate

Lesch and Walter (1996) reviewed outcomes in their trial with reference to their four typologies: type I (social drinking develops into dependence; craving; relief drinking; family history); type II (alcohol consumed to medicate sleep or anxiety; consumption varies with context, behaviour changes with alcohol); type III (alcohol used to self-medicate a psychiatric disorder such as depression; family history positive for alcoholism or psychiatric disorder; impaired behaviours not always related to alcohol); and type IV (brain damage and psychiatric disorders before 14 years; seizures not related to alcohol; mild withdrawal symptoms). they reported that types I and II, but not III and IV, responded to acamprosate.

In the UK trial, Chick and colleagues (2000a) speculated whether the continuous rather than episodic drinker would be more likely to respond because their negative study had more participants with episodic drinking patterns. Kiefer and colleagues (2005) examined predictors in their original trial of acamprosate alone and with naltrexone, and reported that acamprosate was mainly efficacious in patients with low baseline somatic distress (mainly effective in type I) and that craving showed no predictive value.

Mason and Lehert (2010) explored the first US acamprosate trial (Mason et al., 2006) and suggested that acamprosate may reduce the negative impact of subsyndromal anxiety or a past psychiatric history.

In contrast, Verheul and colleagues (2005) examined pooled data from seven RCTs that included 1,485 patients with alcohol dependence. Whilst ‘cumulative abstinence duration’, or continuous abstinence, was predicted by higher levels of craving or anxiety at baseline, this was for all patients and acamprosate showed no differentially efficacy. Other variables that were investigated and showed no significant relationship with outcomes, including severity of dependence (which was nonlinearly associated with cumulative abstinence duration), family history, age of onset and gender. Therefore, they concluded that acamprosate is potentially effective for anyone with alcohol dependence.

Naltrexone

Monterosso and colleagues (2001) reported that those with a family history of alcoholism and high levels of craving were more likely to benefit from naltrexone. Rubio and colleagues (2005) similarly reported from their naltrexone trial that those with a family history of ‘alcoholism’ benefited more, as well as those whose onset of alcohol misuse was before the age of 25 years or those who had history of other substance misuse. Kiefer and colleagues (2005) reported that naltrexone was effective especially in patients with high baseline depression and in types III and IV (Lesch & Walter, 1996).

Several studies have investigated whether genetic variants of the opioid receptors, mu, kappa and delta, are related to naltrexone's efficacy. Several studies have reported an association between greater treatment response and A118G (OPRM1), a functional polymorphism of the μ-opioid receptor gene (Oslin et al., 2003; Anton et al., 2008a; Oroszi et al., 2009; Kim et al., 2009), but not all (Gelernter et al., 2007). In a relatively small sample, Ooteman and colleagues (2009) explored other genotypes and reported effects of GABRA6, GABRA2, OPRM1 and dopamine D2 receptor genes moderated treatment response from acamprosate or naltrexone and subjective and physiological cue reactivity.

It is not clear whether gender influences treatment outcome, with studies of naltrexone in alcohol misuse reporting no gender differences (Anton et al., 2006). Pettinati and colleagues (2008b) reported that in comorbid cocaine/alcohol dependence, men taking naltrexone (150 mg per day) reduced their cocaine and alcohol use whereas women did not; indeed, their cocaine use increased. However, most studies have limited power to detect gender by treatment outcome.

Disulfiram

There is no systematic review and little indication from trials of disulfiram about which type of person might be more likely to benefit from treatment.

7.9.1. Compliance and adherence

Compliance and adherence are related to predictors of efficacy because if a person is not taking their medication as prescribed, then its effectiveness is likely to be reduced. Because acamprosate and naltrexone are generally well-tolerated medications, problematic side effects are unlikely to contribute significantly to reduced compliance.

This issue has only been studied with naltrexone where Rohsenow and colleagues (2000) found that compliance was better in those who believed that the medication would help them stay sober and was not predicted by demographic or pre-treatment alcohol-use variables, commitment to abstinence or self-efficacy about abstinence.

For disulfiram, witnessing or supervision has been shown to be an important component of its effectiveness (Chick et al., 1992; Sereny et al., 1986). People who might do better with unsupervised disulfiram are older (Baekeland et al., 1971; Fuller et al., 1986); more socially stable (Fuller et al., 1986); impulsive (Banys, 1988); and higher in motivation (Baekeland et al., 1971).

7.9.2. When to start pharmacological treatment

People should be abstinent from alcohol at the time of starting medication for relapse prevention. All medications should be used as an adjunct to psychosocial treatment and not prescribed in isolation.

7.9.3. How long to continue with pharmacological treatment

Most trials of medication are between 3 and 6 months long and show efficacy. However, many patients relapse within months to years, but there is very limited evidence to guide how long medication should be continued. People who are doing well may be best advised to remain on medication for at least 6 months. However, some of these people may feel confident enough to stop medication earlier. Alternatively, some may prefer to stay on medication for longer, but continuation beyond 1 year would need to be justified. If a person is not engaging with other aspects of treatment (for example, psychosocial) and is drinking heavily, stopping the medication is appropriate until they engage with treatment. However, if a person is engaged but still drinking, a review of all of their treatment is indicated to assess whether this is optimal, including medication.

There is no evidence currently that long-term use of any of the relapse prevention pharmacotherapy incurs additional adverse consequences, particularly when relapse to heavy drinking will be associated with morbidity and mortality. However, medication is ideally used as an adjunct to support engagement with psychosocial approaches to alter behaviour and attitudes to alcohol.

For acamprosate, Mann (2004) reported from their meta-analysis that the effect sizes increased with time (the effect sizes on abstinence at 3, 6 and 12 months were 1.33, 1.50 and 1.95, respectively). This suggests that a clinically relevant benefit of treatment may be observed as early as 3 months, which gradually increases up to 1 year and possibly beyond. For naltrexone, there is evidence that its effects do not persist when it is stopped (O'Malley et al., 1996).

7.10.1. Systematic review

The literature search identified seven studies that assessed the cost-effectiveness of pharmacological agents for the maintenance phase of treatment of alcohol dependence (Annemans et al., 2000; Mortimer & Segal, 2005; Palmer et al., 2000; Rychlik et al., 2003; Schadlich & Brecht, 1998; Slattery et al., 2003; Zarkin et al., 2008). Full references, characteristics and results of all studies included in the economic review are presented in the form of evidence tables in Appendix 19.

Annemans and colleagues (2000) modelled the healthcare costs of acamprosate compared with no treatment in the prevention of alcoholic relapse over a 24-month time horizon. The patient population started the model following assisted withdrawal in an ambulatory state. Effectiveness data used to populate the model was sourced from several published and unpublished studies. A Belgian health payers' perspective was taken for the analysis. Therefore, only direct medical costs, relating to hospitalisations, psychiatric and GP consultations, and medications, were included in the model. The total expected cost of the acamprosate strategy was €5,255 over the 2-year time horizon compared with €5,783 in the no treatment arm. Therefore, despite the higher drug acquisition costs, acamprosate was shown to be a cost-saving intervention in terms of reduced hospitalisations due to alcohol-related complications. The major limitation of the study was that it was a cost-analysis and did not consider the impact of the interventions on overall clinical effectiveness and patient quality of life. Furthermore the study was from the Belgian health payer's perspective, which may have limited applicability to the UK context.

The study by Mortimer and Segal (2005) conducted a model-based economic analysis of naltrexone plus counselling versus counselling alone amongst detoxified patients with a history of severe alcohol dependence. A lifetime horizon was used for all of the analysis. Clinical effectiveness was measured using QALYs, which were calculated from disability weights derived from a single published source (Stouthard et al., 1997). Clinical-effectiveness data were taken from published studies evaluating interventions targeting heavy drinkers at lower severity levels. These data were used to estimate how people would progress between specific drinking states (problem, moderate or dependent) within the model. The authors did not specify the resource use and cost components included in the model although an Australian health service perspective was adopted for the analysis. The results of the analysis suggested that naltrexone was cost effective in comparison with standard care resulting in an ICER of AUS$12,966.

There are several limitations with the results of the study that reduce their applicability to any UK-based recommendations. Little explanation was given in the article as to how the clinical effectiveness data, which was taken from various sources, was used to inform the health states used in the economic models. The article did not specify the resource use and costs that were included in the analyses although a health perspective was used. The analysis used QALYs as the primary outcome measure, which allows for comparison across interventions, although again there was insufficient description of the utility weights that were applied to the health states within the model.

Palmer and colleagues (2000) modelled the lifetime cost effectiveness of adjunctive acamprosate therapy in conjunction with standard counselling therapy, compared with standard counselling alone, in people with alcohol dependence. The study population comprised men of an average age of 41 years who had been withdrawn from alcohol and had a mixture of alcohol-related complications. The model allowed people to progress through various health states associated with important alcohol-related complications including liver disease, gastrointestinal disease, alcoholic cardiomyopathy and other complications. Clinical effectiveness data was sourced from 28 published studies that were not formally meta-analysed and authors' assumptions. The outcome measure used for the economic analysis was the number of life years gained with adjunctive acamprosate over standard therapy. The perspective of the cost analysis was from German third-party payers. Costs, again reported in Deutschmarks (DMs), included those associated with drug acquisition and treatment of alcohol-related complications.

The results of the cost-effectiveness analysis showed that adjunctive acamprosate therapy was the dominant treatment strategy, resulting in lower costs (DM48,245 versus DM49,907) and greater benefits (15.9 versus 14.6 life years gained) in comparison with standard therapy. Interpretation of the study results is subject to a number of methodological limitations. First, a formal literature review was not undertaken in order to derive effectiveness estimates and no formal meta-analysis of summary data was performed, with the authors using data from studies selectively. Second, cost items used in the analysis were not reported adequately and unit costs and resources were not reported separately. Finally, as noted by the authors, no consideration was given to quality of life in measuring the relative effectiveness of the treatments considered.

The objective of the study by Rychlik and colleagues (2003) was to compare the healthcare costs over 1 year of psychosocial rehabilitation support either alone or with adjunctive acamprosate treatment. The cost-effectiveness analysis was conducted alongside a prospective cohort study across 480 centres in the German primary care setting. Patients who fulfilled DSM-IV criteria for alcohol dependence were included in the study. The primary measure of clinical effectiveness in the study was abstinence rates after 1 year. The perspective of the study was from German health insurance. Direct healthcare costs included medications, hospitalisations, outpatient care and diagnostic and laboratory tests. Total 1-year costs were analysed according to both per-protocol and intention-to-treat analysis due to the expected patient attrition. Within both analyses, the adjunctive acamprosate treatment resulted in lower costs (€1225 to €1254 versus €1543 to €1592) and higher rates of abstinence (32 to 23% versus 20 to 21%) in comparison with no adjunctive treatment. The results of the economic analysis may be of limited applicability to the UK setting due to the cohort study design, the study setting and the short time horizon, as well as the effectiveness measure used.

The study by Schadlich and Brecht (1998) was a model-based cost-effectiveness analysis comparing adjunctive acamprosate therapy (in addition to standard care) with standard care (placebo and counselling or psychotherapy) for alcohol dependence. The population was defined as being alcohol dependent and abstinent from alcohol for up to 28 days prior to entering the study. Data were derived from a single double-blind RCT across 12 outpatient centres in Germany. The primary health outcome measure was the percentage of people remaining abstinent at the end of 48 weeks of medication-free follow-up. Transition probabilities to target events within the model were elicited from clinical expert opinion. The outcome measures used in the cost-effectiveness analysis were cases of target events avoided including cases of alcoholic psychoses, alcohol dependence syndrome, acute alcoholic hepatitis and alcoholic liver cirrhosis. A German healthcare-system perspective was taken for the cost analysis. Costs (reported in Deutschmarks) included in the model related to hospital treatment, acamprosate acquisition and patient rehabilitation for target events.

The ICER of acamprosate versus standard care was −DM2,602 (range: −DM406 to DM8,830) per additional abstinent patient, thus resulting in a net saving in terms of direct medical costs. The results of the study, based on a single RCT in Germany, are of limited relevance to the UK setting. No attempt was made to translate the intermediate outcome of abstinence into final outcomes such as QALYs, which are of greater relevance to decision makers. Another limitation of the study was that resource use quantities were not reported separately from the costs. Costing was also performed retrospectively and was not based on the same patient sample used in the effectiveness analysis, thus limiting the study's internal validity.

The study by Slattery and colleagues (2003) developed an economic model to assess the cost effectiveness of acamprosate, naltrexone and disulfiram compared with standard care within the Scottish health service setting. The populations examined were 45-year-old men and women with a diagnosis of alcohol dependence. The outcome measures used in the economic model were the number of patients who have abstained and number of deaths averted. The clinical effectiveness data was based on a methodologically diverse selection of trials that were not described within the study. Resource use involved in the pharmacological interventions included drug acquisition as well as outpatient and GP consultations. Costs were applied from Scottish health service estimates. Other healthcare costs included in the model were those associated with alcohol-related disease endpoints such as stroke, cancer, cirrhosis and alcohol-related psychoses. Costs were applied according to inpatient length of stay taken from Scottish medical records.

The total costs of pharmacological treatments and any disease endpoints for a hypothetical cohort of 1000 patients were compared with standard care over a 20-year time horizon, to determine any net healthcare cost savings. Acamprosate resulted in net savings of £68,928 whilst naltrexone and disulfiram resulted in net economic costs of £83,432 and £153,189, respectively, in comparison with standard care amongst a hypothetical cohort of 1000 patients. Whilst the results of the study, based on a hypothetical cohort of patients within the Scottish health service, may be applicable to a UK setting there are several problematic methodological issues with the study. First, the sources of the effectiveness data used in the model were not explicitly described by the authors, who suggested that the data was taken from a methodologically diverse selection of trials, thus suggesting a high level of heterogeneity. Second, no attempt was made to translate intermediate clinical endpoints such as abstinence rates into QALYs, which are useful to decision makers when assessing the comparative cost effectiveness of healthcare interventions.

Zarkin and colleagues (2008) evaluated the cost effectiveness of the COMBINE study (Anton et al., 2006) interventions after 16 weeks of treatment. In the study, people with a primary diagnosis of alcohol dependence from across 11 US study sites were randomised to nine intervention groups. In eight groups all participants received medical management and were randomised to receive naltrexone, acamprosate, combination drugs (naltrexone and acamprosate) or placebo or a combined behavioural intervention in addition to naltrexone, acamprosate, combination drugs or placebo. The ninth treatment group received combined behavioural intervention only (without medical management). Three clinical measures were used in the economic analysis: PDA, avoidance of heavy drinking and achieving a good clinical outcome (abstinent, or moderate drinking without problems). Costs were analysed from the treatment provider perspective. Resource use included medication, staff time and laboratory tests.

Each intervention was ranked in increasing order of mean total cost for each of the three effectiveness measures. Only three interventions – medical management and placebo, medical management and naltrexone and naltrexone and acamprosate – were included in the final comparative analysis. This is because the other six interventions were dominated (resulting in higher mean costs but lower effectiveness) by the aforementioned interventions. The ICERs for the comparison of medical management and naltrexone versus medical management and placebo were US$42 per percentage increase in days abstinent, U$2,847 per person avoiding heavy drinking and US$1,690 per person achieving a good clinical outcome. The ICERs for the comparison of naltrexone and acamprosate versus medical management and naltrexone were US$664 per percentage point increase in days abstinent, US$8,095 per person avoiding heavy drinking and US$7,543 per person achieving a good clinical outcome.

This study is the only cost-effectiveness study reviewed that considered combinations of pharmacological and psychosocial interventions. However, there are a number of limitations when interpreting the results of the study. The cost analysis relied on the trial investigators' judgement of best clinical practice, which specifically relates to the US healthcare system and may not be generalisable to the UK health service. Interpretation of the results is further reduced by the short time horizon and the choice of outcome measures used in the analysis. Translation of intermediate outcomes such as rates of abstinence or moderate drinking into final outcomes such as QALYs would also be more helpful to decision makers.

7.10.2. Summary of existing health economic evidence

Of the seven cost-effectiveness studies identified in the literature, four compared acamprosate with standard care (Annemans et al., 2000; Palmer et al., 2000; Rychlik et al., 2003; Schadlich & Brecht, 1998), one compared naltrexone with standard care (Mortimer & Segal, 2005) and one study compared naltrexone, acamprosate and disulfiram with standard care (Slattery et al., 2003). The remaining study compared nine possible treatment combinations including naltrexone, acamprosate, combination drugs (naltrexone and acamprosate) or placebo either alone or in combination with a combined behavioural intervention. Only one study was UK-based (Zarkin et al., 2008) whilst the other studies were based in Belgian, German or US populations. Nearly all of the studies were model-based economic analyses except for Rychlik and colleagues (2003), which was a cohort-based study, and Zarkin and colleagues (2008), which was based on the COMBINE RCT (ANTON2006). Within nearly all of the studies, pharmacological treatments were provided as adjunctive treatments to standard care, which differed across the studies considered.

In summary, the results suggested that acamprosate was either cost saving or the dominant treatment strategy (offering better outcomes at lower costs) in comparison with standard care. Naltrexone plus counselling was cost effective compared with counselling alone in people with a history of severe alcohol dependence (Mortimer & Segal, 2005). The one UK study showed that acamprosate resulted in significant healthcare cost savings whilst naltrexone and disulfiram resulted in significant net economic costs in comparison with standard care (Slattery et al., 2003). Zarkin and colleagues (2008) showed that naltrexone in addition to medical management and combination therapy (naltrexone plus acamprosate) were cost effective over a 16-week period.

Providing an adequate summary of the health economic evidence presented here is difficult, due to the differences across the studies in terms of the comparator treatments considered (that is, definitions of ‘standard care’ differed across studies), study populations, costs and outcomes considered, and other methodological differences. Overall, the evidence reviewed is insufficient to support a single pharmacological treatment over any other.

7.11.1. Introduction

The systematic search of the economic literature identified a number of studies assessing the relative cost effectiveness of pharmacological treatments, either alone or as an adjunct to psychological therapy, in the prevention of relapse in people who are in recovery from alcohol dependence. The studies varied in terms of both methodological quality and applicability to the UK context. The results overall were inconsistent and did not support one pharmacological therapy over another. Therefore, an economic model was developed to answer this question. The objective of the economic model was to explore the relative cost effectiveness of pharmacological treatments for the prevention of relapse in people who are in recovery from alcohol dependence. The aim of the analysis was to reflect current UK clinical practice, using the most relevant and up-to-date information on costs and clinical outcomes. Details on the guideline systematic review of the economic literature on pharmacological interventions for relapse prevention are provided in Section 7.10.1.

7.11.2. Methods

Model structure

A pragmatic decision model was constructed using Microsoft Excel (2007). Within the model a hypothetical cohort of 1000 patients who are in recovery from alcohol dependence can either relapse to heavy drinking (defined as at least five drinks for males; at least four drinks for females) or remain in recovery during a 12-month period. The structure of the decision tree is presented in Figure 9. The time horizon was chosen to reflect current UK guidance and recommendations, which recommend that patients should be maintained on pharmacological therapy for up to 12 months if patients are responding successfully to treatment. Three treatment groups were considered in the model: (1) acamprosate and standard care; (2) naltrexone and standard care; and (3) standard care alone. Standard care was defined as psychological therapy that patients would receive to prevent relapse to heavy drinking. The psychological therapy would be delivered by a community nurse over the 12-month period.

Figure 9

Schematic of model structure.

Clinical input parameters and overview of methods of evidence synthesis

Clinical input parameters consisted of relapse rates associated with each intervention assessed: that is, naltrexone, acamprosate, or placebo. The economic analysis considered all relevant data reported in the studies included in the respective guideline systematic clinical review. To take all trial information into consideration, network (mixed treatment comparison [MTC]) meta-analytic techniques were employed. Network meta-analysis is a generalisation of standard pair-wise meta-analysis for A versus B trials to data structures that include, for example, A versus B, B versus C and A versus C trials (Lu & Ades, 2004). A basic assumption of network meta-analysis is that direct and indirect evidence estimate the same parameter; in other words, the relative effect between A and B measured directly from an A versus B trial, is the same with the relative effect between A and B estimated indirectly from A versus C and B versus C trials. Network meta-analytic techniques strengthen inference concerning the relative effect of two treatments by including both direct and indirect comparisons between treatments and, at the same time, allow simultaneous inference on all treatments examined in the pair-wise trial comparisons while respecting randomisation (Caldwell et al., 2005; Lu & Ades, 2004). Simultaneous inference on the relative effect of a number of treatments is possible, provided that treatments participate in a single ‘network of evidence’ (that is, every treatment is linked to at least one of the other treatments under assessment through direct or indirect comparisons).

Details on the methods and relapse data utilised in the network meta-analysis that was undertaken to estimate clinical input parameters for the economic analysis are presented in Appendix 15. Table 99 provides the mean probability of relapse (as well as the respective 95% credible intervals) at 1 year of treatment for naltrexone, acamprosate and placebo, as estimated by network meta-analysis. Overall, the results of the network meta-analysis are comparable with those obtained in the pair-wise comparisons, reported as the RR of relapse to heavy drinking at 3, 6 and 12 months. These comparisons showed small but significant differences favouring naltrexone and acamprosate versus placebo, but no significant differences between naltrexone and acamprosate over 12 months. The results of the network meta-analysis suggest that acamprosate had the highest probability (63%) of being the best treatment at reducing the probability of relapse over 12 months. However, the wide credible intervals around the mean estimates are indicative of the uncertainty surrounding these mean estimates.

Table 99

Results of network meta-analysis – probability of relapse at 12 months.

Utility data and estimation of quality-adjusted life years

To express outcomes in the form of QALYs, the health states of the economic model were linked to appropriate utility scores. Utility scores represent the health-related quality of life associated with specific health states on a scale from 0 (death) to 1 (perfect health). They are estimated using preference-based measures that capture people's preferences for the health states under consideration. The systematic search of the literature identified one study that reported utility scores for specific health states associated with alcohol-related disorders (Kraemer et al., 2005).

The study by Kraemer and colleagues (2005) directly measured utility scores for a spectrum of alcohol-related health states using different methods of utility measurement including visual analogue scale, time trade-off (TTO) and standard gamble (SG) techniques. The study was based on a cross-sectional interview of 200 adults recruited from one clinic (n = 100) and one community (n = 100) sample in the US. Study subjects completed computerised versions of the utility rating exercises for their current health and six hypothetical alcohol-related health state scenarios presented in random order. Utility ratings were scaled from 0 to 1 and anchored by death (0) and perfect health (1). Table 100 summarises the mean utility scores for the six alcohol-related health states for the three techniques used. As the results in the table show, for each of the techniques used, utility scores decreased as the severity of alcohol use increased.

Table 100

Mean utility scores for alcohol-related health states and utility measurement technique (adapted from Kraemer et al., 2005).

NICE recommends the EQ-5D as the preferred measure of health-related quality of life in adults for use in cost-utility analyses. NICE also suggests that the measurement of changes in health-related quality of life should be reported directly from people with the condition examined, and the valuation of health states should be based on public preferences elicited using a choice-based method, such as TTO or SG, in a representative sample of the UK population. At the same time, it is recognised that EQ-5D utility scores may not be available or may be inappropriate for the condition or effects of treatment (NICE, 2008a). The study by Kraemer and colleagues (2005) did not use the EQ-5D questionnaire to estimate utility scores and was based on a US population sample who did not experience the alcohol-related health states they were asked to rate. Furthermore, the patient sample was not randomly selected but conveniently recruited either from clinic waiting rooms or self-selected within the community after responding to an advertisement. The low sample size (n = 200) also limits the results of the study, contributing to the uncertainty around the mean utility score estimates. However, this was the only study identified in the literature review that applied utility scores to specific alcohol-related health states using appropriate measurement techniques (SG or TTO) as recommended by NICE.

The two health states of interest in the economic model were: (a) in recovery from alcohol dependence; and (b) relapse to alcohol dependence. For these health states, the utility scores for the ‘alcohol dependence’ and ‘alcohol dependence, in recovery’ health states were chosen from Kraemer and colleagues (2005). In the base-case analysis the TTO utility scores were used whilst the SG utility scores were used in the sensitivity analysis.

Drug acquisition costs

Drug acquisition costs were taken from the latest edition of the BNF (British Medical Association & The Royal Pharmaceutical Society of Great Britain, 2010). The recommended daily dosage for acamprosate was 1,998 mg per day and for naltrexone was 50 mg per day. The drug acquisition costs and monthly costs for both drugs included in the analysis are presented in Table 101.

Table 101

Drug acquisition costs and estimated monthly costs of pharmacological interventions included in the economic model.

Other costs of patient management

Estimates on resource use associated with the psychological intervention, outpatient and primary care and blood laboratory tests were based on the expert opinion of the GDG. It was assumed that patients in all three treatment arms would receive the same individual psychological intervention focused specifically on alcohol misuse (for example, CBT, behavioural therapy or social network and environment-based therapy) delivered by a practice nurse. It was assumed that each patient would receive one session per month or 12 sessions over the entire 12-month period if they did not relapse. It was assumed that patients in the three treatment groups would all require one initial 30-minute outpatient consultation with a consultant psychiatrist prior to starting treatment. Patients receiving adjunctive pharmacological interventions would require an additional two visits as part of their medical supervision. The second visit would be a 15-minute outpatient visit with a consultant psychiatrist and the third would be a GP consultation at the end of the 12-month period. At all three visits, it was assumed that patients would require blood tests (liver function test, and urea and electrolytes) to monitor for any potential hepatotoxic effects. It was assumed that patients receiving standard care would not require any further monitoring. Further details of resource use and costs associated with patient management are provided in Table 102.

Table 102

Resource use over 12 months and unit costs associated with patient management for people in recovery from alcohol dependence.

7.11.3. Results of economic modelling

Deterministic analysis

Table 103 provides mean costs and QALYs per 1000 people for the interventions under consideration as well as the results of the incremental analyses. The interventions were ranked from highest to lowest in terms of the number of QALYs gained over 12 months. Acamprosate was associated with the highest costs and the highest number of QALYs whilst standard care was associated with the lowest costs and the lowest number of QALYs. The ICER of acamprosate was £5,043 per QALY versus standard care and £1,899 per QALY versus naltrexone. The ICER of naltrexone versus standard care was £5,395 per QALY, meaning that naltrexone was extendedly dominated. All ICERs lie well below the cost-effectiveness threshold of £20,000 to £30,000 per QALY currently set by NICE (NICE, 2008b).

Table 103

12-month mean costs and QALYs per 1000 patients and ICERs for pharmacological interventions used for relapse prevention in people in recovery from alcohol dependency.

Table 104 shows that the cost-effectiveness results were fairly robust under the scenarios explored in the sensitivity analysis. The ICER of acamprosate versus standard care reached £10,000 per QALY, while naltrexone was extendedly dominated when utility scores estimated from the SG technique were used. The ICER of naltrexone versus standard care was approximately £11,000 per QALY. When the intensity of patient monitoring was increased, the ICER of acamprosate versus naltrexone was £13,323 per QALY and of naltrexone versus standard care was £10,789 per QALY. When the monthly cost of relapse was £0, the ICER of acamprosate compared with standard care increased to approximately £10,000, and naltrexone was extendedly dominated (with an ICER versus standard care of £11,000 per QALY). However, when the monthly cost of relapse was doubled to £300, both pharmacological interventions dominated standard care, resulting in lower costs and higher QALYs over 12 months; acamprosate dominated naltrexone under this scenario. It must be noted that under all scenarios explored in one-way sensitivity analysis, the ICERs of both drugs versus standard care were below the NICE lower cost-effectiveness threshold of £20,000 per QALY.

Table 104

Results of deterministic sensitivity analyses.

Probabilistic analysis

Results of the probabilistic analysis were very similar to those of the deterministic analysis – acamprosate was associated with the highest costs and QALYs and standard care was associated with the lowest costs and QALYs. ICERs were very similar to those calculated in the deterministic analysis. Probabilistic analysis demonstrated that standard care had the highest probability of being cost effective, as well as the highest NMB up to a willingness-to-pay level of £6,000 per QALY. Above this figure, acamprosate had the highest probability of being the most cost-effective treatment option and the highest NMB. Using the current threshold of £20,000 to £30,000 per QALY set by NICE, the probability of acamprosate or naltrexone being the most cost-effective treatment option were approximately 52 to 53% and 44 to 45%, respectively.

Figure 10 shows the cost-effectiveness acceptability curves generated for the three interventions considered whilst Table 105 shows the NMB and probability of each intervention being cost effective at various levels of WTP per QALY gained. Figure 11 shows the CEAF for the three options assessed. It can be seen that acamprosate provides the highest average NMB at any WTP above £10,000 per QALY.

Figure 10

CEACs for three treatment options over 12 months.

Table 105

Net monetary benefit and probability of each intervention being cost effective at various levels of WTP per QALY gained.

Figure 11

CEAF for three treatment options over 12 months.

7.11.4. Discussion of economic model

The results of the economic analysis suggest that acamprosate is potentially the most cost-effective pharmacological treatment, when used as an adjunct to a psychological intervention, for relapse prevention in people in recovery from alcohol dependence. Given the uncertainty characterising the model input parameters, in particular the 12-month probability of relapse, the probability of either acamprosate or naltrexone being the most cost-effective option at the NICE cost-effectiveness threshold of £20,000 was 52% and 44%, respectively.

A major limitation of the analysis was the exclusion of disulfiram, a pharmacological intervention that is currently licensed in the UK for the treatment of relapse prevention in people in recovery from alcohol dependence. Only one open-label RCT was identified in the systematic review that compared disulfiram with naltrexone, which used relapse to alcohol dependence as an outcome measure (De Sousa et al., 2004). The GDG decided it would be inappropriate to include the results of an open-label study in the network meta-analysis. Another limitation was that the RCTs included in the MTC meta-analysis used different definitions of relapse and different baseline psychological therapies, a factor that may undermine the pooled relapse rates considered in both the pair-wise and network meta-analyses.

Another possible limitation of the analysis is the relatively short time horizon of the economic model, although this reflected the time horizon of the RCTs that were included in the systematic review and meta-analyses. Indeed, the majority of the trials included in the network meta-analysis measured rates of relapse up to 3 and 6 months with only three studies actually measuring rates of relapse up to 12 months' follow-up. Ideally, a more comprehensive economic analysis would attempt to model the long-term cost-effectiveness of the three interventions, in terms of exploring the longer-term impact of relapse prevention on future alcohol-related complications and survival. Earlier economic models have attempted to explore the longer-term cost-effectiveness of adjunctive pharmacological therapies over the patients' lifetime, by translating relapse to alcohol dependency into alcohol-related diseases including liver disease, cardiomyopathy, pancreatitis and alcoholic psychoses as well as alcohol-related mortality (Schadlich & Brecht, 1998; Palmer et al., 2000). However, these models required assumptions, often based on limited clinical evidence, about the longer-term prognosis of patients who relapsed to alcohol dependence.

The results of the network meta-analysis are undermined by the heterogeneity between studies in terms of the range of underlying psychological interventions and the study time horizons. All studies included in the analysis were RCTs of pharmacological treatment or placebo as an adjunct to psychological interventions for the prevention of relapse. The RCTs included a wide range of psychological interventions including coping skills, counselling, brief combined behavioural intervention, MET and group therapies. The results of the meta-analyses presented here, including the network meta-analysis, assume that any differences in effectiveness are entirely explained by the adjunctive pharmacological interventions as opposed to the underlying psychological interventions. Whilst the economic model adopted a 12-month time horizon, the majority of the RCTs included in the network meta-analysis, were either of 3 months' (n = 20) or 6 months' (n = 9) duration. The analysis attempted to extrapolate the majority of this data over a 12-month period. If the effectiveness of pharmacological interventions for relapse prevention actually declines over 12 months, the analysis may have over-estimated the cost effectiveness of acamprosate or naltrexone as an adjunct to psychological interventions.

The analysis was based on the perspective of the NHS and personal social services, as recommended by NICE. Costs associated with the interventions considered were estimated from national sources and GDG expert opinion. The results suggested that drug acquisition costs did not determine the relative cost effectiveness of the three interventions. However, the results of the sensitivity analyses suggest that results may be sensitive to the intensity of patient monitoring (for example, specialist visits and blood tests), which were estimated from GDG expert opinion and also the monthly costs of relapse to heavy drinking. However, within both sensitivity analyses, the ICERs for acamprosate and naltrexone were still well below the current NICE cost-effectiveness threshold.

7.11.5. Conclusions

The economic analysis undertaken for this guideline showed that both acamprosate and naltrexone may be potentially cost-effective pharmacological interventions for the prevention of relapse among people in recovery from alcohol dependence. The probability of either drug being the most cost-effective option at the NICE cost-effectiveness threshold of £20,000 was 52% and 44% respectively. However, further research is necessary to establish whether these pharmacological interventions are clinically and cost effective in the longer term, in terms of preventing future alcohol-related diseases. Further clinical data, preferably based on appropriately controlled trials, is also needed to establish the clinical efficacy of disulfiram for relapse prevention.

7.12.1. Clinical review protocol

A systematic search of the literature was conducted to evaluate pharmacological interventions for relapse prevention for children and young people. See Section 7.2 for the aim of the review and the review questions. The clinical review protocol for this section can be found in Section 7.3.

7.12.2. Studies considered

Unlike the adult literature, the GDG was able to identify only three small pilot RCTs in this area for children and young people (Niederhofer & Staffen, 2003a; Niederhofer & Staffen, 2003b; Niederhofer et al., 2003). Due to the limited number of studies and the heterogeneous nature of the outcomes, a narrative synthesis of the available literature was conducted by the review team in order to assess the efficacy of pharmacological interventions for children and young people.

7.12.3. Evidence summary

Niederhofer and Staffen (2003a) conducted a double-blind, placebo-controlled study with 26 participants with a DSM–IV diagnosis of chronic or episodic alcohol dependence. Participants ranged in age from 16 to 19 years. The participants were randomly allocated to treatment with acamprosate (1,332 mg daily) or placebo for 90 days. Participants were assessed at start of treatment, and at 30 and 90 days. Results revealed that the acamprosate group had a significantly higher proportion of days abstinent throughout the 90 days of treatment (p < 0.001), as well as a higher duration of mean cumulative abstinence (p < 0.01). There were no significant differences between the two groups with regards to side effects, and diarrhoea was the only reported side effect.

Niederhofer and colleagues (2003) assessed naltrexone compared with placebo in a double-blind, placebo-controlled study with 30 participants with a DSM–IV diagnosis of chronic or episodic alcohol dependence. Participants ranged in age from 15 to 19 years. All participants received 50 mg of naltrexone daily and were assessed at the start of treatment, and then at 30 and 90 days. At the 90-day assessment point, 60 of the 90 participants completed treatment. Results revealed that the naltrexone group remained abstinent longer during 90 days of treatment (p < 0.01) and had a longer duration of mean cumulative abstinence (69.8 days) than the placebo arm (22.8 days) (p < 0.01). It must be noted that it is not clear from the paper how many participants were randomised to each group; therefore, the findings should be interpreted with caution.

Lastly, Niederhofer and Staffen (2003b) compared disulfiram and placebo in a double-blind, placebo-controlled trial with 26 adolescent participants with DSM–IV chronic or episodic alcohol dependence. Participants ranged in age from 16 to 19 years. Participants received 200 mg of disulfiram daily and were assessed at the start of treatment, and then at 30 and 90 days. Twenty-six of the 49 participants recruited completed the 90 days of double-blind treatment. Results indicated that on day 90 of treatment, two of the placebo-treated patients compared with seven disulfiram-treated patients had been continuously abstinent (p = 0.0063). Additionally, the duration of mean cumulative abstinence was significantly higher in the disulfiram group (68.5 days) than in the placebo group (29.7 days) (p = 0.012).

7.12.4. Clinical evidence summary

Taken together, there is little evidence based on the results of three small pilot RCTs to assess the efficacy of pharmacological interventions in young people. The three studies do, however, provide some preliminary data indicating positive responses in young people to pharmacological interventions when compared with placebo. Due to the poor methodological quality of these studies, however, results should be interpreted with very considerable caution. As a result, recommendations for young people have to rely on extrapolations from the dataset for adults.

Acamprosate

Acamprosate is a well-tolerated medication with minimal side effects, contraindications or cautions associated with its use. The most common side effect is diarrhoea with abdominal pain, nausea, vomiting and pruritus also described. Its contraindications include pregnancy and breastfeeding, renal insufficiency (serum creatinine > 120 micromoles per litre) and severe hepatic failure (Childs-Pugh Classification C). There appear to be no interactions of clinical significance with alcohol.

Naltrexone

Naltrexone is also generally a well-tolerated medication with most trials reporting side effects similar to those reported with placebo or other drugs such as disulfiram or acamprosate. The most common side effects reported for naltrexone included nausea, headache, abdominal pain, reduced appetite and tiredness. However, in some of these studies, 100 mg per day rather than 50 mg per day was used. Nausea has been reported more commonly at the start, particularly in female and lighter drinkers, which can be minimised by starting at 25 mg per day.

Because it is an opioid antagonist, naltrexone cannot be used in people using opioid agonist drugs for analgesia. In addition, if analgesia is required in an emergency, nonopioid medication will be required because naltrexone blockade will last for 48 to 72 hours after taking the last tablet. It is therefore helpful if people carry a card stating that they are taking naltrexone in case of such an emergency. If future analgesia is likely, for example in planned surgery, naltrexone is also therefore not ideal.

Hepatotoxicity was reported in association with use of naltrexone to treat obesity when high doses (> 300 mg per day) were used. Reviews of available data suggest and current US guidelines recommend that hepatic toxicity is very unlikely to occur with doses of 50 mg per day and that continued alcohol use is more likely than naltrexone to cause liver damage (O'Malley et al., 1993). Nevertheless, naltrexone should not be used in those with acute liver failure and caution is suggested when serum aminotransferases are four to five times above normal (Anton et al., 2006; Kleber, 1985). However, naltrexone has been used in people with chronic hepatitis B and/or hepatitis C and no significant difference in liver function test results with naltrexone at the recommended doses has been reported (Lozano Polo et al., 1997).

There is no consistent advice or evidence about monitoring of liver function tests for adverse effects on hepatic function. It is therefore important that the patient understands about the risk of hepatotoxicity and to stop taking naltrexone and promptly seek medical attention if they have any concerns about side effects or start to feel unwell. Deterioration in liver function tests or signs of liver failure have not been widely reported and increases generally normalise on stopping naltrexone. Before ascribing any increases to naltrexone, review other possible contributors such as other medications (prescribed and over-the-counter), complementary treatments and resumption of drinking.

Disulfiram

Given the potential seriousness of the disulfiram–alcohol interaction in addition to the potential adverse effects of disulfiram alone, prescribing needs due care and consideration. Patients must be warned about and have capacity to understand the disulfiram-alcohol reaction and be made aware of the presence of alcohol in foodstuffs, perfumes, aerosols and so on. In addition, they should not have consumed alcohol for at least 24 hours before starting disulfiram and should also be warned that a reaction with alcohol may be experienced for up to 7 days after their last tablet. The alcohol challenge test is no longer recommended (see the SPC and BNF). Fatal disulfiram–alcohol reactions have occurred with high doses of the drug (more than 1 g per day) and were associated with cardiovascular complications such as hypotension or corrected QT interval on the electrocardiograph (Chick, 1999; Kristenson, 1995). With the lower doses now prescribed, more severe reactions after consuming alcohol are less likely to be seen (Malcolm et al., 2008). Indeed, a survey of people taking disulfiram found that for some an interaction only occurred when taking 800 to 1500 mg per day (Brewer, 1984).

The SPC or BNF lists several significant medical and psychiatric contraindications to its use, including cardiovascular problems, severe personality disorder, suicidal risk or psychosis, and contraindications to pregnancy and breast feeding. Caution is also advised in the presence of renal failure, hepatic or respiratory disease, diabetes mellitus and epilepsy. Nevertheless, against this background there is some evidence of its prescribing in a broad range of conditions including possible contraindications such as psychotic disorders or cocaine dependence or people taking methadone with no reports of significant adverse effects (Pani et al., 2010; Petrakis et al., 2005; Petrakis et al., 2000).

Concerning the side effects of disulfiram alone, there have been many fewer trials compared with acamprosate or naltrexone and some are older, hence descriptions may be less comprehensive. Where reported, side effects and adverse events or reactions experienced include drowsiness, fatigue, abdominal pain, nausea and diarrhoea. Psychiatric problems such as dysphoria or psychosis were reported in some studies but the incidence was low. In newer trials comparing disulfiram with acamprosate or naltrexone, the reporting of side effects or adverse events is not dramatically different between the active drugs or placebo. Neuropathy has been reported by some but not all studies, with onset commonly described over months to a year, although onset within days has also been described (see Chick, 1999). From the Danish database, the estimated rate of neuropathy was 1 in 15,000 patient years (Poulsen et al., 1992), although De Sousa and colleagues (2005) reported that three of the 50 (6%) people taking disulfiram in their trial dropped out due to neuropathy.

Use of disulfiram may be associated with the development of an acute hepatitis, which can be fatal. The nature and exact incidence or prevalence of hepatotoxicity is unclear; however, it appears rare with, for example, 30 reports of hepatitis in the previous 40 years (Chick, 1999) and 11 fatal liver reactions in 22 years (1968 to 1991). Based on estimates of number of people taking disulfiram, the estimated the risk of dying from hepatotoxicity caused by disulfiram was 1:30,000 people per year. However, some people received disulfiram for nickel sensitivity, for which they were reportedly at greater risk of hepatitis than those receiving disulfiram for alcohol dependence. Hepatotoxicity at 250 mg per day after 13 days has been described, although a review found disulfiram-related hepatitis starting 16 to 120 days later; however, in one case jaundice appeared within 5 days after taking 1.5 to 2 g per day (that is, up to ten times above the recommended dose) (Chick, 1999). Given the seriousness of hepatitis, a role for monitoring of liver function has been suggested but there is limited evidence to inform guidance. It is therefore important that the patient understands about the risk of hepatotoxicity, and to stop taking disulfiram and promptly seek medical attention if they have any concerns about side effects or start to feel unwell.

Psychiatric complications such as psychosis or confusional states are potentially serious side effects or adverse events and are more likely at higher doses (more than 500 mg per day; Chick, 1999). The Danish and WHO databases report that 4% and 13%, respectively, of all adverse effects of disulfiram were psychiatric (Poulsen et al., 1992). One clinical trial reported over 1 year in over 600 people reported no difference in psychiatric complications between those treated with disulfiram 250 mg per day, or disulfiram 1 mg per day or placebo with the incidence in disulfiram groups at 2.4% (Branchey et al., 1987). Nevertheless, in recent trials disulfiram has been used in people with a variety of psychiatric comorbidities including depression, psychosis or schizophrenia without apparent psychiatric adverse events (see Chick, 1999; Petrakis et al., 2005 and 2006). The rate and quality of adverse events with cocaine and disulfiram are similar to those seen in studies of alcohol dependence (Pettinati, 2005; Carroll et al., 1998). Disulfiram has also been used in people maintained on methadone without reported serious adverse reactions (Ling et al., 1983).

The reader is directed to two comprehensive reviews regarding the safety of disulfiram by Chick (1999) and Malcolm and colleagues (2008).

7.14.1. Aim of review

The aim of this review was to assess the effectiveness of pharmacological interventions not licensed in the UK for the treatment of alcohol misuse. The GDG advised the review team that a large range of pharmacological interventions have been evaluated, often in single trials, with heterogeneous outcomes. Therefore, a narrative synthesis of RCTs was conducted, with no attempt to use meta-analysis. Information about the databases searched and the inclusion/exclusion criteria used for this section of the guideline can be found in Appendix 16e.

7.14.2. Studies considered

A total of nine trials met inclusion criteria for this section of the guideline. There were two trials of extended release injectable naltrexone, two trials of nalmefene, three trials of SSRIs, one current trial of baclofen (which has indicated preliminary findings), and six trials of anticonvulsants (two of topiramate, three of gabapentin and one of pregabalin).

7.14.3. Extended release injectable naltrexone

In addition to oral naltrexone an injectable formulation is available, which has an extended half-life and can overcome poor compliance.

In the US, naltrexone is also available in a once-monthly extended-release injectable formulation (380 mg) and has been used by some in the UK. Two RCTs have been published regarding its efficacy and safety. Kranzler and colleagues (2004) studied a depot formulation in patients who were still drinking but wanted to stop and showed no efficacy on the primary outcome of reduced heavy drinking days. A longer time until first drink and a higher rate of abstinence were reported. The second study compared the 380 mg injectable formulation with one containing 190 mg over 6 months in people still drinking and found reduced heavy drinking in all groups, with the greatest reduction in the higher dose of naltrexone (Garbutt et al., 2005). In addition, greater efficacy was seen in males and in those who had been sober for a week before their injection. A post-hoc analysis revealed that naltrexone reduced alcohol consumption during holiday periods in the US, generally a time of great risk of relapse (Lapham et al., 2009).

Side effects or adverse effects of the extended injectable formulation are reported as similar to oral naltrexone and include abdominal pain, nausea, anorexia and dizziness, although the hepatic safety profile appears similar to placebo (Lucey et al., 2008). However, a greater number of injection site reactions with naltrexone have been reported, which may need medical attention and be due to poor injection technique (Garbutt et al., 2009).

7.14.4. Nalmefene

Like naltrexone, nalmefene is an opioid antagonist but with some kappa partial agonist activity or inverse agonist activity. It was initially proposed as a treatment for alcohol dependence because it has a longer half-life and was thought to have less risk of hepatotoxicity then naltrexone. The first RCT in alcohol dependence reported significantly fewer relapses with nalmefene (20 mg or 80 mg per day; Mason et al., 1999). However, a second multisite RCT comparing 5 mg per day, 20 mg per day with 40 mg per day and placebo reported no efficacy for nalmefene (Anton et al., 2004).

7.14.5. Selective serotonin reuptake inhibitors

The efficacy of SSRIs in treating alcohol misuse without comorbid depression has been studied in three RCTs. They reported that SSRIs may have limited efficacy but importantly may also reduce the impact of psychosocial treatments in improving alcohol misuse in early-onset alcohol dependence. Kranzler and colleagues (1996) reported worse drinking outcomes in early-onset or type B alcohol dependence on fluoxetine compared with placebo. Pettinati and colleagues (2000) found that sertraline had no effect in type B alcohol dependence, whilst improving outcomes in type A. Chick and colleagues (2000a) reported that people with type II alcohol dependence, as defined by Cloninger's Tridimensional Personality Questionnaire, had worse outcomes compared with those on placebo and type I alcohol dependence. Therefore, these three studies suggest that in the absence of a depressive disorder, SSRIs may weaken improvements in alcohol misuse.

One RCT has investigated whether combining naltrexone with sertraline is effective in improving drinking behaviour in native and non-native Alaskan Americans by randomising people to daily naltrexone (50 mg), sertraline (100 mg), naltrexone plus sertraline, or placebo (O'Malley & Nanson, 2002). Naltrexone significantly improved abstinence rates rather than reducing the risk of heavy drinking whilst sertraline had no further benefit.

Overall, given the difficulties in making a diagnosis of depression in such a population and the limited efficacy shown when comorbid depression is present, an SSRI may not be the most appropriate first-line antidepressant to use in alcohol misuse.

7.14.6. Baclofen

According to one trial, baclofen, a GABA-B agonist, increases abstinence rates in people with alcohol-related cirrhosis compared with placebo (Addolorato et al., 2008; 30 mg per day; 12 weeks). It was well tolerated with little contribution to dropouts due to side effects; there were no adverse events reported. There is a large RCT being conducted in the US whose results are yet to be formally published but some of the data have been reported and suggest no efficacy for baclofen (Leggio et al., 2010). Key differences between the studies that may increase the likelihood of efficacy are: goal of abstinence, alcohol dependence requiring assisted withdrawal, and higher anxiety levels.

7.14.7. Anticonvulsants

7.14.8. Clinical summary

A number of pharmacological interventions not licensed for use in the treatment of moderate and severe alcohol dependence were also considered in the review. The evidence indicates that SSRIs do not improve drinking behaviour in non-depressed alcoholics and may worsen drinking-related outcome as well as reduce the efficacy of psychosocial interventions. The initial evidence for the efficacy of injectable naltrexone is encouraging, particularly in those who may not be as compliant with oral naltrexone. However, at the current time there is not enough evidence to support its routine use. The evidence for nalmefene is limited and inconclusive, with only two RCTs with contradictory reported efficacy. There is limited evidence of the efficacy of baclofen as a pharmacological intervention for alcohol dependence. There is limited and contradictory evidence for the efficacy of anticonvulsants in the treatment of alcohol dependence. Gabapentin and pregabalin were found to have some efficacy for reducing drinking. However, for topiramate the evidence was limited and this beneficial effect was not always found after 12 weeks of treatment. Furthermore, the side effects caused by the use of topiramate mean that when it is used the individual most be closely monitored and other pharmacological interventions need to be considered. Overall, there is evidence from a number of trials that some of these pharmacological interventions can reduce drinking and craving and may reduce associated problems with anxiety and insomnia. However, the evidence is limited and there are trials currently underway that will inform their potential role as an adjunct to psychosocial approaches.

7.15.1. Recommendations

7.16.1. Clinical review protocol

A review of pharmacological interventions for individuals who are less severely dependent or non-dependent was conducted. Due to limited literature in this population, a meta-analysis of RCTs could not be conducted. Therefore, a narrative synthesis of the available evidence is presented. Information about the databases searched and the inclusion/exclusion criteria used for this section of the guideline can be found in Appendix 16e. See Table 106 for the clinical review protocol followed for this review.

Table 106

Clinical review protocol for less severely dependent or non-dependent populations.

7.16.2. Evidence summary

In general, psychosocial approaches should be offered to all individuals who misuse alcohol. For those for whom such approaches have not worked or who are mildly dependent, medication may be a treatment option. However the only medication that has been studied in this population is naltrexone. Whilst the majority of participants included in the trials in the meta-analyses were abstinent prior to starting naltrexone, in some of these studies people were still drinking with the aim that naltrexone would help to reduce consumption.

Heinala and colleagues (2001) investigated naltrexone (50 mg) started without assisted withdrawal in people who were dependent and treatment-seeking. They showed that in combination with coping skills but not supportive therapy, naltrexone reduced risk of relapse to heavy drinking but did not improve abstinence or time to first drink. In this study, abstinence was not emphasised as part of coping skills, but was in supportive therapy.

In those less severely dependent and non-dependent, naltrexone (50 mg per day) has been shown to reduce the likelihood of any drinking (Kranzler et al., 2003). Interestingly, if they were taking medication (naltrexone or placebo) in a targeted manner (that is, when anticipating a high-risk situation), greater reductions in heavy drinking days were seen compared with taking medication daily. A follow-up trial confirmed ‘targeted’ naltrexone reduced drinks per day, but only in men (Kranzler et al., 2009). Notably both trials excluded people who had an unsuccessful attempt to reduce their drinking.

Leeman and colleagues (2008) reported in a pilot open study of heavy-drinking young adults (18 to 25 years old) that targeted naltrexone as an adjunct to counselling was well tolerated and reduced drinking, suggesting that this might be a way forward to improve outcomes beyond counselling alone.

Karhuvaara and colleagues (2007) reported that in harmful drinkers experiencing problems controlling their drinking (some may have been dependent), nalmefene (20 mg per day) similarly reduced the number of heavy drinking days.

7.16.3. Clinical summary

The evidence is limited but does support the use of medication (naltrexone) to reduce drinking in non-dependence or mild dependence and does not demonstrate equivalence with psychological interventions for this group. There was no direct evidence for the use of acamprosate in this group.

7.16.4. From evidence to recommendations

The GDG considered that given the limited evidence to support the use of naltrexone to reduce drinking in non-dependence or mild dependence that it should only be used where psychological interventions alone have not been effective. It should be prescribed in conjunction with a psychological intervention.

Although no direct evidence was identified for acamprosate, the GDG considered the equivalence of acamprosate and naltrexone in moderate and severe dependence and decided that it was reasonable to extrapolate from this data and also recommend acamprosate for use in this group.

7.16.5. Recommendation

7.16.5.1.

For harmful drinkers and people with mild alcohol dependence who have not responded to psychological interventions alone, or who have specifically requested a pharmacological intervention, consider offering acamprosate⁵³ or oral naltrexone⁵⁴ in combination with an individual psychological intervention (cognitive behavioural therapies, behavioural therapies or social network and environment-based therapies) or behavioural couples therapy.⁵⁵

7.17.1. Introduction

Individuals presenting for treatment with alcohol misuse may also present with features of other psychiatric disorders, most commonly anxiety or depression. For many, these symptoms will be closely linked to their alcohol misuse and lessen when drinking is reduced or stopped. For this reason, it is important target their alcohol misuse rather than just starting treatment for a comorbid psychiatric disorder. Such comorbidity is associated with a poorer prognosis (Bradizza et al., 2006; Mason & Lehert, 2010; Verheul et al., 1998), increased rates of relapse (Driessen et al., 2001), poorer medication compliance, lower treatment attendance rates and higher rates of self-harm and suicidal behaviours (Martinez-Raga et al., 2000).

There are a variety of treatment approaches for people with comorbid alcohol dependence and a psychiatric disorder but they all emphasise integrated treatment for both disorders. However, this is not always easy to achieve with thresholds for referral to ‘addiction services’ and ‘psychiatric services’ differing and a lack of dedicated dual-disorder services. In addition, addiction services vary in their psychiatric expertise. Provision varies considerably across the UK despite initiatives (Mental Health Policy-DH, 2002). A NICE guideline covers psychosis and co-existing substance misuse (NICE, 2011a).

Psychological treatment approaches aimed at addressing Axis 1 and Axis 2 disorders have been increasingly developed, but in many cases alcohol dependence remains a diagnosis of exclusion even though for many the comorbid psychopathology has preceded the diagnosis of alcohol dependence. On the basis of this, one might question whether or not relapse rates could be influenced were treatment for comorbid disorders provided at the same time as treatments for alcohol dependence.

7.17.2. Clinical review protocol

The review of pharmacological interventions where there is significant comorbidity is considered in this section. A systematic search and the expertise of the GDG were used to identify RCTs or meta-analyses of medication in non-psychotic psychiatric disorders. A further synthesis of the data was not undertaken because, apart from in depression, the number, nature and quality of the studies did not permit this. Two meta-analyses of treating comorbidity of alcohol dependence and depression were drawn on. The expertise of the GDG was used to focus on key trials of relevance to current practice in the UK. Information about the databases searched and the inclusion/exclusion criteria used for this section of the guideline can be found in Appendix 16e. A summary of the available evidence is described below.

7.17.3. Alcohol misuse comorbid with a psychiatric disorder

This section considers two approaches for using pharmacotherapy and psychological interventions. First, its use for treating the alcohol misuse in the context of a nonpsychotic psychiatric disorder, and second, for treating the comorbid psychiatric disorder.

Pharmacological interventions

There are limited studies of disulfiram, acamprosate or naltrexone in people with a psychiatric disorder and alcohol dependence. The largest RCT assessed the efficacy and safety of disulfiram and naltrexone in 254 people who misused alcohol and had an Axis I psychiatric disorder (Petrakis et al., 2005). It was a heterogenous group with some individuals having more than one diagnosis. Individuals were randomised to naltrexone (50 mg per day) or placebo (double blind), but openly randomised to disulfiram (250 mg per day) or nothing, resulting in four groups: naltrexone alone, placebo alone, naltrexone and disulfiram, placebo and disulfiram. There was no overall advantage of one medication over the other and no advantage of the combination of both medications over placebo. However, the abstinence rate is very high at 77%.

Table 107

Clinical review protocol for pharmacological interventions in the present of comorbidities.

A series of secondary analyses were then conducted to compare people with and without particular Axis 1 disorders within the group. In those with PTSD (37%) compared with those without (63%), either naltrexone or disulfiram alone or together improved alcohol outcomes (Petrakis et al., 2006). PTSD symptoms also improved, with those in the disulfiram group showing the greatest improvement. Those with PTSD were more likely to report gastrointestinal, emotional or neurological side effects. By comparison, the presence or absence of current depression did not influence outcomes (Petrakis et al., 2007).

In people with depression and alcohol dependence, Pettinati and colleagues (2010) reported that the combination of sertraline and naltrexone resulted in better abstinence rates than with use of either medication alone or placebo (23.8%; c2 = 12.9, degrees of freedom = 1, p = 0.001). Notably, there was no difference between the groups in improvements in depressive symptoms, although a trend was reported favouring the combination (83% versus 58%; c2 = 6.1, degrees of freedom = 1, p = 0.014).

7.17.4. Treatment of the comorbid psychiatric disorder

This section focuses on the pharmacological and combined pharmacological and psychological treatment of comorbid disorders. The issue of psychological interventions alone for harmful or dependent alcohol misuse has been considered in Chapter 6.

7.17.5. Comorbid alcohol and drug misuse

This section covers pharmacotherapy of comorbidities where it either plays a significant role in management, for example opioid dependence, or where pharmacotherapy has not been shown to be generally efficacious (for example, cocaine). It does not cover comorbidity with drugs of misuse where psychosocial approaches are preferable and pharmacotherapy does not play a significant role (for example, cannabis, ecstasy and ketamine).

7.17.6. Clinical evidence summary

Whilst comorbidity with a psychiatric disorder or another substance is common, there were few studies investigating pharmacological treatments for these problems. Some studies were identified but were and therefore diagnostic criteria differed from those undertaken more recently; a significant proportion were of poor quality with small numbers.

In the RCTs that included people with alcohol dependence and a variety of psychiatric disorders, no benefit of medication (naltrexone, disulfiram or combination) on improving alcohol consumption was found. However, the abstinence rate was much higher than would normally be seen in routine clinical practice. Secondary analyses reported no advantage of medication in improving alcohol consumption when comparing those currently depressed versus non-depressed, but did show a beneficial effect in those with PTSD compared with those without. This emphasises the importance of treatment targeted at alcohol misuse rather than hoping an antidepressant will reduce drinking by improving mood. Whilst there were no adverse effects on their psychiatric disorder, no significant benefits were apparent either. A more recent trial in comorbid alcohol dependence and depression found that naltrexone but not sertraline improved alcohol outcomes, with mood similarly improving in all groups. There are no studies of acamprosate in comorbidity; however, it could be considered given its good safety profile. There is little consistent evidence for the use of psychological interventions for the treatment of alcohol dependence in people with comorbid psychiatric disorders. Where evidence of benefit from some psychological interventions was identified, it was often from mixed drug and alcohol populations from small single studies and was not judged sufficient evidence on which to base a recommendation.

The two meta-analyses of treatment of comorbid depression broadly came to the same conclusion that antidepressants had a modest to no effect on improving depressive symptoms in those who are not at least 1 week abstinent. The effect of the antidepressant on alcohol use was also of limited benefit, and where there was some abstinence was not sustained. In those with severe depression, antidepressants may improve mood, but alcohol-focused treatment is still required. There is little evidence to suggest which antidepressant is best, although one meta-analysis suggested that SSRIs were less effective than those with a mixed serotonergic-noradrenergic pharmacology. However, some of these medications also carry adverse safety profiles with alcohol and there is insufficient evidence about the newer antidepressants. In the few studies in those with an anxiety disorder, whilst antidepressant medication may improve anxiety symptoms this was not associated with a beneficial effect on alcohol consumption. The evidence for those with either comorbid depression or anxiety suggests that focusing on managing their alcohol misuse at the start is key because whilst medication may help their anxiety or depression, improvements in their alcohol misuse will not necessarily follow.

There were only a few studies about the role of pharmacotherapy in those with alcohol and illicit drug misuse. Treatment of illicit drug misuse must be optimised using psychosocial and/or pharmacological approaches as appropriate, whilst monitoring the effect this has on alcohol consumption to ensure alcohol does not substitute for reducing illicit drug misuse. Alcohol misuse must also be specifically addressed. Many individuals with alcohol misuse smoke heavily and should be offered support to stop. There is limited evidence to suggest whether alcohol and nicotine should be given up simultaneously or sequentially, therefore patient preference should guide the decision.

7.17.7. From evidence to recommendations

The GDG noted that symptoms of anxiety and depression are common in people with harmful alcohol-use or alcohol dependence. However, for many people the symptoms remit once abstinence or a significant reduction in alcohol consumption has been achieved. In addition, treatment for comorbid disorders (in particular depression and anxiety) whilst people are consuming significant levels of alcohol does not appear to be effective. However, a number of patients have comorbid disorders that do not remit when alcohol consumption is reduced. The GDG therefore recommend that the first step in treating people presenting with alcohol misuse and comorbid depression/anxiety is to treat the alcohol misuse. Given that the presence of a comorbid disorder following a reduction in alcohol consumption is associated with a poorer long-term prognosis, an assessment of the presence and need for treatment for any comorbid depression or anxiety should be considered 3 to 4 weeks after abstinence is achieved. Some people with depressive disorders will require immediate treatment (for example, those at significant risk of suicide) and the recommendations below should not on any way stand in the way of immediate treatment being provided in such a situation. In reviewing the evidence for comorbid disorders, the GDG did not find any treatment strategies or adjustments that should be made because of the comorbid problem and, in view of this, decided to refer to the relevant NICE guidelines (see also the NICE guideline on common mental health problems; NICE, 2011b). Given the high prevalence of smoking in people with alcohol-related problems, the GDG considered it to be important to emphasise the need for effective treatment in this population. For people with comorbid drug and alcohol misuse and psychotic disorders, see the relevant NICE guideline (NICE, 2011a).

7.17.8. Recommendations

7.17.8.1.

For people who misuse alcohol and have comorbid depression or anxiety disorders, treat the alcohol misuse first as this may lead to significant improvement in the depression and anxiety. If depression or anxiety continues after 3 to 4 weeks of abstinence from alcohol, assess the depression or anxiety and consider referral and treatment in line with the relevant NICE guideline for the particular disorder⁵⁶.

7.17.8.2.

Refer people who misuse alcohol and have a significant comorbid mental health disorder, and those assessed to be at high risk of suicide, to a psychiatrist to make sure that effective assessment, treatment and risk-management plans are in place.

7.17.8.3.

For the treatment of comorbid mental health disorders refer to the relevant NICE guideline for the particular disorder, and:

• for alcohol misuse comorbid with opioid misuse actively treat both conditions; take into account the increased risk of mortality with taking alcohol and opioids together⁵⁷
• for alcohol misuse comorbid with stimulant, cannabis⁵⁸ or benzodiazepine misuse actively treat both conditions.

Service users who have been dependent on alcohol will need to be abstinent, or have very significantly reduced their drinking, to benefit from psychological interventions for comorbid mental health disorders.

7.17.8.4.

For comorbid alcohol and nicotine dependence, encourage service users to stop smoking and refer to Brief Interventions and Referral for Smoking Cessation in Primary Care and Other Settings (NICE Public Health Guidance No. 1, 2006).

7.17.9. Research recommendation

7.17.9.1.

For people with alcohol dependence which medication is most likely to improve concordance and thereby promote abstinence and prevent relapse?

This question should be answered by: (a) an initial development phase in which a series of qualitative and quantitative reasons for non-adherence/discontinuing drugs used in the treatment of alcohol are explored; (b) a series of pilot trials of novel interventions developed to address the problems identified in (a) undertaken to support the design of a series of definitive trials; (c) a (series of) definitive trial(s) of the interventions that were successfully piloted in (b) using a randomised controlled design that reports short-term (for example, 3 months) and longer-term (for example, 18 months) outcomes. The outcomes chosen should reflect both observer and service user-rated assessments of improvement and the acceptability of the intervention. Each individual study needs to be large enough to determine the presence or absence of clinically important effects, and mediators and moderators of response should be investigated.

7.18.1. Aim of review

The following section draws on the review of Wernicke-Korsakoff syndrome (WKS) and is developed as part of the NICE (2010b) guideline on the management of alcohol-related physical complications including the management of acute withdrawal. The GDG failed to identify any evidence for specific interventions in WKS beyond prevention strategies using thiamine – these are covered in the guideline mentioned above (NICE, 2010b). The GDG therefore adopted a consensus-based approach to the review of the literature, synthesising previously published narrative reviews to assist in development of the recommendations for this guideline.

7.18.2. Narrative summary

Wernicke's encephalopathy (WE) is traditionally thought of as a disorder of acute onset characterised by nystagmus, abducent and conjugate gaze palsies, ataxia of gait, and a global confusional state, occurring together or in various combinations (Victor et al., 1989). Wernicke first described the disorder in 1881 and the symptoms he recorded included disturbances of eye movement, ataxia of gait, polyneuropathy and mental changes including apathy, decreased attention span, and disorientation in time and space. Work by Alexander (1940) and then Jolliffe and colleagues (1941) established that a deficiency in thiamine (vitamin B1) was central to causation and potential treatment of the disorder (Lishman, 1998). Korsakoff gave the first comprehensive account of the amnestic syndrome now known as Korsakoff's psychosis (KP) in 1887, which includes features such as delirium characterised by recent memory loss with confabulation but with relative preservation of other intellectual functions. More recent work has highlighted a retrograde memory impairment with a ‘temporal gradient’, such that earlier memories are recalled better than more recent ones (Kopelman et al., 2009). The two disorders were brought together by Victor and colleagues in 1971 (Victor et al., 1971), and WKS is now considered to be a unitary disorder comprising acute WE, which proceeds in a proportion of cases to KP. A major complicating factor is that the pathology of WE may not be associated with the classical clinical triad (see above) in up to 90% of patients (Harper et al., 1986). Therefore, it has been suggested that a presumptive diagnosis of WE should be made for any patient with a history of alcohol dependence who may be at risk. This includes anyone showing evidence of ophthalmoplegia, ataxia, acute confusion, memory disturbance, unexplained hypotension, hypothermia, coma or unconsciousness (Cook, 2000).

Untreated, WE leads to death in up to 20% of cases (Harper, 1979; Harper et al., 1986), or KP in up to 85% of the survivors. A quarter of the latter group may then require long-term institutionalisation (Victor et al., 1989). Furthermore, the incidence of KP has been reported to be rising in some parts of the UK (Ramayya & Jauhar, 1997). For the reasons mentioned above it is probable that WE is under-diagnosed and inadequately treated in hospital, let alone in the community (Thomson & Marshall, 2006). It is therefore not known how often people with alcohol dependence in the community unnecessarily suffer brain damage.

Cognitive impairment is common in people with chronic alcohol-use disorders, with between 50 and 80% experiencing mild to severe cognitive deficits (Bates et al., 2002). The clinical and neuropsychological features of alcohol-related brain damage are well described, and the deficits appear to centre on visuospatial coordination, memory, abstract thinking and learning new information, with general knowledge, over-rehearsed information and verbal skills largely spared (Lishman, 1998). Attempts have been made to describe the unique features of ‘alcoholic dementia’ (Oslin & Cary, 2003), but there is a lack of evidence linking any specific neuropathology with heavy alcohol intake (Joyce, 1994). A range of potential factors have been implicated in the causation of alcohol-related brain damage, including direct alcohol neurotoxicity, thiamine deficiency, traumatic brain injury, familial alcoholism, childhood psychopathology, age and education (Bates et al., 2002). Studies in people with features suggestive of WE have shown that their memory and general intellectual function are roughly equivalent (Bowden, 1990). Therefore, the effects of thiamine deficiency on cognition are more widespread than amnesia, with effects on visuospatial and abstracting functions being indicated (Jacobson et al., 1990).

The mechanism by which chronic, heavy alcohol consumption causes thiamine deficiency is by increasing metabolic demand, decreasing dietary intake and reducing hepatic storage capacity due to liver damage (Cook et al., 1998, Thomson et al., 1987). Brain cells require three thiamine-dependent enzymes to metabolise glucose (transketolase, pyruvate dehydrogenase complex and α-ketoglutarate dehydrogenase) (Butterworth, 1989), and a deficiency of thiamine reduces the activity of these enzymes leading to brain cell death and reduced cognitive function (Butterworth, 1989). Cognitive impairment due to subclinical WKS in alcohol dependence may therefore be responsive to thiamine therapy. Abstinence can also improve cognition and therefore it remains the mainstay of any effective prevention programme. This is important because apart from thiamine there are no established pharmacotherapeutic strategies to specifically prevent impairment of or improve cognition once a deficit has been established.

For those with established WKS, appropriate rehabilitation, usually in supported accommodation for those with moderate and severe impairment, is the correct approach because there is some evidence to suggest that people with WKS are capable of new learning, particularly if they live in a calm and well-structured environment, and if new information is cued (Kopelman et al., 2009). There have been a few case reports of using medications to treat dementia in WKS with mixed results (Cochrane et al., 2005; Luykx et al., 2008). In an open study, the noradrenergic antidepressant reboxetine appeared to improve cognitive performance in those who had had WKS for less than 1 year (Reuster et al., 2003). Fluvoxamine has been shown to improve memory consolidation and/or retrieval in patients with WKS (Martin et al., 1995b).

The NICE (2010b) guideline on the management of alcohol-related physical complications made recommendations about people who did not have clinical features of WE, but were at high risk of developing it. They identified a high-risk group who may be characterised by the following features:

• alcohol-related liver disease
• medically-assisted withdrawal from alcohol (planned or unplanned)
• acute alcohol withdrawal
• malnourishment or risk of malnourishment; this may include:

  -

  weight loss in past year

  -

  reduced BMI

  -

  loss of appetite

  -

  nausea and vomiting

  -

  a general impression of malnourishment

• homelessness
• hospitalised for acute illness
• hospitalised for comorbidity or another alcohol issue.

From the perspective of acute inpatient care, the NICE guideline on alcohol-related physical complications (NICE, 2010b) also recommended the use of intramuscular thiamine because there were concerns about the absorption of oral thiamine in a group undergoing assisted withdrawal. There is also a problem of lack of compliance with oral preparations in people drinking heavily in the community, and so some authors advocate a choice between intravenous and intramuscular thiamine therapy (Thomson & Marshall, 2006). Intramuscular Pabrinex has a lower incidence of anaphylactic reactions than the intravenous preparation at 1 per 5 million pairs of Pabrinex ampoules, which is far lower than many frequently used drugs that carry no special warning in the BNF (Thompson & Cook, 1997; Thompson & Marshall, 2006).

Relatively little is also known about the outcomes of the treatment of alcoholic Korsakoff syndrome. The large case study by Victor and colleagues (1971) reported that 25% recovered, 50% showed improvement over time and 25% remained largely unchanged. Other authors also believe that some improvement does occur in approximately 75% of patients over a number of years if they remain abstinent from alcohol (Kopelman et al., 2009). There is little evidence from research studies to design and inform effective rehabilitation specifically in WKS (Smith & Hillman, 1999) although strategies developed in cognitive rehabilitation for a range of cognitive impairments may be of value (Cicerone et al., 2005).

7.18.3. From evidence to recommendations

The GDG accepted the evidence that thiamine has a key preventative role in WKS, and adapted the recommendations from the NICE guideline on alcohol-related physical complications (NICE, 2010b) to take account of the use of thiamine in a community-based population. Due to the high risk of long-term brain injury and the potentially serious consequences of WE, a high index of suspicion for WE should be adopted and thiamine prescribed accordingly. A number of at-risk groups are specified in the recommendation. The GDG also considered the care of people with established WKS and subsequent cognitive impairment. The limited data available suggested that continued abstinence from alcohol and a supportive and structured environment may have some beneficial effects for people with WKS and, given the high morbidity and mortality in this group, the GDG concluded that supported independent living for those with mild impairment and 24-hour care for those with severe impairment should be made available.

7.18.4. Recommendations

7.18.4.1.

Follow the recommendations in NICE clinical guideline 100⁵⁹ on thiamine for people at high risk of developing, or with suspected, Wernicke's encephalopathy. In addition, offer parenteral thiamine followed by oral thiamine to prevent Wernicke-Korsakoff syndrome in people who are entering planned assisted alcohol withdrawal in specialist inpatient alcohol services or prison settings and who are malnourished or at risk of malnourishment (for example, people who are homeless) or have decompensated liver disease.

7.18.4.2.

For people with Wernicke-Korsakoff syndrome, offer long-term placement in:

• supported independent living for those with mild cognitive impairment
• supported 24-hour care for those with moderate or severe cognitive impairment.

In both settings the environment should be adapted for people with cognitive impairment and support should be provided to help service users maintain abstinence from alcohol.