1.2.1.1. Comparators

The comparators examined in the model were treatment sequences chosen by the GDG economic subgroup and agreed by the GDG. The GDG elected to consider the cost-effectiveness of one-off induction treatment strategies in the induction of remission model, to reflect the nature of the treatment and the data that could be extracted from the clinical trials. The GDG were satisfied that, having established the most cost-effective induction sequence, longer term costs and effects could be captured in the maintenance model, where relapses from maintenance treatment are then assumed to be treated with the most cost-effective one-off induction sequence found from this analysis. The economic subgroup comprised two consultant gastroenterologists, one specialist IBD nurse and one patient representative. The GDG noted that, although 5-ASAs were considered as a class in the clinical review, the different preparations may have different costs and side-effect profiles, and therefore it was decided to separate them for the economic analysis. The drug sequences were those included in a network meta-analysis of induction of remission. It was decided that, in line with the recommendations made in TA 187³⁷ and GDG consensus view of clinical practice, biologics should be offered as the last therapy for each of the treatment strategies looked at. The specific treatment sequences were chosen based on the clinical practice of the economic subgroup members, and what they deemed to be the most likely treatment pathway for a patient experiencing an acute exacerbation of Crohn's disease. These are shown in Table 1.

1.2.1.2. Population

The population entering the model comprises people with an acute exacerbation of Crohn's disease, defined by a Crohn's Disease Activity Index (CDAI) score of > 150. Biologics are only recommended for people with severe Crohn's disease³⁷; an assumption was made that people whose exacerbation failed to respond to two lines of treatment would be regarded as falling under the aegis of the TA, though it was noted that this may not always be the case. Strategy 9 in Table 1 is only relevant for people where the Crohn's disease has progressed to severe before initiation of biologic treatment, despite only failing one line of induction.

1.2.1.3. Time horizon

The time horizon considered in the base-case model was 30 weeks. This time horizon was chosen to reflect the longest drug treatment sequence based upon clinical practice. This also reflects the assumptions made in the conditional logistic regression conducted to derive efficacy inputs. In this analysis it was assumed, based on information from the trials and GDG opinion, that the trials were of sufficient duration such that remission or withdrawal would occur by a certain time, or not at all. For example, it was assumed that in a glucocorticosteroid trial lasting 16 weeks, the proportion of people entering remission would not be significantly higher than in a trial lasting eight weeks, since those people who are reported to be in remission at 16 weeks were likely to have entered remission by eight weeks. The effect of treatment durations that were closer to those used in the trials was explored in sensitivity analysis. Treatment durations for the base case and sensitivity analysis are shown in Table 2.

Table 2

Drug treatment durations.

1.2.2.1. Model structure

A decision tree was constructed, whereby the QALY gain was driven by the proportion of people in whom remission was successfully induced. Remission was defined as not withdrawing due to an adverse event and a CDAI score of < 150. People who withdrew due to an adverse event or did not respond to treatment moved on to the next line of treatment. Although the GDG noted it was unlikely that all treatments would have the same side-effect profile, they accepted that the reporting of specific adverse events in the RCTs was not sufficient to model specific treatment-related adverse events. On that basis, they agreed that withdrawals from treatment could be used as a proxy for adverse events, and that costs and disutilities pertaining to adverse events for each treatment would be captured by both the additional cost of further treatment, and by patients still having the utility weight associated with active disease. Sensitivity analyses were conducted which explored the effects of including drug-related adverse events for glucocorticosteroid monotherapy; observational data was used to conduct this analysis. To capture the benefits of inducing remission early, people in whom remission is induced on the first-line treatment gained more QALYs than those who respond on second-, third- or fourth- line treatment. The structure of the model is shown in Figure 1.

Figure 1

Model structure.

Key assumptions:

• Treatment continues to the end of the treatment cycle regardless of whether people enter remission.
• Utility is assumed to improve in the middle of the treatment cycle for those entering remission.
• For time spent in active disease, people incur more contacts with the health service than they do in remission (see Table 11).
• Withdrawals are assumed to occur at the end of a treatment cycle.
• All people who do not enter remission by the end of the time horizon are assumed to undergo surgery.

1.2.3.1. One-way sensitivity analysis

One-way sensitivity analyses were also conducted to test the robustness of model results to changes in key parameters. In one-way sensitivity analysis, one parameter is varied while all other parameters are kept constant and the effects of changing this parameter on the model results are explored. The one-way sensitivity analyses conducted are described in Table 20.

Table 20

One-way sensitivity analyses in the model.

1.2.3.2. Probabilistic analysis

A Monte Carlo simulation¹⁶ was conducted to explore the uncertainty in model results. In a Monte Carlo simulation, each parameter is assigned a distribution reflecting its uncertainty; random draws are then taken from this distribution and propagated through the model, to calculate costs and QALYs. This process is repeated 10,000 times and a model result which represents an average of the simulations is computed.

1.2.4.1. Summary table of model inputs

Model inputs were based on clinical evidence identified in the systematic review undertaken for the guideline, supplemented by additional data sources as required. Model inputs were validated by the GDG. A summary of the model inputs used in the base-case analysis is provided in Table 3, Table 4, and Table 5 below. More details about sources, calculations and rationale for selection can be found in the sections following this summary table.

Table 3

Cost inputs.

Table 4

Clinical inputs- probabilities of withdrawal and remission.

Table 5

Utility weights in the model.

1.2.4.2. Treatment effects (remission and withdrawal)

The results of conventional meta-analyses of direct evidence alone make it difficult to determine which intervention is the most effective treatment. The challenge of interpretation has arisen for two reasons:

• Some pairs of alternative strategies have not been directly compared in a randomised controlled trial
• There are frequently multiple overlapping comparisons that could potentially give inconsistent estimates of effect.

This is particularly problematic for probabilistic analysis. To overcome these problems, a Bayesian network meta-analysis (NMA) was conducted in WinBUGS, using code and assistance provided by the NICE technical support unit.

Conventional meta-analysis assumes that for a fixed-effect analysis, the relative effect of one treatment compared to another is the same across an entire set of trials. In a random-effects model, it is assumed that the relative effects are different in each trial but that they are from a single common distribution and that this distribution is common across all sets of trials.

Network meta-analysis requires an additional assumption over conventional meta-analysis. The additional assumption is that intervention A has the same relative effect across all trials of intervention A compared to intervention B as it does across trials of intervention A versus intervention C, and so on. Thus, in a random-effects network meta-analysis, the assumption is that intervention A has the same effect distribution across all trials of A versus B, A versus C and so on.

The aim of the NMA was to calculate treatment-specific probabilities for withdrawal and remission conditional on non withdrawal. It is assumed that people who withdraw cannot go into remission, and similarly people counted as ‘a remission’ have not withdrawn due to adverse events, in other words, the two events are mutually exclusive. Treatment effects for the model had to be accounted for such that the number of withdrawals and remissions could not exceed the number of people in the trial. This negative correlation in outcomes is taken account of by carrying out a conditional logistic regression NMA.

For the network meta-analysis, treatment effects were calculated so as to reflect the clinical review as closely as possible in terms of pooling of studies. In order to obtain treatment effects for remission conditional on non withdrawal, the number of withdrawals was removed from the denominator when entering data for remission into WinBUGS. Baseline log odds of withdrawal and remission conditional on non-withdrawal were calculated using a logistic regression conducted on the placebo arms in the trials and then adjusted by the treatment specific log odds ratios calculated by the NMA using the following logic:

Let BO_w, BO_r∣w^c, θ_w , θ_r∣w^c and OR_w, OR_r∣w^c denote the baseline odds (from the placebo arms),treatment-specific odds and treatment-specific log odds ratio for withdrawal and remission given no withdrawal respectively. Then:

And:

This approach has the advantage that baseline and relative effects are both modelled on the same log odds scale. It also ensures that the uncertainty in the estimation of both baseline and relative effects is accounted for in the model.

To reflect the populations explored in the clinical review, two separate analyses were conducted; one for people on first-line monotherapy treatment, and one for people on second-line treatment in combination with a glucocorticosteroid, having failed first-line glucocorticosteroid monotherapy.

1.2.4.3. First-line NMA

The schematic of trials compared in the first line NMA is shown in Figure 2.

Figure 2

Network of trials compared in the first-line NMA.

In the trial network in Figure 2, glucocorticosteroid treatment featured in the most trials (10) followed by budesonide (9), mesalazine (8), placebo (7), sulfasalazine (2) and azathioprine monotherapy (1). For ease of interpretation, placebo was chosen as the baseline treatment for comparisons in the NMA. The data used in the first-line induction NMA can be found in Table 6. Please note that although azathioprine monotherapy was included in the NMA for completeness, it was not compared in the cost-effectiveness analysis. First-line azathioprine monotherapy was not included because of the lack of evidence and because it has a slower response than the other treatments considered. (3 – 4 months i.e. longer than the 2 month time horizon considered)

Table 6

Data for the first-line NMA.

A logistic regression was run on the placebo arms of the trials shown in Table 6; this produced a baseline odds, BO which was used in the NMA to derive treatment-specific probabilities as described above. The baseline odds for withdrawal and remission had lognormal distributions parameterised as:

The model was run for 50,000 iterations with a burn in period of 50,000. Vague uninformative priors were combined with data-driven likelihood functions to produce posterior probability estimates. The final treatment-specific probability estimates and their associated confidence intervals can be seen in Table 7.

Table 7

Probabilities from 1st line NMA.

It can be seen from Table 7, that among first-line treatments, sulfasalazine was associated with the highest probability of withdrawal- 35%- but with the 95% confidence interval ranging from 5% to 80%. Glucocorticosteroid treatment was associated with the highest probability of remission conditional on non-withdrawal- 66% with 95% confidence interval ranging from 53% to 79%. These estimates were used to parameterise treatment effects in the model; it should be noted that there is a large amount of imprecision in these estimates, particularly for withdrawal. This is often the case when calculating treatment effects for withdrawal due to small event numbers.

Model fit was assessed by calculating the total residual deviance and comparing with the number of unconstrained data points. In the withdrawals NMA, the total residual deviance was 38.6 which, when compared to 43 unconstrained data points, shows that the model fitted the data reasonably well. In the remission conditional on no-withdrawal NMA, the total residual deviance was 35.94 which again, when compared to 31 unconstrained data points, showed that the model fitted the data reasonably well. DIC statistics of 154.9 and 200.1 were calculated for the withdrawal and remission conditional on no-withdrawal NMAs respectively.

Posterior estimates of heterogeneity- between trial variance- were calculated and values of 0.29 and 0.22 were found for the withdrawal and remission trials. This shows there was a large amount of variation in treatment effects calculated from different trials.

Inconsistency in the network was assessed by fitting an ‘inconsistency model’¹⁵

1.2.4.4. Second-line NMA

The schematic of trials compared in the second line NMA is shown in Figure 3.

Figure 3

Network of trials compared in the second-line NMA.

In the trial network in Figure 3, glucocorticosteroid monotherapy and azathioprine + a glucocorticosteroid featured in the most trials (eight); methotrexate + a glucocorticosteroid featured in five trials. For ease of interpretation, glucocorticosteroid treatment- which can be thought of here as the placebo comparison- was chosen as the baseline treatment for the NMA. The data used in the second-line induction NMA can be found in Table 8.

Table 8

Data for the first-line NMA.

It should be noted that since Present was a crossover trial, the denominators reported for safety and efficacy are different; this is reflected in the clinical review. To account for this, the numerator for remission given no withdrawal was not adjusted by removing withdrawals, since it is not possible to determine exactly which of the people eligible to be assessed for the remission outcome withdrew. This is a conservative approach since it means the number of people being assessed for the remission outcome is bigger and thus the overall probability of achieving remission becomes smaller. Denominators of 36 were used for both arms of the Present study for the remission outcome.

A logistic regression was run on the glucocorticosteroid monotherapy arm of the trials shown in Table 8; this produced a baseline odds, BO which was used in the NMA to derive treatment-specific probabilities as described above. The baseline odds for withdrawal and remission had lognormal distributions parameterised as:

The model was run for 50,000 iterations with a burn in period of 50,000. Vague uninformative priors were combined with data-driven likelihood functions to produce posterior probability estimates. The final treatment-specific probability estimates and their associated confidence intervals can be seen in Table 9.

Table 9

Probabilities from first-line NMA.

It can be seen from Table 9 that out of the two second-line treatments, methotrexate + a glucocorticosteroid was associated with the highest probability of withdrawal- 11%- with 95% confidence interval ranging from 0% to 70%. Azathioprine + a glucocorticosteroid was associated with a higher probability of remission conditional on no withdrawal- 66% with 95% confidence interval ranging from 33% to 92%. These estimates were used to parameterise treatment effects in the model; it should be noted that there is a large amount of imprecision in these estimates, particularly for withdrawal. This is often the case when calculating treatment effects for withdrawal due to small event numbers, but the remission conditional on no withdrawal outcome is also associated with large imprecision.

Model fit was assessed by calculating the total residual deviance and comparing this with the number of unconstrained data points. In the withdrawals NMA, the total residual deviance was 20.94 which, when compared to 21 unconstrained data points, shows that the model fitted the data well. In the remission conditional on no withdrawal NMA, the total residual deviance was 23.71 which, when compared to 21 unconstrained data points, shows that the model fitted the data reasonably well. DIC statistics of 70.4 and 105.5 were calculated for the withdrawal and remission conditional on no-withdrawal NMAs respectively.

Posterior estimates of heterogeneity- between trial variance- were calculated and values of 0.67 and 1.22 were found for the withdrawal and remission trials. This shows there was a very large amount of variation in treatment effects calculated from different trials.

Posterior estimates of heterogeneity were calculated for the ‘inconsistency model’ and values of 0.61 and 1.53 were found for the withdrawal and remission trials. This shows that the incorporation of the consistency equations did not force the between trial variance to increase in the remission conditional on no-withdrawal and therefore it is unlikely that there is inconsistency within the network. There was, however, an increase in heterogeneity in the withdrawal NMA (0.61 vs 0.67), which prompted further investigation of inconsistency within the network. Since there were only three treatments in the network the Bucher method⁷ can be used to detect whether the amount of inconsistency in the network is statistically significant or a chance finding. A p-value of 0.46 was calculated to test the null hypothesis that there was no inconsistency in the network; this shows that there is insufficient evidence to reject the null and therefore it cannot be said that there is inconsistency.

Figure 4 shows how the model probabilities interact for one treatment cycle in the patient pathway. Figure 4 shows that from the number of people entering the model, n, the number of people entering remission n_rem after one treatment cycle and the number of people failing to enter remission n_{no rem} for a given drug can be calculated using the equations:

Figure 4

Patient pathway and probabilities (one treatment cycle).

1.2.4.5. Utilities

For economic evaluation, a specific measure of Health Related Quality of Life (HRQoL) known as utility is required to calculate QALYs. Utilities indicate the preference for health states on a scale from 0 (death) to 1 (perfect health). The NICE reference case³⁶ specifies that the preferred way for this to be assessed is by the EQ-5D instrument.

A systematic search identified four studies^9,21,23,54 with appropriate data to use as utility weights for the model. A description of these studies is shown in Table 10

Table 10

Selecting utility weights for the model.

Utility weights derived by Stark et al⁵⁴ were used due to the comparative lack of limitations and the directness of the population. In particular, the Stark data was favoured due to:

• Use of EQ-5D to elicit utility weights directly from patients
• UK EQ-5D tariff used
• Use of CDAI thresholds that mirrored those used in most of the papers in the clinical review.

1.2.4.6. Resource use and cost

The GDG thought it likely that, as well as the treatment-specific tests and consultations, people with an exacerbation of Crohn's disease would have regular consultations and tests regardless of the drug. These consultations and tests are summarised in Table 11 and the weighted average cost calculated is shown in Table 3.

Table 11

Consultations and tests for people not in remission (regardless of induction treatment).

Weighted average costs of drug preparations were used in the model; weights were calculated using prescription-cost-analysis data³⁹ in order to calculate drug costs which reflect how they are prescribed in clinical practice.

Based on the costs, weights and dosing schedule described in Table 12, a cost of £38.10 was calculated for an eight week course of prednisolone drug treatment in the base case. The total cost including tests and consultations was £362.17 (Table 3).

Table 12

Prednisolone costs in the model including tests.

Based on the costs, weights and dosing schedule described in Table 13, a cost of £167.94 was calculated for an eight week course of budesonide drug treatment in the base case. The total cost including tests and consultations was £492.01 (Table 3).

Table 13

Budesonide costs in the model including tests.

Based on the costs, weights and dosing schedule described in Table 14, a cost of £98.01 was calculated for an eight week course of mesalazine in the base case. The total cost including tests and consultations was £416.32 (Table 3).

Table 14

Mesalazine costs in the model including tests.

Based on the costs, weights and dosing schedule described in Table 15, a cost of £44.16 was calculated for an eight-week course of sulfasalazine in the base case. The total cost including tests and consultations was £362.48 (Table 3).

Table 15

Sulfasalazine costs in the model including tests.

Based on the cost and dosing schedule described in Table 16, a cost of £6.54 was calculated for an eight week course of methotrexate + a glucocorticosteroid drug treatment in the base case. The total cost including tests and consultations was £355.67 (Table 3).

Table 16

Methotrexate costs in the model including tests.

Based on the cost and dosing schedule described in Table 17, a cost of £42.11 was calculated for an eight-week course of azathioprine + a glucocorticosteroid treatment in the base case. The total cost including tests and consultations was £429.34 (Table 3).

Table 17

Azathioprine costs in the model including tests.

Based on the cost and dosing schedule described in Table 18, a cost of £1,056.36 was calculated for biologic drug treatment in the base case. The total cost including tests and consultations was £1,426.09 (Table 3).

Table 18

Biologic costs in the model including tests.

1.2.4.7. Cost of surgery

The cost of surgery was calculated in close collaboration with the surgeon on the GDG as described below and in Table 19.

Table 19

Surgery costs in model.

1. The five most common operations in Crohn's disease- pan proctocolectomy, colectomy, right hemicolectomy, small intestine resection and strictureplasty- were chosen and matched to their closest fitting OPCS and HRG codes.
2. Weights were calculated using HES data⁴¹, selected OPCS codes and assuming that 10% of all operations would be associated with a ‘major complication or comorbidity’
3. An average cost per operation was calculated by multiplying these weights by the costs attached to selected HRG codes, and adding in a pre-operative and post-operative consultation for each operation.

So, assuming pre-operative and post-operative costs for a consultation of £94 and £57 (Table 3), and a re-operation rate of 10%, the overall cost of surgery assumed in the model was assumed to be £5,351.24.

1.2.5.1. Calculating QALYs

To calculate QALYs for a given treatment sequence, both the probability of inducing remission for each individual treatment, and the time spent in remission over the course of the model for a given treatment strategy are considered. To do this, the treatment strategy is partitioned into individual treatments and the number of weeks of remission and active disease that occur as a direct result of each treatment are calculated. These are then aggregated over the duration of the strategy and QALYs for a given strategy are calculated by multiplying the number of weeks of remission and active disease by the appropriate utility weights.

Figure 5 is a visual representation of how weeks of remission and active disease are calculated in the model. Note that, in the absence of data, it was assumed that for all people in whom remission is successfully induced, remission occured half-way through treatment, and people do not relapse.

Figure 5

Calculating weeks of remission and active disease for a given treatment strategy.

It can be seen from the diagram that QALYs are calculated as follows:

Let p₁, p₂, p₃ and p₄ be the probabilities of successfully inducing remission with treatments 1, 2, 3 and 4 respectively.

Let t₁, t₂, t₃ and t₄ be the durations of treatment in weeks associated with treatments 1, 2, 3 and 4 respectively.

Let W_R1, W_R2, W_R3, and W_R4 denote the expected number of weeks of remission associated with treatments 1, 2, 3 and 4 respectively.

Let U_R and U_A be the utility weights associated with remission and active disease respectively.

Let H denote the time horizon in the model, which will be equal to the length of the longest treatment strategy and is 30 weeks in the base case.

Treatment one:

Treatment two:

Treatment three:

Treatment four:

Note that the term p₄ × (H − t₁ − t₂ − t₃ − t₄) is added to the equation for cases when the overall length of the treatment strategy is less than the time horizon. In the event that the duration of the treatment strategy is equal to the time horizon this term is equal to zero, since H = t₁ + t₂ + t₃ + t₄.

Then the total number of weeks of remission, W_R and number of weeks of active disease W_A are given by:

And the total treatment specific QALYs, Q, are calculated as:

1.2.5.2. Probabilistic analysis in the model

In the probabilistic analysis, distributions were assigned to treatment effects, utilities and, where possible, costs in order to account for the uncertainty in model inputs and capture the effect of this uncertainty on model outputs.

Treatment effects:

To capture the uncertainty in treatment effects, a sample of 1000 random sets of treatment effects was taken from the NMA using the CODA function in WinBUGS. This has the advantage of preserving the correlation between variables, which would not be accounted for if they were sampled from their individual distributions. For the probabilistic cost-effectiveness analysis, for each simulation a random set of treatment effects was chosen from the sample using random number generation.

Reference costs:

To assign a distribution to reference costs, it was assumed that they followed a lognormal distribution and used the inter-quartile range to calculate an approximate standard error on the log scale.

Let X be the cost for which a distribution is required, i.e. ln (X) ∼ Normal (μ, σ²)

Let M be the mean associated with the cost.

Let IQR be the inter-quartile range associated with the cost.

Note that for normally distributed data:

and the standard error, s, is related to the standard deviation by:

Then the standard error on the log scale can be calculated as:

And random draws can be taken from the distribution:

Utilities:

Utilities were sampled probabilistically by assigning lognormal distributions to utility decrements as described in (ref Briggs). Normal distribution parameters were converted to lognormal parameters by method of moments, as defined below:

Let E[X] and Var[X] be the mean and variance respectively, of the utility decrement U

Then the parameters of the lognormal distribution, μ and σ² are found by:

1.2.5.3. Calculating cost effectiveness

It is possible, for a particular cost-effectiveness threshold, to express cost-effectiveness results in term of net benefit (NB). This is calculated by multiplying the total QALYs for a comparator by the threshold cost per QALY value (for example, £20,000) and then subtracting the total costs. The decision rule then applied is that the comparator with the highest NB is the most cost-effective option at the specified threshold. That is the option that provides the highest number of QALYs at an acceptable cost. For ease of computation NB is used to identify the optimal strategy in the probabilistic-analysis simulations.

Let C_t and Q_t denote the mean costs and mean QALYs respectively, associated with a given treatment. Then mean net benefit NB_t is calculated as:

where £20,000 per QALY represents the cost-effectiveness threshold in the NICE reference case.

The net benefit for each of the 1000 simulations in the probabilistic analysis is calculated. This allows the probability that a given treatment would be optimal, based on the number of times it has the highest net benefit, can be estimated.

However, the strategy that is optimal overall is the one that has the highest net benefit calculated using the mean costs and QALYs, where means were the average of the 1,000 simulated estimates.

1.3.2.1. One-way sensitivity analysis

One-way sensitivity analyses were also conducted in order to test the robustness of model results to changes in key parameters. In all deterministic sensitivity analyses, the same strategy namely glucocorticosteroid treatment, glucocorticosteroid + azathioprine then a biologic was the most cost-effective strategy.

Table 24

One-way sensitivity analysis – mean net monetary benefit.

2.2.1.1. Comparators

The comparators examined in the model were the same as in the clinical review and approved by the GDG economic subgroup. The economic subgroup comprised two consultant gastroenterologists, one specialist IBD nurse and one patient representative. Methotrexate could not be included since withdrawals from treatment - a key driver of treatment effects in the model- were not reported in the only methotrexate study found in the clinical review. The comparators explored in the model were:

• No treatment
• Azathioprine
• Mesalazine
• Olsalazine
• Budesonide
• Glucocorticosteroids

It should be noted that although they were combined in the clinical review, mesalazine and olsalazine were separated in the economic analysis due to potential differences in costs and side-effect profiles.

2.2.1.2. Population

The population entering the model comprises people in medically-induced remission of Crohn's disease. In most cases, remission was defined by a Crohn's Disease Activity Index (CDAI) score of ≤ 150, though due to paucity of evidence, trials using other definitions, for example physician's objective assessment was used.

2.2.1.3. Time horizon

The time horizon considered in the base case model was two years; this time horizon was chosen to reflect the duration of the longest trial explored in the clinical review for maintenance of remission. A longer time horizon of 10 years was explored in sensitivity analysis. A lifetime time horizon was not explored, since the incremental QALYs vs no treatment at two and ten years were not substantially different, and therefore using a longer time horizon was unlikely to have a significant impact on model results.

2.2.2.1. Model structure

A Markov model was constructed, wherein the QALY gain is driven by the amount of time people spend in remission and active disease. A cycle length of two months was chosen to reflect the duration of induction treatments for people in relapse.

Due to the way withdrawals were reported in the RCTs, two separate analyses were conducted for the clinical review, a non-conservative analysis where only the ‘relapse’ outcome was analysed, and conservative analysis where ‘relapse + withdrawals’ was analysed. Treatment effects in the economic model were parameterised so as to account for these two different methods. For the non-conservative analysis in the economic model, withdrawals and relapses were treated separately so that people who withdrew from treatment were still assumed to be in remission (although from this point their risk of relapse reverts back to the risk associated with no treatment). For the conservative analysis in the economic model, people who withdrew were assumed to be in relapse.

The GDG advised that people in relapse should be treated with the induction sequence that was found to be most cost-effective in the induction of remission model (a glucocorticosteroid, followed by azathioprine + a glucocorticosteroid then a biologic). The GDG were uncertain as to what the induction sequence should be for people who relapse while on azathioprine maintenance treatment. People who relapse on azathioprine treatment are likely to have a glucocorticosteroid or biologic induction therapy added to their azathioprine regimen, and therefore initiation of azathioprine induction therapy in these people is not relevant as they are already taking azathioprine. One plausible alternative was to assume a three-line induction sequence for azathioprine (a glucocorticosteroid – a biologic – surgery) but a four-line sequence (a glucocorticosteroid – azathioprine + a glucocorticosteroid- a biologic – surgery) for the other maintenance strategies but this three-line sequence is less cost effective and this may potentially bias the assessment. This scenario was explored, but an analysis where there was a three-line induction sequence (a glucocorticosteroid – a biologic – surgery) for all maintenance strategies was also conducted in order to address this potential imbalance. In this analysis only the maintenance treatment varies between comparators and not the induction sequence but this is probably only a reasonable comparison for people who have had a recent history of severe disease and so would necessitate more urgent treatment.

• Conservative treatment effects - three lines of induction treatment (including surgery) for people relapsing on azathioprine, four lines of induction treatment for all other people (Cons 4L).
• Non- conservative treatment effects - three lines of induction treatment (including surgery) for people relapsing on azathioprine, four lines of induction treatment for all other people (Non-cons 4L).
• Conservative treatment effects - three lines of induction treatment (including surgery) for all people in relapse (Cons 3L).
• Non-conservative treatment effects - three lines of induction treatment (including surgery) for all people in relapse (Non-cons 3L).

The model structure is shown in Figure 7; note that in the first two analyses described above, the circled health state is omitted for people relapsing on azathioprine maintenance treatment. It is also omitted from the Cons 3L and Non-cons 3L models.

Figure 7

Health states in the maintenance of remission economic model.

Key assumptions:

• People enter the model in the remission maintenance treatment state
• People who relapse enter the acute induction treatment sequence
• People in whom remission is successfully re-induced on first- or second-line induction treatment go back on their initial maintenance treatment
• People who fail induction on biologics undergo surgery
• If remission is successfully induced on biologics, people stay on biologics until either:
  • Failure: where they undergo dose escalation (equivalent to re-induction using biologics) and have then can either :
    • responding and being put again on maintenance dose or
    • not responding and go to surgery
  • Completion of 12 months: where they are reassessed and
    • if in remission they go on to no maintenance treatment
    • if not they undergo dose escalation (i.e. re-induction) and go back to the start of the sequence.

2.2.2.2. Uncertainty

2.2.3.1. Summary table of model inputs

Model inputs were based on clinical evidence identified in the systematic review undertaken for the guideline, supplemented by additional data sources as required. Model inputs were validated with clinical members of the GDG. Summaries of the model inputs used in the base-case analysis are provided in Table 25, Table 26, Table 27 and Table 28. More details about sources, calculations and rationale for selection can be found in the sections following this summary table.

Table 25

Cost inputs.

Table 26

Clinical inputs- probabilities of relapse + withdrawal (conservative analysis) per cycle.

Table 27

Clinical inputs- probabilities of withdrawal and remission conditional on not withdrawing (non-conservative analysis).

Table 28

Utility weights in the model.

The costs and probabilities associated with induction treatments for people in relapse are those associated with the optimal strategy in the induction of remission model and can be found in the induction of remission model write-up.

2.2.3.2. Baseline events (withdrawal, relapse and relapse + withdrawal)

Baseline events were modelled in WinBUGS using a random effects logistic regression on the placebo arms of the RCTs from the clinical review. The aim of the logistic regression was to calculate the baseline log odds of withdrawals due to adverse events and remission conditional on non withdrawal.

2.2.3.3. Relative treatment effects (withdrawal, relapse and relapse + withdrawal)

Relative treatment effects were calculated so as to reflect the clinical review as closely as possible in terms of pooling of studies. As described above, treatment effects were parameterised in two different ways, to facilitate a conservative and non-conservative analysis. Calculation of these treatment effects is described in the following section.

Conservative analysis

For the conservative analysis, with the exception of 5-ASA treatment and glucocorticosteroid treatment, all of the treatment effects could be extracted directly from the clinical review. As discussed previously in the induction of remission model write-up,5-ASA compounds are treated separately for economic analysis within this guideline due to differences in costs and side-effect profile. For glucocorticosteroid treatment, only one trial ⁵³ reported withdrawals and was used to obtain a relative risk for relapse + withdrawal of 1.15 at one year. In the meta-analysis for the clinical review, five mesalazine studies ^{1,2,45,53,57,62} and one olsalazine ²⁹ study were pooled. These were separated and a new meta-analysis was conducted using the mesalazine studies from the clinical review to obtain a relative risk of 0.81 at one year for relapse + withdrawal with mesalazine. The olsalazine study ²⁹ was then used to calculate a relative risk 1.23 for relapse + withdrawal.

Baseline log odds were converted to the natural scale to obtain a baseline risk of relapse + withdrawal of 52% at one year using the following equation:

$${\mathit{p}}_{\mathit{pl}}=\frac{{\mathit{e}}^{({\mathit{\theta }}_{\mathit{pl}})}}{\mathbf{1}+{\mathit{e}}^{({\mathit{\theta }}_{\mathit{pl}})}}$$

where θ_pl and p_pl denote the baseline log odds and risk of relapse + withdrawal respectively, on placebo.

All the relative risks, the baseline risk and associated confidence intervals for relapse + withdrawal are shown in Table 29. Note that in the model these are converted from yearly to two-monthly probabilities to reflect the cycle length.

Table 29

Conservative analysis treatment effects.

Equation 1- Calculating probability of relapse conditional on non-withdrawal

$$\mathit{P}(\mathit{R}\mid {\mathit{W}}^{\mathit{c}})=\frac{\mathit{P}(\mathit{R})}{\mathbf{1}-\mathit{P}(\mathit{W})}$$

Treatment effects used in the non-conservative analysis are shown in Table 30. Please note that treatment effects for withdrawal were calculated as odds ratios.

Table 30

Non-conservative analysis treatment effects.

Once the probabilities of relapse, withdrawal and relapse conditional on non withdrawal had been calculated, yearly probabilities were calculated as described in this section, and then transformed to two- monthly probabilities using the equation:

$${\mathit{p}}_{\mathbf{2}\mathit{month}}=\mathbf{1}-\mathbf{\text{exp}}[-\frac{\mathbf{1}}{\mathbf{6}}(-\mathbf{\text{ln}}(\mathbf{1}-{\mathit{p}}_{\mathbf{1}\mathit{year}}))]$$

All of the probabilities calculated using the methods described in this section can be found in Table 26 and Table 27.

Figure 4 shows how the probabilities described above and shown in Table 26 and Table 27 operate in the model for the conservative and non-conservative analyses. Note that Figure 8 describes only the patient pathway up to relapse, and a full representation of the patient pathway can be found in Figure 7.

Figure 8

Conservative and non-conservative analysis.

In the conservative analysis, people who either relapse or withdraw are assumed to have active disease and move into the active disease health states where they receive treatment for induction of remission.

In the non-conservative analysis, people who withdraw from treatment are still assumed to be in remission, but not receiving drugs for maintenance of remission; their probability of relapse is then equal to the baseline relapse probability- i.e. the same as people in the no treatment arm. People who do not withdraw from treatment may relapse according to the probability calculated in Equation 1. People relapsing from either the remission- maintenance treatment or remission- no maintenance treatment health states move into the active disease health states where they receive treatment for induction of remission.

2.2.3.4. Utilities

For economic evaluation, a specific measure of HRQoL known as utility is required to calculate QALYs. Utilities indicate the preference for health states on a scale from 0 (death) to 1 (perfect health). The NICE reference case³⁶ specifies that the preferred way for this to be assessed is by the EQ-5D instrument.

The utility weights used in the model for people with active disease and people in remission were the same as those used in the induction model and are shown in Table 28. The rationale for selecting these utility weights can be found in the induction model write-up.

2.2.3.5. Resource use and cost

Please note that all of the drug-specific information in this section refers to people undergoing treatment for maintenance of remission. For information pertaining to induction of remission please see the induction of remission model write-up.

The GDG advised that, as well as the treatment-specific tests and consultations, people in remission of Crohn's disease would have regular consultations and tests regardless of treatment. The GDG noted that these consultations would differ according to how long people were in remission for, and agreed on yearly frequencies of consultations for people who have been in one, two and three years of remission. This was accounted for in two different ways in the model. In the base case, the two-year frequencies were used and an overall cost of £25.75 applied per cycle to people in remission over the two-year time horizon. In a sensitivity analysis, an average cost of consultations over three years was calculated, and a cost of £26.51 applied per cycle to people in remission over the two-year time horizon. These consultations and tests are summarised in Table 31.

Table 31

Consultations and tests for people not in remission (regardless of induction treatment).

Drug acquisition costs

Drug costs in the model were calculated based on weighted average drug acquisition costs were calculated based on prescribing patterns for all preparations³⁹. This can be seen in Table 32.Error! Reference source not found.

Table 32

Base case weekly drug acquisition costs in the model.

Drug-specific tests

Drug-specific tests were decided by the GDG economic subgroup and chair. The GDG noted that the frequency of these drug specific tests would differ according to how long people were in remission for, and agreed on yearly frequencies of consultations for people who have been in one, and two years of remission. This was accounted for in two different ways in the model. In the base case, the two-year frequencies were used and an overall cost applied per cycle to people in remission over the two-year time horizon. In a sensitivity analysis, an average cost of tests over the two years was calculated, and a cost applied per cycle to people in remission over the two-year time horizon. These consultations and tests are summarised in Table 33, Table 34 and Table 35. Note that the consultations are expressed in two-monthly frequencies, so as to reflect the cycle length in the model.

Table 33

Drug-specific tests for mesalazine and olsalazine in the model.

Table 34

Drug-specific tests for glucocorticosteroid treatment and budesonide in the model.

Table 35

Drug-specific tests for azathioprine in the model.

2.2.4.1. Calculating QALYs

In order to calculate the QALYs associated with a given treatment, for each cycle the prevalence of people in each health state was multiplied by the utility weight associated with that health state and divided by an adjustment factor to reflect the cycle length. A worked example of the utility calculation is shown below; please note this is a simplified calculation and the full calculation would take account of all the health states shown in Figure 7.

Table 36

Example calculation for model QALYs.

Thus the total QALY for the cycle described above would be 0.083 + 0.045 = 0.128. These QALY contributions were then aggregated over the two-year model time horizon to calculate the total number of QALYs associated with each treatment.

2.2.4.2. Probabilistic analysis in the model

In the probabilistic analysis, distributions were assigned to treatment effects, utilities and, where possible, costs in order to account for the uncertainty in model inputs and capture the effect of this uncertainty on model outputs. Please see the induction of remission model write-up for more details on how inputs pertaining to induction of remission were made probabilistic.

Treatment effects:

Treatment effects were sampled probabilistically by assigning lognormal distributions as described elsewhere to quantities in Table 25 and Table 27.

Reference costs:

In order to assign a distribution to reference costs, it was assumed that they followed a lognormal distribution and the interquartile range was used to calculate an approximate standard error on the log scale.

Let X be the cost assigned to a distribution to, i.e. ln (X) ∼ Normal(μ,σ²)

Let M be the mean associated with the cost.

Let IQR be the interquartile range associated with the cost.

Note that for normally distributed data:

And noting that the standard error s, is related to the standard deviation by:

Then the standard error on the log scale can be calculated as:

And therefore random draws from the distribution can be taken:

Utilities:

Utilities were sampled probabilistically by assigning lognormal distributions to utility decrements as described elsewhere ⁶. Normal distribution parameters were converted to lognormal parameters by method of moments, as defined below:

Let E[X] and Var[X] be the mean and variance respectively, of the utility decrement U

Then the parameters of the lognormal distribution, μ and σ² are found by:

2.2.4.3. Calculating cost effectiveness

It is possible, for a particular cost-effectiveness threshold, to express cost-effectiveness results in terms of net benefit (NB). This is calculated by multiplying the total QALYs for a comparator by the threshold cost per QALY value (for example, £20,000) and then subtracting the total costs. The decision rule then applied is that the comparator with the highest NB is the most cost-effective option at the specified threshold. That is the option that provides the highest number of QALYs at an acceptable cost. For ease of computation NB is used to identify the optimal strategy in the probabilistic analysis simulations.

Let C_t and Q_t denote the mean costs and mean QALYs respectively, associated with a given treatment. Then the mean net benefit is then calculated as NB_t:

Where £20,000 per QALY represents the cost-effectiveness threshold in the NICE reference case.

This net benefit is calculated for each of the 1000 simulations in the probabilistic analysis. This means that the probability that a given treatment would be optimal, can be estimated based on the number of times it has the highest net benefit.

However, the strategy that is optimal overall is the one that has the highest net benefit calculated using the mean costs and QALYs, where means were the average of the 1,000 simulated estimates.

2.3.1.1. Conservative analysis- four-line model

Base case results

The results for the model run with conservative treatment effects and with relapses from all treatments (except azathioprine) being treated with four lines of induction therapy- a glucocorticosteroid, azathioprine + a glucocorticosteroid, a biologic, surgery- are shown in this section. In this analysis, the induction sequence for people relapsing from azathioprine is: a glucocorticosteroid, a biologic, surgery.

Table 38 shows that, in the base case conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, azathioprine was associated with the most weeks in remission and hence the highest number of QALYs.

Table 38

Mean clinical outcomes in the base case (per patient).

Table 39 shows that, in the base case conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, the ‘no treatment’ arm was associated with the lowest costs for maintenance of remission- £2,200 per patient.

Table 39

Mean costs in the base case (per patient).

Table 40 shows that in the base case conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, the ‘no treatment’ arm was most cost effective at a willingness-to-pay threshold of £20,000 per QALY. Olsalazine and glucocorticosteroid treatment were dominated by no treatment. Budesonide, mesalazine and azathioprine were associated with ICERs of £40,000, £25,000 and £21,000 per QALY gained respectively compared with no treatment. This is represented visually in Figure 9, the cost-effectiveness plane depicting the mean costs and QALYs associated with each treatment. Note that in this diagram, the blue line represents a willingness-to-pay threshold of £20,000 per QALY and therefore any treatments that fall to the right of the line are cost effective compared to no treatment at this threshold.

Table 40

Cost-effectiveness in the base case (mean).

Figure 9

Cost-effectiveness plane conservative four-line analysis.

Probabilistic sensitivity analysis

Cost effectiveness ranks at £20,000 per QALY and their associated 95% confidence intervals were calculated by Monte-Carlo simulation. These are shown in Figure 10.

Figure 10

Conservative four-line model, forest plot of cost-effectiveness rankings.

It can be seen from Figure 10 that although no treatment is most cost-effective in this case- as shown in the deterministic analysis- there is a lot of uncertainty around the relative ranks of the different treatments. The probability that each treatment was the most cost-effective in terms if incremental net benefit at a willingness to pay of £20,000 per QALY is shown in Table 41.

Table 41

Probability of each treatment being most cost-effective.

2.3.1.2. Conservative analysis - three-line model

The results for the model run with conservative treatment effects and with all treatments having three lines of induction therapy- a glucocorticosteroid, a biologic and surgery- are shown in this section.

Table 42 shows that, in the base case conservative analysis where relapses from all treatments are treated with three lines of induction therapy, azathioprine was associated with the greatest number of weeks in remission and hence the highest number of QALYs.

Table 42

Mean clinical outcomes in the base case (per patient).

Table 43 shows that, in the base case conservative analysis where relapses from all treatments are treated with three lines of induction therapy, azathioprine was associated with the lowest costs for maintenance of remission- £3,021 per patient.

Table 43

Mean costs in the base case (per patient).

Table 44 shows that in the base case conservative analysis where relapses from all treatments are treated with three lines of induction therapy, azathioprine was the most cost-effective treatment at a willingness–to-pay threshold of £20,000 per QALY. Olsalazine and glucocorticosteroid were dominated by no treatment. Budesonide was associated with an ICER of £15,000 per QALY gained compared with no treatment and both mesalazine and azathioprine were dominant vs no treatment. Note that even though mesalazine and budesonide were cost-effective vs no treatment, they were still substantially dominated by azathioprine in terms of costs and QALYs. This is represented visually in Figure 11, the cost-effectiveness plane depicting the mean costs and QALYs associated with each treatment. Note that in this diagram, the blue line represents a willingness-to-pay threshold of £20,000 per QALY and therefore any treatments that fall to the right of the line are cost effective at this threshold.

Table 44

Cost-effectiveness in the base case (mean).

Figure 11

Cost-effectiveness plane conservative three-line analysis.

Probabilistic sensitivity analysis

Cost-effectiveness ranks at £20,000 per QALY and their associated 95% confidence intervals were calculated by Monte-Carlo simulation. These are shown in Figure 12.

Figure 12

Conservative three-line model, forest plot of cost-effectiveness rankings.

It can be seen from Figure 12 that, although azathioprine is most cost effective in this case- as shown in the deterministic analysis- there is a lot of uncertainty around the relative ranks of the different treatments.

The probability that each treatment was the most cost-effective in terms if incremental net benefit at a willingness to pay of £20,000 per QALY is shown in Table 45.

Table 45

Probability of each treatment being most cost-effective.

2.3.2.1. Non-conservative analysis- four-line model

The results for the model run with non-conservative treatment effects and with relapses from all treatments except azathioprine being treated with four lines of induction therapy-glucocorticosteroid treatment, azathioprine + a glucocorticosteroid, a biologic, surgery- are shown in this section. In this analysis, the induction sequence for people relapsing from azathioprine is: a glucocorticosteroid, a biologic, surgery.

Table 46 shows that in the base case non-conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, azathioprine was associated with the greatest number of weeks in remission and hence the highest number of QALYs.

Table 46

Mean clinical outcomes in the base case (per patient).

Table 47 shows that in the base case non-conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, azathioprine was associated with the lowest costs for maintenance of remission- £1,671 per patient.

Table 47

Mean costs in the base case (per patient).

Table 48 shows that in the base case non-conservative analysis where relapses from all treatments except azathioprine are treated with four lines of induction therapy, azathioprine was the most cost-effective treatment at a willingness-to-pay threshold of £20,000 per QALY. Olsalazine was dominated by no treatment. Budesonide and mesalazine were associated with ICERs of £20,000 and £88,000 per QALY gained respectively compared with no treatment and glucocorticosteroid treatment and azathioprine were dominant vs no treatment. This is represented visually in Figure 13, the cost-effectiveness plane depicting the mean costs and QALYs associated with each treatment. Note that in this diagram, the blue line represents a willingness-to-pay threshold of £20,000 per QALY and therefore any treatments that fall to the right of the line are cost effective at this threshold.

Table 48

Cost effectiveness in the base case (mean).

Figure 13

Cost-effectiveness plane non-conservative four-line analysis.

Probabilistic-sensitivity analysis

Cost effectiveness ranks at £20,000 per QALY and their associated 95% confidence intervals were calculated by Monte-Carlo simulation. These are shown in Figure 14.

Figure 14

Non-conservative four-line model, forest plot of cost-effectiveness rankings.

It can be seen from Figure 14 that, although azathioprine is most cost-effective in this case- as shown in the deterministic analysis- there is a lot of uncertainty around the relative ranks of the different treatments.

The probability that each treatment was the most cost-effective in terms if incremental net benefit at a willingness-to-pay of £20,000 per QALY is shown in Table 49.

Table 49

Probability of each treatment being most cost-effective.

2.3.2.2. Non-conservative analysis- three-line model

The results for the model run with non- conservative treatment effects and with all treatments having three lines of induction therapy- a glucocorticosteroid, a biologic, surgery- are shown in this section.

Table 50 shows that in the base case non-conservative analysis where relapses from all treatments are treated with four lines of induction therapy, azathioprine was associated with the greatest number of weeks in remission and hence the highest number of QALYs.

Table 50

Mean clinical outcomes in the base case (per patient).

Table 51 shows that in the base case non-conservative analysis where relapses from all treatments are treated with three lines of induction therapy, azathioprine was associated with the lowest costs for maintenance of remission- £1,671 per patient.

Table 51

Mean costs in the base case (per patient).

Table 52 shows that in the base case non-conservative analysis where relapses from all treatments are treated with three lines of induction therapy, azathioprine was the most cost-effective treatment at a willingness-to-pay threshold of £20,000 per QALY. Olsalazine was dominated by no treatment. Budesonide was associated with an ICER of £65,000 per QALY gained compared with no treatment and glucocorticosteroid treatment, mesalazine and azathioprine were dominant vs no treatment. Note that even though the ICERs associated with mesalazine and glucocorticosteroid treatment are under the cost-effectiveness threshold of £20,000 per QALY, both are significantly dominated by azathioprine. This is represented visually in Figure 15, the cost-effectiveness plane depicting the mean costs and QALYs associated with each treatment. Note that in this diagram, the blue line represents a willingness-to-pay threshold of £20,000 per QALY and therefore any treatments that fall to the right of the line are cost effective at this threshold.

Table 52

Cost effectiveness in the base case (mean).

Figure 15

Cost-effectiveness plane non-conservative three-line analysis.

Probabilistic sensitivity analysis

Cost effectiveness ranks at £20,000 per QALY and their associated 95% confidence intervals were calculated by Monte-Carlo simulation. These are shown in Figure 16.

Figure 16

Non-conservative four-line model, forest plot of cost-effectiveness rankings.

It can be seen from Figure 16 that azathioprine is most cost-effective in this case- as shown in the deterministic analysis. In addition, there is much more certainty around its cost-effectiveness ranking.

The probability that each treatment was the most cost-effective in terms if incremental net benefit at a willingness-to-pay of £20,000 per QALY is shown in Table 53.

Table 53

Probability of each treatment being most cost-effective.