Quality assessment of studies included in clinical effectiveness review

Janine Dretzke; Deirdre Blissett; Chirag Dave; Rahul Mukherjee; Malcolm Price; Sue Bayliss; Xiaoying Wu; Rachel Jordan; Sue Jowett; Alice M Turner; David Moore

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Dretzke J, Blissett D, Dave C, et al. The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation. Southampton (UK): NIHR Journals Library; 2015 Oct. (Health Technology Assessment, No. 19.81.)

Cover of The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation

The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation.

Show details

Contents

< Prev Next >

Appendix 2Quality assessment of studies included in clinical effectiveness review

TABLE 50

Quality assessment RCTs (NLIV vs. control)

Study	Random sequence generation	Allocation concealment	Blinding of patients	Blinding of outcome assessment	Incomplete outcome data^a		Selective outcome reporting
Study	Random sequence generation	Allocation concealment	Blinding of patients	Blinding of outcome assessment	NIV group	Control group	Selective outcome reporting
Bhatt et al. 2013⁸³	LOW	LOW	HIGH	UNCLEAR	LOW	HIGH	LOW
Bhatt et al. 2013⁸³	Random number generator	Opaque sealed envelopes which were opened during screening visits	No sham NIV arm	No details	No loss to follow-up	3/15 (20%) early withdrawals; analysis in 12/15 (no ITT and no reasons given for withdrawal or on similarity to completers)	No apparent selective reporting
Casanova et al. 2000⁸⁵	LOW	LOW	HIGH	UNCLEAR	UNCLEAR	LOW	LOW
Casanova et al. 2000⁸⁵	Random numbers table	Randomisation by independent office, so likely that concealment adequate	No sham NIV arm	No details	6/26 (23%) withdrawals: 5/6 because of ‘pressure being too high’, 1/6 after diagnosis of significant aortic stenosis. Results for completers only. No details on baseline differences between dropouts and completers. Stated that ‘inclusion of the patients who did not complete the trial (intent-to-treat) did not affect any of the outcomes’⁸⁵	2/26 (8%) withdrawals because of abnormal echocardiographic findings detected during routine follow-up. Results for completers only. No details on baseline differences between dropouts and completers. Stated that ‘inclusion of the patients who did not complete the trial (intent-to-treat) did not affect any of the outcomes’⁸⁵	Not all results reported at all time points, although it was mentioned in the text whether or not there were any significant differences
Cheung et al. 2010⁹⁰	LOW	LOW	UNCLEAR	UNCLEAR	UNCLEAR	UNCLEAR	LOW
Cheung et al. 2010⁹⁰	Computer-generated random numbers	Drawing of sequentially numbered and sealed opaque envelopes by non-study personnel	CPAP as ‘placebo NIV’. This was an open-label study, but ‘care had been taken to avoid biasing the patients into believing either mode was superior’⁹⁰	No details	Withdrawals reported/accounted for (8/23, 35%). The main results in both arms of the study were analysed by an ITT approach. Non-completers were included in the final analysis, with their timed data censored on the withdrawal dates. Varying (reducing) numbers of patients included for arterial pH and PaCO₂. No details on characteristics of dropouts and completers	Withdrawals reported/accounted for (4/24, 17%). The main results in both arms of the study were analysed by an ITT approach. Non-completers were included in the final analysis, with their timed data censored on the withdrawal dates. Varying (reducing) numbers of patients included for arterial pH and PaCO₂. No details on characteristics of dropouts and completers	No apparent selective reporting
Clini et al. 2002⁹⁹	LOW	LOW	HIGH	LOW	UNCLEAR	UNCLEAR	LOW
Clini et al. 2002⁹⁹	Centralised block randomisation	Centralised randomisation likely to ensure allocation concealment	No sham NIV arm	All physiological measurements were performed by personnel blind to treatment and not involved in the study	Numbers and reasons given for dropouts in both NPPV and control groups. The numbers of those lost to follow-up in each group were also recorded. Similar number of dropouts/losses in both groups [12/43 (28%) NIV, 15/47 (32%) LTOT] if early dropouts were included. Slightly more patients lost to follow-up from LTOT group compared with NIV group (7/47 vs. 1/43) and more non-compliers in NIV group (7/43 vs. 1/47). Baseline characteristics of dropouts stated to be similar to those of completers. ‘The main parameters were evaluated both in terms of patient completers and in terms of the ITT approach. The last observation carried forward was used as a method of ITT and data are presented accordingly. Data on patients’ compliance were evaluated only in terms of patient completers in order to document “per protocol” analysis’⁹⁹		It appears that all of the study’s prespecified outcomes have been reported
De Backer et al. 2011⁹¹	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	UNCLEAR	UNCLEAR	UNCLEAR
De Backer et al. 2011⁹¹	Stated only that patients were randomised	No details	No sham NIV arm	No details	No details on loss to follow-up	No details on loss to follow-up	Additional measurements were taken but not reported. Blood gases at 1, 3 and 12 months, 6MWD at 3 months. Hospital admission, morbidity/mortality and compliance recorded but not reported (these were not predefined outcomes)
Duiverman et al. 2008⁷⁹	LOW	UNCLEAR	HIGH	UNCLEAR	HIGH	UNCLEAR	LOW
Duiverman et al. 2008⁷⁹	Computerised randomisation (with minimisation for FEV₁, PaCO₂ and body mass index)	Randomisation performed by independent statistician	No sham NIV arm	No details	6/37 early dropouts before baseline measurements (two died, two withdrew and two had other diseases); seven further dropouts during 3 months’ study (5 intolerance to NIV, 1 non-compliant with rehabilitation, 1 death). Total 35% dropouts. Non-completers had a lower FEV₁ (p < 0.05), lower vital capacity (p < 0.05) and higher residual volume as a percentage of total lung capacity (p < 0.05). Stated that main outcomes were evaluated for completers (not clear how many patients assessed for each outcome)	3/35 (8.6%) dropouts because of non-compliance. Non-compliers had a higher total lung capacity and residual volume than completers (p < 0.01). Stated that main outcomes evaluated for completers (not clear how many patients were assessed for each outcome)	No apparent selective reporting
Duiverman et al. 2011⁸⁰	LOW	UNCLEAR	HIGH	UNCLEAR	HIGH	HIGH	LOW
Duiverman et al. 2011⁸⁰	Computerised randomisation (with minimisation for FEV₁, PaCO₂ and body mass index)	Randomisation performed by independent statistician	No sham NIV arm	No details (but analyses performed by an independent statistician)	6/37 early dropouts (during in-hospital rehabilitation programme, Duiverman 2008⁷⁹); further seven dropouts before start period of this follow-on study. 24 started home-based follow-up period, 9/24 dropouts (3 withdrew, 1 aorta dissection, 5 deaths); total dropout 59% (22/37). significantly lower baseline PaO₂ in dropouts compared with completers. ‘All data of all patients available at the start of the home-based period included for analyses and all available data used for analyses until patients dropped out.’⁸⁰ Patient numbers stated for different outcomes at different time points	3/35 dropouts (during in-hospital rehabilitation programme, Duiverman 2008⁷⁹). 32 started home-based follow-up period, 12/32 (37%) dropouts (3 non-compliant, 1 lung transplantation, 1 stroke, 1 deterioration in condition, 1 treated with CPAP, 5 deaths); significantly worse CRQ score and 6MWD in dropouts compared with completers. ‘All data of all patients available at the start of the home-based period included for analyses and all available data used for analyses until patients dropped out.’⁸⁰ Patient numbers stated for different outcomes at different time points	No apparent selective reporting
Garrod et al. 2000⁸⁴	UNCLEAR	LOW	HIGH	UNCLEAR	UNCLEAR	UNCLEAR	LOW
Garrod et al. 2000⁸⁴	Randomisation using sealed envelopes. No further details	Sealed envelopes suggest that allocation was likely concealed	No sham NIV arm	No details	3/24 (12%) withdrawals (one transient ischaemic attack, two non-compliance). Available for assessments: 17/23 (after 4-week run-in), 18/23 at 8 weeks, 17/23 at 12 weeks (between 22% and 27% loss to follow-up). There were no significant differences in baseline variables between patients who completed all assessments compared with those who withdrew or were unable to attend an assessment	1/22 (4%) withdrawal (refusal to attend training sessions). Available for assessments: 18/22 (after 4-week run-in), 21/22 at 8 weeks, 20/22 at 12 weeks (between 9% and 18% loss to follow-up). There were no significant differences in baseline variables between patients who completed all assessments compared with those who withdrew or were unable to attend an assessment	It appears that all of the study’s prespecified outcomes have been reported
Gay et al. 1996¹⁰⁰	UNCLEAR	UNCLEAR	LOW	UNCLEAR	HIGH	LOW	LOW
Gay et al. 1996¹⁰⁰	Stated only that patients were randomised	No details	Sham NIV (same equipment, but ‘ventilated’ with lowest EPAP level and had no added IPAP or timed breaths). All patients were told that they may be randomised to a ‘low-pressure’ setting	No details	3/7 (43%) discontinued after a median of 1 month. Significantly more than in sham group (main reason was difficulty sleeping). Results based on completers only	6/6 completed study; no losses to follow-up	No apparent selective reporting
Kaminski et al. 1999¹⁰¹	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	UNCLEAR	LOW	LOW
Kaminski et al. 1999¹⁰¹	Stated that allocated randomly	No details	No sham NIV arm	No details	2/7 (29%) discontinued NIV and crossed over to control arm, four deaths, no further losses to follow-up. Last assessment before death included	Five deaths (5/12), no further losses to follow-up. Last assessment before death included	No apparent selective reporting
Köhnlein et al. 2014⁷⁶	LOW	LOW	HIGH	LOW	LOW for survival, HIGH for QoL; UNCLEAR for remaining outcomes	LOW for survival; HIGH for QoL; UNCLEAR for remaining outcomes	LOW
Köhnlein et al. 2014⁷⁶	Computer-generated block randomisation	Randomisation hotline, so assume allocation concealed	No sham NIV arm	Outcome assessors unaware of treatment assignment throughout the study	2/102 lost to follow-up. ITT for primary outcome survival. Patient numbers not always clear for other outcome assessments at different time points. HRQoL assessments in subgroups of patients only	No losses to follow-up. ITT for primary outcome survival. Patient numbers not always clear for other outcome assessments at different time points. HRQoL assessments in subgroups of patients only	No obvious selective reporting. QoL and compliance reported for only a subset of patients but made explicit
McEvoy et al. 2009⁷⁴	LOW	LOW	HIGH	UNCLEAR	UNCLEAR (LOW for survival)	UNCLEAR (LOW for survival)	LOW
McEvoy et al. 2009⁷⁴	The central study co-ordinator generated a random sequence of treatment assignments that were stratified by centre	Sealed opaque envelopes; central co-ordinator verified that the patient met all eligibility criteria before the site research nurse broke the envelope seal	No sham NIV arm	Sleep studies were scored by experienced sleep scorers who were blinded to treatment allocation. No details for other outcomes	4/72 (5%) lost to follow-up (not contactable or withdrawal of consent). Varying number of patients attended for repeat measurements (high mortality rate and reluctance of patients to attend; therefore, not ITT and for first 12 months only). ITT and PP analysis for survival	4/72 (5%) lost to follow-up (not contactable or withdrawal of consent). Varying number of patients attended for repeat measurements (high mortality rate and reluctance of patients to attend; therefore, not ITT and for first 12 months only). ITT and PP analysis for survival	Main outcomes appear to be reported. Results for FVC appear not to be reported
Murphy et al. 2011⁷⁸ (abstract, interim trial report)	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	UNCLEAR	UNCLEAR	UNCLEAR
Murphy et al. 2011⁷⁸ (abstract, interim trial report)	Stated only that patients were randomised	No details	No sham NIV arm	No details	No details – data on 20 (of 36 randomised) that have been followed up for 3 months at time of writing	No details – data on 20 (of 36 randomised) that have been followed up for 3 months at time of writing	Results reported only for sleep-related outcomes and compliance. However, blood gases and HRQoL measures also mentioned in methodology
Sin et al. 2007⁸²	UNCLEAR	UNCLEAR	LOW	LOW	HIGH	LOW	LOW
Sin et al. 2007⁸²	Randomisation occurred at a central site	Randomisation undertaken at central site by one individual who was unaware of patients’ clinical status	Subjects blinded by using sham therapy; authors state that ‘complete blinding may not have been present and we cannot completely eliminate the possibility of a “placebo effect”, although this seems unlikely in view of the excellent compliance observed in those assigned to sham therapy’⁸²	All outcome measurements performed and interpreted by personnel who were blinded to treatment allocation	2/13 (15%) refused NIV after randomisation. Not included in analysis. No details on whether or not patient characteristics were similar	No loss to follow-up/no dropouts	No apparent selective reporting
Struik et al. 2014⁷⁵	LOW	UNCLEAR	HIGH	UNCLEAR	LOW for survival, HIGH for blood gases and QoL, UNCLEAR for remaining outcomes	LOW for survival, HIGH for blood gases and QoL, UNCLEAR for remaining outcomes	LOW
Struik et al. 2014⁷⁵	Computer-generated randomisation with minimisation	No details	No sham NIV arm	No details	25/101 dropouts. Lack of motivation (15/25), discomfort associated with treatment (8/25), dementia (1/25), cerebrovascular accident (1/25). ITT analysis for survival, unclear for hospital admissions and exacerbations, completers only for QoL and blood gases	24/100 dropouts. Lack of motivation (14/24), unable to come for testing (6/24), switch to NIV (4/24). ITT analysis for survival, unclear for hospital admissions and exacerbations, completers only for QoL and blood gases	No apparent selective reporting
Xiang et al. 2007⁹²	LOW	UNCLEAR	HIGH	UNCLEAR	UNCLEAR	UNCLEAR	LOW
Xiang et al. 2007⁹²	Random number table used to generate randomisation sequence	No details	No sham NIV arm	No details	All results appear to be based on all patients (ITT) but no details on how missing values dealt with	All results appear to be based on all patients (ITT) but no details on how missing values dealt with	No apparent selective reporting
Zhou et al. 2008⁸¹	LOW	UNCLEAR	HIGH	UNCLEAR	LOW (primary), HIGH (secondary)	LOW (primary), HIGH (secondary)	LOW
Zhou et al. 2008⁸¹	Random number table used to generate randomisation sequence	No details	No sham NIV arm	No details	Results for primary outcomes appear to be based on all patients. Between 7% and 14% loss to follow-up for secondary outcomes (results for completers only, no details on similarities)	Results for primary outcomes appear to be based on all patients. Between 7% and 14% loss to follow-up for secondary outcomes (results for completers only, no details on similarities)	No apparent selective reporting

: PP, per protocol.
a: For attrition bias in RCTs, the following criteria were applied to classify the risk as LOW, UNCLEAR or HIGH: LOW, no loss to follow-up, or < 10% loss to follow-up and details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers); UNCLEAR, loss to follow-up less than 10% and no details on ITT/effect of missing data given OR loss to follow-up > 10%, but details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers); HIGH, loss to follow-up > 10% and no details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers) OR loss to follow-up > 30%. These cut-off points are arbitrary and have not been used in any sensitivity analyses. Any likely impact of quality on results has been discussed narratively.

TABLE 51

Quality assessment crossover RCTs (NIV vs. control)

Study	Random sequence generation	Allocation concealment	Blinding of patients	Blinding of outcome assessment	Incomplete outcome data NIV group^a	Selective outcome reporting	Is it clear that the order of receiving treatments was randomised?	Can it be assumed that the trial was not biased from carry-over effects?	Dropouts after first treatment period (how incorporated into analysis?)	Are data available from both treatment periods?	Was a form of paired analysis used?
Meecham-Jones et al. 1995⁷⁷	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	HIGH	LOW	No Cochrane guidelines for rating risk of bias
Meecham-Jones et al. 1995⁷⁷	Stated that randomisation was achieved with a previously generated randomised sequence	No details	No sham NIV arm	No details	4/18 (22%) did not complete all stages of the study. 1/4 was withdrawn because of lung transplantation (during second study period), 1/4 because of development of bronchial carcinoma, 1/4 died at home during acute exacerbation 2 weeks after entering second study period (NIV), 1/4 was withdrawn because of inability to tolerate equipment. Time point of withdrawal was not clear for all. Results based on 14/18 completers	No apparent selective reporting	Yes	No statistical tests for carryover performed	Analysis based on completers only. Of the 4 withdrawals, 2/4 were during the second treatment period and it was unclear for the other 2/4	Yes	Yes
Strumpf et al. 1991³⁴	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	HIGH	LOW	No Cochrane guidelines for rating risk of bias
Strumpf et al. 1991³⁴	Stated only that patients were randomised	No details	No sham NIV arm	No details	23 initially enrolled, 4/23 of whom did not meet eligibility criteria. 7/23 could not tolerate the mask (complaints included intolerable nasal mucosal irritation unresponsive to corticosteroids or humidification, inability to sleep, excessive anxiety associated with ventilator use). Unclear how many withdrew during first/second treatment period. 5/23 patients withdrew because of other illnesses (3/5 during the NIV treatment period and 2/5 during the control period). Results presented for 7 patients who completed both treatment periods. Total dropout 70%. Stated that baseline pulmonary functions did not differ significantly between the 7 patients who completed both arms and the 23 patients initially enrolled	No apparent selective reporting	Yes	ANOVA performed to determine whether or not results may have been affected by sequence effects-no significant trends revealed	Unclear when patients dropped out, but dropouts not included in analysis	Yes, but only for 7/19 randomised patients	Yes

: ANOVA, analysis of variance.
a: For attrition bias in RCTs, the following criteria were applied to classify the risk as LOW, UNCLEAR or HIGH: LOW, no loss to follow-up, or < 10% loss to follow-up and details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers); UNCLEAR, loss to follow-up less than 10% and no details on ITT/effect of missing data given OR loss to follow-up > 10%, but details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers); HIGH, loss to follow-up > 10% and no details given of an ITT analysis or an account of how missing data are unlikely to affect the results (e.g. similarity of characteristics of dropouts and completers) OR loss to follow-up > 30%. These cut-off points are arbitrary and have not been used in any sensitivity analyses. Any likely impact of quality on results has been discussed narratively.

TABLE 52

Quality assessment controlled non-randomised studies (NIV vs. control)

Study	Prospective or retrospective	How were NIV and control groups selected (e.g. from the same source, at the same time)?	Were NIV and control groups similar at baseline?	Blinding of outcome assessment	Was blinding of outcome assessment the same for both groups?	NIV group incomplete outcome data	Control group incomplete outcome data	Was follow-up time and method of follow-up the same in both groups?	Selective reporting
Budweiser et al. 2007⁹⁶	Prospective	NIV initiated or attempted in most patients. Those who refused NIV from the beginning or could not tolerate NIV during hospital stay (mostly because of mask intolerance) formed the control group	Most baseline characteristics appear to be similar. There was a difference in LTOT at discharge (95% NIV group and 81% in control group); a subgroup analysis was performed for patients on LTOT (with or without NIV) only; also used as variable in adjusted HR	No details. Only survival as an outcome measure (objective measure, blinding not as relevant)	No details. As survival is an objective measure, blinding is not as relevant	Not specifically stated. Details on those who died or discontinued but no mention of losses to follow-up. 12/99 (12%) discontinued NIV (3/12 because of mask intolerance, 3/12 because of decreased motivation, 4/12 reported improvement of symptoms, 1/12 had lung transplantation, 1/12 was not specified). No details on discontinuation rates of LTOT	Not specifically stated. Details on those who died given but no mention of losses to follow-up. No details on discontinuation rates of LTOT	Follow-up time was slightly longer in the NIV group (19.8 months vs. 12.9 months); as a result of earlier deaths in non-NIV group? Patients undergoing long-term NIV were regularly admitted for re-evaluation to hospital and thus may have had more intense contact than the control group	No apparent selective reporting
Clini et al. 1998⁹³	Prospective	Those not complying with NIV during in-hospital adaptation period. Lack of compliance defined as the patient’s inability to use NIV properly for at least 5 hours for even 1 night (subjective intolerance, excessive air leaks)	Stated that the two groups were not different for anthropometric and functional characteristics; similar severity of airway obstruction and hyperinflation; previous smoking habit and medical therapy did not differ between the two groups, neither did numbers of acute exacerbations over previous 2 years and rates of endotracheal intubation	6MWT performed and recorded under supervision of a nurse not involved in the study	Appears to be	21/49 did not tolerate NIV in adaptation period and formed the control group. No further losses to follow-up reported apart from deaths. Unclear how many patients are contributing to results at different time points. No mention of ITT analysis or how missing data were handled		Yes	No apparent selective reporting
Clini et al. 1996⁹⁴	Prospective (and also a historical control – data not extracted)	Patients matched for anthropometric, functional and blood gas data. Patients in NIV group had suffered from at least one episode of acute respiratory failure needing non-invasive mechanical ventilation or had undergone at least two admissions to respiratory units for severe exacerbations not requiring ventilatory support; 7/17 patients included in the control group had undergone ICU admissions needing mechanical ventilation but were not able to perform long-term NIV	No significant differences at baseline	No details	No details	No details (except deaths). Blood gases based on varying numbers of patients. Numbers not stated for other hospital-related outcomes		Yes	No apparent selective reporting
Heinemann et al. 2011⁹⁷	Retrospective	Those not meeting criteria for NIV (i.e. PaCO₂ > 52.5 and/or pH 7.35) formed the control group	No. Those discharged without NIV were significantly older and had higher SAPS-II (simplified acute physiology score-II) scores (simplified acute physiology score-II) at admission, but better pulmonary function and showed a trend towards lower severity of hypercapnia	No details	No details	No details (only on deaths)		No. Those on NIV received more intensive medical care as they went for additional check-ups	No apparent selective reporting
Laier-Groeneveld and Criee 1995⁸⁸	Retrospective	Control group were treated during same time period in same clinic	No. Control group were normocapnic, NIV group were hypercapnic. Control group patients would not have been able to receive NIV. Higher pO₂ and FEV₁ in control group (although not clear if statistically significant difference)	Stated that investigator was blinded regarding NIV but in context of measuring arrhythmias, so not relevant. Not relevant for survival		No details	No details	Appears yes	No apparent selective reporting
Lu et al. 2012⁹⁸	Retrospective	All groups are selected from the patients who were in hospital from January 2009 to December 2010 and with stable COPD (PaCO₂ ≥ 55 mmHg) after treatment. No details on how control group was selected vs. NIV group	All baseline characteristics appear to be similar	No details	No details	No losses to follow-up	Not specifically stated. Details on those who were lost to follow-up and died	Yes	No apparent selective reporting
Milane and Jonquet 1985⁸⁶	Retrospective	Group selected from patients hospitalised during 1973–1983 because of an exacerbation (same centre). Blood gas measurements determined eligibility for NIV or not	Stated that similar for age. Slightly better blood gas values in those not receiving home NIV	No details	No details	No details	No details	NIV patients received additional home visits to check medication and ventilator technique	No apparent selective reporting
Pahnke et al. 1997⁸⁷	Retrospective	Control group – those who refused NIV a priori or within first 3 months	No details	No details	No details	No details	No details	No details	No apparent selective reporting
Paone et al. 2014⁸⁹	Prospective	Patients allocated to NIV or control group on basis of compliance (during an NIV trial) and/or willingness to be trained	Propensity matched scores obtained and used for adjusted analyses. No obvious difference between groups at baseline	No details (only hospitalisations and survival as outcome measures, both objective so blinding less relevant)	No details	No details on loss to follow-up during the 24-month follow-up period. Four patients crossed over to NIV, but they are included in the main analysis		Appears to be. All had regular clinical evaluations every 2 months	No apparent selective reporting
Tsolaki et al. 2008⁹⁵	Prospective	Those who had good compliance with ventilator during hospital stay but refused to continue NIV at home on a long-term basis	No statistically significant differences between groups for baseline characteristics; trend towards higher BMI in NIV group	No details	No details	3/27 early dropouts because of poor compliance with ventilator (< 5 hours/day). Appear to be no further dropouts (except deaths). Numbers assessed for outcomes at different time points not specifically stated. Early dropouts not included in analysis	Appear to be no losses to follow-up (except deaths). Numbers assessed for outcomes at different time points not specifically stated	Stated that all patients followed up in an identical pattern and closely supervised for adherence to medical treatment	Not all time points presented for PaO₂ and HCO₃. Otherwise no selective reporting

: BMI, body mass index.

TABLE 53

Quality assessment crossover RCTs (NIV vs. NIV)

Study	Random sequence generation	Allocation concealment	Blinding of patients	Blinding of outcome assessment	Incomplete outcome data NIV group	Selective outcome reporting	Is it clear that the order of receiving treatments was randomised?	Can it be assumed that the trial was not biased from carry-over effects?	Dropouts after first treatment period (how incorporated into analysis?)	Are data available from both treatment periods?	Was a form of paired analysis used?
Dreher et al. 2010³²	UNCLEAR	UNCLEAR	HIGH	UNCLEAR	HIGH	LOW	No Cochrane guidelines for rating risk of bias
Dreher et al. 2010³²	Randomised two-period crossover design – no further details	No details	No. Open label	No. Described as open-label study. Investigators not blinded	Two dropouts from low intensity treatment arm during first period (1/2 refused treatment in hospital because of intolerance, another stopped at home despite successful establishment in hospital). 2 patients refused to swap over to low intensity after first treatment period with high intensity. Four fewer patients during treatment period 2 (23%). No details on whether or not characteristics of dropouts were similar to completers. 17 patients during treatment period 1, 13 patients during treatment period 2	No apparent selective reporting	Yes (patients randomised to receive a specific sequence)	No effect found based on period effect tests and test for carryover effects	4/17 during first treatment period. Only patients included who received both treatments (?)	Yes	Yes. Paired analyses for all patients who received the respective treatments
Oscroft et al. 2010¹¹²	UNCLEAR	LOW	LOW	LOW	UNCLEAR	LOW	No Cochrane guidelines for rating risk of bias
Oscroft et al. 2010¹¹²	No details on method of randomisation	Sealed opaque envelopes	Trial subjects were blinded to the ventilators’ mode	Spirometry, assessment of lung volumes, gas diffusion and shuttle walk test assessments performed by technicians blinded to the treatment group	1/25 lost to follow-up (4%). 1/25 withdrew consent after exacerbation (during volume-assured NIV period). Not all patients were able to participate in the shuttle walk test (results based on 18/24). FEV₁ and FVC were based on 23 participants. Unclear for SGRQ and SF-36	No apparent selective reporting	Yes	No details	One withdrew during volume assured NIV (not clear if treatment period 1 or 2). Not included in analysis	Yes	Yes
Murphy et al. 2012¹¹³	UNCLEAR	LOW	UNCLEAR	UNCLEAR	HIGH	LOW	No Cochrane guidelines for rating risk of bias
Murphy et al. 2012¹¹³	Randomised, two–treatment crossover design	‘Randomisation via sealed envelope allocation’¹¹³	Trial described as single blind (no further details)	Trial described as single blind (no further details)	5/12 (42%) patients withdrew during trial period, 4/5 of those during high-pressure ventilation. Stated that there were no significant differences between completers and withdrawers in terms of age, sex, anthropometrics, gas exchange, spirometry or ventilator settings, with the exception of FVC (46% completers, 69% withdrawers)	No apparent selective reporting	No details	Carryover effect could not be excluded owing to a relatively short washout period	No details, it appears dropouts were not included in analysis	Yes	Yes

Copyright © Queen’s Printer and Controller of HMSO 2015. This work was produced by Dretzke et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK321582

Contents

< Prev Next >

PubReader
Print View
Cite this Page
Dretzke J, Blissett D, Dave C, et al. The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation. Southampton (UK): NIHR Journals Library; 2015 Oct. (Health Technology Assessment, No. 19.81.) Appendix 2, Quality assessment of studies included in clinical effectiveness review.
PDF version of this title (2.6M)

Other titles in this collection

Health Technology Assessment

Recent Activity

Clear Turn Off Turn On

Quality assessment of studies included in clinical effectiveness review - The co...
Quality assessment of studies included in clinical effectiveness review - The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation
Details of non-invasive ventilation used in studies included in clinical effecti...
Details of non-invasive ventilation used in studies included in clinical effectiveness review - The cost-effectiveness of domiciliary non-invasive ventilation in patients with end-stage chronic obstructive pulmonary disease: a systematic review and economic evaluation

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

Bookshelf