Systematic review and meta-analysis of randomised controlled trials of gastro-oesophageal reflux interventions for chronic cough associated with gastro-oesophageal reflux

Appendix [posted as supplied by author]

Search strategy

The search was performed by Elizabeth Arnold (from Cochrane Airway Group) using the topic strategy: "gastro-oesophageal reflux" OR "gastroesophageal reflux" OR "gastro-esophageal reflux" OR "reflux" OR "ger" OR "gerd" OR "acid" OR "esophagus" OR "oesophagus", all as [textword] or [MeSH] AND "cough" as [textword] or [MeSH]. Trials were identified from the following sources: (a) Cochrane Controlled Trials Register (CENTRAL) which includes the Airways Collaborative Review Group Specialised Trials Register, (b) MEDLINE 1966-April 2004; (c) OLDMEDLINE; and (d) EMBASE 1997-2004. In addition (e) list of references in relevant publications were searched and (f) written communication with the authors of trials were undertaken.

Selection

Studies in adults and children were eligible if they were randomised controlled trials of any GORD treatment and chronic cough (>3-weeks) where cough was an outcome and not primarily related to an underlying respiratory disorder (cystic fibrosis, asthma, chronic obstructive airway disease, suppurative lung disease etc) or medication use (angiotensin converting enzyme inhibitor). Treatment regimes evaluated were classified by type; anti-reflux conservative measures, H₂ receptor antagonists, PPI and surgical therapy.

Study characteristics

Trials that satisfied the inclusion criteria were reviewed and the following information recorded: study setting, year of study, source of funding, patient recruitment details (including number of eligible subjects), inclusion and exclusion criteria, criteria used for GOR diagnosis, GOR symptoms, randomisation and allocation concealment method, numbers of participants randomised, blinding (masking) of participants, care providers and outcome assessors, dose and type of GORD therapy, duration of therapy, co-interventions, numbers of patients not followed up, reasons for withdrawals from study protocol (clinical, side-effects, refusal and other), details on side-effects of therapy, and whether intention-to-treat analyses were possible. Data was extracted on the outcomes described previously. Further information was requested from the authors where required.

The primary outcome was (a) proportions of participants who were not cured or not substantially improved at follow up (failure to cure). Secondary outcomes were: (b) proportions of participants who not substantially improved at follow up; (c) mean difference in cough indices (cough diary, cough frequency, cough scores); (d) proportions experiencing adverse effects of the intervention, (e.g. rash, surgical morbidity etc); (e) proportions experiencing complications eg requirement for medication change, repeat surgery etc. The proportions of participants who failed to improve on treatment and the mean clinical improvement were determined using the following hierarchy of assessment measures (i) Objective cough indices (cough frequency, cough receptor sensitivity); (ii) Symptomatic (Quality of life (QOL), Likert scale, visual analogue scale, level of interference of cough, cough diary) as assessed by patient; (iii) Symptomatic (QOL, Likert scale, visual analogue scale, level of interference of cough, cough diary) as assessed by parents/carers; (iv) Symptomatic (Likert scale, visual analogue scale, level of interference of cough, cough diary) as assessed by clinicians; (v) Relevant airway markers consistent with inflammation.

Data extraction and quality assessment

Two reviewers (ABC, LAG) independently reviewed literature searches to identify potentially relevant trials for full review and independently selected trials for inclusion. Quality assessment was performed whereby four components of quality were assessed:

1. Allocation concealment. Trials will be scored as; Grade A: Adequate concealment, Grade B: Unclear, Grade C: Clearly inadequate concealment. (Grade A = high quality).
2. Blinding. Trials will be scored as: Grade A: Participant and care provider and outcome assessor blinded, Grade B: Outcome assessor blinded, Grade C: Unclear, Grade D: No blinding of outcome assessor (Grade A, B = high quality).
3. Reporting of participants by allocated group. Trials will be scored as: Grade A: The progress of all randomised subjects in each group described, Grade B: Unclear or no mention of withdrawals or dropouts, Grade C: The progress of all randomised subjects in each group clearly not described. (Grade A = high quality).
4. Follow-up. Trials will be scored as: Grade A: Outcomes measured in >90% (where withdrawals due to complications and side-effects are categorised as treatment failures), Grade B: Outcomes measured in 80-90%, Grade C: Unclear, Grade D: Outcomes measured in <80%. (Grade A = high quality).

Each study was also assessed using the Jadad 5 point scale.¹ Inter-reviewer reliability for the quality of studies was measured by Kappa. Data were independently extracted by 2 reviewers using a standard performa and cross checked.

Results from studies that met the inclusion criteria and reported any of the outcomes of interest were included in the meta-analyses. In cross-over trials,^2,3 when data was combined with parallel studies only data from the first arm was used.⁴ There are methodological arguments of handling cross over data,^4,5 and we considered it was invalid to use second arm data given the known period effect of cough (ie cough tends to resolve with time⁶ and carry-on effect.

Data were pooled with a fixed effect model. Random effects modelling was applied where we observed significant statistical heterogeneity (I² >20%). For continuous outcomes (e.g. symptom scores) we pooled data using a weighted mean difference (WMD), unless data were only available from different metrics, in which case a generic, standardised scale was used (SMD) which expresses a difference in terms of the standard deviations for each trial. Both methods generate a mean difference and 95% confidence interval for each trial and a summary estimate for the underlying treatment effect across the studies. We extracted a mean and a standard deviation (SD) for the two treatment groups. For crossover we extracted a mean difference between the two groups based upon the published means, and we calculated a standard error for the mean (SEM) based upon a p value from paired t tests. Where individual patient data were available for data pooled with a SMD, we calculated a mean difference and 95% CI for the study, and converted the mean difference to SD units. Altman’s recommendations was used to calculate number needed to treat.⁷

An a priori subgroup analysis was planned for age (adults or children aged <18 years and <12 months); definition of GORD used (acid GORD defined by pHmetry or oesophageal biopsy or non acid reflux; and intervention type (medical or surgical intervention). Medical intervention further subgrouped H₂ antagonist, PPI and conservative therapy). Sensitivity analyses were planned to assess the impact of the potentially important factors on the overall outcomes study quality, study size, variation in the inclusion criteria, differences in the medications used in the intervention and comparison groups, differences in outcome measures, analysis using random effects model, analysis by "treatment received", analysis by "intention-to-treat" and analysis by study design-parallel and cross over studies. Publication bias was tested using a funnel plot.

References

1. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996;17:1-12.

2. Eherer AJ, Habermann W, Hammer HF, Kiesler K, Friedrich G, Krejs GJ. Effect of pantoprazole on the course of reflux-associated laryngitis: a placebo-controlled double-blind crossover study. Scand J Gastroenterol 2003;38:462-7.

3. Kiljander TO, Salomaa ER, Hietanen EK, Terho EO. Chronic cough and gastro-oesophageal reflux: a double-blind placebo- controlled study with omeprazole. Eur Respir J 2000;16:633-8.

4. Elbourne DR, Altman DG, Higgins JPT, Curtin F, Worthington HV, Vail A. Meta-analyses involving cross-over trials: methodological issues. Int J Epidemiol 2002;31:140-9.

5. Cochrane Reviewers' Handbook. Chichester, UK: John Wiley & Sons Ltd., 2004.

6. Chang AB. State of the Art: Cough, cough receptors, and asthma in children. Pediatr Pulmonol 1999;28:59-70.

7. Altman DG. Confidence intervals for the number needed to treat. BMJ 1998;317:1309-12.