Research

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

BMJ 2006; 332 doi: http://dx.doi.org/10.1136/bmj.38677.559005.55 (Published 05 January 2006) Cite this as: BMJ 2006;332:11
  1. A B Chang, consultant in paediatric respiratory medicine (annechang{at}ausdoctors.net)1,
  2. T J Lasserson, review group coordinator2,
  3. T O Kiljander, consultant in respiratory medicine3,
  4. F L Connor, consultant in paediatric gastroenterology4,
  5. J T Gaffney, research nurse1,
  6. L A Garske, consultant in respiratory medicine5
  1. 1 Department of Respiratory Medicine, Royal Children's Hospital, Herston, Brisbane, Qld 4029, Australia
  2. 2 Cochrane Airways Group, St George's Hospital Medical School, London SW17 0RE
  3. 3 Medical Center Mehiläinen, Kauppiaskatu 8, FIN-20100 Turku, Finland
  4. 4 Department of Gastroenterology, Royal Children's Hospital, Herston, Brisbane
  5. 5 Princess Alexandra Hospital, Brisbane, Woolloongabba, Qld 4102, Australia
  1. Correspondence to: A Chang
  • Accepted 27 October 2005

Abstract

Objective To evaluate the efficacy of treatment for gastro-oesophageal reflux disease (GORD) on chronic cough in children and adults without an underlying respiratory disease.

Design Systematic review and meta-analysis.

Data sources Cochrane, Medline, and Embase databases, references from review articles.

Included studies Randomised controlled trials on GORD treatment for cough in children and adults without primary lung disease. Two reviewers independently selected studies and extracted paediatric and adult data on primary (clinical failure) and secondary outcomes.

Results 11 studies were included. Meta-analysis was limited to five studies in adults that compared proton pump inhibitors with placebo. All outcomes favoured proton pump inhibitors: the odds ratio for clinical failure (primary outcome) was 0.24 (95% confidence interval 0.04 to 1.27); number needed to treat (NNT) was 5 (harm 50 to ∞ to benefit 2.5). For secondary outcomes, the standardised mean difference between proton pump inhibitors and placebo was −0.51 (−1.02 to 0.01) for mean cough score at the end of the trial and −0.29 (−0.62 to 0.04) for change in cough score at the end of the trial. Subgroup analysis with generic inverse variance analysis showed a significant mean change in cough (−0.41 SD units, −0.75 to −0.07).

Conclusion Use of a proton pump inhibitor to treat cough associated with GORD has some effect in some adults. The effect, however, is less universal than suggested in consensus guidelines on chronic cough and its magnitude of effect is uncertain.

Introduction

Cough is the most common symptom presenting to general practitioners.1 Chronic cough considerably impairs quality of life in adults and worries parents of children with cough. Prolonged or chronic cough has been variously defined as a cough that persists for more than three to eight weeks and non-specific cough defined as non-productive cough in the absence of identifiable respiratory disease or known cause.2

Gastro-oesophageal reflux (GOR)—that is, reflux of gastric contents into the oesophagus—can be acid or non-acid. Reflux may be physiological and is associated with a range of gastrointestinal symptoms (abdominal pain, halitosis, etc) and extraoesophageal symptoms (cough, hoarseness, etc).3 Cohort studies in adults suggest that GOR disease (GORD) related to acid causes 21–41% of chronic non-specific cough.1 Guidelines on chronic cough suggest use of empirical treatment for GOR,4 5 including a therapeutic trial of three to six months of treatment for GORD.6 Although laboratory studies have shown a temporal relation between acid in the oesophagus and cough, some studies have shown that the cough resolves only after a mean of 169–179 days after treatment.6 Other studies have shown that acid GORD is associated with, but is not the cause of, cough.7

Current treatments for GORD include conservative measures (diet, positioning, etc), pharmaceuticals (acid suppressants such as histamine H2 receptor antagonists, and proton pump inhibitors; prokinetic agents such as domperidone, metoclopramide, and cisapride), and surgical approaches (fundoplication). These well established treatments for GOR, however, may not be beneficial for associated cough or may increase respiratory morbidity.8 We examined the efficacy of treatments for GOR on non-specific chronic cough in adults and children in a systematic review. This review is based on a Cochrane systematic review.9

Methods

We used QUOROM guidelines, Cochrane collaboration method, and software (RevMan 4.2) (see bmj.com). Studies in adults and children were eligible if they were randomised controlled trials of any GORD treatment for chronic cough (lasting more than three weeks) where cough was an outcome and not primarily related to an underlying respiratory disorder. We classified the evaluated treatment regimens by type: anti-reflux conservative measures (for example, positioning, diet), H2 receptor antagonists, proton pump inhibitor, and surgical therapy. Our primary outcome was proportion of participants who were not cured at follow-up (failure to cure). Secondary outcomes were proportion of participants not substantially improved at follow-up, mean difference in cough indices (frequency of cough, scores, sensitivity), proportion who experienced adverse effects (such as rash, surgical morbidity, etc), and proportions who experienced complications (requirement for change in medication, repeat surgery, etc). We determined the proportions of participants who failed to improve on treatment using a hierarchy of assessment measures (see bmj.com).

We use the search strategy standardised by the Cochrane Airways Group as well as references in relevant publications and written communication with the authors of papers. Two reviewers independently reviewed literature searches, selected articles, and extracted data. We used the κ statistic to assess agreement between reviewers. Details of other statistics including a priori, subgroup, and sensitivity analyses are on bmj.com. When we combined data with parallel studies we used only data from the first arm of crossover trials.

Fig 1
Fig 1

Details of included and excluded trials

Results

We identified 763 potentially relevant titles and reviewed 84 papers for inclusion (fig 1). There was 92% agreement for inclusion of the 11 studies (three in children, eight in adults, n = 383) that met criteria for the systematic review (table). All but one10 were single centre studies; the only multicentre study was also the only study supported by industry.10 All but two studies were in English.11 12 Additional data were sought from all authors of English articles, and two groups provided raw data.13 14 Jadad and quality assessment scores varied (table). Agreement for quality of studies was excellent; the weighted κ score was 0.71 for Jadad score and 0.89 for quality assessment.

Table 1

Description of studies included in review

View this table:

Paediatrics

Two studies in children found some improvement in GORD symptoms referring to the gastrointestinal system but data for the effect on cough were inconsistent.10 15 One study found no significant effect of cisapride or domperidone on cough associated with GORD.11 There were insufficient data in infants and children for meta-analysis and no randomised controlled trials on the use of proton pump inhibitors or surgery.

Adults

Of the eight studies in adults, six were published full articles,1214 1618 one was an abstract,19 and one was duplicated in a conference report and abstract.20 Five were parallel studies, and three were crossover studies. Five studies compared proton pump inhibitors with placebo but used varying doses and frequency (studies based in otolaryngology clinics14 16 17 used higher doses of proton pump inhibitors, table). Although all studies included patients with cough that was presumed to be associated with GORD, criteria for entry into the studies varied. In three studies, participants enrolled through the otolaryngology department had “laryngitis” symptoms.14 16 17 In these studies, the exclusion criterion of primary lung disease (for example, asthma defined by hyper-responsiveness) was not as stringently applied as in the studies that enrolled participants through medical outpatients.12 13 18 20 In all but one study,12 the presence of GORD was confirmed objectively. No studies used non-acidic reflux as an entry criterion.

Outcome measures for all studies were various subjective cough scales; none used objective monitoring of cough. One study compared proton pump inhibitors with ranitidine12 and another was a 2x2 factorial design with cisapride and dietary intervention.19 We could not extract data from the single study on H2 receptor antagonists compared with placebo; trialists reported an improvement in cough scores with intervention in all participants and the effect was significant at two weeks.20 In the six studies that compared proton pump inhibitors or cisapride with placebo, all but one study showed no difference in improvement in cough scores in the active and placebo arms.12 14 1619 Participants took proton pump inhibitors for two to three months, and only one study followed up patients after the trial.18

In the meta-analysis, the primary outcome (clinical failure) was the only outcome where we could use “intention to treat” data for all included studies displayed in the forest plot. The pooled odds ratio effect estimate was 0.24 (95% confidence interval 0.04 to 1.27 (fig 2; 49 participants)). Number needed to treat for benefit was 5 (harm 50 to ∞ to benefit 2.5).21 Evaluation of the mean cough score at the end of the trial (secondary outcome), in a pooled (62 participants) analysis of three studies13 14 16 showed a non-significant difference between groups (standardised mean difference −0.51, −1.02 to 0.01). When we restricted this pooled analysis to crossover studies,13 14 the difference was also not significant (−0.29, −0.62 to 0.04). The outcome of change in cough scores at the end of the intervention pooled from four studies (77 participants)13 14 16 17 was not significant (−0.29, −0.62 to 0.04 (fig 3)). Data from the two crossover trials13 14 that assessed mean change in symptoms showed a significant difference (−0.41 SD units, −0.75 to −0.07 (fig 4, 35 subjects per arm)).

Fig 2
Fig 2

Meta-analysis of primary outcome (clinical failures—that is, patients still had cough at the end of the trial or reporting period), analyses by intention to treat (49 participants included in meta-analysis)

Fig 3
Fig 3

Meta-analysis of standardised cough scores at end of intervention (77 participants included in meta-analysis)

Fig 4
Fig 4

Meta-analysis of data from the two crossover trials13 14 assessing mean change in symptoms (subanalysis). Using generic inverse variance for crossover studies, use of proton pump inhibitors was significantly more likely than placebo to reduce cough scores at end of intervention, −0.41 SD units (−0.75 to −0.07)

Few studies determined “time to response” and duration of treatment effect was limited; after four weeks of intervention change in cough score favoured proton pump inhibitors but was not significantly different (P = 0.09, standardised mean difference −0.51, −1.08 to 0.06, from two studies,13 16 51 participants). The effect at eight weeks compared with four weeks of intervention13 16 was also not significant (−0.44, −1.04 to 0.16). The single study that specifically reported time to response18 was also the only study that followed up participants after the trial (open study followed by randomised controlled trial), and the effect of treatment occurred within 5–14 days.18

None of the studies reported any relevant adverse events with the interventions, and it was not possible to look for an association between level of risk and benefit. Sensitivity analyses did not alter any results.

All studies that provided sufficient data reported significant improvement in gastrointestinal symptoms. In three studies, however, there was no difference in improvement between the intervention and placebo groups.14 16 17 All three studies on otorhinolaryngological symptoms reported significant improvement in symptoms of laryngitis over time.14 16 17 In two studies, the difference in subjective and semi-objective (laryngoscopy) improvement was similar in treatment and placebo groups.14 17 One study found significant differences between proton pump inhibitors and placebo in hoarseness and throat clearing but not in throat pain.16

Discussion

In this systematic review of the effects of treatment for GOR for cough associated with GORD, we identified 11 randomised controlled trials. In the meta-analysis of five studies, subgroup analysis indicated that treatment of cough associated with GORD improves subjective cough in participants with non-specific cough. There was no effect in the pooled analysis of the main outcomes, although all favoured proton pump inhibitors. The NNT to achieve cough resolution was 5 (the confidence interval included infinity). This review also highlights the large placebo and time period effect of treatment for chronic cough.

Differences between trial types

European5 and American guidelines4 on the management of chronic cough recommends the use of GORD therapy (including as empirical therapy). The meta-analyses for main outcomes were not significant but all favoured treatment with proton pump inhibitors and subanalyses yielded significant results. Thus, proton pump inhibitors are probably effective in this population. Nevertheless, the small effect (considering other meta-analyses on cough22 and large confidence intervals) of GORD treatment on cough that we found contrasts with results of non-controlled trials, which almost invariably report high rates of improvement: 100% with proton pump inhibitors alone, 86% with proton pump inhibitors with motility agents, and cure rates of 81% by surgical intervention.9 The difference between the effect of treatment of GORD on cough seen in randomised controlled trials and uncontrolled trials may be related to the period or placebo effect, which is as high as 85%.23

Other possible contributors are selection of patients, the degree of acid inhibition, length of therapy, outcomes of cough measured, role of non-acid reflux, and the presumed GORD related cough which in fact may not have been caused by GORD. Both GOR and cough are common and often coexist, but an association does not imply cause and effect.7 Our finding of an effect of treatment with proton pump inhibitors to improve cough in subgroup analysis, however, does suggest that in a proportion of adults with chronic cough and GORD, GORD is a contributing cause. The degree of acid inhibition achieved with treatment is probably a small factor, if any, given that most studies achieved a resolution of GORD gastrointestinal symptoms. It is possible, however, that different degrees of acid suppression are required to control the different manifestations of GORD. Results from the meta-analysis on length of therapy were inconclusive.

Limitations of this review

The validity of this systematic review is also hindered by the disparate nature of the interventions. In all studies participants were selected for gastrointestinal symptoms or objective evidence of GORD, and most, but not all, had cough. GORD criteria also varied between studies, which may have influenced results. Most studies did not use criteria for GORD specified by international guidelines.3 24 There are no data from randomised controlled trials on the efficacy of GORD treatment for chronic cough in the absence of GORD symptoms. In view of the possibility of period or placebo effect leading to misdiagnosis of chronic non-specific cough as due to GORD, we believe that objective confirmation of GORD is preferable to empirical therapy in patients with chronic non-specific cough and no gastrointestinal symptoms. In those with gastrointestinal symptoms, however, an empirical trial of proton pump inhibitors may be justified.

Proton pump inhibitors are currently the most potent non-surgical intervention for acid-GORD, and no adverse events were reported in these studies. The use of these drugs as a trial of therapy is probably low risk but rational prescription is required.8 Proton pump inhibitors may also cause cough and the package insert for omeprazole includes cough as an adverse event (1.1%).9 In contrast with pharmaceutical interventions, we found no studies on surgical intervention, which is associated with a small but measurable rate of serious adverse events.

Conclusion and future research

From this systematic review and meta-analysis we conclude that proton pump inhibitors for cough associated with GORD probably have some effect in some adults, though the effect is less universal than reported in cohort studies. There are insufficient data to support or refute common recommendations to change the diet or use motility agents either as a single agent or in conjunction with proton pump inhibitors. In children, the absence of data makes specific recommendations impossible, and other causes of cough should be considered in children before any empirical treatment with a prolonged course of GORD medications or interventions. Sufficiently powered parallel placebo controlled randomised controlled trials that include an examination of “time to response,” optimal duration of treatment, effects after treatment, and valid cough outcomes (objective and validated subjective scales) are required to justify international guidelines.4 5 Examinations of different treatments for acid and non-acid GOR are also required.

What is already known on this topic

Gastroesophageal reflux disease (GORD) is common in adults with chronic cough

International guidelines on cough recommend empirical treatment for GORD in those with chronic cough, though evidence for an effect is conflicting

What this study adds

Drugs to reduce the effect of gastric acid in GORD are beneficial in some adults with cough associated with GORD

The effect is less universal than suggested in cohort studies and international guidelines on chronic cough, and the magnitude of the clinical effect is uncertain

Acknowledgments

Embedded Image Details on the search strategy, study selection and characteristics, and data extraction and quality assessment can be found on bmj.com

We thank Chris Cates and Michael McKean for their advice, support, and comments on the protocol and review; Elizabeth Arnold for performing the relevant searches and obtaining the articles; Charlotta Pisinger and Gianni Ferrara for translations; and A J Eherer, A Ing, S E Kopec, S R Orenstein, and J A Vanderhoof for responding or providing additional data.

Contributors: ABC was responsible for initiation, design and direction of review, wrote the protocol and review, selected studies, extracted data, analysed and interpreted results, and is guarantor. LAG selected studies and extracted data, wrote the primary protocol and review, and is guarantor. TJL extracted and analysed data, interpreted results, and wrote the review. TOK wrote the review. JTG extracted data. FLC reviewed the manuscript.

Footnotes

  • Funding ABC is supported by the Australian National Health and Medical Research Council and the Royal Children's Hospital Foundation. This review was supported by a bursary from the Cochrane Child Health Field.

  • Competing interests ABC has received a small grant from Glaxo for a study on asthma. TOK has received lecturing fees from AstraZeneca, Wyeth, Pfizer, and Boehringer Ingelheim, and is a consultant for AstraZeneca. LAG has received lecturing fees from AstraZeneca.

  • Ethical approval Not required.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
View Abstract