|Lansoprazole compared with histamine2-receptor antagonists in healing gastric ulcers: a meta-analysis
|Tunis S R, Sheinhait I A, Schmid C H, Bishop D J, Ross S D
To compare the gastric ulcer healing rates of lansoprazole with histamine2-receptor antagonists (H2RAs).
MEDLINE (1966 - 16 January 1996) including all languages, using the subject heading 'stomach ulcer' and text terms 'lansoprazole', 'lanzor', 'ogast', 'takepron', 'Prevacid', 'AG-1749', and 'EN-123065'. In addition, Current Contents was searched weekly, and bibliographies were handsearched. Unpublished trials were obtained from the drug's manufacturer.
Study designs of evaluations included in the review
Randomised controlled trials (RCTs) comparing lansoprazole with a H2RA, recruiting at least ten patients.
Specific interventions included in the review
Lansoprazole 30 mg/day compared with H2RAs (ranitidine, famotidine, cimetidine, and roxatidine).
Participants included in the review
Studies recruiting patients with a gastric ulcer of at least 3 mm in diameter at entry to the study, confirmed by endoscopy, were included. The range of mean ulcer diameters across studies was 7-21 mm, and the range of mean ages was 49-64 years.
Outcomes assessed in the review
Ulcer healing evaluated by upper endoscopy after 4 or 8 weeks of treatment, or both.
How were decisions on the relevance of primary studies made?
Decisions for selection were made independently by two reviewers who were blinded to source of studies and treatment identifiers.
Assessment of study quality
Assessment of randomisation, patient and investigator blinding, and accounting for withdrawals and drop-outs. Each included study was assessed using the scoring system of Jadad et al. (see Other Publications of Related Interest). This system assigns a score from 0 to 5 for study quality, with 5 indicating the highest quality.
Categories of data extracted included demographic information, withdrawals, and drug dose and schedule. The rates of endoscopic healing after 4 and 8 weeks of treatment were evaluated for each trial on both an intention-to-treat and an evaluable patient basis. The intention-to-treat analysis included all randomised patients who met the study's stated eligibility criteria and initiated treatment; patients who were lost to follow-up were considered as treatment failures, and patients with gastric cancer and active gastrointestinal haemorrhage were excluded. The evaluable patient protocol excluded patients who had no follow-up endoscopy. Data were extracted independently by two reviewers who were blinded to source of studies and treatment identifiers, with discrepancies resolved by discussion. Data were entered into electronic spreadsheets, and all computerised data entries were checked against the extraction forms for accuracy.
Methods of synthesis
How were the studies combined?
Healing data were combined and analysed by three different statistical methods: a fixed-effect model with individual studies weighted by the inverse of the variance, a random-effects model weighted by the inverses of the sums of the within-study variances and the between-study variances, and a Bayesian model using the assumption that results were based entirely on the collected data from primary studies (see Other Publications of Related Interest), with analysis stratified by trial so that comparisons between treatment groups always involved patients within the same trial. In each analysis, the log risk ratio and risk differences were calculated, together with associated 95% confidence intervals (95% CI) and Bayesian posterior probability.
How were differences between studies investigated?
Regression models were constructed by introducing covariates into the population stage of the hierarchical Bayesian model using the log risk ratio as the outcome. A separate analysis was carried out including only trials with a quality score of at least 3. Results of unpublished trials were also considered separately.
Results of the review
Thirteen RCTs were included overall, eight unpublished and five published. Twelve trials were included in the intention-to-treat analysis (n=1655) and 13 in the evaluable patient analysis (n=1527).
The following results are those generated from using the more conservative Bayesian model for combining data. Similar results were found using the fixed-effect and random-effects models.
Intention-to-treat analysis (12 trials):
All trials favoured lansoprazole over the histamine2-receptor antagonists (H2RAs.) For ulcer healing at 4 weeks, the risk ratio (RR) with associated 95% confidence intervals (95% CI) was 1.33 (95% CI: 1.19, 1.49) for all trials, and 1.23 (95% CI: 1.09, 1.39) for six trials of higher methodological quality. For ulcer healing at 8 weeks, the RR was 1.12 (95% CI: 1.06, 1.19) for all trials, and 1.08 (95% CI: 1.01, 1.16) for six trials of higher methodological quality.
Evaluable patient analysis (13 trials):
For ulcer healing at 4 weeks, the RR was 1.37 (95% CI: 1.23, 1.52) for all trials, and 1.27 (95% CI: 1.14, 1.42) for six trials of higher methodological quality. For ulcer healing at 8 weeks, the RR was 1.10 (95% CI: 1.05, 1.15) for all trials, and 1.07 (95% CI: 1.03, 1.11) for six trials of higher methodological quality.
Analysis of six co-variates in 13 trials used in the analysis of evaluable patients demonstrated a tendency for studies conducted in Japan to show greater relative advantages for lansoprazole compared with H2RAs. No other covariates, including smoking and alcohol use, accounted for any difference in results. Unpublished trials tended to report higher healing rates for both lansoprazole and the H2RAs compared with published trials.
Lansoprazole heals ulcers more quickly compared with H2RAs and also achieves higher overall rates of healing.
Overall, this is a well conducted systematic review. The selection criteria for primary studies were appropriate and clearly explained. The literature search was limited to one database. Had other sources been accessed, it is possible that other, relevant material, could have been identified. The authors obtained unpublished material from the drug manufacturer; attempts to locate unpublished studies from other sources would have further enhanced the search strategy. Adequate details of included trials were presented in tables. Trials were systematically assessed for methodological quality, but the scoring system used only addressed three methodological aspects. The methods used for pooling data were appropriate and clearly described; it is possible that a sensitivity analysis on the basis of the different histamine2-receptor antagonists could have been useful. Useful details were given relating to the process of the review. The author's conclusions appear to follow on from the presented evidence. It should be noted that this review was commissioned by the manufacturer of lansoprazole. However, the authors state that the protocol, analysis, and report, were prepared independently.
Implications of the review for practice and research
Practice: The authors do not state any implications for practice.
Research: The authors state that the studies in this meta-analysis did not address the possibilities that duration of therapy may be shortened with lansoprazole compared with a H2RA or that lansoprazole may be more effective than a H2RA combined with antibiotics in the eradication of Helicobacter pylori infection associated with gastric ulcers. Therefore, the inference is that these may be worthwhile as further areas of research.
Tunis S R, Sheinhait I A, Schmid C H, Bishop D J, Ross S D. Lansoprazole compared with histamine2-receptor antagonists in healing gastric ulcers: a meta-analysis. Clinical Therapeutics 1997; 19(4): 743-757
Other publications of related interest
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. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:19-48.
3. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88. 4. DuMouchel W. Meta-analysis for dose-response models. Stat Med 1995;14:679-85. 5. DuMouchel W. Bayesian meta-analysis. In: Berry D, editor. Statistical methodology in the pharmaceutical sciences. New York (NY): Marcel Dekker; 1990. p. 509-29.
Subject indexing assigned by NLM
2-Pyridinylmethylsulfinylbenzimidazoles; Anti-Ulcer Agents /therapeutic use; Helicobacter Infections /drug therapy; Histamine H2 Antagonists /therapeutic use; Lansoprazole; Omeprazole /analogs & Randomized Controlled Trials as Topic; Stomach Ulcer /drug therapy; Wound Healing /drug effects; derivatives /therapeutic use
Date bibliographic record published
Date abstract record published
This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.