CRD summary

This review concluded that non-steroidal anti-inflammatory drug (NSAID) use had a protective effect from gastric cancer. Given that only epidemiological studies of unknown quality were included and that there were potential issues with study synthesis, the authors’ conclusions should be interpreted with some caution.

Authors' objectives

To quantify the association between non-steroidal anti-inflammatory drug (NSAID) use and gastric cancer.

Searching

MEDLINE, EMBASE and Web of Sciences databases were searched from January 1966 to March 2009; search terms were provided. Bibliographies of retrieved articles were also reviewed for additional relevant studies. Only English language studies were included.

Study selection

Cohort or case-control studies that reported an association between non-steroidal anti-inflammatory drugs (NSAIDs) and gastric cancer were eligible for inclusion.

Included studies administered both aspirin and non-aspirin based NSAIDs. Exposure assessment was either by interview, questionnaire or according to database records. Study durations ranged from two to nine years. Most studies were described as population-based; three were described as hospital-based. Study populations were not described. The incidence of any gastric cancer or cardia/non-cardia gastric cancer was assessed. Where adjusted results were reported, between one and 11 factors were adjusted for including age, sex, race, smoking, alcohol, dietary factors, symptoms, and social class.

Two authors reviewed all studies.

Assessment of study quality

The authors did not state that they assessed validity.

Data extraction

Crude and adjusted data were extracted. Crude data referred to the number of people with or without gastric cancer in the exposed and control groups. For the adjusted data, risk ratios (RR) or odds ratios (OR), with 95% confidence intervals (CIs), that had been adjusted for one or more potential confounders were extracted. Standardised incidence ratios (the ratio of observed number of cancers to expected number of cancers based on reference cancer incidence rates for the general population) were accepted as adjusted data (as reported in some cohort studies).

Data extraction was performed independently by two authors using a standardised form. Any discrepancies were resolved by consensus.

Methods of synthesis

Between study heterogeneity was examined used I². Where I² was 50% or less, a fixed-effect Mantel-Haenszel model was used to estimate pooled odds ratios. Separate analyses were conducted for crude and adjusted data. For I² over 50%, a random-effects model was adopted.

Subgroup analyses according to study design (case-control versus cohort), type of drug (all NSAIDs, aspirin only and non-aspirin NSAIDs), site of cancer (cardia and non-cardia) and sample source (hospital-based versus population-based samples) were carried out.

Funnel plots were used to identify any publication bias; funnel plot asymmetry was tested for using linear regression. Duval and Tweedie’s trim-and-fill method was used to adjust the pooled odds ratio if publication bias was identified.

Results of the review

Twenty-one studies were included in the review, comprising 12 case-control studies, one nested case-control and eight cohort studies. The number of study participants was reported in such a way that the overall number of participants across studies could not be calculated.

Non-steroidal anti-inflammatory drug (NSAID) use was found to reduce the incidence of gastric cancer by 21% for the unadjusted analyses (RR 0.79, 95% CI 0.73 to 0.86; I²=71.8%; 16 studies) and by 29% for the adjusted data (RR 0.71, 95% CI 0.64 to 0.79; I²=59.8%; 18 studies). Give the heterogeneity, random-effects models were used to pool the data. Significant publication bias was identified (p<0.01); following application of the trim-and-fill method, smaller reductions in risk of gastric cancer were produced (summary crude RR 0.89, 95%CI 0.83 to 0.97 and adjusted RR 0.81, 95%CI 0.73 to 0.89).

The subgroup analyses showed similar effects for study design (for both crude and adjusted data) and type of NSAID (crude data only). The pooled analysis of adjusted data showed a stronger protective effect from aspirin (RR 0.69, 95%CI: 0.55, 0.86), although the seven contributing studies were highly heterogeneous (I²=80.6%). The analyses by cancer site found a stronger protective effect for non-cardia cancers with NSAIDs for both crude and adjusted data (reductions in incidence of 24% (95% CI 16 to 31) for the crude data and 37% (95% CI 30 to 43) for the adjusted data, although the crude data were fairly heterogeneous (I²⁼60.5%). Stronger effects were identified for hospital-based studies, with reduction in cancer incidence of 35% (95% CI 19 to 48; four studies) for crude data and 39% (95%CI 24 to 51; three studies) for adjusted data, but the number of studies was small.

Authors' conclusions

Non-steroidal anti-inflammatory drug use reduced the risk of gastric cancer. The effect was strongest for data adjusted for one or more confounding risk factors.

CRD commentary

The ain of the review was clearly set out. However, minimal inclusion criteria were described and few study details were provided, which made it difficult to judge how objectively the inclusion criteria were applied and how generalisable the study results were. For example, no description of study participants was provided and no assessment of the comparability of cases and controls appear to have been carried out. Three major databases were searched, but no attempt was made to identify grey literature; only English language studies were included, so potentially relevant studies may have been missed and publication and language bias was possible. Significant publication bias was found in the analysis. Two reviewers assessed studies for inclusion and carried out data extraction, which reduced the risk of reviewer bias.

No assessment of study quality was reported, so the reliability of individual study results was unclear. In the synthesis of studies, the authors combined cohort and case-control studies in the same analysis (appearing to have combined relative risks and odds ratios - different measures which should not be used interchangeably); they also included studies which had been adjusted for a variety of different factors. This made it difficult to be confident that the studies were sufficiently similar to allow meta-analysis.

Given that only epidemiological studies were included, there was no study quality assessment and that there were potential problems with study synthesis, the authors’ conclusions should be interpreted with some caution.

Implications of the review for practice and research

The authors did not state any implications for practice or further research

Funding

Not stated

Bibliographic details

Tian W, Zhao Y, Liu S, Li X. Meta-analysis on the relationship between nonsteroidal anti-inflammatory drug use and gastric cancer. European Journal of Cancer Prevention 2010; 19(4): 288-298. [PubMed: 20386312]

Original Paper URL

http://journals.lww.com/eurjcancerprev/Abstract/2010/07000/Meta_analysis_on_the_relationship_between.6.aspx

Indexing Status

Subject indexing assigned by NLM

MeSH

Anti-Inflammatory Agents, Non-Steroidal /therapeutic use; Case-Control Studies; Cohort Studies; Humans; Risk Factors; Stomach Neoplasms /epidemiology /prevention & control

AccessionNumber

12010006018

Database entry date

14/09/2011

Record Status

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.