Meta-analytic approaches to dose-response relationships, with application in studies of lung cancer and exposure to environmental tobacco smoke

Methods to systematically review and meta-analyse observational studies: a systematic scoping review of recommendations

Background

Systematic reviews and meta-analyses of observational studies are frequently performed, but no widely accepted guidance is available at present. We performed a systematic scoping review of published methodological recommendations on how to systematically review and meta-analyse observational studies.

Methods

We searched online databases and websites and contacted experts in the field to locate potentially eligible articles. We included articles that provided any type of recommendation on how to conduct systematic reviews and meta-analyses of observational studies. We extracted and summarised recommendations on pre-defined key items: protocol development, research question, search strategy, study eligibility, data extraction, dealing with different study designs, risk of bias assessment, publication bias, heterogeneity, statistical analysis. We summarised recommendations by key item, identifying areas of agreement and disagreement as well as areas where recommendations were missing or scarce.

Results

The searches identified 2461 articles of which 93 were eligible. Many recommendations for reviews and meta-analyses of observational studies were transferred from guidance developed for reviews and meta-analyses of RCTs. Although there was substantial agreement in some methodological areas there was also considerable disagreement on how evidence synthesis of observational studies should be conducted. Conflicting recommendations were seen on topics such as the inclusion of different study designs in systematic reviews and meta-analyses, the use of quality scales to assess the risk of bias, and the choice of model (e.g. fixed vs. random effects) for meta-analysis.

Conclusion

There is a need for sound methodological guidance on how to conduct systematic reviews and meta-analyses of observational studies, which critically considers areas in which there are conflicting recommendations.

Electronic supplementary material

The online version of this article (10.1186/s12874-018-0495-9) contains supplementary material, which is available to authorized users.

Empirical evaluation of meta-analytic approaches for nutrient and health outcome dose-response data

The objective of this study is to empirically compare alternative meta-analytic methods for combining dose-response data from epidemiological studies. We identified meta-analyses of epidemiological studies that analyzed the association between a single nutrient and a dichotomous outcome. For each topic, we performed meta-analyses of odds ratios with five approaches: using extreme exposure categories only, two-step approach (first calculated study-specific effects then combined across studies) using unadjusted data, two-step approach using adjusted data, one-step approach (analyzed all data in one regression model) using unadjusted data, and one-step approach using adjusted data. Meta-analyses including only extreme exposure categories gave consistently bigger effects and wider confidence intervals than meta-analyses using all data. Confidence intervals of effect sizes were generally wider in meta-analyses with the two-step approach, compared with the one-step approach. Meta-analyses using unadjusted data and adjusted data differed, with no consistent pattern of discordance in direction, statistical significance, or magnitude of effect. We discourage using meta-analysis approaches that only use data from extreme exposure categories. The one-step approach generally has higher precision than the two-step approach. Sensitivity analysis comparing results between meta-analyses of adjusted and unadjusted data may be useful in indicating the presence of confounding.

Quantification of ETS exposure in hospitality workers who have never smoked

BACKGROUND

Environmental Tobacco Smoke (ETS) was classified as human carcinogen (K1) by the German Research Council in 1998. According to epidemiological studies, the relative risk especially for lung cancer might be twice as high in persons who have never smoked but who are in the highest exposure category, for example hospitality workers. In order to implement these results in the German regulations on occupational illnesses, a valid method is needed to retrospectively assess the cumulative ETS exposure in the hospitality environment.

METHODS

A literature-based review was carried out to locate a method that can be used for the German hospitality sector. Studies assessing ETS exposure using biological markers (for example urinary cotinine, DNA adducts) or questionnaires were excluded. Biological markers are not considered relevant as they assess exposure only over the last hours, weeks or months. Self-reported exposure based on questionnaires also does not seem adequate for medico-legal purposes. Therefore, retrospective exposure assessment should be based on mathematical models to approximate past exposure.

RESULTS

For this purpose a validated model developed by Repace and Lowrey was considered appropriate. It offers the possibility of retrospectively assessing exposure with existing parameters (such as environmental dimensions, average number of smokers, ventilation characteristics and duration of exposure). The relative risk of lung cancer can then be estimated based on the individual cumulative exposure of the worker.

CONCLUSION

In conclusion, having adapted it to the German hospitality sector, an existing mathematical model appears to be capable of approximating the cumulative exposure. However, the level of uncertainty of these approximations has to be taken into account, especially for diseases with a long latency period such as lung cancer.

Recent developments in meta-analysis

The art and science of meta-analysis, the combination of results from multiple independent studies, is now more than a century old. In the last 30 years, however, as the need for medical research and clinical practice to be based on the totality of relevant and sound evidence has been increasingly recognized, the impact of meta-analysis has grown enormously. In this paper, we review highlights of recent developments in meta-analysis in medical research. We outline in particular how emphasis has been placed on (i) heterogeneity and random-effects analyses; (ii) special consideration in different areas of application; (iii) assessing bias within and across studies; and (iv) extension of ideas to complex evidence synthesis. We conclude the paper with some remarks on ongoing challenges and possible directions for the future.

Challenging the epidemiologic evidence on passive smoking: tactics of tobacco industry expert witnesses

OBJECTIVE

To analyse the statements given by tobacco industry defence witnesses during trial testimonies and depositions in second-hand smoke cases and in parallel, to review criticisms of epidemiology in industry-funded publications in order to identify strategies for discrediting epidemiologic evidence on passive smoking health effects.

METHODS

A collection of depositions and trial testimony transcripts from tobacco industry-related lawsuits filed in the United States during the 1990s, was compiled and indexed by the Tobacco Deposition and Trial Testimony Archive (DATTA). Statements in DATTA made by expert witnesses representing the tobacco industry relating to the health effects of passive smoking were identified and reviewed. Industry-supported publications within the peer-reviewed literature were also examined for statements on exposure misclassification, meta-analysis, and confounding.

RESULTS

The witnesses challenged causation of adverse health effects of passive smoking by citing limitations of epidemiologic research, raising methodological and statistical issues, and disputing biological plausibility. Though not often cited directly by the witnesses, the defence tactics mirrored the strategies used in industry-funded reports in the peer-reviewed literature.

CONCLUSION

The tobacco industry attempted to redirect the focus and dialogue related to the epidemiologic evidence on passive smoking. This approach, used by industry experts in trial testimony and depositions, placed bias as a certain alternative to causation of diseases related to passive smoking and proposed an unachievable standard for establishing the mechanism of disease.

Synthesizing medical evidence: systematic reviews and metaanalyses

Systematic reviews and metaanalyses have become increasingly popular ways of summarizing, and sometimes extending, existing medical knowledge. In this review the authors summarize current methods of performing meta-analyses, including the following: formulating a research question; performing a structured literature search and a search for trials not published in the formal medical literature; summarizing and, where appropriate, combining results from several trials; and reporting and presenting results. Topics such as cumulative and Bayesian metaanalysis and metaregression are also addressed. References to textbooks, articles, and Internet resources are also provided. The goal is to assist readers who wish to perform their own metaanalysis or to interpret critically a published example.

Nonsteroidal Antiinflammatory Drug Use and Lung Cancer

STUDY OBJECTIVE

Studies done both in laboratory animals and humans suggest that nonsteroidal antiinflammatory drug (NSAID) use may reduce the risk of developing lung cancer. Many epidemiologic studies exploring this association lacked sufficient power to draw definitive conclusions. We conducted a metaanalysis to examine the effect of NSAID use on the risk of lung cancer.

DESIGN

We searched the literature using MEDLINE, CANCERLIT, related conference abstracts, and bibliographies of selected studies. The estimators of relative risk (RR) and associated variances, adjusted for the greatest number of confounders, were abstracted and included in the metaanalysis. Combined estimators of RR were calculated using either fixed or random-effect models. Metaanalyses were performed on 14 studies (the number of cases ranged from 81 to 2,560) that examined the association between lung cancer and NSAIDs. Further, subgroup analyses were performed on the nine studies that had adjusted for the effects of smoking.

RESULTS

The combined estimate of RR was 0.79 (95% confidence interval [CI], 0.66 to 0.95) when all the studies were included in the analysis, and was 0.68 (95% CI, 0.55 to 0.85) when the analysis was limited to the subgroup of studies adjusted for smoking. The combined estimates for case-control studies and cohort studies within this subgroup were 0.63 (95% CI, 0.47 to 0.86) and 0.78 (95% CI, 0.62 to 0.98), respectively. We also observed that small cell lung cancer was more inversely associated with NSAID use (RR, 0.48; 95% CI, 0.30 to 0.75) than non-small cell lung cancer (RR, 0.66; 95% CI, 0.56 to 0.79).

CONCLUSION

The findings of this study support an inverse association between NSAID use and risk of lung cancer, but do not suggest a causal relationship.

Evaluation of the exposure-response relationship of lung cancer mortality and occupational exposure to hexavalent chromium based on published epidemiological data

Some have suggested a threshold mechanism for the carcinogenicity of exposure to hexavalent chromium, Cr(VI). We evaluated the nature of the exposure-response relationship between occupational exposure to Cr(VI) and respiratory cancer based on results of two recently published epidemiological cohort studies. The combined cohort comprised a total of 2,849 workers employed at two U.S. chromate production plants between 1940 and 1974. Standardized mortality ratios (SMRs) for lung cancer in relation to cumulative Cr(VI) exposure categories were reported using regional mortality rates. Linear additive and multiplicative relative risk regression models were fit to the SMRs of the individual and combined studies. Both models fit the data from the individual studies reasonably well; however, the fit was somewhat less adequate for the pooled data. Meta-analysis of the slope estimates obtained from the multiplicative relative risk model showed substantial heterogeneity between the two epidemiological studies. In conclusion, these data indicate that a linear dose response describes the relationship between Cr(VI) and lung cancer reasonably well, and therefore these analyses do not necessarily support the threshold hypothesis for the lung carcinogenicity of Cr(VI). However, these results must be interpreted with recognition of the limitations of the use of epidemiological data in the evaluation of nonlinear exposure-response patterns.

Evidence that the SSRI dose response in treating major depression should be reassessed: a meta-analysis

The limitations in design and analysis of currently available dose-response studies of SSRI treatment of major depression have led to the conclusion that dose response is flat. We applied concepts from our companion article to determine if currently available data is consistent with a "potential" and an "expressed" dose response. Using these concepts, we performed a meta-analysis on all identifiable published fixed-dose and dose-escalation studies that reported the effect of different SSRI oral doses on efficacy. "Potential" dose response in fixed-dose studies with categorical response outcomes equaled a significant meta-analyzed slope of 3.1%/100 SSRI mg equivalents (SMEs) (SE=1.2%) or 7.8% across the dose range. Similar analysis in dose-escalation studies that reported categorical response data yielded a non-significant meta-analyzed slope of 3.7%/100 SMEs (SE=2.3%) or 9.3% across the dose range. Analyses of the "expressed" dose response demonstrated in the studies indicated a slope statistically equal to zero. The current analysis suggests a "potential" dose response can be demonstrated for SSRIs in treating major depression. The analysis suggests an "expressed" dose response could exist in best clinical practice. Study designs better tailored to address the relevant clinical question would test these hypotheses more appropriately than previous studies.

Effect of smoking cessation on major histologic types of lung cancer

STUDY OBJECTIVES

It is well-recognized that the risk of lung cancer declines after smoking cessation. However, the degree of decline in different histologic types of lung cancer is not well understood. We conducted a meta-analysis of peer-reviewed studies to assess the effect of smoking cessation on rates of major histologic types of lung cancer.

DESIGN

Studies published in English between 1970 and 1999 were identified through searches of computerized databases (ie, MEDLINE and CANCERLIT). Combined estimates of relative risk and 95% confidence intervals were calculated for 27 studies using fixed and random effects models. Separate analyses were conducted for men and women.

RESULTS

Smoking cessation was associated with a reduction in the risk of all the major histologic types of lung cancer. The highest reduction was in small cell lung carcinoma (SCLC) and squamous cell carcinoma (SQC), and the lowest reduction was seen in large cell cancer and adenocarcinoma. In women, the combined risks for SQC and SCLC were higher than those in men. The dose-response curve for intensity of smoking was steeper in women.

CONCLUSION

The findings of this study suggest that smoking cessation results in the greatest reductions for SCLC and SQC. This effect is most marked in heavy smokers, particularly among women.

Synthesis of evidence from epidemiological studies with interval-censored exposure due to grouping

We describe a method for assessing dose-response effects from a series of case-control and cohort studies in which the exposure information is interval censored. The interval censoring of the exposure variable is dealt with through the use of retrospective models in which the exposure is treated as a multinomial response and disease status as a binary covariate. Polychotomous logistic regression models are adopted in which the dose-response relationship between exposure and disease may be modeled in a discrete or continuous fashion. Partial conditioning is possible to eliminate some of the nuisance parameters. The methods are applied to the motivating study of the relationship between chorionic villus sampling and the occurrence of terminal transverse limb reduction.

An illustrated guide to the methods of meta-analysis

Meta-analysis is now accepted as a necessary tool for the evaluation of health care. Such analyses have been carried out in virtually every area of medicine to evaluate a wide spectrum of health care interventions and policies. This paper has three broad aims: (1) to describe the basic principles of meta-analysis, using a meta-analysis of interventions intended to reduce hospital re-admission rates for illustration; (2) to consider threats to the internal validity of meta-analysis, and the measures which can be taken to minimize their impact; and (3) to present an overview of more specialist and developing methods for synthesizing data, with the intention of outlining the directions meta-analysis may take in the future. The methods used to synthesize studies, which take 'weighted averages' of effect sizes have been refined to a high degree, while the methods for dealing with threats to the validity of meta-analyses such as publication bias, and variations in quality of the primary studies, are at a less advanced stage. However, many consider this standard 'weighted average' approach to meta-analysis not to be 'state of the art' in at least some situations, where the use of more sophisticated methods, generally to explain variation in estimates from different studies and synthesize a broader base of evidence, would be advantageous. Currently, approaches which attempt to do this are mainly still in the experimental stage and, unfortunately, ideas which sound natural and appealing are often difficult to implement in practice. Clearly, it will be some time before they are used routinely, but significant steps have been made.

Effect of cigarette smoking on major histological types of lung cancer: a meta-analysis

Although cigarette smoking is an established risk factor for lung cancer, the strength of association with different histologic types is not well understood. This meta-analyses of peer-reviewed studies was conducted to assess the effect of cigarette smoking on major histologic types of lung cancer. Studies were identified through MEDLINE and CANCERLIT searches. A total of 48 studies published between 1970 and 1999 were identified. Combined estimates of relative risks (OR) and 95% confidence intervals (CI) were calculated using fixed and random effect models. Separate analyses were conducted by study design and gender. A linear dose-response was fit to studies reporting data on intensity and duration of smoking. All histologic types of lung cancer were significantly associated with cigarette smoking. The association was stronger with squamous cell carcinoma (SQC) and small cell carcinoma (SCLC) than with large cell cancer (LGC) and adenocarcinoma (ADC). The combined OR for heaviest smoking intensity (30+ cigarettes per day) ranged from 4.10 (CI 3.16-5.31) for ADC to 18.3 (CI 9.26-36.4) for SCLC. The combined OR for longest duration of smoking (40+ years) ranged from 3.80 (CI 2.35-6.16) for ADC to 38.6 (CI 11.9-125) for SCLC. In women, the combined OR for SQC and SCLC were higher than those in men. The dose response curve for intensity of smoking was steeper in women. The findings of this study provide additional evidence for a causal relationship between smoking and all histologic types of lung cancer.