Chris Bain

Systematic Reviews in Health Care: Frontmatter

What do we do if diVerent clinical studies appear to give diVerent answers? This user-friendly introduction to this diYcult subject provides a clear, unintimidating and structured approach to systematic reviews and incorporates several key features: • A practical guide to meta-analysis and the fundamental basis of evidence-based medicine • A step-by-step explanation of how to undertake a systematic review and the pitfalls to avoid • Liberally illustrated with explanatory examples and exercises • A review of the available software for meta-analysis

Systematic Reviews in Health Care: General methods

What do we do if diVerent clinical studies appear to give diVerent answers? This user-friendly introduction to this diYcult subject provides a clear, unintimidating and structured approach to systematic reviews and incorporates several key features: • A practical guide to meta-analysis and the fundamental basis of evidence-based medicine • A step-by-step explanation of how to undertake a systematic review and the pitfalls to avoid • Liberally illustrated with explanatory examples and exercises • A review of the available software for meta-analysis

Oral contraceptives and risk of ovarian cancer

Among a large cohort of married, female, registered nurses under 55 years of age, oral contraceptive (OC) use was examined in women with ovarian cancer and 470 age‐matched controls. Use of OCs before the diagnosis of cancer was reported by 28% of the women with ovarian cancer and 33% of the controls, yielding a relative risk of 0.8 (95% confidence limits 0.4–1.5). This apparent inverse relationship was attributable to a large effect in women 35 years of age or younger (relative risk = 0.2, 95% confidence limits 0.1–1.0). Patients with ovarian cancer were 2.2 times more likely than controls to be nulliparous. These data provide reassurance that OC use is not likely to be associated with any major increase in risk of ovarian cancer, but suggest that future studies of this relationship need to consider the possible confounding effect of infertility.

A case-control study of risk indicators among women with premenopausal and early postmenopausal breast cancer.

Among 714 premenopausal and 130 postmenopausal breast cancer cases matched with 8440 controls for age in years and menopausal status, risk indicators for breast cancer were similar, although most associations were stronger in the premenopausal women. Compared with nulliparous women, the relative risk (RR) for those with first birth before age 25 years was 0.7 (95% confidence limits [CL] from 0.5 to 0.9) among premenopausal women, and 0.7 (0.4–1.4) for postmenopausal women. In the premenopausal cases, a history of breast cancer in a sister gave a RR of 3.0 (2.1–4.1) and in a mother 1.9 (1.4–2.5), whereas for the postmenopausal women the RRs were 1.4 (0.6–3.1) and 13 (0.6–2.6), respectively. Fibrocystic breast disease was also a significant predictor of subsequent breast cancer in the premenopausal and postmenopausal women. In relation to women having a single birth, premenopausal women with six or more births had a risk of breast cancer of 0.6 (0.4–1.0), which was present even after adjustment for age at first birth.

Systematic Reviews in Health Care: Appraising and selecting studies

Readers will naturally wish to know how good the reviewed research is and why you have excluded some studies that address the question at issue. In both situations you need to explain your judgments, which will usually be based on your assessment of study quality and applicability. The process will usually need to be done in two stages; firstly, an initial screen for basic eligibility criteria and secondly, a detailed appraisal of quality. The eligibility screen might ask whether the study addresses the question and achieves some minimal quality criteria. For example, for an intervention question this might be evidence of a control group. This process is outlined in Figure 3.1. Standardizing the appraisal Providing an explicit and standardized appraisal of the studies that have been identified is important for two reasons. Firstly, a systematic review should try to base its conclusions on the highest-quality evidence available. To do this requires a valid and standardized procedure to select from the large pool of studies identified so that only the relevant and acceptable quality studies are included in the review. Secondly, it is important to convey to the reader the quality of the studies included as this indicates the strength of evidence for any recommendation made. What study features should be assessed? Overall, the study features that are most important to assess are those that involve selection and measurement bias, confounding and follow-up of participants. In Part 2 these features are examined for each question type under the following headings: Has selection bias (including allocation bias in randomized controlled trials (RCTs)) been minimized? Have adequate adjustments been made for residual confounding? Have the final outcomes been adequately ascertained? Has measurement or misclassification bias been minimized?