Presentation on theme: "Case-control study. Introduction for case-control study Patients with a disease (case) are compared with controls who do not have the disease. – 以 ‘ 有病."— Presentation transcript:
Introduction for case-control study Patients with a disease (case) are compared with controls who do not have the disease. – 以 ‘ 有病 ’ 及 ‘ 無病 ’ 做為主要分組 – 估計 ‘ 暴露因子 (exposure)’ 或 ‘suspected factors’ 的相對危險性 (relative risk) 。 優點： cheaper than cohort study 限制： difficult in the choice of appropriate controls
An efficient method of retrospective sampling within a cohort study. By definition, exposure (dietary intakes) must be established retrospectively, and this poses problems in nutritional studies, especially when a disease has a long pre-clinical phase and the effects of diet only manifest after a corresponding latent interval. Controls should be similar to case subjects in all ways other than their exposure to the risk factors under investigation.
Study population and selection of case Formulating the research questions. Then select the best population to address the questions. –The study population must be large enough and have a high enough incidence of disease –To provide a sufficient numbers of case over the course of the investigation. –In practice, the choice of a study population is often constrained by operational requirements. –e.q. The investigator may only have resources to conduct the study in the area where he/she normally works.
4. Research questions of interest Occasionally the study population is the starting point of an investigation, and the choice of risk factors for examination is decided secondarily. The trigger to such an investigation might be an unexplained focus of disease. –E.g.: The observation of high rate of stomach cancer in South Louisiana prompted a case- control study to look for a possible dietary explanation.
Case ascertainment and selection To ensure that cases were ascertained by uniform criteria. Another consideration in the selection cases is the specificity of the diagnosis. E.g. 研究 vitamin E 與肺癌的關係 – 肺癌有幾種 subtypes – 選其中一種 subtype 做研究，較能得到顯著的結果 –Vitamin E 只對患有 squamous carcinoma 有保護效 E.g. Chilli pepper –Being stronger related to intestinal-type gastric cancer than diffuse-type gastric cancer.
Ascertainment of exposure factors Exposure factors must be measured –risk factors of interest (dietary or nutritional) 可選擇與特定營養素含量豐富的食物為 exposure factors E.g.: 研究鈣質攝取量與大腿骨折發生率的關係：選擇富 含鈣質的食物做研究 ---- milk, cheese, cake, biscuits, puddings –Confounding factors Associated with both causal agent and resulting event E.g. 研究 dietary factors and lung cancer confounding factors ： smoking habits, air pollution.
3. Factors affecting exposure assessment If the disease has a long pre-clinical phase, the relevant exposure may have occurred many years before diagnosis, and in nutrition studies this can pose particular difficulty –E.g. liver cancer and unfresh peanut consumption. Exposure recall may be affected by the disease status –Changing dietary pattern due to pre-clinical disease –Recall bias e.q. 對於腎結石患者 --- 較無法真正得知其發病前真正的 飲水量，所以較不易探討「發病前的飲水量」和「腎結 石」之關係 解決方式 – 以 pre-clinical phase 的病人為病例 – 以 Blinding 的方式做資料的取得
Bias encountered in the case-control study 常見 case-control study 的偏差 –Selection bias: sampling the case and control groups 解決方式： – 訂出一套標準的 inclusion and exclusion criteria –e.q. 選取 case and control groups 時，除了疾病條件不 同外，其他條件儘量相同 –Information bias: observation bias and recall bias –Confounding factors
Selection of controls The aim –From controls, one should get reliable estimate of exposure to risk factors and confounders among members of defined or theoretical study population who are at risk of becoming cases during the period of study. The objects leads to two requirements: –The exposure of controls should be representative of that in members of study population who are at risk of become cases. –The exposure of controls should be ascertainable with the same accuracy as for cases.
Control subjects selected from the general population Usually there is no perfect control group, and the choice of controls must be a compromise. Three general rules are worth bearing mind: –Controls should always come from the study population (or from the same catchment area) –Where possible, the method used to ascertain exposure should be similar for cases and controls. –If information is to be obtained at interview or by physical examination, this should, where possible, be carried out blind to the case/control status of the subjects.
Matching Matching is used in case-control studies –To permit allowance for confounders which are complex or difficult to define. statistically more efficient: efficient analysis requires that there be a similar ratio of cases to controls at each level of exposure to the confounding variable. –To reduce biases in the ascertainment of exposure. Unlike in a cohort study, matching in a case-control study does not in itself eliminate the effects of a confounders.
Nested case-control studies 結合 Cohort 及 case-control study 的研究法 The case-control approach –An efficient sampling within a theoretical cohort study –It can be ‘nested’ within real cohort studies. 例如： – 研究 colon cancer 與 vitamin D 關係 – 受試者若前數十年曾抽血 – 此血液保存於良好的條件下 – 直接挑選目前有 colon cancer 患者 – 並挑選出相當人數的 controls – 再從血庫中挑出這些人的血樣本做 vitamin D 的分 析 – 以探索 colon cancer 與 vitamin D 關係 優點：經濟、資料易得
Study size and statistical power Sample size : 參考 page 82 Case : control ratio 約 1:1 – 1:4
Analysis and interpretation The main measure of association commonly derived from case-control studies is the odds ratio (OR). The OR approximates closely to the relative risk.
OR = ad / bc ; 95% CI = OR Unmatched analysis
Matched analysis OR = s / t ; 95% CI = OR
Interpreting an association as causal relationship Assessment of whether of an observes association is likely to be directly causal and not the result of unrecognized confounding depends upon several consideration: