Breslow and Day Statistical Methods in Cancer Research Vol 1 WHO IARC 1980 is the "bible". A simpler clear exposition is in dos Santos Silva Cancer Epidemiology principles and methods WHO IARC 1999.
Most common practice is now some form of logistic regression.
I think estimation of odds ratios using logistic regression modalities ( simple or multivariate) is useful . Make sure you compare adjusted and unadjusted ORs. but overall this is a reasonable approach to determine the relevance of risk factors.
There is a systematic way to analyze data obtained in case control studies. Regardless the medium of analysis, the basic steps are:
1.Tabulation of the results in the form of 2x2 or 2Xn tables
2. Calculation of the percentage of exposed to the risk factor among cases and controls
3. Calculation of the Odds Ratio
4. Calculation of the attributable risk
5. Calculation of the CHi-squared test
6. If multiple risk factors are identified then logistic regression analysis is also done
All these can be handled by moderate expertise in using SPSS computer programme.
Case Control Study Analysis:
1. Comparison between proportion of diseases with history of exposure and proportion of non-diseases with exposure. If proportion is found to be different, then it can be tested statistically by z test for proportion. Other option is to test by Pearson's Chi-square Test.
2. Calculation of odds ratio from 2 X 2 contingency table. Then interpretation of odds ratio.
3. For multivariate analysis with many independent risk factors - logistic regression with categorical variable i.e. disease - present / absent. If matched case control study, then conditional logistic regression.
I recommend Logistic regression. For a case-control study with individual matching using conditional logistic regression and frequency mating non-conditional one is the best strategy. And I suggest you to include all variables in the final model in order to find confounders.
- Badges
- Science topic