Colleagues,
A bit of a puzzle, I am reviewing the effect of an intervention against coronary heart disease (CHD) in a large population (at least 250 000 subjects drawn from North American (USA and Canada), Western Europe (UK, Germany, Netherlands), Southern Europe / The Mediterranean, India, Taiwan and Japan). The challenge is that some of the studies state RR values while others use OR values. Is there a concise way to convert RR values to OR values?
I ran into an argument that was raised in Grant (2014) Converting an odds ratio to a range of plausible relative risks for better communication of research findings, BMJ, 2014, doi: 10.1136/bmj.f7450. The author argues that one can convert OR values to RR, the converse can be true - one can convert RR values to OR?
RR = OR / (1 - p + (p x OR)) (Grant, 2014)
To use the equation, how can one determine the baseline risk of CHD (p) for each of the populations above (North American (USA and Canada), Western Europe (UK, Germany, Netherlands), Southern Europe / The Mediterranean, India, Taiwan and Japan)? Can we assume that baseline risk of such study populations is zero (0) since the subjects recruited into each study were CHD free (obviously there is always a baseline risk greater than 0!)?
Can someone suggest a way to treat such data (RR / OR)?
Thank you in advance.
This may help: http://www.r-bloggers.com/how-to-convert-odds-ratios-to-relative-risks/
Many thanks Kent. I ran into the same post before, still some questions remain unanswered. Regards.
In the case of the studies you are considering, p is the value of risk in the control group and it is determined at the end of the study period. Its numerical value is a fraction of people in the control group (no intervention)who developed CHD during the study time. The baseline risk is indeed always different from zero and can be calculated in multiple ways. However, p value in the referenced formula represents the risk for the control group within the study, not at the baseline. The original paper admits simplification in a nomenclature, and references "baseline risk", even for the cases like your studies, when the comparison is made to the control group rather than some historical estimate of the risk.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis