Hi,

You may find this excerpt from the Cochrane handbook helpful: http://handbook.cochrane.org/chapter_9/9_4_4_4_which_measure_for_dichotomous_outcomes.htm

In essence, relative risk measures such as odds ratios and risk ratios are preferred to absolute measures like risk differences. As far as relative risk measures go, risk ratios are considered easier to interpret than odds ratios. You can transform odds ratios reported in published papers to risk ratios using the method described in the Cochrane handbook: http://handbook.cochrane.org/chapter_12/12_5_4_4_computing_risk_ratio_from_an_odds_ratio_or.htm

That said, I have generally worked with pooled odds ratios.

Hope that helps.

Puja

"In selecting among the risk ratio, odds ratio, and risk difference the researcher needs to consider both substantive and technical factors.

The risk ratio and odds ratio are relative measures, and therefore tend to be relatively insensitive to differences in baseline events. By contrast, the risk difference is an absolute measure and as such is very sensitive to the baseline risk. If we wanted to test a compound and believed that it reduced the risk of an event by 20 % regardless of the baseline risk, then by using a ratio index we would expect to see the same effect size across studies even if the baseline risk varied from study to study. The risk difference, by contrast, would be higher in studies with a higher base rate.

At the same time, if we wanted to convey the clinical impact of the treatment, the risk difference might be the better measure. Suppose we perform a meta analysis to assess the risk of adverse events for treated versus control groups. The risk is 1/1000 for treated patients versus 1/2000 for control patients, for a risk ratio of 2.00. At the same time, the risk difference is 0.0010 versus 0.0005 for a risk difference of 0.0005. These two numbers (2.00 and 0.0005) are both correct, but measure different things.

Because the ratios are less sensitive to baseline risk while the risk difference is sometimes more clinically meaningful, some suggest using the risk ratio (or odds ratio) to perform the meta-analysis and compute a summary risk (or odds) ratio. Then, they can use this to predict the risk difference for any given baseline risk."

The OR overestimates RR. 

If the prevalence is low use OR is a good estimatoria for RR.

in metaanalysis use fixed effects model or ramdom effects model to estimate a resume meassure. Use RR in prospective ( cohort, ECA) reports and use OR in retrospective designs ( case-control, cross-sectional).

Also see my other answer. 

Statistically: see Deeks, J. J. (2002) ‘Issues in the selection of a summary statistic for meta-analysis of clinical trials with binary outcomes’, Statistics in Medicine, 21(11), pp. 1575–1600. doi: 10.1002/sim.1188.

Also interesting: Cummings, P. (2009) ‘The relative merits of risk ratios and odds ratios.’, Archives of pediatrics & adolescent medicine, 163(5), pp. 438–445. doi: 10.1001/archpediatrics.2009.31.

Interpretation wise: see Zhang, J. and Yu, K. F. (1998) ‘What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes.’, JAMA : the journal of the American Medical Association, 280(19), pp. 1690–1. doi: 10.1001/jama.280.19.1690.