I have a question regarding pooling hazard ratios derived through different means for a meta-analysis. Software: RevMan 5; Method: generic inverse variance (log hazard ratios).

My issue is that some studies calculate hazard ratio while treating the parameter as a continuous variable (e.g.: HR of 1.3, which means risk is increased by 1.3x per unit increase in SUVmax). However, some studies calculate the hazard ratio while treating SUVmax as a dichotomous variable by using a cut-off (e.g.: cut-off of 5; HR is 4.0, which means SUVmax > 5 is four times riskier compared to SUVmax < 5)

Is it safe to simply pool these hazard ratios together as they are (continuous and dichotomous) for meta-analysis? It does not seem statically sound to me, but 3 other published meta-analyses that I've found on PET-parameters pooled continuous and dichotomous HRs together so I don't know who's right or wrong.

Thank you.

  • $\begingroup$ Definitely you cannot pool HR from different studies (and not even from the same) if cut-offs vary substantially. You should try to standardized them. For instance, a logHR of 0.1 for a 1-year change might (but this is not necessarily true) translate into a logHR for a 10-year change. Otherwise try to pool p values, this has been addressed in detail in several books. $\endgroup$ – Joe_74 Mar 9 '19 at 14:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.