Home > Community > Can hazard ratios and odds ratio be used interchangeably in...
Upvote

26

Downvote
+ Hazard assessment
+ Meta-analysis
Posted by
Neil Morrison

Can hazard ratios and odds ratio be used interchangeably in...

Avun Jahei  Follow

If there was an extremely low proportion of subjects with an event in all experiments (let's say <10%) and the hazard and odds ratios are vey close to 1, then hazard, odds and relative risk ratios will be relatively close to each other.

More

Upvote

VOTE

Downvote
Brian Bi  Follow

Can RR be used interchangebly with HR while doing the meta-analysis? 

More

Upvote

VOTE

Downvote
Muhammad Zulfiqar  Follow

I have no idea.

More

Upvote

VOTE

Downvote
Dalton Mendeleve  Follow

Another 2008 reference by Ian Scott addressed this issue, specifically in pages 14-15. Scott, I. Interpreting risks and ratios in therapy trails Australian Prescriber, 2008;31:12-16. http://www.australianprescriber.com/magazine/31/1/12/6

More

Upvote

VOTE

Downvote
Andre Jackson  Follow

Yes they can be used.

More

Upvote

VOTE

Downvote
Demi Blye  Follow

Dear Muhammad Ali,
Risk (hazard) ratios and odds ratios cannot be used interchangeably in meta-analysis. Like euro and pound they have to be converted into the same value e.g. Swedish crowns. You find the formula here:
http://hiv.cochrane.org/sites/hiv.cochrane.org/files/uploads/Ch09_Analysing.pdf
Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 9: Analysing data and undertaking meta-analyses. In: Higgins JPT, Green S (editors).
Cochrane Handbook for Systematic Reviews of Interventions. Version 5.0.1 [updated September 2008]. The Cochrane Collaboration, 2008. Available from www.cochrane-handbook.org.

More

Upvote

VOTE

Downvote
David Sandbrook  Follow

The widely cited 2007  paper by Jayne Tierney and colleagues "Practical methods for incorporating summary time-to event data into meta-analysis" addresses these issues, including methods for approximating hazard ratios using count data. The article is Open Access: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1920534/

More

Upvote

VOTE

Downvote
Dare Too  Follow

Dear Mohammed Ali
The answer to your question is no - hazards and risks or odds are not interchangable! Regardsing the specific statistical differnces you would have to consult a statistician. But i can try to give you a lay-physician explanation.
Risks refer to absolute numbers of an event (i.e. disease) in a popualation - we have no consideration of time in a risk.
Hazards refers only to the "speed" of specific events in a population. A hazard is therefore a time to event estimate and will never reflect the absolute risk of an event in a population.
Hazard risk and risk ratios are therefor two different measures of events in a population. They are based on two different infernential statistics and most likely also based on two different types of studies (propesctive cohort vs. interventional studies). The statistical question asked in a hazard rate is "does a specific exposure cause outcome quicker than to not being exposed" where the question in a risk ratio is "does exposure cause outcome more often compared to not being exposed".   
My advice to you is therefore not to include hazard ratios if you are making a meta analysis based on risks or odds ratios.  

More

Upvote

VOTE

Downvote
Bob Given  Follow

Dear Muhammad Ali,
Risk (hazard) ratios and odds ratios cannot be used interchangeably in meta-analysis. Like euro and pound they have to be converted into the same value e.g. Swedish crowns. You find the formula here:
http://hiv.cochrane.org/sites/hiv.cochrane.org/files/uploads/Ch09_Analysing.pdf
Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 9: Analysing data and undertaking meta-analyses. In: Higgins JPT, Green S (editors).
Cochrane Handbook for Systematic Reviews of Interventions. Version 5.0.1 [updated September 2008]. The Cochrane Collaboration, 2008. Available from www.cochrane-handbook.org.

More

Upvote

VOTE

Downvote
Alisha Talks  Follow

Thank a lot Ali Bhai. 

More

Upvote

VOTE

Downvote
Brenda Barlow  Follow

Dear Mohammed Ali
The answer to your question is no - hazards and risks or odds are not interchangable! Regardsing the specific statistical differnces you would have to consult a statistician. But i can try to give you a lay-physician explanation.
Risks refer to absolute numbers of an event (i.e. disease) in a popualation - we have no consideration of time in a risk.
Hazards refers only to the "speed" of specific events in a population. A hazard is therefore a time to event estimate and will never reflect the absolute risk of an event in a population.
Hazard risk and risk ratios are therefor two different measures of events in a population. They are based on two different infernential statistics and most likely also based on two different types of studies (propesctive cohort vs. interventional studies). The statistical question asked in a hazard rate is "does a specific exposure cause outcome quicker than to not being exposed" where the question in a risk ratio is "does exposure cause outcome more often compared to not being exposed".   
My advice to you is therefore not to include hazard ratios if you are making a meta analysis based on risks or odds ratios.  

More

Upvote

VOTE

Downvote
Dave Fuller  Follow

Not really, and without assumption on baseline hazard ( hazards for each group in absolute terms, not just the ratio, and how it changes over time) the answer can be very different. For example, assuming an exponential distribution for survival function (one of the simplest), S(t) = exp(- lambda * t), hazard ratio of 2 would mean that the second group has lambda 2x that of the first. Then, if all population is of the same "age", odds ratio at time t would be [1-S2(t)]/S2(t) / [1-S1(t)/S1(t) ] = (1-exp(-2*L1*t) / (1-exp(-L1*t). So as Mohamad mentioned below, if lambdas are small, this is close to 2L / L, or 2. While playing with this a bit, I found that depending on distribution of age in the population and how prevalent a condition is, hazard of 2 can correspond to odds ratios of anything from 2 till 10 (realistically) and even higher, at the same time risk ratios for very prevalent conditions would be biased towards one if population is rather "old" and close to 2 is young. In short, one can not infer OR from HR without further assumptions, but the smaller OR is the closer these should be

More

Upvote

VOTE

Downvote