Dear Ding Ding,

You may try to search the literature on the association between drug and liver injury (rather than drug-induced liver injury). I would suggest to exclude patients with viral hepatitis from both case and control sets, as these are the major risk factors of liver injury which may mask the association you are going to study.

Hwai-I

A review article for your reference.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2903197/

 i don't agree with this. Please note whether all the times everyone develops hepatotoxicity with antitubercular drugs, probably no, only minority develops either due to idiosyncracy or, due to overdosage. Therefore in this case odds ratio is not 100%. Odds ratio will be different for different drugs. Therefore please search previous reports regarding incidence of hepatotoxicity in individual hepatotoxic drugs and calculate sample size, by doing that you  can eliminate sampling error and may get correct odds ratio for individual hepatotoxic medications

GKP is right, hte OR is not 100%. Get in touch with a statistician and a drug safety expert

OR=100 is almost absurd. OR=100% is nonsense.  Either the OR or the rate must be picked, not both.

For sample size calculations, the main as well as difficult question  -as I understand your problrm - is to estimate and specify for the calculations the exposure rate in controls. This can be obtained from literaure serch, as Hwai-I Yang says. Of course, this requires to pick a certain drug, since -as our colleagues have stated- a "global toxicity" or including any kind of hepatic toxicity would lead to flawed results.

Regards,