Choice of the denominator in case population studies: event rates for registration for liver transplantation after exposure to NSAIDs in the SALT study in France


  • Author contributions: Nicholas Moore was the overall study supervisor, intervening when necessary for the smooth operation of the study. Sinem Ezgi Gulmez and Séverine Lignot were the scientific and operational study coordinators, devising the initial document generation under the control of the scientific committee, organizing negotiations with transplant centers and study data retrieval, writing the study reports and draft article and verifying all contents. Patrick Blin was the study epidemiological overseer, contributing to study design, operations, analysis and understanding. Dominique Larrey and Georges-Philippe Pageaux were our referent hepatologists, and Antoine Pariente, Jacques Bénichou and Bernard Bégaud contributed to the methodological conceptualization. All authors contributed comments to the final version of this paper.

Correspondence to: S. E. Gulmez and N. Moore, Department of Pharmacology, Univ Bordeaux Segalen, 33076 Bordeaux cedex, France. E-mail:,



The effect of denominator options on event rates was tested on the French part of the Study of Acute Liver Transplant (SALT).


SALT is a case population study of acute liver failure registered for transplantation (ALFT), exposed to non-steroidal anti-inflammatory drugs (NSAIDs) or non-overdose paracetamol, from 2005 to 2007. Population exposure was computed from the Intercontinental Medical Services' (IMS) and the French national healthcare insurance system's data as the number of defined daily doses (DDDs) sold or dispensed and the number of exposed patients.


Nine ALFT cases were exposed to 10 NSAIDs and 49 to non-overdose paracetamol. NSAID sales ranged from 0.04 billion (niflumic acid) to 0.5 billion (ibuprofen) DDDs, amounting to 2.5 billion DDDs for all NSAIDs. The mean per-person exposure ranged from 13.1 (niflumic acid) to 43.2 (ketoprofen) DDDs, reaching 60.5 DDDs for any NSAID. The number of users ranged from 2 million (niflumic acid) to 13 million (ibuprofen), which was 26.6 million for all NSAIDs. The ALFT rates per billion DDDs ranged from 0 to 26 for individual NSAIDs and amounted to 4.6 for all NSAIDs. The ALFT rates per billion DDDs were inversely correlated with the average per-patient exposure (R2 = 0.935, p = 0.0016). The ALFT rates per million users ranged from 0.31 to 0.49, which was 0.41 for all NSAIDs, with no difference between drugs and no effect of mean per-patient exposure (R2 = 0.01, p = 0.9). Whether measured per DDD or per user, the event rate with paracetamol was three to five times higher than with NSAIDs.


ALFT risk with NSAID seems to be user dependent rather than person-time (exposure) dependent. Choosing the wrong denominator in case population studies might give erroneous results. Copyright © 2012 John Wiley & Sons, Ltd.