historical comparison of survivals why not but only under certain conditions

Background: Meanings, properties and possible implications of specific survival (SS) are worth exploring when computed from mortality tables, particularly by virtue of the relationship with excess mortality (ExcM) and expected mortality (ExpM). Design: The variability of specific mortality (SM) and ExpM reflects the whole range of patients’ fate, from the worst scenario (observed mortality higher than ExpM) to the best (SM equal to ExpM: a condition corresponding to full clinical recovery of the disease). Obviously, SM of patients and ExpM of reference population are conditioned by the same social and clinical factors, which act in parallel on both groups. So, the ExpM of reference subjects corresponds to the average age limit for a given histori cal period, to which the survival of patients can be normalized, also giving the possibility of post-hoc analysis of survival. SS, which results from the difference between observed and expected mortality, is also an index of the years of survival “stolen” by the disease with respect to the expected. All these times can be statistically compared by simple inferential tests. Results: As an example of achievable result, we report a comparative analysis of two randomized trials, conducted in different times and closed several years ago, with identical target patients (advanced-stage Hodgkin lymphoma) and similar treatment (MEC ten-drug chemotherapy in the first study, and the similar CEC in the second, in which cyclophosphamide 650 mg/m2 re placed mechlorethamine 6 mg/m2). The analysis of SS revealed a previously underestimated statistical difference in favor of the patients treated with CEC: the difference seems tom be primarily due to the drug substitution, in spite of their presumed equiva -lence. Years of life lost per year of follow-up in CEC arm: 0.148 ± 0.228; in MEC arm: 0.411± 0.528 (P= 0.02314). Conclusions: The comparative evaluation of SS offers more reliable results and is feasible in post-hoc investigations, even be tween different timeframes.