CALCULATION, MEANING, AND INTERPRETATION OF THE NUMBER NEEDED TO TREAT (NNT) IN RANDOMIZED PHARMACOLOGICAL TRIALS FOR THE PRIMARY AND SECONDARY PREVENTION OF OSTEOPOROTIC FRACTURES.

Abstract

Osteoporotic fractures are a substantial problem for public and private health sys tems. The severity of its consequences varies depending on the type of fracture,but all of them to some degree cause disturbances in the quality of life, incapacity, and economic expenses. The efforts to decrease their occurrence are basically oriented to identify the populations with the greatest probability of having fractures, and investigating the impact of interventions (generally pharmacological) for primary and secondary prevention. The common denominator that underlies and gives a theoretical ground to these efforts is the concept of risk. A risk factor is any characteristic or circumstance detectable in a person or group of persons that is known to be associated with an increase in the probability of having, developing, or being especially exposed to a morbid process. A specific variable/characteristic can be a risk factor, a “surrogate”, or a clinical outcome, depending on how it is analyzed. Before analyzing the impact of different pharmacological treatments addressed to reduce osteoporotic fracture risk, some considerations have to be made. Among the parameters used to compare the treated and the control groups, we find two measures of the risk of having osteoporotic fractures: the absolute Risk (AR), that measures the real incidence of fractures in the treated and nontreated populations, and the Relative Risk (RR), that compares the frequency of fractures that occur among those that were treated and not treated with the study drug. From them derives the relative risk reduction (RRr) and the absolute risk reduction (ARr) from which the number of patients that need to be treated with a drug in order to prevent an event (fracture) can be calculated: the NNT. On top of its sound mathematic-statistical basis, the NNT offers the clinician a quick intuitive interpretation of the impact of a certain treatment, for which its application has become an influential instrument at the moment of taking therapeutic decisions. In general, therapeutic trials give “raw” NNT values. Since these trials have variable durations, a problem arises when the time variable –besides the basal AR and RRr– is introduced in the calculation and interpreta tion of the NNT. The calculated NNT makes no reference to which was the NNT in the 1st, 2nd, etc., years of the study. The annualized NNT would be a correct way –but not a practical one– of referring to the NNT in fixed treatment times.