Cumulative Incidence and Incidence Rate Ratio for Estimation of Risk of Tuberculosis in Patients With Cancer

To the Editor—In their systematic review and meta-analysis of the risk of active tuberculosis (TB) in patients with cancer, Cheng et al evaluated the cumulative incidence of TB among patients with cancer and used this information together with annual country-specific TB incidence rates from the World Health Organization to estimate the incidence rate ratio of TB in patients with cancer compared to the general population [1]. The analysis raises some methodological concerns that have implications for the interpretation of the results. The authors use the term "cumulative incidence rate," when indeed cumulative incidence is a proportion, not a rate (events expressed per unit time) [2, 3]. For included studies that only contained information on cumulative incidence, but not an actual incidence rate, the authors used cumulative incidence (over the study period) divided by the incidence rate per year in the general population to calculate an incidence rate ratio. This calculation is methodologically incorrect [3] and will result in overestimation of the incidence rate ratio. The included study by Kamboj and Sepkowitz that evaluated the risk of TB among patients with cancer at the Memorial Sloan-Kettering Cancer Center in New York City between 1980 and 2004 illustrates the problem [4]. In this study, there were 103 cases of TB among 186843 patients with cancer during the study period, resulting in a cumulative incidence of 55 cases of TB per 100000 patients (over 25 years). In Table 1 of their article, Cheng et al presented the cumulative incidence of 55 and divided this number by the TB incidence per 100000 persons/year in the general population during the study period (8.6/100000/year), resulting in an incidence rate ratio (calculated by the authors) of 6.3 (55/8.6). The problem is that the cumulative incidence among cancer patients accumulated over 25 years whereas the incidence in the general population represented events occurring within 1 year. Assuming a regular distribution of these TB cases over time, the TB incidence rate would have been 2.2/100000 per year (55 cases of TB per 100000 patients over 25 years), but of course this estimate is not correct either, because it wrongly assumes that every patient survived and was followed up for 25 years. Thus, the available study data do not allow calculation of an incidence rate ratio, and the analysis should have been limited to studies that provided a true incidence rate of TB.