Clinical studies often employ restrictive inclusion and
Clinical studies often employ restrictive inclusion and exclusion criteria that inevitably lead to patient selection. Indeed, in line with Mengis et al.  the most frequently noted reason for non-inclusion into a clinical trial was, in addition to patient choice, the exclusion criteria of the clinical trial. The presence of (serious) comorbidity is a major exclusion criterion in most clinical trials. In this study, patients treated with ICT within a clinical trial had less comorbidities, as compared to patients receiving ICT outside a clinical trial. Moreover, the multivariable analysis showed that high-risk HCT-CI—which is a measure of serious comorbidity—was the only factor associated with trial participation in our study. Interestingly enough, however, in our cohort, the adjusted risk of mortality was not affected by trial participation. In AS1517499 to our results, a significant difference in OS between patients treated within and outside a clinical trial was shown by Mengis et al.  However, their study population consisted of both younger and older AML patients, thereby leaving aside the impact of trial participation on survival in different age groups .
Our study showed that cytogenetic and molecular testing were not performed in the great majority of elderly AML patients who received NIT. Similar results were found among this patient group in a Swedish population-based study . Both assessments play an essential role in the risk stratification of AML, which, in turn, is needed to select the most appropriate therapy. Older patients with a favorable risk profile who are medically fit benefit most from induction and consolidation chemotherapy . On the other hand, most elderly patients with a poor risk profile will likely benefit from less intensive approaches such as hypomethylating agents . In addition, therapeutic advances with specific inhibitors, such as FLT3-, IDH1- and IDH2-inhibitors, are currently in progress [, , ]. These agents could be applied in combination with other therapies or as a single treatment, especially in elderly patients not eligible for ICT.
A limitation of our study was that in many patients the WHO performance status was poorly documented in the medical records (81%), and cytogenetic and molecular diagnostics were not often performed, especially among recipients of NIT. Therefore, these factors, which are considered as important factors for outcome [12,26,34], could not be included in the multivariable survival analysis. As a result, the effect estimates of age might have some uncontrolled confounding. In addition, hospital type of diagnosis was not included in the multivariable logistic analyses, because non-university hospitals often refer patients who are eligible for ICT to university hospitals.
Conflict of interest
Introduction Cancer incidence projections are of major interest to identify healthcare and research needs and to allocate resources accordingly . Due to demographic change in many Western countries, it is presumed that overall more cancer cases will occur in future. Expected cancer incidence rates are commonly estimated using statistical models fitted to observed, often long-term, incidence data from cancer registries. In order to estimate future case numbers, additional data on the expected population size are required. Since all estimates of future incidence and population size are extrapolations based on statistical models, they naturally come along with uncertainty; that is, stochastic uncertainty, but also uncertainty with respect to the model specification and perhaps other forms of uncertainty. Uncertainty concerning expected cancer incidence (and similarly any other population-level health indicator) should ideally be considered in the reporting of the results, in their interpretation and ultimately in health policy decision making. Historically, reports of cancer incidence projections have often been deterministic, i.e. lacking quantification of uncertainty associated with reported estimates [, , ]. However, there is an increasing demand for uncertainty quantification. For example, in the Guidelines for Accurate and Transparent Health Estimates Reporting (the GATHER statement) it is now considered essential for best practice that authors report a quantitative measure of uncertainty, such as uncertainty intervals . It is also requested to report which sources of uncertainty were, and were not, accounted for. The authors of the guideline remark: “Best practices for calculation of uncertainty intervals, and especially for combining multiple sources of uncertainty, are an area of active research. By requiring that researchers report a quantitative measure of uncertainty, and that they state which sources of uncertainty are accounted for, we aim to advance science in this area.” .