We are extremely grateful for the effort of Dr Kuendgen et al.1 in validating our recently reported prognostic score for patients with lower risk (low or intermediate-1) myelodysplastic syndrome (MDS).2 First, we were happy to see that implementation of the model in the German population resulted in the identification of a group of patients with low-risk disease (N=164, 35%, OS 80 months) from another group (N=311, 65%) with significantly worse outcomes (OS from 30 to 35 months). These compared to a median survival of 80 months for our low-risk category 1 group (although this group represented only 21% of our patients), and 27–14 months for the two higher risk categories, which represented close to 80% of patients in our model. Indeed, one of the main objectives of our study was to precisely establish the point that ‘lower risk’ MDS is a misconception and that a large fraction of patients in this category have relatively poor prognosis and would benefit from early intervention. It should be noted that, for simplicity, we arbitrarily divided our patients into three subgroups for graphic display, but, indeed, the model divided the patients into seven different categories or scores, and, therefore, Kuendgen et al. have used an oversimplified version of the score. In their analysis, the authors show that the model was not a good predictor of time to acute myeloid leukaemia transformation, whereas the recent WPSS3 score was able to separate patients both in terms of OS and time to progression to acute myeloid leukaemia. It should be noted that the WPSS was validated in the group of patients included in the Düsseldorf registry and, therefore, the analysis of Kuendgen et al. in that regard is of limited value. Second, the reasons for differences in the risk to transformation to acute myeloid leukaemia are probably related to specific characteristics of the patient populations studied and specific referral patterns.
We agree that the differences observed between both models are related not only to intrinsic differences in the two patient populations studied, but also to the application of the model itself. First, it is clear that our patient population constitutes a group of patients referred to a tertiary care center, whereas the German group represents both patients referred as well as patients not referred but belonging to their community. It would be anticipated that the prognosis of our patients is worse, justifying the referral from the community physician to our center. As pointed out by Kuendgen et al., in our study, we could not control for a very important characteristic: time from initial diagnosis to referral. Indeed, our model was developed as a platform to incorporate therapeutic interventions for patients referred to our center as opposed to a predictive model at the time of initial diagnosis. Kuendgen et al. have also cited other characteristics as important. Among these was the fact that we could not incorporate the number of transfusions or ferritin levels (although by univariate analysis the latter was significant). Kuendgen also criticized our oversimplified use of cytogenetic information. As described in the manuscript, in this analysis of ‘lower risk patients,’ a large fraction was diploid with not even a single dominant group that could be analyzed independently. Other characteristics cited included LDH (not analyzed by us) and β2-microglobulin (predictive by univariate analysis).
This is a preview of subscription content, access via your institution