In MDS Del(5q), Factors Other Than Age and Sex Distribution Contribute to the Prognostic Advantage, Which Diminishes over Time

Introduction

The WHO classification of myelodysplastic syndromes defines several subtypes. Prognosis of MDS del(5q) is considered favourable, which is reflected in the WHO prognostic scoring system (WPSS). Compared to other subtypes like refractory cytopenia with unilineage dysplasia (RCUD), refractory cytopenia with multilineage dysplasia (RCMD) and refractory anaemia with ringed sideroblasts (RARS), MDS del(5q) patients are however younger and the percentage of female patients is higher.

Thus, the aim of this work was to determine to which extent the prognostic advantage of MDS del(5q) over other MDS subtypes is due to the more favourable age and sex distribution in this group.

Patients and Methods

Between 1982 and 2013, 5383 patients were included in the Duesseldorf MDS registry. We identified 2011 patients who had no more than 5% blasts in the bone marrow and were classified either as RCUD, RCMD, RARS or MDS del(5q).

Besides WHO type and blast count, sex, age at diagnosis as well as the classification into primary and secondary MDS were regarded. 99 patients were omitted from the analysis due to missing values. Thus, the final data set contained 1912 patients, of whom 1576 patients had received best supportive care only as well as 336 patients who had received one or more therapies during the observation period, the most frequent being lenalidomide.

Kaplan-Meier curves and Cox proportional hazards models were estimated for mortality. Interaction effects between WHO type and sex were tested. Models were compared using the Akaike information criterion (AIC). Additionally “progression to AML” was analysed with the competing event “death without progression”.

Results

According to the WHO criteria, 125 (7%) were classified as MDS del(5q), 265 patients (14%) as RCUD, 206 (11%) as RARS and 1316 (69%) as RCMD. Median age was 68 years (range: 32-89) for MDS del(5q), 70 years (17-91, RCUD), 73 years (18-88, RARS) and 71 years (18-93, RCMD). In our sample 911 patients (48%) were women, but with considerable differences between the different WHO subtypes (MDS del(5q):67%, RCMD: 44%). The median observation time was 9.1 years (range: 0.01-31 years). During this period, 1312 deaths were observed, 194 patients progressed to AML.

Median survival was 6.17 years (95%-CI: [5.07-8.09]) for MDS del(5q) patients, 5.20 years (95%-CI: [4.41-5.98]) for those with RARS, 4.48 years (95%-CI: [3.70-5.49]) for RCUD patients and 2.96 years (95%-CI: [2.64-3.22]) for those with RCMD. Survival curves gave strong hints of non-proportionality.

Thus we had to incorporate a time-dependent effect for the WHO type in the multivariate cox model considering all covariates mentioned above, Best results were achieved with an effect of the type β₁+β₂*log(time). Among blasts and type of MDS (primary vs. secondary), age (hazard ratio (HR): 1.040, 95%-confidence interval (CI): [1.035-1.046]) and sex (HR: 0.856, 95%-CI: [0.766-0.956] for women vs. men) remained in the model as well. No interactions between age, sex and WHO type were found.

According to this model, at diagnosis HRs were 1.812 (95%-CI: [1.199-2.740], RCUD vs MDS del(5q)), 1.167 (95%-CI: [0.742-1.837], RARS vs MDS del(5q)) and 2.284 (95%-CI: [1.550-3.366], RCMD vs MDS del(5q)). Interestingly this advantage of MDS del(5q) disappeared over time, the hazard ratio became 1 after about 3.5 years for RCUD, 7.2 years for RCMD.

Regarding progression to AML no time-dependent coefficients were needed and we did not find any significant differences concerning the WHO subtypes. Age and sex did not seem to play a major role. The risk of progression to AML was mainly determined by the blast count and the type of MDS (primary vs. secondary).

Conclusions

Our analyses indicate that the higher survival probabilities are not only due to the different age and sex distributions. However, we found strong hints that the survival advantage of MDS del(5q) is vanishing over time, especially in comparison to RCUD and RCMD. For the next revision of the WPSS it should be discussed to introduce a time-dependent component at least for the different WHO subtypes, since it seems likely that the hazard ratios are not constant over time.