Essential thrombocythemia (ET) is classified as a Philadelphia chromosome–negative myeloproliferative neoplasm and it carries a heterogeneous clinical course, encompassing an increased risk of both thrombotic and hemorrhagic complications. It may, over time, evolve into myelofibrosis or, less commonly, transform into acute leukemia 1, 2. The primary therapeutic goals in ET are symptom control and prevention of thrombosis. Decisions regarding cytoreductive therapy in ET are guided by risk stratification, with age, history of thrombotic events, and JAK2 mutational status serving as the primary determinants, while cardiovascular risk factors play a key role in determining the use of aspirin 3-5. Thus, robust and widely applicable risk models are critical for optimizing treatment decisions in ET. Currently, prognostic scoring systems for survival assessment in ET are limited. The international prognostic score for essential thrombocythemia (IPSET), which consists of age, leukocytosis, and history of thrombosis, has been widely used for over a decade 6. Furthermore, revised IPSET-T and IPSET-T models were employed for the prediction of thrombosis in ET 7, 8. More recently, the Triple A (AAA) model, which integrates absolute neutrophil count, absolute lymphocyte count, and age, has been proposed and shown to provide superior predictive performance on survival than IPSET 9. Our group has also validated this Triple A score recently 10. We read with great interest the recent article by Tefferi et al. 11 introducing the AAA+ model for predicting survival in patients with ET. The AAA+ model represents an advancement over the initial AAA model by integrating additional variables including absolute monocyte count (≥ 0.5 × 109/L), sex, arterial thrombosis history, and hypertension. Through this approach, the model integrates host-related factors, cardiovascular risk factors, and disease-related variables. Notably, the model defines an ultra-low risk group with a median survival of 42.7 years. When comparing model-predicted survival with observed outcomes and population-matched expected survival, the ultra-low risk group demonstrated survival comparable to that of the general population, whereas higher-risk groups showed a clear divergence, underscoring the impact of disease-related mortality. Survival outcomes (overall (OS), thrombosis- (TFS), bleeding-, and myelofibrosis-free survivals) were estimated using the Kaplan–Meier method and compared by log-rank test. Median survival with 95% confidence interval (CI) was reported, and restricted median survival time (RMST) was calculated when median survival was not reached. The prognostic value of AAA+ risk stratification was assessed using Cox regression, expressed as hazard ratio (HR) with 95% CIs, referencing the ultra–low-risk group. Model discrimination and fit were evaluated using the C-index and Akaike Information Criterion (AIC), with a higher C-index and lower AIC indicating better performance. To test the utility of the AAA+ score, we applied this score in our cohort of 565 patients that was previously employed for the validation of the Triple A score 10. Demographic, clinical, and laboratory data of this cohort were provided in Table S1. Over a median follow-up of 5.99 years (IQR, 2.66–10.19), 62 patients (10.9%) died. Thrombosis occurred in 58 patients (10.3%), bleeding in 25 (4.4%), progression to myelofibrosis in 29 (5.1%), and transformation to leukemia in 6 (1%). According to the AAA+ system, 24.8% (n = 140), 43.7% (n = 247), 12.6% (n = 71), and 18.9% (n = 107) were classified ultra-low-, as low-, as intermediate-, and high-risk, respectively. The median OS was 22.9 years for individuals in the low-risk group and 9.8 years for those in the high-risk group. The RMST for the ultra-low and intermediate-risk groups were 28.9 years (95% CI: 27.4–30.4) and 16.9 years (95% CI: 16.5–17.4), respectively. No deaths were observed at 5 and 10 years in the ultra-low risk group, and 20-year OS was 93.8% (95% CI: 86–100). In the low-risk group, OS was 98.4% at 5 years (95% CI: 96.6–100), 95.3% at 10 years (95% CI: 91.6–99), and 63.2% at 20 years (95% CI: 44–89.9). In the intermediate-risk group, OS was 91.6% at 5 years (95% CI: 84.8–99) and 77.4% at 10 years (95% CI: 65.4–91.7). In the high-risk group, the 5- and 10-year OS rates were 71.7% (95% CI: 61.9–83.1) and 43.2% (95% CI: 28.9–64.6), respectively. According to Kaplan–Meier analysis, there were significant differences in OS across the risk groups (p < 0.0001) (Figure 1). The AAA model demonstrated a C-index of 0.82, while the AAA+ model showed a slightly higher C-index of 0.84. The AIC values were 582.37 and 580.85 for the AAA and AAA+ models, respectively. Using the ultra-low risk group as reference, patients in the low-risk group had an approximately 6.5-fold higher risk of death (HR: 6.53; 95% CI: 1.42–29.96; p = 0.02). In the intermediate-risk group, this risk was markedly higher (HR: 19.56; 95% CI: 4.09–93.61; p < 0.001), and for the high-risk group, the risk of mortality was extremely elevated (HR: 68.76; 95% CI: 15.19–311.20; p < 0.001) (Figure 2). In the ultra-low-risk group, TFS was 95.1% (95% CI: 91%–99.4%) at 5 years, 91% (95% CI: 85%–98%) at 10 years, and 89% (95% CI: 81%–97.2%) at 20 years. Corresponding TFS rates were 94.9% (95% CI: 91.8%–98.0%) and 88.6% (95% CI: 83.2%–94.3%) in the low-risk group, 91.5% (95% CI: 84.5%–99%) at both 5 and 10 years in the intermediate-risk group, and 70.8% (95% CI: 60%–83.4%) and 59.5% (95% CI: 45.9%–77.1%) in the high-risk group, respectively. The RMST was 28.9 years (95% CI: 27.0–30.7) for the ultra-low-, 17.4 years (95% CI: 16.2–18.6) for the low-, 14.8 years (95% CI: 13.9–15.8) for the intermediate-, and 11.1 years (95% CI: 9.4–12.8) for the high-risk group (Figure 3). Compared to the ultra-low-risk group, the risk of thrombosis was higher but not statistically significant in the low- (HR: 1.84; 95% CI: 0.80–4.23; p = 0.10) and intermediate-risk groups (HR: 1.53; 95% CI: 0.49–4.77; p = 0.50). In contrast, the high-risk group demonstrated a markedly increased thrombotic risk (HR: 8.01; 95% CI: 3.44–18.66; p < 0.001) (Figure S1). There were no significant differences in terms of bleeding- and myelofibrosis-free survivals among risk groups (p = 0.71 and p = 0.17, respectively) (Figure S2). We demonstrated that the AAA+ score is a strong predictor of survival. Patients in the ultra-low risk group have markedly better long-term survival, whereas early mortality is substantially higher in the high-risk group. Moreover, both the AAA+ model and the AAA model demonstrated C-index values above 0.80 (0.84 and 0.82, respectively). Tefferi and colleagues 11 reported results from three cohorts: the Mayo Clinic cohort, serving as the discovery cohort with 658 patients, and two validation cohorts consisting of 5968 patients from Israel and 682 patients from Italy. Except for the Mayo Clinic cohort, where the C-index of the AAA model was 0.79, the C-index values for both the AAA and AAA+ models exceeded 0.80 in the other two cohorts. Consistent with these findings, the AAA+ model yielded higher C-index values than the AAA model across all three cohorts, suggesting that the AAA+ model may provide slightly better discrimination of patient outcomes. However, it should be noted that statistical significance was reached only in the Mayo Clinic discovery cohort. In contrast, the differences in C-index between the AAA+ and AAA models in the Israeli and Italian validation cohorts did not achieve statistical significance. On the other hand, the C-index of the AAA+ model was consistently higher than that of the IPSET score in all three cohorts. Therefore, although the AAA+ model demonstrates some statistical superiority over the AAA model, caution is warranted in generalizing these findings, and further validation in independent cohorts may be required. The original AAA model is valued for its simplicity and objectivity, as it relies exclusively on complete blood count parameters and age. In comparison, the AAA+ model incorporates a history of arterial thrombosis, which may be difficult to ascertain consistently due to reliance on medical records and patient reporting, and does not define explicit thresholds for arterial hypertension, potentially necessitating clinical judgment. Despite the challenges in interpreting thrombosis history, it remains a well-established risk factor in ET. Additionally, monocytosis, a parameter that we previously investigated, has also been shown to carry prognostic significance. The AAA + A model, which incorporates absolute monocyte count into the original AAA framework, was first proposed by Tefferi et al. 12 and subsequently validated by an external group 13 and then by us 10. Thus, the integration of monocyte count appears to provide improved prognostic value. We also showed that the AAA+ score is a predictor of TFS and can meaningfully distinguish between clinical risk groups. Therefore, the AAA+ model could be alternatively used for thrombosis prediction. Similar to the AAA model 9, 10, this model also did not show a significant difference across risk groups in terms of both bleeding- and myelofibrosis-free survivals. Consequently, despite the retrospective design and relatively small sample size, our study demonstrated that increasing AAA+ scores were strongly associated with poorer survival and effectively distinguished between risk groups. The AAA+ model clearly introduces further refinements, and its validation in a cohort of 7308 patients represents a substantial advantage. Our findings are consistent with their observations. The authors have nothing to report. The authors have nothing to report. This study was approved by the local ethical committee. The authors declare no conflicts of interest. Research data are not shared. Figure S1: Hazard ratios for thrombosis risk across AAA+ risk groups estimated using Cox regression. Figure S2: Kaplan–Meier analysis of bleeding- (A) and myelofibrosis- (B) free survivals in different AAA+ risk groups. Table S1: Demographical, clinical and laboratory features of the patient cohort at diagnosis (IQR: Interquartile range) *CALR and MPL mutations were analyzed in 424 patients. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
Sönmez et al. (Sat,) studied this question.