The triple A model (age, absolute neutrophil count, absolute lymphocyte count‐<scp>AAA</scp>) predicts survival and thrombosis in polycythemia vera

Abstract

BCR::ABL1-negative chronic myeloproliferative neoplasms (MPNs), polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis are characterized by constitutive activation of the JAK-STAT signaling pathway caused by mutations in the Janus Kinase 2 (JAK2), carelticulin, or thrombopoetin receptor genes. These neoplasms share persistent chronic inflammatory state and high cardiovascular risk. The main treatment goal in ET and PV is to prevent thrombotic events; PV patients older than >60 years or those with thrombosis history are considered to be at high risk of future thrombotic events according to European Leukemia Network (ELN) criteria, whereas prior thrombosis or age >60 years along with the presence of JAK2 mutation classifies ET patients at high risk. To mitigate thrombotic risk, all PV patients receive aspirin, as well as low-to-high ET patients. Additionally, high-risk patients are treated with hydroxyurea or interferons.1 Advanced age, leukocytosis, abnormal karyotype, and adverse mutations are associated with inferior overall survival (OS) in PV,2 while leukocytosis and prior thrombosis are the main determinants of OS in ET patients.3 In recent years, significant efforts have been made in order to refine the current risk stratification in MPNs. A major breakthrough has been the development of the molecularly defined risk stratification in PV and ET (MIPSS-PV and MIPSS-ET)4; however, molecular analyses are costly, require specialized infrastructure, and are not widely available. Therefore, attemps have been made to identify more accessible parameters which may help the clinicians worldwide to improve the prognostication of these patients, most often by using parameters within the complete blood cell count (CBC), that is, red blood cell distribution width (RDW),5 absolute neutrophil and lymphocyte counts,6, 7 or the ratios of neutrophil/lymphocytes (NLR)8 and platelets/lymphocytes (PLR).9 The rationale behind it is that stronger clonal myeloproliferation (i.e., neutrophilia and thrombocytosis) and more pronounced immune dysregulation (i.e., lymphopenia) in MPNs may differently affect distinct CBC compartments (i.e., in opposite directions) and be associated with inferior outcomes in patients with an increased tumor burden. In 2023 Tefferi et al.7 examined the individual prognostic contribution of age, absolute neutrophil count (ANC), and absolute lymphocyte count (ALC) on OS in ET patients; rising age, high (>8 × 109L) ANC, and low (<1.7 × 109L) ALC were individually associated with an inferior OS. Using hazard ratio (HR)-weighted scores the authors developed and externally validated (University of Florence) a novel 4-tiered prognostic which they termed triple A (Age, Absolute neutrophil count, Absolute lymphocyte count-AAA), assigning 4 points for age >70 years, 2 points for age 50–70 years, and 1 point each for ANC > 8 × 109L and ALC < 1.7 × 109L; AAA risk categories included low-risk (0–1 points), intermediate-1 risk (3 points), intermediate-2 risk (4 points), and high-risk (5–6 points) and were able to excellently discriminate OS. Finally, the prognostic impact of the AAA model was also independent of abnormal karyotype and adverse mutations.7 Therefore, this cheap and easily applicable prognostic model may have the potential to be used globally for prognostication of ET patients. Here, we investigated whether AAA model may also predict OS and thrombotic risk in PV patients. This multicenter study was conducted at four hematological centers in Croatia and Serbia in the period from 2002 to 2021 and was approved by the Ethics Committees from all participating centers. Patients with PV whose disease diagnosis was reassessed according to 2016 World Health Organization criteria were retrospectively included. Clinical and laboratory data at the time of disease diagnosis were collected through medical chart review. Comorbidities were assessed cumulatively with the Charlson Comorbidity Index (CCI); specific cardiovascular comorbidities of interest were arterial hypertension, diabetes mellitus, hyperlipidemia, and smoking. Statistics were performed with MedCalc Statistical Software® (Ostend, Belgium, version 22.016). Shapiro–Wilk's test was used to check for data distribution. Categorical variables were compared using the chi-squared test and continuous variables were compared with Kruskall–Wallis test. The Jonckheere–Terpstra trend test was used to test trends of increase in different continuous variables across AAA risk categories. Survival analyses were perfomed with the Kaplan–Meier method, the log-rank test, and the Cox regression analysis. Time to thrombosis (TTT) was measured from the time of disease diagnosis until a thrombotic event or last follow-up visit with death being a censoring event. Arterial thrombotic events considered were acute myocardial infarction, transitory ischemic attack, ischemic stroke, or acute peripheral arterial occlusion; venous thrombotic events of interest were pulmonary embolism and/or deep vein thrombosis. OS was measured from the time of diagnosis until death or the last follow-up visit. A total of 279 PV patients were included; the median age was 66 (range 20–92), 131 (47%) were females, and the median CCI was 3 (range 0–8). High-risk disease was recorded in 201 (72%) of patients and 231 (82.7%) had at least one cardiovascular risk factor. A total of 27 (9.7%), 141 (50.5%), 43 (15.4%), and 68 (24.4%) were classified as low, intermediate-1, intermediate-2, and high risk according to the AAA model,7 respectively. The overall patient characteristics and stratified according to AAA risk categories are summarized in Table S1. As presented, higher AAA categories were enriched in older patients, with high-risk disease, arterial hypertension, and with higher total leukocyte and ANC (p < .050); these associations may, at least partly, be due to the intrinsic design of the AAA risk model which accounts for older age and higher ANC. On the other hand, smokers were more frequent in lower AAA category (p for trend = .009), most probably due to higher tobacco use among the younger population. There were no statistically significant differences with respect to other clinical and laboratory variables. The median follow-up time was 61.8 months (range 1.8–213); 121 (43.4%) deaths and 48 (17.3%) thrombotic events (arterial 33 and 15 venous) were recorded during this time. Death causes included heart failure (n = 45, 37.2%), thrombotic events (n = 18, 14.8%), other cancers (n = 17, 14%), infections (n = 2, 1.6%), PV progression (n = 2, 1.6%), other/unknown (n = 37, 30.5%). Univariately, rising age according to the AAA model was associated with an inferior OS with the corresponding median survival not reached, 164, and 76 months for age <50, 50–70, and >70 years, respectively (p < .001). High (>8 × 109/L) ANC was also associated with an inferior OS (median survival 164 vs. 100 months, HR = 1.8, p = .001) whereas low (<1.7 × 109/L) ALC did not have an impact on OS (p = .700). In the multivariate Cox regression analysis, rising age (HR = 18.1, p < .001), high ANC (HR = 5.50, p = .019), and higher CCI (HR = 13.52, p < .001) were independently associated with an inferior OS when being additionally adjusted for sex, low ALC, cytoreductive treatment, and prior thrombosis (p > .050 for all analyses). Finally, as shown in Figure 1A, the AAA model was able to excellently differentiate OS in PV patients with the median OS being not reached, 172, 108, and 76 months for the low, intermediate-1, intermediate-2, and high-risk patients, respectively (p < .001). In a similar multivariate Cox regression analysis, the AAA model (HR = 23.94, p < .001) and CCI (HR = 18.12, p < .001) were both independently associated with an inferior OS when being controlled for sex, prior thrombosis, and cytoreductive treatment (p > .050 for all analyses). With regard to future thrombotic risk, rising age according to the AAA model (median TTT not reached vs. 208 vs. 153 months for age <50 years, age 50–70 years age >70 years, respectively; HR = 2.7 for both older age groups vs. the youngest age group, p < .001) was the strongest determinant of thrombotic risk, along with high ANC (HR = 1.82, p = .040), whereas high-risk ELN status (HR = 1.77, p = .070) and low ALC (HR = 1.76, p = .091) were of borderline significance; prior thrombosis was not associated with an inferior TTT (p = .312). In the multivariate Cox regression model, rising age (HR = 6.73, p = .009) and high ANC (HR = 3.6, p = .050) were both independently associated with an inferior TTT when being additionally adjusted for low ALC, cytoreductive treatment, prior thrombosis, and sex (p > .050 for all analyses). Finally, the AAA model was also able to differentiate TTT in PV patients (p < .001) as shown in Figure 1B, with patients in low-to-intermediate risk categories having comparable thrombotic risk over time and patients in the high-risk category having the shortest TTT (high-risk vs. other HR = 5.88, p < .001). The negative prognostic impact of high AAA was evident for both arterial (HR = 3.58, p = .007) and venous thrombotic events (HR = 28.79, p < .001), as illustrated in Figure S1. We further investigated how do ELN and AAA risk models mutually compare regarding thrombotic risk stratification. Two scores were not mutually independent regarding TTT probably due to overlapping contribution of advanced age that was highly valued in both scores; however, the Harrell's C-index for the prognostication of future thrombosis was higher for the AAA model (0.627) than for the ELN risk stratification (0.538). It should be pointed out that cytoreductive treatment was mostly driven by the ELN risk score in this real-life cohort of patients which could have diminished its prognostic performance. Considering that the AAA model does not account for prior thrombosis, we also investigated whether it may predict future thrombosis regardless of thrombosis history. Importantly, the negative prognostic impact of high AAA was evident both in patients with (HR = 3.54, p = .042) and without prior thrombosis (HR = 6.99, p < .001), as shown in Figure S2. This observation may suggest a good applicability of the AAA model for thrombotic risk prognostication in different PV patient subpopulations, as it additionally accounts for ANC and ALC, variables associated with increased myeloproliferation and immune dysregulation which were recently shown to bear prognostic significance in both ET and PV with respect to thrombotic risk.6-9 Finally, when simultaneously stratifying patients according to both, ELN and AAA high-risk categories as shown in Figure S3, the AAA low-to-intermediate risk group had similar thrombotic risk regardless of ELN risk status, whereas only patients classified as high risk according to both AAA and ELN risk categories experienced shorter TTT in comparison to other risk categories (HR = 5.88, p < .001). Notably, there were no AAA high- and ELN low-risk patients, since it is not possible to achieve high-risk AAA status without points given for age due to intrinsic design of the AAA score. Notably, both ANC and ALC were not prognostic of an inferior TTT in ELN low-risk PV patients (p > .050 for both analyses), suggesting that the AAA model may not refine the ELN low-risk thrombotic risk stratification. On the other hand, high-risk AAA category performed well in both patients with and without thrombosis history as mentioned previously. In fact, a significant proportion (n = 133/201, 51.2%) of ELN high-risk patients may actually have lower thrombotic risk according to their ANC and ALC status, as shown in Figure S3. This interesting observation may be interpreted in two ways. It may suggest that the use of the AAA model may potentially help to identify a proportion of ELN high-risk patients actually having a lower risk of thrombosis despite being classified as ELN high-risk allowing for more personalized management. On the other hand, it may also indicate that some ELN high-risk patients treated with cytoreduction may experience attenuation of thrombotic risk only when being classified as non-high risk according to the AAA model and that the use of this particular scoring system could, therefore, help to identify a subset of ELN high-risk patients with an unmet clinical need for more intensive (i.e., twice-daily aspirin) or alternative therapeutic approaches (i.e., different cytoreductive treatments or combination therapies). Nevertheless, additional validation of our results is needed before application of the AAA model could be implemented into routine clinical practice. Finally, collaborative international studies on larger datasets and among demographically different populations are needed to identify the optimal and PV-specific cut-off values of ANC and ALC for prognostication of clinically relevant outcomes in PV. As only two (1.1%) PV patients (of 179 with available data) transformed to myelofibrosis and none transformed to acute leukemia we were unable to analyze the potential impact of the AAA model on the risk of disease transformation. This low incidence of disease transformation probably corresponds to a relatively short follow-up time (~5 years) where the median time to acute leukemia and secondary myelofibrosis transformation in PV patients is 4.6–19 and 8.5–20 years, respectively.10 In addition, approximately half of the patients included in the study were managed in the community setting suggesting that some events may have not been captured. Therefore, future studies are needed to elucidate whether the AAA model may also predict the risk of disease transformation. Limitations of this study are its retrospective design, limited number of patients included, and the relatively short follow-up period, as well as the absence of cytogenetic and molecular prognostic informations. Therefore, additional studies are needed to analyze whether the incorporation of high-risk somatic mutations (i.e, SRSF2) into the AAA model may help to further refine this prognostic model and to test its performance when compared to newer risk stratification models such as MIPSS-PV.4 Nevertheless, this study demonstrated that AAA model may be used for prognostication of PV patients with respect to two most important disease outcomes separately, thrombosis and death; this observation may be especially relevant when considering that cardiovascular reasons were the main causes of death. Moreover, the AAA model may also have the ability to further refine the prognosis of ELN high-risk patients. For this reason, additional studies on larger number of patients are needed to validate our results, as well as to prospectively evaluate whether AAA score may be used to guide therapeutic decisions. Finally, whether the AAA model may also have a prognostic role in myelofibrosis remains to be elucidated in future studies. The authors declare no conflicts of interest. Data used for the generation of this study is available upon reasonable request directed at the corresponding author (Ivan Krecak). Table S1. Patient characteristics. The chi-squared, the Kruskall–Wallis test, and the Jonckheere–Terpstra trend test were used. Figure S1. Time to arterial- and time to venous thrombosis stratified according to AAA risk categories. The Kaplan–Meier method and the log-rank test were used. Figure S2. Time to thrombosis stratified according to AAA risk categories in patients with and without prior thrombosis. The Kaplan–Meier method and the log-rank test were used. Figure S3. Time to thrombosis stratified according to European Leukemia Network (ELN) and AAA risk categories. The Kaplan–Meier method and the log-rank test were used. 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.

Topics & Concepts