Comparative clinical and molecular landscape of primary and secondary myelofibrosis: Superior performance of MIPSS70+ v2.0 over MYSEC‐PM

Comparative clinical characteristics, molecular landscape and prognosis scoring for primary (PMF) and secondary myelofibrosis (SMF). Among myeloproliferative neoplasms (MPNs), myelofibrosis (MF) can occur de novo in primary MF (PMF) or secondary (SMF) to polycythemia vera (PV) or essential thrombocythemia (ET). For patient risk stratification and appropriate treatment allocation, many clinico-biological prognostic scores have been developed, such as the international prognostic scoring system (IPSS), dynamic IPSS (DIPSS) or DIPSS-plus. In the last decades, several studies showed the prognostic relevance of molecular alterations in PMF, including mutations in driver or High Molecular Risk (HMR) genes (ASXL1, EZH2, IDH1/2, SRSF2, or U2AF1). Based on these studies, genetic and molecular alterations were integrated in the latest prognostic scores, such as the mutation-enhanced IPSS 70 (MIPSS70),1 MIPSS70-plus and the MIPSS70+ version 2.0. As opposed to PMF, few studies explored mutations prognosis significance in SMF. JAK2-mutated and triple negative SMF have been associated with a higher AML incidence, while CALR-mutated SMF showed an increased survival compared with JAK2-mutated patients, independently of type-1/type-2 mutational status. Thus, the myelofibrosis secondary to PV and ET prognostic model (MYSEC-PM)2 was specifically developed for SMF patients and displays a more accurate prognostic performance than IPSS and DIPSS. Concerning HMR mutations, only SRSF2 mutations have been associated to poorer survival in post-ET MF (PET-MF) patients; however, no molecular prognostic models accounting for additional mutations in SMF have yet been described. We therefore set to describe and compare the clinical and molecular landscapes of PMF and SMF, as well as patients outcome, across a monocentric cohort of 455 MF patients, including 227 PMF patients, 14 pre-PMF (hence forth considered as PMF), 86 PPV-MF, and 128 PET-MF, followed between 2011 and 2021 in Saint-Louis hospital MPN comprehensive reference center. Figure S1 depicts patient flowchart. Patient characteristics are presented in Table S1. Interestingly, at the time of diagnosis, median age was 58 for PMF and 62 for SMF patients (p < .001). The PMF cohort was composed of less females than the SMF cohort (33.2% vs. 57.9% respectively, p < .001). As shown in many studies, age and gender influence MPN presentation and outcome: as females are more represented in ET, they are also more represented in PET-MF. However, we also found that female prevalence in post-PV MF (PPV-MF) (58%) is higher than female prevalence in PV (around 33%–35%). Thus, gender seems to play a role also on myelofibrotic evolution. As younger PV and ET patients tend to be females, it seems logical that their risk of fibrotic progression is increased. PMF and SMF patients had similar complete blood counts, except for platelets, higher in SMF, driven by the PET-MF subgroup. Symptoms, transfusion need, and cytogenetic characteristics were similar between groups. No statistical difference was observed in response to first line treatment and at last news between PMF and SMF patients. PPV-MF and PET-MF differed in median age at MF diagnosis (64 vs. 60 years respectively, p = .001) and sex (54.7% vs. 60.2% females respectively, p < .001). As expected, driver gene mutations and median platelet levels were also different. Interestingly, MYSEC-PM classification was significantly different, including more PET-MF patients in the low-risk category in comparison with PPV-MF (30% vs. 15% respectively, p = .008) (Table S2). Within our cohort, 315 patients had an available NGS molecular evaluation (160 PMF and 155 SMF). Figure 1A depicts PMF and SMF mutational landscapes. Out of 241 PMF patients, the driver mutation was JAK2V617F in 139 (57.7%), MPL in 18 (7.5%) and CALR in 63 (26.1%) patients; no driver was found in 20 (8.3%) "triple negative" patients. In PPV-MF, the driver mutation was JAK2V617F in 85 (98.8%) patients. In PET-MF, the driver mutation was JAK2V617F in 59 (46.1%), MPL in 9 (7.0%) and CALR in 49 (38.3%) patients; while 10 (7.8%) patients were "triple negative." Interestingly, PMF patients displayed more additional mutations than SMF patients, with a median of 2 mutations (interquartile range [IQR]: 1–3) compared to 1 (IQR: 0–2) respectively (p = .003). Adverse mutations, including HMR and TP53, were predominant in PMF in comparison with SMF patients (38.2% and 28.0% respectively, p = .001). Similarly, "signaling pathway" mutations were more frequent in PMF than SMF (14.1% vs 7% respectively, p = .008). We did not find any significant difference between PMF and SMF in the mutational frequency of the following gene categories: "transcription factors" (10.8% vs. 10.3% respectively), "spliceosome genes" (8.7% vs. 5.1% respectively), and "epigenetic genes" (27.4% vs. 28.0% respectively) (Table S1). PPV-MF and PET-MF did not show any significant differences in additional mutations number and gene type (Table S2). Pairwise co-occurrence of mutations according to PMF or SMF status is represented in Figure S2. JAK2, TET2, DNMT3A, and ASXL1 were the most frequently associated mutations. Using a logistic regression analysis to find covariates associated with PMF or SMF, we highlighted that female gender (odds ratio [OR] = 2.77, 95% confidence interval [CI]: 1.89–4.06, p < .0001), higher IPSS, fewer additional mutations (OR = 0.80, 95% CI: 0.70–0.93, p = .003), fewer HMR-TP53 mutations (OR = 0.47, 95% CI: 0.30–0.73, p = .001), and fewer signaling pathway mutations (OR = 0.40, 95% CI: 0.21–0.76, p = .006) were associated with SMF status. Multivariate analysis confirmed the independent statistical impact of female status (OR = 2.64, 95% CI: 1.45–4.83, p = .002), higher IPSS and fewer presence of HMR-TP53 (OR = 0.45, 95% CI: 0.22–0.90, p = .025), and signaling pathway mutations (OR = 0.34, 95% CI: 0.14–0.84, p = .020) in SMF (Table S3 and Figure S3). We then compared PET-MF with PPV-MF using the univariate model described above, excluding driver mutation status, since near all PPV-MF patients are JAK2V617F. In the univariate analysis, lower hematocrit (OR = 0.94, 95% CI: 0.89–1.00, p = .039), lower WBC count (OR = 0.95, 95% CI: 0.92–0.98, p = .003), lower ANC (OR = 0.92, 95% CI: 0.87–0.98, p = .005), an abnormal karyotype excluding complex/monosomal (OR = 0.37, 95% CI: 0.15–0.90, p = .028), and lower MYSEC-PM were significantly associated with PET-MF. In the multivariate analysis, only a lower MYSEC-PM remained significantly associated with PET-MF (Table S4). Interestingly, SMF patients harbored significantly less additional mutations compared to PMF, despite an older age at diagnosis. This finding challenges the idea that PV and ET fibrotic evolution is due to the accumulation of additional mutations in the driver-mutated clone. An alternative hypothesis is that very few additional mutations would be sufficient to induce fibrotic progression in the PV and ET long-term bone marrow niche inflammatory setting. To our knowledge, only 3 other studies directly compared PMF and SMF mutational status, but in smaller cohorts. Two3, 4 did not find any differences in the number of additional mutations, while the third5 seemed to report less mutations in SMF but did not perform any statistical comparative analysis. Furthermore, we report here that SMF patients are significantly less mutated in "HMR-TP53" and "signaling pathway" gene categories. While HMR gene mutations prognosis impact was established in PMF cohorts, it may play a different role in SMF pathogenesis. Li et al.3 and Courtier et al.4 reported a significantly lower proportion of "spliceosome" and higher proportion of TP53 mutations in the SMF cohort compared with PMF. TP53 prevalence in our study (4.2% for both groups) was similar to the literature for PMF patients, whereas it is lower for SMF than the 14% prevalence reported by Courtier et al. We then evaluated the impact of PMF or SMF status on patient outcome. Within a median follow-up of 4.7 years (IQR: 2.2–9.6) in the PMF cohort and 3.5 years (IQR: 1.6–6.7) in the SMF patients' cohort, a total of 49 (20.3%) and 46 (21.5%) deaths occurred among PMF and SMF patients, respectively. Overall survival (OS) was not statistically different between groups (5-year survival: 85.2% in PMF vs. 79.4% in SMF, p = .065) (Figure 1B). Most frequent causes of death were progression or graft versus host disease for PMF, and infection for SMF (Figure S4). Twenty-seven (11.2%) PMF and 17 (7.9%) SMF patients transformed into AML/MDS and transformation-free survival (TFS) was not different between PMF and SMF patients (5-year TFS: 90.9% in PMF vs. 92.6% in SMF, p = .640) (Figure 1C). Outcomes comparison between PET-MF and PPV-MF did not show any difference in terms of OS (p = .600), nor in terms of TFS (p = .814) (Figure S5). These results should be put in perspective with the SMF groups' lower prevalence of "HMR-TP53" mutations. Indeed, despite SMF displaying a lower mutational risk profile, it is not associated with a better OS and TFS. Although being responsible for the same outcomes, PMF and SMF physiopathological history seems to differ. Finally, we set to evaluate classical PMF prognostic scores performance in our SMF cohort, using Harrell's C-index (Figure 1D,E). IPSS and DIPSS concordance indexes were estimated at 0.725 and 0.694, respectively, while MYSEC-PM, the score specifically developed for SMF, had a better performance (C-index = 0.731). Interestingly, although developed for PMF, MIPSS70 and particularly MIPSS70+ v 2.0 displayed a high-performance-index (C-index = 0.731 and 0.794 respectively), therefore more accurately stratifying SMF patients according to their risk. These results confirm that adding cytogenetic and molecular data improves prognostic scores, in line with a recent study highlighting DIPSS+ better performance compared with MYSEC-PM in SMF.6 To conclude, we report here the largest comparison of PMF and SMF cohorts in the literature, highlighting their clinical and molecular differences. We show a differential molecular landscape of each disease subtype suggesting diverse physiopathologies, despite similar OS and TFS. Finally, our study validates MIPSS70 and MIPSS70+ v2.0 as the first molecular scores accounting for additional mutations and cytogenetic abnormalities in SMF, therefore improving accurate patients' prognosis stratification. C.M., L.-P.Z., R.D.O., N.G., and M.C. collected the data. S.G., N.M., E.V., and B.C. performed molecular analyses. J.S.-D, R.D.O., W.V., N.P., E.R., S.G., J.-J.K., and L.B. provided patients care. H.P. performed statistical analysis. L.B. and M.G. analyzed the data, performed statistical analysis and wrote the manuscript. L.B. designed and supervised the study. All co-authors reviewed, edited and critically discussed the manuscript. The authors thank the clinical care team of the Comprehensive Myeloproliferative Neoplasms Center for samples and data collection, and the staff of the cellular biology laboratory for excellent technical assistance. The authors also thank the French Intergroup for Myeloproliferative neoplasms (FIM) for insightful discussions. Lina Benajiba is an "ATIP-Avenir" grant recipient. This work was supported by an "Association Laurette Fugain," a "Féderation Leucémie Espoir," a "Fondation ARC pour la recherche sur le cancer," an "INCa Prev-Bio N° 2021-164," and a "CCA-INSERM-Bettencourt" funding. L.B. has received research support from Gilead and Pfizer for research projects unrelated to the current study. The remaining authors declare no competing financial interests related to this work. The data that support the findings of this study are available from the corresponding author upon reasonable request. Data S1. Supporting information. 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.