A Prognostic Model to Predict Ruxolitinib Discontinuation and Death in Patients with Myelofibrosis

Simple Summary Despite significant clinical activity, 50% to 70% of patients discontinue ruxolitinib within 3 to 5 years. The identification of patients who are more likely to discontinue it early has now become of paramount practical importance, given the availability of new drugs that are either approved (i.e., fedratinib, pacritinib, and momelotinib) or undergoing advanced clinical investigation for patients with a suboptimal response or ruxolitinib resistance (i.e., navitoclax, pelabresib, and imetelstat). A retrospective, real-world analysis was performed on 889 MF patients treated with ruxolitinib from the observational “RUX-MF” Italian study. We investigated predictors of early ruxolitinib discontinuation and death on therapy in 889 MF patients. Results confirm that prolonged ruxolitinib administration is associated with improved OS when compared to earlier discontinuation. While this outcome may be expected at the 5-year timepoint, interestingly, the survival advantage was observed also categorizing patients at earlier timepoints. Abstract Most patients with myelofibrosis (MF) discontinue ruxolitinib (JAK1/JAK2 inhibitor) in the first 5 years of therapy due to therapy failure. As the therapeutic possibilities of MF are expanding, it is critical to identify patients predisposed to early ruxolitinib monotherapy failure and worse outcomes. We investigated predictors of early ruxolitinib discontinuation and death on therapy in 889 patients included in the “RUX-MF” retrospective study. Overall, 172 patients were alive on ruxolitinib after ≥5 years (long-term ruxolitinib, LTR), 115 patients were alive but off ruxolitinib after ≥5 yrs (short-term RUX, STR), and 123 patients died while on ruxolitinib after <5 yrs (early death on ruxolitinib, EDR). The cumulative incidence of the blast phase was similar in LTR and STR patients (p = 0.08). Overall survival (OS) was significantly longer in LTR pts (p = 0.002). In multivariate analysis, PLT < 100 × 109/L, Hb < 10 g/dL, primary MF, absence of spleen response at 3 months and ruxolitinib starting dose <10 mg BID were associated with higher probability of STR. Assigning one point to each significant variable, a prognostic model for STR (STR-PM) was built, and three groups were identified: low (score 0–1), intermediate (score 2), and high risk (score ≥ 3). The STR-PM may identify patients at higher risk of failure with ruxolitinib monotherapy who should be considered for alternative frontline strategies.


Introduction
Ruxolitinib was the first JAK1/2 inhibitor approved for the treatment of patients with myelofibrosis (MF) with splenomegaly and/or symptoms.In large prospective and retrospective cohorts, ruxolitinib has been shown to reduce splenomegaly and MF-associated symptoms in around 50% of the patients and improve quality of life in most patients, regardless of response.In addition, the introduction of ruxolitinib into clinical practice has been associated with a significant increase in life expectancy [1,2].This beneficial effect was magnified in patients who achieved a spleen response, particularly if this occurred within the first 6 months of therapy.Indeed, in a pooled analysis of the COMFORT trials, a 5 dL increase from baseline in spleen volume correlated with a worse outcome; also, in the ruxolitinib phase 1/2 trial, MF patients with ≥50% reduction in spleen length were found to have significantly better survival [3,4].Finally, we have previously shown that MF patients who did not achieve a stable spleen response to ruxolitinib by 6 months had significantly worse overall and progression-free survival compared to stable and unstable responders [5].However, the durability of spleen response is limited, with more than 50% of responding patients losing such response within 3 to 5 years of therapy.Additionally, treatment-related toxicities, including on-target anemia and thrombocytopenia, infections, and second cancers, can substantially compromise the beneficial effect of ruxolitinib.As a result, despite significant initial clinical activity in many cases, 50% to 70% of patients discontinued ruxolitinib within 3 to 5 years [6,7].After ruxolitinib discontinuation, the therapeutic possibilities have so far been very scarce, and the outcome is very poor, with median survival ranging between 13 and 23 months [8,9].
Identifying patients who are more likely to respond well to ruxolitinib is a major goal of clinical research because it would also have important implications in clinical practice.Indeed, it would also guide the choice of firstline therapy.We previously showed that patients with higher disease burden (large splenomegaly, low platelet count, and transfusion dependency), higher risk category, and a longer time (>2 years) interval between diagnosis and start of therapy have a lower probability of spleen response at 6 months [3].Recently, the response to ruxolitinib after 6 months (RR6) model has been developed [4].The RR6 model is based on four parameters, including RUX dose <20 mg twice daily at baseline, months 3 and 6; palpable spleen length reduction from baseline ≤30% at months 3 and 6; red blood cell (RBC) transfusions at months 3 and/or 6; RBC transfusions at all time points.This model identifies a high-risk category of patients with impaired survival who might benefit from a prompt treatment shift.
Ruxolitinib has been the only JAK1/JAK2 inhibitor available in clinical practice for over a decade.Recently, new-generation JAK inhibitors have expanded treatment options for patients with splenomegaly and/or symptoms.Fedratinib is a selective JAK2/FLT3/BRMD4 inhibitor approved by the FDA in 2019 based on the results of the JAKARTA and JAKARTA-2 studies that included ruxolitinib naïve and experienced patients, respectively [10,11].After ruxolitinib failure, 55% and 21% of the patients achieved spleen and symptom response, respectively, at 6 months.Comparably to ruxolitinib, fedratinib may induce on-target hematological toxicity, with grade 3-4 anemia and thrombocytopenia occurring in 38% and 22% of patients, respectively [11].Pacritinib is a JAK1-sparing inhibitor of JAK2/FLT3/IRAK1/activin A receptor type 1 (ACVR1) that has received accelerated FDA approval for patients with splenomegaly and/or symptoms and severe thrombocytopenia (platelet counts <50 × 10 9 /L) [12].In addition to spleen and symptom improvement in patients with MF, results from retrospective analyses of subpopulations from the PERSIST-2 trial suggest pacritinib may also provide anemia benefits [13].The JAK1/JAK2/ACVR1 inhibitor momelotinib has been studied in three phase 3 trials [14 -16].Across these trials, momelotinib showed consistent spleen, symptom, and anemia benefits for patients with MF.
The addition of drugs with a different mechanism of action than JAK inhibition to ruxolitinib monotherapy is currently under clinical investigation in order to improve responses and possibly minimize hematological toxicity.Investigational clinical trials comparing combination therapy with ruxolitinib plus new agents and ruxolitinib alone in the frontline setting are also ongoing.
Overall, the identification of patients who are more likely to discontinue ruxolitinib early has now become of paramount practical importance, given the availability of new drugs that are either approved (i.e., fedratinib, pacritinib, and momelotinib) or undergoing advanced clinical investigation for patients with suboptimal response or ruxolitinib resistance (i.e., navitoclax, pelabresib, and imetelstat) [17][18][19].
The aims of this study were to investigate predictors of early ruxolitinib discontinuation and death on therapy in 889 MF patients and to create a prognostic score which could help to identify those patients at higher risk of failure of ruxolitinib monotherapy.

Study Setting and Definitions
After IRB approval, the "RUX-MF" retrospective study collected data from 889 chronic phase MF patients who received ruxolitinib outside clinical trials in 26 hematology centers.All centers were asked to report, in an electronic case report form (e-CRF), their consecutive MF patients who received ruxolitinib according to standard clinical practice.The total number of medical files was reported by each center by data input into an electronic database developed to record all study data after the de-identification of the patients with an alphanumeric code to protect personal privacy.Any treatment decision, including starting ruxolitinib doses and dose adjustments over time, was at the physician's discretion, based on patients' characteristics and independent from participation in this study.After the first data entry, the follow-up information was validated with a revision of clinical data and specific queries were addressed to the participating center in case of inconsistent data.All patients were followed from 2013 until death or to data cut-off (28 June 2022), with a median follow-up time of 4.4 years.
Risk category was assessed according to the Dynamic International Prognostic Score System (DIPSS) in primary MF (PMF) and to Myelofibrosis Secondary to Polycythemia Vera and Essential Thrombocythemia Prognostic Model (MYSEC-PM) in secondary MF (SMF) [20,21].MF-related symptoms were assessed using the 10-item Myeloproliferative Neoplasm Symptom Assessment Form Total Symptom Score (MPN10-TSS).Spleen and symptoms responses were assessed by palpation and by routine MPN-TSS evaluation, respectively, according to 2013 IWG-MRT/European LeukemiaNet (ELN) criteria [22].Progression to blast phase (BP) was defined according to WHO criteria [23].

Ethical Aspects
The RUX-MF study was performed in accordance with the guidelines of the institutional review boards of the participating centers and the standards of the Helsinki Declaration.All patients provided written informed consent.The promoter of this study was the IRCCS Azienda Ospedaliero-Universitaria S. Orsola-Malpighi, Bologna, which obtained approval from the Area Vasta Emilia Centro Ethics Committee (approval file number: 048/2022/Oss/AOUBo).The study was also approved by the local ethics committee of all participating centers (protocol code: RUX-MF).The study had no commercial support.

Statistical Analysis
Statistical analysis was carried out at the biostatistics laboratory of the MPN Unit at the Institute of Hematology "L. and A. Seràgnoli", IRCCS Azienda Ospedaliero-Universitaria, Bologna.
Continuous variables were summarized by their median and range, and categorical variables were summarized by the count and relative frequency (%) of each category.Comparisons of quantitative variables between groups were carried out by a two-sample Wilcoxon rank-sum (Mann-Whitney) test; association between categorical variables was tested by the χ 2 test.
The cumulative incidence of progression to the blast phase was calculated, treating death as a competing event.Comparison of overall survival (OS) was carried out with the Cox regression model in multivariable analysis with DIPSS or MYSEC-PM, with adjustment for delayed entry and evaluation of the model's performance in terms of goodness of fit.
Regressors associated with ruxolitinib discontinuation with p <0.05 in univariate analysis were jointly tested in a multivariate prognostic model.A rounded weight was associated with each risk factor based on its hazard ratio (HR).Sums of scores recognizing patients with a similar probability of early discontinuation were unified in risk categories.
Cumulative incidence of the blast phase with death as a competing event was similar in LTR and STR patients (2.4% and 4.8% at 5 years, p = 0.08).
Overall survival (OS) was significantly longer in LTR patients (p = 0.002) (Figure 1).Specifically, the median OS was 84.5 and 63.0 months in LTR and STR patients, respectively.S1).

Expanding the STR-PM to Earlier Timepoints
An additional analysis was performed, looking at different timepoints of ruxolitinib discontinuation.
First, the association between LTR/STR status and overall survival was investigated.Notably, OS improvements were also observed in patients who continued ruxolitinib over patients who discontinued the therapy within 3 years (Figure 3A), with a median OS of 64.7 and 54.0 months (p < 0.001).Accordingly, survival probability was higher for patients alive and on ruxolitinib after 1 year from therapy start (LTR-1 year and within 1

Expanding the STR-PM to Earlier Timepoints
An additional analysis was performed, looking at different timepoints of ruxolitinib discontinuation.
First, the association between LTR/STR status and overall survival was investigated.
Notably, OS improvements were also observed in patients who continued ruxolitinib over patients who discontinued the therapy within 3 years (Figure 3A), with a median OS of 64.7 and 54.0 months (p < 0.001).Accordingly, survival probability was higher for patients alive and on ruxolitinib after 1 year from therapy start (LTR-1 year and within 1 year, with a median OS of 45.9 and 32.3 months (p < 0.001)) (Figure 3B).
First, the association between LTR/STR status and overall survival was investigated.Notably, OS improvements were also observed in patients who continued ruxolitinib over patients who discontinued the therapy within 3 years (Figure 3A), with a median OS of 64.7 and 54.0 months (p < 0.001).Accordingly, survival probability was higher for patients alive and on ruxolitinib after 1 year from therapy start (LTR-1 year and within 1 year, with a median OS of 45.9 and 32.3 months (p < 0.001)) (Figure 3B).Second, the STR-PM was applied to both cohorts of LTR/STR 3 years and 1 year.Notably, the STR-PM was also applied considering earlier timepoints for ruxolitinib discontinuation.
In the STR-PM 1-year, the probability of early discontinuation was 12.6%, 18.0%, and 26.8% in low, intermediate, and high-risk patients (accounting for 45.2%, 34.6%, and 20.2% of the population, respectively) (p = 0.001) (Figure 4B).Second, the STR-PM was applied to both cohorts of LTR/STR 3 years and 1 year.Notably, the STR-PM was also applied considering earlier timepoints for ruxolitinib discontinuation.

Discussion
The therapeutic scenario of MF has undergone rapid changes in the last five years, thanks to the introduction of new JAK2 inhibitors that are already or about to be available in clinical practice (e.g., fedratinib, pacritinib, and momelotinib) [13,[24][25][26].Furthermore, multiple studies are being conducted to assess both the efficacy and safety of new drugs with different mechanisms of action, either as monotherapy or in combination with ruxolitinib.Therefore, the identification of patients who are more likely to respond poorly to

Discussion
The therapeutic scenario of MF has undergone rapid changes in the last five years, thanks to the introduction of new JAK2 inhibitors that are already or about to be available in clinical practice (e.g., fedratinib, pacritinib, and momelotinib) [13,[24][25][26].Furthermore, Cancers 2023, 15, 5027 9 of 12 multiple studies are being conducted to assess both the efficacy and safety of new drugs with different mechanisms of action, either as monotherapy or in combination with ruxolitinib.Therefore, the identification of patients who are more likely to respond poorly to ruxolitinib monotherapy or to discontinue early the drug is becoming crucial in guiding clinical practice.
This study confirms that prolonged ruxolitinib administration is associated with an improvement of the OS when compared to earlier discontinuation [6,27].While this outcome may be expected at the 5-year timepoint, interestingly, the survival advantage was observed also categorizing patients at earlier timepoints (namely, 1 and 3 years from ruxolitinib start).This advantage is unrelated to decreased leukemic progression, as the incidence of the blast phase was similar in patients who discontinued ruxolitinib and patients who are continuing it.This finding further suggests the absence of the significant disease-modifying activity of ruxolitinib monotherapy [28].
It is acknowledged that the outcome after ruxolitinib treatment is poor, regardless of the reason for discontinuation.Additionally, we have recently shown that the OS of patients who discontinued ruxolitinib in the chronic phase varied according to the type of salvage treatment after it, with patients receiving ruxolitinib rechallenge or novel agents having significantly improved survival compared to patients allocated to standard treatments (i.e., hydroxyurea, danazol, and splenectomy) [8,9].To date, data on the impact of second-generation JAK2 inhibitors on the survival of patients after ruxolitinib failure are scant.
The probability of remaining on ruxolitinib may be predicted by the STR-PM, which consists of several factors, including ruxolitinib starting dose ≥10 mg BID, baseline hemoglobin ≥10 g/dL, baseline platelet count ≥100 × 10 9 /L, diagnosis of post-Polycythemia Vera/Essential Thrombocythemia MF, and achievement of a spleen response at 3 months.
Notably, the presence of cytopenia is closely correlated with a higher probability of early ruxolitinib discontinuation and death in the STR-PM score.The association between the cytopenic phenotype and a worse outcome has been repeatedly observed and is likely to be related to multiple factors.Particularly, cytopenic MF is characterized by molecular alterations that have been found to correlate with worse outcomes, including low JAK2 variant allele frequencies, triple negativity, and the presence of high-risk subclonal mutations.In addition, cytopenic MF represents a great therapeutic challenge, with reduced options and a lower probability of clinical benefit that is related to the difficulty of administering adequate ruxolitinib doses and to a higher propensity to develop drug-related toxicities [17][18][19]29].
We recognize the limitations of this study, in particular its retrospective nature and the need to validate our findings in additional cohorts.In addition, immortalization bias may play a role when analyzing (conditional after 5 years of follow-up) survival in patients with longer and shorter duration of treatment.Finally, competing risks other than death, but probably leading to subsequent death (i.e., disease progression), may exist.Nonetheless, these results have been generated within a cohort of very well-characterized MF patients, followed in dedicated hematology centers, and data have been thoroughly verified after the first collection.
Additionally, these findings, although entirely innovative, are aligned with many previous clinical observations from real-world and prospective studies that have indicated cytopenia, low ruxolitinib dose, and absence of a response as markers for poorer outcomes of MF patients [3,4,8,30].

Conclusions
The STR-PM suggests, in particular, that the optimization of ruxolitinib dose right from the start of therapy and an early evaluation of the spleen response may be crucial to minimize the risk of early discontinuation and to promptly intercept patients that are more likely to fail ruxolitinib.Accordingly, low baseline ruxolitinib doses and the absence of a spleen response at 6 months are negative predictors of survival in the RR6 model [4].

Figure 2 .
Figure 2. Probability of early ruxolitinib discontinuation based on the STR-PM considering patients who discontinued at 5 years.

Figure 2 .
Figure 2. Probability of early ruxolitinib discontinuation based on the STR-PM considering patients who discontinued at 5 years.

Figure 3 .
Figure 3. Overall survival (OS) adjusted for delayed entry according to STR/LTR phenotype at 3 years (A) and 1 year (B).

13 Figure 3 .
Figure 3. Overall survival (OS) adjusted for delayed entry according to STR/LTR phenotype at 3 years (A) and 1 year (B).

Figure 4 .
Figure 4. Probability of early ruxolitinib discontinuation based on the STR-PM considering patients who discontinued at 3 years (A) and 1 year (B).

Figure 4 .
Figure 4. Probability of early ruxolitinib discontinuation based on the STR-PM considering patients who discontinued at 3 years (A) and 1 year (B).