We conducted a single-center retrospective study of engraftment dynamics in patients transplanted for primary or secondary MF. All patients undergoing allogeneic HCT at University Hospital Basel between 09/2000 and 09/2019 were screened. One single patient transplanted for MF with cord-blood as a stem cell source was excluded. Of 60 patients, 11 underwent a second HCT within the study period. Data was extracted from electronic patient files. The study was approved by the ethical committee Nordwest- und Zentralschweiz (EKNZ, no.2016-01930) and informed consent was obtained from all patients.

We evaluated determinants of time to engraftment for neutrophil, platelet and red cell lineages. Neutrophil engraftment was defined as first of 3 consecutive days with ≥0.5x10⁹/l neutrophils in peripheral blood, platelet engraftment as first of 3 consecutive days with ≥20x10⁹/l platelets in the absence of transfusions within the 7 preceding days, and red cell engraftment as first day with ≥30x10⁹/l reticulocytes. Conditioning was classified as reduced intensity (RIC) or myeloablative (MAC) conditioning according to Bacigalupo (20). Since thiotepa-busulfan-fludarabin (TBF) conditioning (thiotepa 10mg/kg, busulfan 6.4 mg/kg and fludarabine 150 mg/m²) was not included in this classification and consistently associated with haploidentical HCT at our center, it was evaluated separately. TBF was followed by haploidentical HCT, subsequent PTCy 50 mg/kg on days +3/+4 for GvHD prophylaxis and granulocyte colony-stimulating factor (G-CSF) support at 0.5 Mio units/kg from day+5 (TBF-PTCy/G-CSF) in this cohort. G-CSF support at 0.5 Mio units/kg/day was also applied to 5 patients with RIC and 1 patient with MAC regimens at variable time-points in the post-transplant course at the discretion of the treating physicians to promote reconstitution. Donor cell chimerism was assessed at 1, 3 and 6 months after HCT in peripheral blood and/or bone marrow. Poor graft function was evaluated as previously described (12, 21–24) with cytopenia in at least two hematopoietic lineages incl. neutrophils ≤ 1.5 G/l, platelets ≤ 30 G/l and/or Hb ≤ 85 g/l lasting for > 2 consecutive weeks following documented engraftment beyond d+14 in the presence of full donor chimerism and the absence of severe acute or chronic GvHD, relapse, drug-related or CMV reactivation-related myelosuppression. Patients’ disease risk category was determined according to respective prognostic scoring systems, primarily DIPSS plus (25), DIPSS (26), IPSS (27), and MYSEC (28) scores, as well as Cervantes (29) and MIPSS70 plus (30) scores in one patient each. Infections in the post-transplant period as well as thrombocyte concentrate (TC) and erythrocyte concentrate (EC) transfusion requirements were assessed in the first six months after allogeneic HCT.

Descriptive statistics were used for patient and transplantation characteristics. Time to engraftment, time to full donor cell chimerism and overall survival were estimated by Kaplan-Meier method and log-rank test used to assess statistical significance between groups. Estimates are given as median with 95% confidence interval (CI). Patients were censored at death, secondary graft failure or second transplantation. Univariate Cox regression was used to evaluate spleen size as a continuous variable. Key disease characteristics such as splenomegaly and BM fibrosis and factors significant in univariate analysis were implemented in multivariate forward-conditional Cox regression. Results are given as hazard ratio (HR) with 95% CI. Cumulative incidence function was used for the estimation of non-relapse mortality (NRM) as well as for acute and chronic GvHD with relapse calculated as the competing event. P-values of ≤0.05 were considered statistically significant. Data were analyzed and figures generated with SPSS 28.0.0 (IBM, Armonk, NY, USA), GraphPad Prism 9.2.0 (GraphPad Inc., San Diego, CA, USA) and NCSS 2020 package (NCSS, LLC, Kaysville, Utah, USA).

We retrospectively evaluated 71 allogeneic HCTs conducted in 60 MF patients in 2000–2019 with >50% performed after 2015 ( Table 1 , Supplementary Table 1 ). Median age at transplantation was 60 (31–72) years, increasing over time from 55.5 (31–61) to 58 (40–70), and 61 (34–72) years in the periods 2000–2009, 2010–2014, and 2015–2019, respectively, with a majority of patients being males. Median follow-up of survivors was 61 (24–198) months. Most patients (62.0%) suffered from primary MF (PMF), while 22.5% and 12.7% were transplanted for secondary MF evolving from polycythemia vera (PPV-MF) or essential thrombocythemia (PET-MF) and one patient each for unclassifiable myeloproliferative neoplasm (MPN-U) and MDS/MPN overlap syndrome ( Figure 1A ). JAK2 V617F, calreticulin (CALR), and thrombopoietin receptor (MPL) driver mutations were detected in 66.2%, 19.7%, and 2.8% of patients, respectively, while 4.2% were triple negative for either mutation ( Figure 1B ). A majority of patients had intermediate-2 or high-risk disease (46.7% and 30%, respectively) as assessed by prognostic risk scoring, similarly in patients with PMF and secondary MF including PPV-/PET-MF ( Supplementary Figure 1 ). BM fibrosis was advanced in most patients with 56.3% showing grade 3 fibrosis. Splenomegaly was prevalent with a median spleen size of 19 (11–32) cm at transplantation as assessed by sonography. Transplantations after previous splenectomy were uncommon with only 3 transplantations in 2 patients occurring in the setting of absent spleen. Ruxolitinib therapy before HCT was administered in a majority of 38/71 cases (53.5%) of the overall cohort with 30 patients receiving ruxolitinib up to transplantation and 8 patients with ruxolitinib discontinuation >1 month before HCT. Median time on ruxolitinib therapy before HCT was 10 months (range 1-78 months) and median spleen size at HCT did not significantly differ between ruxolitinib-treated and untreated patients (p=0.67). In our cohort, 7 patients received ruxolitinib after allogeneic HCT, mostly (5/7) after engraftment as a treatment of acute GvHD, while two patients with extensive splenomegaly were on ruxolitinib through the peri-transplant period. Myeloablative (MAC) or reduced-intensity (RIC) conditioning followed by cyclosporine and methotrexate or mycophenolate-mofetil with or without anti-thymocyte globuline (ATG) for GvHD prophylaxis was applied in 32.4% and 56.3% of allogeneic HCT, respectively. The novel preparatory regimen of haploidentical HCT with PTCy as GvHD prophylaxis, which beyond AML is also increasingly used in specific, rare entities such as in MF patients at our center, accounted for 15.6% of patients from 2015, but just 9.9% from 2000 ( Figure 1C ). In these patients, haploidentical HCT followed conditioning with thiotepa-busulfan-fludarabine, while subsequent PTCy on day +3/+4 was given for GvHD prophylaxis and granulocyte colony-stimulating factor (G-CSF) from day+5 for regeneration support (TBF-PTCy/G-CSF). Median number of infused CD34+ cells was 7.2 (1.94–18.8) x10⁶/kg body weight. Peripheral blood stem cells (PBSC) represented the prevalent CD34+ cell source (87.3%), while a minority of non-haploidentical (6.3%) and a majority of haploidentical HCTs (57%) used BM as source. Second allogeneic HCT was performed in 11 patients to manage graft failure, relapse or persistent disease.

Given that hematopoietic engraftment, which is known to be delayed in MF vs. AML (4, 10), impacts on the susceptibility for infectious and bleeding complications, we evaluated factors influencing engraftment in the neutrophil, megakaryocytic and erythroid lineages. Neutrophil engraftment occurred at a median of 20 (11–36) days post-transplant, while one patient experienced graft failure and one deceased before engraftment. Time to neutrophil engraftment was similar over the different time periods 2000-2009, 2010-2014 and 2015-2019 (p=0.374, Figure 2A ) and was analogous in first and second HCTs (p=0.72, Figures 3A, B ). Median time to platelet engraftment was 26 (0–121) days, including 6 patients maintaining platelets ≥20x10⁹/l throughout the peri-transplantation period. Platelet engraftment was faster in 2000–2009 vs. later periods (p<0.001), but remained stable 2010–2014 and 2015–2019 (p=0.931, Figure 2B ), and was analogous in first and second transplantations (p=0.535, Figures 3C, D ). Reticulocyte engraftment occurred at a median of 21 (12–236) days with no relevant difference among the three time periods (p=0.687, Figure 2C ) or between first and second transplantations (p=0.64, Figures 3E, F ). Donor cell chimerism 1, 3 and 6 months after HCT was similar in 2010-2014 and 2015-2019, while lower 2000-2009 at 1-3 months, and was analogous in first and second HCTs ( Supplementary Figures 2 , 3 ). Given that hematopoietic reconstitution times in all three lineages were analogous between first and second transplantations, as well as the limited number of second transplantations, first and second transplants were pooled for further analyses.

To explore potential determinants of engraftment, we evaluated MF disease characteristics including primary vs secondary MF, driver mutations in JAK2, CALR and MPL, prognostic risk group, splenomegaly and BM fibrosis as well as ruxolitinib therapy before HCT. Transplant-related factors with potential impact on engraftment dynamics were also assessed including the type of conditioning regimen, GvHD prophylaxis, type of donor, CD34+ cell number and source, administration of G-CSF support, and donor-recipient relation for blood group and CMV status. PMF vs secondary MF (median 20 vs 19 days; p=0.65), driver mutation status (p=0.57) as well as prognostic risk group (p=0.21) did not influence kinetics of neutrophil regeneration. In line with the notion that alterations of the microenvironment could interfere with engraftment (11), we observed that BM fibrosis significantly influenced neutrophil engraftment (median 15 vs 20 days, grade 1 vs 3, p=0.017; median 15 vs 19 days, grade 1 vs 2, p=0.054; Supplementary Table 2 ). In addition, splenomegaly impacted on engraftment of the neutrophil lineage as reflected by reconstitution times significantly relating to spleen size by univariate Cox-regression analysis (p=0.032, Supplementary Table 2 ). The shortened engraftment times observed in the setting of splenectomy should be interpreted with caution given that only a small minority of 3 transplantations in 2 MF patients of the entire cohort were performed after a splenectomy based on the rarity of this intervention. Ruxolitinib therapy before HCT did not significantly impact on neutrophil engraftment in our cohort (p=0.27). CD34+ cell dose significantly influenced neutrophil engraftment with a median of 18 days if >8x10⁶ cells/kg were infused versus 20 days with low cell numbers <6x10⁶ cells/kg (p=0.037) consistent with effects observed in similar cohorts (31, 32). We explored a potential impact of conditioning schemes used as preparatory regimens for allogeneic HCT in MF. We did not observe a difference in time to neutrophil engraftment for RIC vs MAC protocols (p=0.819). However, a potential effect of TBF conditioning, which was used exclusively for haploidentical HCT in our cohort, was detectable with TBF associated with later neutrophil engraftment (median 21 vs 19 days, p=0.017; Figure 4A ). This finding should be cautiously noted given the low absolute number of TBF-conditioned haploidentical HCT in this longitudinal cohort, and should promote evaluations in larger, multicentric cohorts of MF undergoing HCT. While ATG as GvHD prophylaxis related to shortened engraftment (p=0.028), PTCy which was exclusively used for TBF-conditioned haploidentical HCT in our cohort, associated with longer neutrophil reconstitution times. This was similarly reflected in HCT from haploidentical as compared to HLA matched donors (21 vs 20 days, p=0.019), while matched related and unrelated donors behaved analogously (p=0.176). BM as CD34+ cell source, which was used in the majority of haploidentical TBF-PTCy HCTs, and post-transplant G-CSF consistently used in TBF-PTCy HCTs in our cohort, analogously associated with later engraftment of neutrophils (median 26 vs 19 days, p=0.007 for BM vs. PBSC; median 21 vs 19 days, p=0.004 for post-transplant G-CSF; Figures 4B, C ). Given that TBF conditioned haploidentical transplants overlap with the use of BM as a stem cell source and administration of G-CSF support, these findings may reflect an overall different behavior of engraftment dynamics with this procedure.

Similar to the findings for neutrophil engraftment, reconstitution of platelets did not reveal differential effects in primary vs secondary MF, with JAK2, CALR and MPL driver mutations or different prognostic risk group, but splenomegaly significantly prolonged platelet engraftment (p=0.002, HR: 0.930, Supplementary Table 2 ). Consistent with an impact of splenomegaly, the few HCT after previous splenectomy associated with earlier platelet engraftment (p<0.001). Similar to the influence of BM fibrosis on neutrophil regeneration, fibrosis grade also showed a trend for later platelet engraftment (p=0.088). Ruxolitinib therapy before HCT did not show a significant effect (p=0.07). As for transplant-related factors, RIC and MAC regimens led to comparable engraftment dynamics (p=0.579), while later platelet engraftment was observed upon TBF conditioning with PTCy as GvHD prophylaxis (p=0.015; Figure 5A ) and administration of post-transplant G-CSF (p=0.001), while BM vs PBSC showed a similar trend (p=0.148; Figures 5B, C ). Higher CD34+ cell doses significantly correlated with faster engraftment (p=0.021, Supplementary Table 2 ) as described (11), while blood-group barrier and CMV serology status did not influence platelet reconstitution. Overall, red cell engraftment as reflected by reticulocytes ≥30x10⁹/l was less dependent on the evaluated determinants as compared to neutrophil and platelet lineages including prognostic risk groups and pre-transplant ruxolitinib treatment (p=0.97). However, red cell engraftment still related to splenomegaly, which prolonged engraftment time (p=0.008, HR: 0.942; Supplementary Table 2 ). In line with the notion that not only regeneration but also consumption by e.g. hemolysis might affect reconstitution of the red cell mass, we observed later reticulocyte engraftment upon major (p=0.032), minor (p=0.013) and bidirectional (p=0.047) blood group barriers.

Development of full donor cell chimerism established at a median of 35 days after HCT (range 18-156 days), did not show significant differences relating to CD34+ cell source and number, type of conditioning regimen, use of ATG or PTCY as well as G-CSF, donor/recipient constellation or pre-transplant ruxolitinib therapy ( Figure 6 ). Similarly, disease-specific factors including MF subtypes PMF and PET-/PPV-MF, driver mutation status or splenomegaly did not significantly impact on the dynamics of full chimerism development, whereas higher fibrosis grade associated with increased time to full chimerism (median 58 days, 33 days and 18 days for grade 3, grade 2 and grade 1 fibrosis, respectively, p<0.001, Figure 6 ).

To consolidate these findings, we performed multivariate analyses, including age at transplantation, gender, MF disease characteristics, and factors with significant effects in univariate analysis. Upon correction for splenectomy, fibrosis grade, conditioning, use of ATG, G-CSF, donor type, CD34+ cell dose and source, age and gender, an impact of splenectomy on neutrophil engraftment was confirmed, although only 3 transplantations in 2 patients were performed after previous splenectomy in our cohort (p<0.001, HR: 107.67). Given the rarity of splenectomy in our cohort (4.2%), conflicting data on its effects on outcome (33, 34), and decreasing relevance of this procedure, we omitted splenectomy as a factor from further multivariate analyses. Consequently, post-transplant G-CSF support (p=0.02, HR: 0.415) and BM as CD34+ cell source (p=0.039, HR: 0.421) were confirmed to associate with later neutrophil engraftment ( Supplementary Tables 3 , 4 ). However, this finding should be cautiously interpreted considering the substantial overlap of these factors with TBF-conditioned haploidentical HCT in our cohort and the limited number of patients, although substantial for a single center longitudinal cohort of higher risk, transplant-eligible MF.

Similarly, multivariate analysis for determinants of platelet engraftment confirmed the effects of post-transplant G-CSF administration (p<0.001, HR: 0.273), grade 3 vs grade 1 BM fibrosis (p=0.013, HR: 0.145) and splenectomy (p=0.011, HR: 7.044) as seen in univariate analysis. When splenectomy, which was only performed in 2 patients, was omitted from the model, post-transplant G-CSF (p=0.002, HR: 0.301) and grade 3 vs grade 1 BM fibrosis (p=0.026, HR: 0.179) maintained significant effects along with splenomegaly (p=0.021, HR: 0.94, Supplementary Tables 3 , 4 ) in line with similar cohorts (11, 32). For erythroid reconstitution, multivariate analyses confirmed the influence of a major blood group barrier (p=0.019, HR: 0.449) and splenomegaly (p=0.012, HR: 0.942) on reticulocyte engraftment dynamics. Factors associated with haploidentical TBF-conditioned HCT were not evident, confirming the finding from univariate analyses that red cell reconstitution is not as much impacted by novel preparatory regimens ( Supplementary Tables 3 , 4 ).

To explore potential long-term effects of delayed engraftment, we assessed overall survival (OS) in MF after first allogeneic HCT, which was 6.7 years (2460 days) in our cohort in line with previous studies (35–38). Patients with neutrophil engraftment later than the median showed a trend towards reduced OS without statistical significance (median OS 2.7 years, 95% CI 0-5.8 years, p=0.099), while platelet or red cell engraftment after the median did not impact on OS in our cohort. A potential impact of delayed engraftment on OS after second allogeneic HCT was not assessed given the limited number of patients. The cumulative incidence of non-relapse mortality (NRM) in our cohort at 5 years after first HCT was 19.3% and not significantly different in patients with delayed engraftment when assessed as engraftment time greater than the median of the cohort (95% CI 11.3-33.2%, p=0.84, Supplementary Figure 4 ). Cumulative incidence of relapse in the overall cohort was 26.9% at 5 years (95% CI 17.7-40.9%). Acute graft versus host disease (GvHD), as assessed for grades 2-4, occurred at a cumulative incidence of 33.1% at 180 days after HCT without a significant difference upon delayed engraftment (95% CI 23.7-46.2%, p=0.48, Supplementary Figure 5 ), while chronic GvHD was documented in 46.7% (95% CI: 35.7-61.3%) at 5 years after HCT and was similar in patients with delayed neutrophil engraftment (p=0.28, Supplementary Figure 6 ). Infections within 6 months post-transplant were assessed from 2010 onwards, as only a limited number of transplantations (n=6) were performed 2000-2009. While a majority of patients suffered from post-transplant infections in 2010-2014 and 2015-2019 (80.0% and 64.4%, respectively), patients with delayed neutrophil engraftment when assessed as engraftment time greater than the median of the cohort, showed similar rates of post-transplant infections as compared to patients with engraftment time shorter than the median in both time periods (75.0 vs. 83.3% and 56.5% vs. 70.0%, respectively). Within 6 months after HCT, a median of one infection per patient occurred with a range of 0-5 post-transplant infections per patient similarly in patients with delayed engraftment ( Supplementary Figure 7 ).

Notably, delayed engraftment impacted on transfusion requirements after allogeneic HCT. We observed that thrombocyte concentrate (TC) transfusion dependence at day 100 after HCT was > 5-fold more prevalent upon delayed platelet engraftment when assessed as engraftment time longer than the median of the cohort (20% vs. 3% of patients). While delayed neutrophil engraftment also related to somewhat higher TC transfusion dependence at day 100 (15.2% vs. 9.7%), delayed reticulocyte engraftment did not have an impact. For erythrocyte concentrate (EC) transfusion dependence at day 100 after HCT, we observed that it was not affected by delayed neutrophil engraftment (35.5% vs. 36.4%), but increased upon prolonged reticulocyte and platelet engraftment (50.0% vs. 21.2% EC transfusion dependent upon delayed red cell engraftment and 42.9% vs. 24.2% upon delayed platelet engraftment, respectively). Given these findings, we specifically assessed the total number of TC and EC transfusions until engraftment and found that the number of required TC transfusions was significantly higher upon delayed platelet engraftment (median 29 vs. 6 TC, p < 0.0001, Supplementary Figures 8A-C ). The number of required EC transfusions was significantly increased in settings of delayed reticulocyte (median 18 vs. 8 EC, p < 0.0001) and also delayed platelet engraftment (median 16.5 vs. 8 EC, p = 0.0006, Supplementary Figures 8D-F ). We confirmed these findings over a more extended time period of 6 months after HCT suggesting a relevance of engraftment dynamics for transfusion requirements for such a longer period. We found that the number of TC transfusions within 6 months after HCT was significantly increased in settings of delayed platelet (median 32 vs. 7 TC, p <0.0001) and also neutrophil engraftment (median 23.5 vs. 15 TC, p = 0.04, Supplementary Figures 9A-C ). Four patients did not engraft platelets and required 86-200 TC transfusions within 6 months after HCT (not shown). For EC, the number of required transfusions within 6 months after HCT was higher upon delayed reticulocyte (median 24 vs 11 EC, p < 0.0001) as well as platelet engraftment (median 23 vs. 10.5 EC, p < 0.0001), but not affected by neutrophil engraftment dynamics ( Supplementary Figures 9D-F ). Three patients did not engraft reticulocytes and required 24-58 EC transfusions within 6 months after HCT (not shown).

In addition, poor graft function as previously defined (12, 21–24) was observed in 8 patients within 6 months after HCT. Median onset was at 39 days (range 29-95 days) after HCT and lasted for a median duration of 58 days (range 21-203 days). In all but one patient all three hematopoietic lineages incl. neutrophils, platelets and reticulocytes were affected ( Supplementary Table 5 ). Median age of patients with poor graft function was similar to the entire cohort (64 years, range 46-70 years). A majority of 7/8 patients were transplanted for PMF (88%) and calreticulin mutations were overrepresented as driver mutations in 5/8 (63%) patients. Splenomegaly was observed in all patients with poor graft function at median spleen size of 18 cm (range 15-27 cm) at HCT similar to the overall cohort. Bone marrow fibrosis was pronounced with grade 3 fibrosis in most patients. Infused CD34+ cells were predominantly of peripheral origin and median dose was somewhat lower as compared to the entire cohort at 6.1 Mio/kg body weight. A major donor-recipient blood group barrier was present in 5/8 patients with poor graft function and thus more prevalent than in the overall cohort (63%).