Early‐onset severe infections in allogeneic hematopoietic stem cell transplantation recipients with graft failure

Graft failure (GF) is a rare but severe event after allogeneic hematopoietic stem cell transplantation (HSCT), leading to increased mortality through infections, disease relapse and drawbacks of marrow aplasia.1 Retrospective studies have estimated the overall incidence of GF around 5.5%.1, 2 Hematological risk factors such as non-malignant underlying diseases, partial remission at transplant, HLA mismatch, use of a cord blood source and low graft cellularity are consistently associated with GF.1-3 In most studies, infections are cited as circumstantial events but to date, no study has specifically documented early-onset severe infections (ESIs) associated with GF. The present study focuses on the incidence, the chronology, the type and the outcome of ESIs during the expected engraftment period in the setting of GF after allogeneic HSCT. We conducted a retrospective, observational, multicentric matched case-control (1:2) study among adult allogeneic HSCT recipients transplanted at three French tertiary-care Hematology departments (Saint-Louis, Paris; Lyon-Sud, Lyon; Oncopôle, Toulouse) between January 2008 and December 2017. GF recipients were identified by cross-referencing the databases of the Hematology departments and by additional chart reviews. Eligible GF recipients matched the following criteria1-3: (i) primary GF was failure to achieve donor-derived ANC ≥0.5 G/L for more than three consecutive days by day 42 post-HSCT, including neutropenic GF with persisting aplasia and non-neutropenic GF with autologous recovery, without evidence of disease relapse; (ii) early secondary GF was the loss by day 42 post-HSCT of a previously functioning graft (donor-derived sustained ANC ≥0.5 G/L for more than 3 days) associated with loss of full donor chimerism without evidence of disease relapse. Each case was matched with two controls according to stem cell source, underlying hematological disease, temporal proximity of HSCT (±5 years), age (±10 years) and gender. Death before day 20 after HSCT was an exclusion criterion. Ethics Committees of each hospital approved the study. ESIs were defined as life-threatening fungal, viral, parasitic or bacterial infection occurring upon conditioning (day 7 before HSCT) until day 42 post-HSCT using the most recent consensus definitions from international groups provided in Supporting Information Table S1. Cumulative incidence and survivals were calculated using the Fine and Gray competing risk regression model. Competing events were death from all causes for the cumulative incidence of ESIs, and infection-free death for the cumulative survival rate of infection-related death. Analyses were based on two-sided P-values, with statistical significance defined by P < 0.05 and conducted with R software version 3.4.3. Over the study period, 2094 allogeneic HSCT were performed. Forty-nine GF were ultimately selected, including 45 (91.8%) primary GF and 4 (8.2%) early secondary GF. Full baseline characteristics are provided in Supporting Information Table S2. In univariate analysis, ESIs were strongly associated with GF (OR 11.04; 95%CI [3.86-31.61]; P < 0.0001). Infections associated with ESIs were toxoplasmosis (OR 29.44; 95%CI [1.29-671.65]; P = 0.034), prolonged undocumented sepsis-like syndrome (OR 24.35; 95%CI [1-592.07]; P = 0.050), invasive fungal infection (IFI) (OR 11.13; 95%CI [2.49-49.72]; P = 0.002), bacterial blood stream infections (BSIs) (OR 8.29; 95%CI [1.78-38.69]; P = 0.007) and viral infections (OR 2.84; 95%CI [1.28-6.27]; P = 0.010) with the subset including BK virus, adenovirus and influenza A infections significantly associated with GF (OR 11.02; 95%CI [1.251-97.16]; P = 0.031). Significantly higher frequency of CMV serodiscordance (both positive recipient (R+) with negative donor (D-) and R−/D+) was observed among cases (OR 2.17; 95% CI [1.06-4.42]; P = 0.033) with a prominence of R+/D- (31.3%) vs R−/D+ (18.8%). Median delay to first ESI episode was 12 (IQR, 7-22) vs 19.5 (IQR, 8.8-26.8) days for cases and controls, respectively (P = 0.201). The delay according to the graft source was shorter in cases than in controls (CB, 9.5 [IQR, 7.3-17.3] vs 14 [IQR, 10-20.3], PB 9.5 [IQR, 2.5-20.5] vs 25 [IQR, 9.3-33], BM 20 [IQR, 13-25] vs 29 [IQR, 19-32] days, respectively). The median time to ESI diagnosis was 8 [7-11], 8.5 [5-19] and 11.5 [7.75-22.75] days for bacterial BSI, IFI and viral infections, respectively, and later for sepsis-like syndrome and toxoplasmosis at 20 [8-25] and 20 [10.5-28] days, respectively. In multivariate analysis, ESIs (OR 14.35; 95% CI [3.58-57.49]; P < 0.0001), total nucleated cells ≤2.5 (OR 4.23; 95% CI [1.21-14.74]; P = 0.024) and CMV serodiscordance (OR 2.83; 95%CI [1.08-7.41]; P = 0.035) were independently associated with GF. The cumulative incidence of ESIs upon conditioning until day 42 post-HSCT was significantly higher for cases vs controls (75.5%; 95%CI [64.4-88.6] vs 26.5%; 95%CI [19.1-36.9]; Gray test: P < 0.0001) (Figure 1A). When addressing infection-related mortality, the 5-year cumulative survival rate was higher in controls than in cases (75.2% [64.7-82.7] vs 47.3% [27.2-61.9] respectively; Gray test: P = 0.002) (Figure 1B). Estimated 5-year overall survival was significantly lower in cases (19.7%, 95% CI [5.2-34.2]) than in controls (52.9%, 95% CI [42.4-63.4]) (HR 2.475; 95% CI [1.48-4.137]; P < 0.0001) (Figure 1C). When addressing GF rescued by a second transplantation (n = 24), haplo-identical HSCT resulted in a better outcome (P = 0.023 for unrelated donor and P = 0.0035 for CB) (Figure 1D). These results show that ESIs are frequently associated with GF in adult HSCT recipients. Association with toxoplasmosis has previously been reported in high seroprevalence areas.4 A potential cofounder is trimethoprim/sulfamethoxazole treatment, with regards to its myelosuppressive effect, but the rather late delay for treatment initiation likely pleads for a direct effect of T. gondii on engraftment. Consistently, a study has shown that T. gondii is able to induce apoptosis of human umbilical cord mesenchymal stem cells.5 So far, no direct link has ever been evidenced between GF and bacterial BSI or IFI. Innate and adaptive immune impact of sepsis could impair engraftment in different ways including induced immune cells apoptosis or T-cell exhaustion and shift to a TH2 phenotype.6 Most GF studies consistently acknowledge the primordial role of viral infections in GF, notably human herpes virus 6 in double cord blood HSCT.1-3 In cases, although the most frequent ESIs were of viral origin, the odd of GF occurrence upon viral infections was the lowest among pathogenic agents. The relative high frequency of viral reactivations, together with monitoring of viral load guiding pre-emptive treatment options may explain this result. CMV serodiscordance remained independently associated with GF in our multivariate model. Ljungman et al. have already reported a negative impact of CMV serodiscordance on overall survival, in case of unrelated donor (for R−/D+) or myeloablative conditioning (for R+/D-).7 The benefit of early prophylactic strategies with well-tolerated anti-CMV drugs may further reduce the risk of such a complication. Whether ESIs occurring in the window of engraftment are causes or consequences of GF is a matter of debate. Literature-based mean time to engraftment is 16, 20 and 25 days for PB, BM and CB grafts, respectively. Based on the temporality of ESIs reported here, together with pathophysiological arguments, we suggest that ESIs may contribute to graft rejection. A second rescue graft is considered the treatment of choice for primary GF and persisting aplasia, with a 5-year overall survival ranging from 11% to 50%.1-3 In our study, haplo-identical grafts generated the best outcome, although caution is advised given the low number of patients. To conclude, ESIs are strongly associated with GF. More efficient prevention and treatment strategies are key issues during the critical window of engraftment after HSCT. Further studies should address associations between hematopoietic stem and progenitor cell engraftment and infections, particularly the mechanisms involved in GF or poor graft function. The authors would like to thank Fabien Subtil for his help for the biostatistical analysis, Dr. Franck-Emmanuel Nicolini and Pr. Mauricette Michallet. The authors gratefully acknowledge Myriam Renault, Sylvie Lengay and Elodie Colonnese. The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. VA and AC contributed to conception and design of the study, acquisition of the data, interpretation of the data, drafted the manuscript and approved the final version; FV participated in the design of the study and performed statistical analysis, revision of the article for important intellectual content, and approved the final version; RPL, AH, EB, GS, HLW contributed to acquisition of the data, interpretation of the results, revision of the article for important intellectual content, and all approved the final version; FA conceived and designed the study, participated in analyses and interpretation of the data, drafted the manuscript and approved the final version. All authors have read and approved the final manuscript. 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.