Multiple sclerosis (MS) is a chronic immune-mediated disease of the central nervous system, which can cause severe physical and cognitive disability and inflammatory and neurodegenerative mechanisms are considered its 2 major pathological components.¹ The suppression of inflammatory disease activity with disease-modifying treatments (DMTs) represents the cornerstone of MS therapeutic management, although the sustained long-term disease remission remains often elusive,² and safety concerns in the context of long term drug exposure represents an important limitation. Autologous hematopoietic stem cell transplantation (AHSCT) has gained increasing interest as a treatment option, particularly for MS patients with limited response to DMTs. The procedure aims to remove all adaptive immunity, which in MS includes pathogenetic autoreactive lymphocytes, promoting a non-inflammatory immune reconstitution.³ The refinement of patients selection toward earlier stages of MS, the evolution of AHSCT protocols, and clinical experience have contributed to improve its efficacy, to substantially reduce treatment-related mortality and to improve management of adverse events.⁴ The European Bone Marrow Transplantation Society guidelines are now recommending AHSCT as an option for treating patients with aggressive forms of MS, even without prior exposure to DMTs.⁵

After AHSCT, in most MS patients, remission of inflammatory parameters (relapses and magnetic resonance imaging [MRI] activity), prevention of disability worsening, and long-term disease stabilization have been reported.^6-13 Although AHSCT is believed to exert higher efficacy than standard treatments,⁸ real-world evidence of its effectiveness, compared with highly active treatments, remains scarce. Previous observational studies were carried out on small cohorts enriched with patients presenting with severe disease features.^{11, 14, 15} In addition, 2 randomized controlled trials showed superiority of AHSCT, but their small size and the choice of DMT comparators represent important limitations. One study compared AHSCT against mitoxantrone,¹⁶ which is no longer in use because of its toxicity. The other trial compared AHSCT with a treatment arm of heterogenous DMTs, including highly efficacious and first-line treatments.¹⁷ A recent analysis of the MSBase registry demonstrated better outcomes after AHSCT compared with fingolimod and natalizumab treatment, but similar clinical outcomes when compared with patients receiving ocrelizumab (OCR) infusions. Limitations of the study included different AHSCT protocols across centers, the short follow up of the OCR cohort, and the lack of MRI data.¹⁸ We previously reported the efficacy and safety outcomes in a cohort of MS patients, who were selected for AHSCT under homogenous criteria, and had their procedure and were followed up at 2 collaborating London centers.¹⁹ The dataset was recently expanded, and by using the overlap weighting propensity score method (PSOW),²⁰ we set out to perform comparative analyses among patients receiving AHSCT in the real world, and 2 highly active DMTs, alemtuzumab (ATZ)²¹ and OCR.²²

Methods

Data were retrospectively collected from a consecutive series of patients, with a diagnosis of relapsing–remitting (RR) MS, who were followed up at the MS tertiary center of Imperial College NHS Trust (Charing Cross Hospital, London, UK), where they were routinely clinically evaluated by 5 neurologists expert in MS, and where they had regular MRI assessments. Patients included in the analysis received at least 2 courses of ATZ (cycle 1: 12 mg i.v. daily for 5 days, cycle 2: 12 mg for 3 days) from July 2010 to November 2021 or were treated with OCR (2 300-mg i.v. doses initially administered 14 days apart, followed by 600 mg every 6 months) from January 2017 to November 2021. The AHSCT was carried out initially at the hematology center of King's College Hospital between February 2012 and November 2021, and subsequently also of Hammersmith Hospital from April 2016 to January 2020, both in London (UK). Once approved for AHSCT, patients were assigned to 1 of the 2 hematology centers based on the residence address, referring clinicians, and waiting times. Ethical approval from research committee and consent from patients was not required, as data were collected anonymized. The dataset was locked for analysis in November 2021.

The indication for AHSCT was initially (2012–2015) based on the agreement of at least 2 neurologists and 1 hematologist with expertise in AHSCT that the treatment was in the patient's best interest. In September 2015, AHSCT eligibility criteria (Table 1) were formally defined to select patients with clinically and radiologically active forms of RRMS, refractory to standard DMT, and each case's eligibility had to be reviewed and approved by the Pan-London multidisciplinary team panel of neurologists and hematologists. The multidisciplinary team panel assessed the eligibility based on clinical information collected on referral forms and on MRI results, which had previously been reviewed with the neuroradiologist. For some cases, who did not strictly meet inclusion criteria, but convincingly fulfilled the overall eligibility profile, AHSCT was offered, documenting the specific basis for the approval.

Results

Data on 1,035 RRMS patients, who were treated with AHSCT, ATZ, and OCR, were available, and 621 (AHSCT n = 103; ATZ n = 204; OCR n = 314) patients met the inclusion criteria (Fig 1). The cohort baseline demographics, and clinical and radiological characteristics before and after PSOW are shown in Tables 2 and 3. Before PSOW, at treatment initiation, the patients in the AHSCT group were older and had longer disease duration compared with the ATZ group, and they had more relapses and MRI activity in the preceding 2 years, and higher disability score, compared with both the ATZ and OCR groups. The median follow up was 43 months (interquartile range [IQR] 29–53), 43 months (IQR 35–58), and 32 months (IQR: 24–36) in the AHSCT, ATZ, and OCR cohorts, respectively. The clinical and radiological outcomes during the observation period of the 3 cohorts without PSOW are summarized in Table S1; 9.7%, 25.5%, and 15.7% had new relapses, 9.7%, 27.5%, and 4.8% had new MRI activity, and 31.1%, 28.9%, and 17.1% experienced EDSS progression, in the AHSCT, ATZ, and OCR group, respectively. Clinical and radiological features of patients experiencing relapses/MRI activity and EDSS progression while on treatment are outlined in Tables S2 and S3.

Comparison of HSCT versus ATZ

Table 2 shows the clinical and radiological baseline features in pairwise AHSCT-ATZ cohorts, after PSOW. The ARR over 5 years was significantly lower among patients who received AHSCT compared with those who were treated with ATZ (ARR mean: 0.020 vs 0.078; ARR ratio: 0.26, 95% CI 0.12–0.57; p < 0.001; Fig 2A). In addition, being treated with AHSCT compared with ATZ was associated with a significantly lower weighted cumulative incidence of clinical relapses (HR 0.24, 95% CI 0.10–0.55; p = 0.001; Fig 2B). After 3 years, the weighted cumulative KM estimate of remaining free of relapses was 95% (95% CI 87–99) versus 77% (95% CI 71–82), whereas at 5 years, it was 89% (95% CI 75–96) versus 63% (95% CI 54–71), in the AHSCT and ATZ groups, respectively. The E-value RR for unmeasured confounding was 7.2 for the ARR ratio (2.9 for its CI lower limit) and 7.8 (3.0 for its CI lower limit) for relapse-free survival.

In line with the relapse analysis, the use of AHSCT compared with ATZ was associated with a significantly lower weighted risk of new MRI activity (HR 0.34, 95% CI 0.15–0.79; p = 0.012). The weighted cumulative KM estimate of remaining free of new MRI activity after 3 years of follow up was 92% (95% CI 86–98) in the AHSCT group and 74% (95% CI 69–80) in the ATZ group, and after 5 years this dropped to 85% and 71%, respectively (Fig 2C). The E-value for unmeasured confounding was 5.3 (1.8 for its CI lower limit).

During the observation period, in the 2 weighted groups, patients had similar risk of experiencing confirmed disability accumulation (HR 0.92, 95% CI 0.50–1.71; p = 0.80; Fig 2D). Among AHSCT-treated patients compared with ATZ, the weighted cumulative KM estimate of EDSS progression after 3 years was 22% (95% CI 13–31) and 23% (95% CI 17–29), whereas after 5 years it was 38% (95% CI 24–54) and 30% (95% CI 22–40), respectively. At the last follow up, the mean EDSS change from baseline was +0.068 (SD 1.10) for AHSCT patients, and +0.56 (SD 1.13) for those who were treated with ATZ, whereas the median EDSS change was of 0 (IQR 0–0.5) and 0 (IQR 0–1), respectively.

Treatment with AHSCT was associated with a significantly higher weighted probability of experiencing EDSS improvement (HR 2.68, 95% CI 1.02–7.00; p = 0.045; Fig 2E), which occurred in a larger proportion of AHSCT-treated patients (19%) compared with ATZ-treated patients (4%). The weighted cumulative KM estimate of disability improvement after 3 years, was 16% (95% CI 7–26) versus 4% (95% CI 2–6), and after 5 years it was 21% (95% CI 13–32) versus 12 (95% CI 8–21) in the AHSCT and ATZ groups, respectively.

Overall, a significantly smaller proportion of patients who received HSCT compared with ATZ (36.9% vs 52.0%) experienced NEDA status failure (HR 0.49, 95% CI 0.31–0.80; p = 0.004; Fig 2F). The weighted cumulative KM estimate of maintaining NEDA status at 3 years was 79% (95% CI 70–88) and 53% (95% CI 46–60), whereas at 5 years was 55% (95% CI 40–68) versus 33% (95% CI 24–42), in the AHSCT and ATZ groups, respectively. The E-value RR for the unmeasured confounder of the NEDA estimate and its CI lower limit were 3.5 and 1.8, respectively.

Comparison of AHSCT versus OCR

Table 3 shows clinical and radiological baseline features in pairwise AHSCT and OCR cohorts after PSOW. In the OCR group, patients received a mean of 5 (SD 1.5) infusions and were followed up for a shorter time (median 34 vs 43 months) compared with the AHSCT group. The use of AHSCT compared with OCR was associated with a significantly lower mean ARR (mean 0.026 vs 0.067; ARR ratio 0.38, 95% CI 0.17–0.87; p = 0.021; Fig 3A), and with a lower, albeit not significantly, weighted risk of clinical relapse (HR 0.40, 95% CI 0.15–1.05; p = 0.063; Fig 3B). After 3 years from baseline, the weighted cumulative KM estimate of remaining free of relapse was 91% (95% CI 83–95) and 81% (95% CI 72–87) in the HSCT and in the OCR groups, respectively. The E-value RR for unmeasured confounding was 4.7 for the ARR ratio and 1.6 for its CI lower limit.

During the mean 3 years of follow up, the proportion of patients without new MRI activity was very high in both groups. After 3 years from therapy initiation, the weighted cumulative KM estimate of remaining free of new MRI activity was 93% (95% CI 88–98) versus 93% (95% CI 91–96) in the AHSCT and OCR groups, respectively (Fig 3C). Patients in the AHSCT and OCR groups had similar risk of experiencing new MRI activity (HR 1·03, 95% CI 0.32–3.25; p = 0.96).

The proportion of patients experiencing disability accumulation and disability improvement was also similar in the AHSCT and OCR groups. The KM analysis showed a similar weighted cumulative estimate of 6-month confirmed EDSS worsening (at 3 years: 24%, 95% CI 17–35 vs. 19%, 95% CI 15–25; p = 0.86; Fig 3D) and of 6-month EDSS improvement (at 3 years: 13%, 95% CI 7–23 vs. 13%, 95% CI 7–17; p = 0.77) (Fig 3E) in the AHSCT and OCR groups, respectively. At the last follow up, the mean EDSS change from baseline was 0.068 (SD 1.10) for AHSCT patients and 0.12 (SD 0.78) for those receiving OCR, whereas the median EDSS change was 0 (IQR 0–0.5) and 0 (IQR 0–0.5), respectively.

Finally, patients treated with AHSCT had a lower, but not statistically significant, weighted risk of NEDA-3 status failure compared with the OCR group (HR 0.62, 95% CI 0.35–1.09; p = 0.095; Fig 2F). At 3 years from baseline, the estimated proportion of patients experiencing NEDA was 73% (95% CI 63–81) and 65% (95% CI 58–70) in the AHSCT and OCR groups, respectively.

Discussion

Case series, observational studies, and trials have demonstrated that AHSCT can effectively halt clinical and subclinical inflammatory disease activity in MS patients not responding to conventional DMTs.²⁸ Its use has been recommended by the American Society for Blood and Marrow Transplantation as “standard of care,” even for untreated patients with negative prognostic features and at high risk of developing severe disability.²⁹ Advances in patients selection, treatment procedures, and management of risks and adverse events have improved the transplant safety profile,⁶ imposing a reassessment of the AHSCT position in the MS treatments algorithm. In 2 previous trials, the ASTIMS and MIST, AHSCT was shown to be more effective in suppressing the disease inflammatory parameters compared with mitoxantrone,¹⁶ which is no longer used in MS management due to its toxicity, or compared with a heterogenous group of MS patients treated with a wide range of DMTs, among which only 54% received high-efficacy medications.¹⁷ Overall, clear indications as to whether and when clinicians should favor AHSCT as an alternative to highly active DMTs remain unclear, and only clinical trials can definitively answer the question.

In the present retrospective real-world study, we implemented PSOW, which is a robust method for controlling confounders in observational studies, ensuring homogeneity across treatment groups. We demonstrated, among patients with relapsing forms of MS, superior efficacy of AHSCT in preventing clinical relapses compared with 2 high-efficacy DMTs, ATZ and OCR. During the 5-year follow-up period, patients who received AHSCT, compared with those who were treated with ATZ, had significantly lower mean ARR (0.020 vs 0.078), and in significantly larger proportions remained free of new relapses (89% vs 63%) and of new MRI activity (85% vs 71%). The difference between the 2 groups became mainly evident after the first 2 years from baseline, when in the ATZ group, following the completion of the second course of treatment, an increasingly larger number of patients started to experience new clinical and subclinical inflammatory activity. Similarly, during the shorter follow-up period (3 mean years), in the AHSCT group, compared with the OCR group, we observed significantly lower mean ARR (0·026 vs 0·067), and, albeit not statistically significant, a larger proportion of patients free of relapses (91% vs 81%), suggesting a trend to reduced probability of acute attacks. Both the AHSCT and the OCR groups performed similarly in preventing new MRI activity. The robustness of the superior effect of AHSCT on ARR is confirmed by the high E-value RR for unmeasured confounding in both the ATZ (7.2) and OCR (4.7) analyses, indicating that a potential unknown confounder, to nullify results, should have a strong association with both the treatment and the outcome, which is an unlikely scenario.

Reassuringly, sensitivity analyses of unweighted data (Figs. S1A–E and S2A–E) and of weighed data estimated from generalized boosted model-based propensity scores (Tables S4 and S5) were in complete agreement with our main analysis results, indicating little bias affecting our data. Notably, these also showed a statistically significant greater effect of AHSCT compared with OCR, not only on the ARR, but also on the risk of clinical relapses (Table S5), which further confirms the superior efficacy of the transplant in preventing mechanisms determining the occurrence of inflammatory attacks.

In contrast to the analyses of inflammatory (clinical relapses and MRI activity) outcomes, AHSCT was not associated with better outcomes on disability accumulation. Confirmed EDSS worsening occurred in similar proportions in the AHSCT and in the 2 DMTs comparators groups, indicating that, among patients who received the transplant, the more profound relapse suppression compared with ATZ and OCR did not significantly impact on the accumulation of permanent disability. The therapeutic effect of AHSCT might be exclusively related to the prevention of mechanisms underlying the relapses-associated worsening, with little influence on neurodegenerative processes accounting for the progression independent of relapsing activity, which gradually becomes more prominent late in the disease course.^{30, 31} However, due to referral practice and treatment selection criteria, most of the patients who received AHSCT had already failed responding to highly active treatments and, thus, had already accumulated severe disability (mean EDSS 5) at baseline. This implies that, although patients were still experiencing predominantly relapsing symptoms, they may have entered a subtle progressive course, albeit not fulfilling the formal definition of secondary progressive MS, and they were at high risk of disability progression unrelated to relapses, which is more likely to occur in the more advanced disease stage. Therefore, selection bias might have hindered any effect of AHSCT in preventing disability worsening. In earlier studies, younger age, shorter disease duration, and mild to moderate disability were shown to be features distinguishing patient groups more likely to benefit from AHSCT.⁶ Whether the transplant compared with highly effective DMTs can more effectively prevent the long-term disability accumulation when administered early in the disease course and at younger age remains an open question. Presently, randomized controlled trials (RCTs) are ongoing (BEAT-MS, NCT04047628;STAR-MS, EudraCT: 2019-001549-42; RAM-MS NCT03477500) comparing AHSCT with ATZ, OCR, natalizumab, cladribine, and rituximab, which will provide higher-grade RCT evidence.

Interestingly, we observed that the use of AHSCT was associated with a significantly higher probability of experiencing EDSS improvement when compared with ATZ, although such difference was not observed in the comparison with OCR. This is in line with previous studies showing that AHSCT-treated patients are more likely to experience disability improvement when compared with other DMTs, including ATZ.^32-34 Mechanisms underlying tissue repair remain largely unclear, but it reasonable to assume that the suppression of chronic inflammatory activity, which is induced by central nervous system penetrant chemotherapy used during the AHSCT conditioning regime, has the potential to promote remyelination and neural plasticity.

The superior efficacy of AHSCT compared with ATZ is consistent with results from several previous studies analyzing smaller cohorts with shorter follow up.^32-34 In contrast, a recent analysis from the MSBase provided comparative evidence of AHSCT versus OCR, showing similar efficacy of the 2 compounds in preventing new relapses and disability worsening.¹⁸ However, in this study, a large proportion of OCR-treated patients (72%) dropped out of the analysis after 2 years from baseline because of short follow up. In addition, the lack of information on pre-baseline therapy in a large number of patients (58.2% of AHSCT, and 57.5% of OCR cohorts), the considerable heterogeneity of AHSCT protocol (67% of the transplanted patients received intermediate-intensity conditioning regimens, and in 16% of the patients, ATG was not used), and the lack of MRI data among propensity score matching criteria, and as an endpoint, represent important limitations, which might explain differences compared with the present study results.

We acknowledge that, in our analysis, the AHSCT group's observed superiority in suppressing inflammatory disease activity, compared with OCR, was less prominent than in comparison with ATZ. The ARR was significantly lower among those who had AHSCT compared with OCR, but the difference between the 2 groups in the cumulative proportion of patients free of relapses (91.3% vs 80.7%) did not reach statistical significance, although in sensitivity analyses, the transplant was shown to exert a statistically significant greater effect also on the time to new relapses (Table S5). In addition, when assessing the MRI outcomes, we observed a “flooring effect,” with only 6.7% of patients experiencing MRI activity in both the AHSCT and OCR groups. This is consistent with a strong efficacy of both treatments on MRI lesions development, and it is further explained by the relatively short follow-up period in the OCR cohort, as most of the patients in our analysis were censored 3 years after therapy initiation. With extended follow up, it is reasonable to expect that the AHSCT superiority, compared with OCR, in preventing focal inflammatory activity could emerge more clearly, but further confirmatory studies with longer observation period are needed.

Different pharmacokinetic properties must also be taken into consideration, as AHSCT is a one-off procedure able to induce immune changes sustained over time, while OCR requires regular treatment administrations to exert its therapeutic effect. Although formal cost-effectiveness analyses are lacking, the cost of a one-off therapy with comparable effectiveness may be lower than long-term treatment with B-cell depleting licensed agents.³⁵

In addition, over the past years there has been a substantial improvement of AHSCT-related mortality and management of adverse events.⁴ Among our AHSCT-treated patients, the incidence of immediate and late adverse events was not in excess of the rates reported in literature, including the observed (1) death, indicating a treatment-related mortality of 1.1%, which is within the range of previous reports. Indeed, in a recent meta-analysis including 4,831 patients from 50 studies, the pooled treatment-related mortality was 4.0% (95% CI 2–6%), a number influenced by the high mortality reported in earlier studies.³⁶ In contrast, safety concerns have been raised related to the long-term depletion of B cells, and consequent hypogammaglobulinemia and neutropenia, with associated risk of impaired response to vaccines,³⁷ and of serious infections and neoplasms, especially among older patients,^{38, 39} which could be potentially mitigated by improving the selection of patients and by adopting an extended dosing schedule.⁴⁰ However, factors accounting for B-cell suppression related complications remain largely unknown, hindering the possibility of propensity matching patients based on the risk of side effects.

Arguably, the different degree of effectiveness of AHSCT compared with ATZ and OCR might be explained by the markedly different mechanisms of action of the 3 compounds. Patients undergoing AHSCT following the lympho/myeloablation experience an immune reset through the generation of new self-tolerant T and B precursors from the reinfused self-hematopoietic stem cells,⁶ whereas after administration of ATZ, complete T-cell recovery tends to occur within 3–4 years, and B-cell recovery within the first year, without renewing the adaptive immune compartment.⁴¹ The OCR depletes B cells and 20% of circulating T cells expressing CD20 within 2 weeks from the first course,⁴² and data on immune reconstitution after rituximab suspension showed that the restored B-cell compartment is mainly composed of naïve B cells, the levels of pro-inflammatory cytokine is decreased, whereas anti-inflammatory cytokines are increased.^{43, 44} The effect of AHSCT on the B-cell compartment is largely unknown, and further studies are required to understand if its mechanism of action could, wholly or in part, depend on depletion and reconstitution of the B-cell compartment.

Notably, AHSCT induced sustained disease stability in the majority of our cohort's patients, with 63.1% maintaining NEDA-3 status and 93.3% not requiring further treatment within 5 years from the transplant (Table S1). Among patients who experienced disease activity after AHSCT, 7 subsequently received other DMTs, including natalizumab (n = 3), cyclophosphamide (n = 2), ocrelizumab (n = 1), and a second AHSCT (n = 1). In addition, 5 years after AHSCT, 85% of patients were still free of inflammatory disease activity (Table S1). Although it is reasonable to expect a few more “non-responders” emerging in the longer term, it is unlikely that this will significantly impact the overall cost–benefit of the transplant.

In our cohort, all but 1 patient received cyclo-ATG conditioning and 1 (the first patient) BEAM-ATG. Previous European Bone Marrow Transplantation Society studies reported no significant differences in the efficacy or safety of cyclo-ATG versus BEAM-ATG used for autografting in MS.^{15, 45} Although different conditioning intensities may reveal differential efficacy or safety in other study settings (e.g. RCTs), the current evidence does not support major differences to be expected between lymphoablative (cyclo-ATG) versus more intensive, partly myeloablative conditioning (BEAM-ATG). It is possible that high-intensity, fully myeloablative regimes, such as busulfan–cyclophosphamide, could be more effective, but their higher risk profile has so far limited their utilization to 1 experienced center in Canada only.⁴⁶

The retrospective design, lack of true randomization, and varied follow-up durations among treatment groups may restrict the generalizability of our results. We have attempted to reduce treatment–indication bias by including the most relevant variables in the PSOW calculation, but we cannot entirely rule out the possibility that other unmeasured prognostic factors and sources of bias may have affected the results. In addition, although PSOW is considered a robust method for controlling confounders in observational studies, we acknowledge the limitation of our statistical approach, which implied analyzing modelled data rather than directly comparing the crude real-world data. However, results from analyses of unweighted crude data (Figs. S1A–E and S2A–E) were reassuringly in agreement with our main analyses of weighted data. Although some analyses may have had low statistical power mainly due to the relatively small sample size of the AHSCT group compared with the majority of previously published studies, we are here reporting data from 1 of the largest cohorts of AHSCT treated patients.^{11, 14-16, 18, 32} In addition, unlike many other studies, in our cohort, AHSCT was prescribed to patients who fulfilled predefined inclusion criteria, which were evaluated by a multidisciplinary team panel for each single case, ensuring homogeneity of access to the transplant. Further strengths of our study include the uniform AHSCT protocol administered at 2 European Bone Marrow Transplantation Society accredited centers, the inclusion of MRI data at baseline and on disease monitoring, the clinical assessments performed by 5 experienced MS physicians at a single MS tertiary center, covering a well ascertained geographic area (southwest London), which reduces the possibility of ascertainment bias and of interrater variability of EDSS scores. Importantly, in our analysis of non-responders (Tables S2 and S3), a larger proportion in the AHSCT group compared with the ATZ and OCR groups previously had disease activity while being on highly active DMTs (66.7%, 45.7%, and 41.3%, respectively), indicating a more aggressive disease course and, therefore, higher risk of experiencing further clinical and radiological breakthrough. Overall, this might have downsized the AHSCT suppressive effect on relapses and MRI activity.

The results allow us to draw conclusions separately about the effectiveness of AHSCT against 2 high-efficacy DMTs in a typical clinical scenario of patients experiencing the relapsing form of MS with highly active disease previously not adequately controlled by treatments. Our data showing that the superior efficacy of AHSCT suggested by a number of indirect comparisons is not clear-cut, and corroborates the rationale for RCTs to confirm their equipoise. The RCTs are expected to provide the definitive, high-quality evidence to support clinical decision-making for treatment of individuals with highly active MS.

Author Contributions

P.A.M., A.Z., R.N., and A.Sc. contributed to the conception and design of the study. All authors contributed to acquisition and analysis of data. P.A.M., A.Z., A.Si., R.N., and A.Sc. contributed to drafting the text or preparing the figures.

Potential Conflicts of Interest

Nothing to report.