Lipid levels increase to the normal range after nonmyeloablative hematopoietic cell transplantation for sickle cell disease

Abstract

Background:

Individuals with sickle cell disease (SCD) have a unique type of dyslipidemia characterized by low total cholesterol (TC), low low-density lipoprotein cholesterol (LDL-c), low high-density lipoprotein cholesterol (HDL-c), and normal triglycerides (TG). This lipid state is theorized to be cardioprotective against atherosclerosis. In SCD, hematopoietic cell transplant (HCT) offers a potentially curative therapy. Long-term survivors of HCT for hematologic malignancies are at increased risk for dyslipidemia and atherosclerosis long-term. The effects of HCT on SCD dyslipidemia are unknown.

Objective:

This retrospective cohort study characterizes lipid profiles at baseline and after nonmyeloablative allogeneic HCT for SCD.

Study Design:

We analyzed data from 116 patients after nonmyeloablative HLA-matched sibling or haploidentical HCT for SCD at the NIH from 2009 to 2021. Total cholesterol, HDL-c, LDL-c, and TG were collected pre-HCT, one year post-HCT, and annually thereafter. Data were analyzed using linear generalized estimating equation regression modeling.

Results:

Successful HCT was associated with a rise in TC, LDL-c, and HDL-c and a decline in TG post-HCT. After HCT, previously low lipid levels increased to the normal range. These changes occurred within the first year of HCT and were maintained thereafter. In patients with graft failure, TC and LDL-c levels remain unchanged from their pre-HCT baseline. Sirolimus use for graft versus host disease prophylaxis was associated with higher TG levels.

Conclusions:

These findings suggest that SCD dyslipidemia resolves with reversal of the SCD phenotype. The normalization of lipid parameters suggests SCD patients are not at increased risk for atherosclerosis after successful HCT compared to their peers; further studies with longer follow-up are required.

Introduction

Sickle cell disease (SCD) is associated with increased morbidity and early mortality^{1, 2}; the life expectancy of patients with SCD in the US is nearly half that of the general population ^{3, 4}. While human and murine data suggest that hyperlipidemia and atherosclerosis are not major contributors to cardiac comorbidities in SCD^5–7, cardiopulmonary complications remain the most common cause of death^{1, 2}. Indeed, patients with SCD display a unique type of dyslipidemia characterized by low levels of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), high-density lipoprotein cholesterol (HDL-c), and normal levels of triglycerides (TG) ⁸, which may be protective against atherosclerosis ⁵.

The mechanism behind this SCD-associated hypocholesterolemia remains unknown. Early work exploring cholesterol levels in patients with various causes of anemia demonstrated a close relationship between hematocrit and cholesterol levels⁹. In addition, elevated TG levels correlate with markers of hemolysis and inflammation¹⁰. Interestingly, patients with other hemolytic anemias have similar dyslipidemia, characterized by low TC and normal TG^11–15. One proposed mechanism is that increased erythroid marrow activity is associated with increased cholesterol absorption and utilization to create more red blood cells¹³.

Lipid measures have been proposed as markers of SCD disease severity. An elevated TG: HDL-c ratio (>2) has been associated with increased endothelial dysfunction in SCD¹⁰. A higher TG: HDL-c ratio (≥2.93) was associated with clinical complications such as acute chest syndrome and vaso-occlusive crises^{16, 17}.

Nonmyeloablative hematopoietic cell transplantation (HCT) is potentially curative for patients with SCD, even those with severe disease^18–20. Long-term survivors of HCT for hematologic malignancies are at increased risk for hypercholesterolemia²¹. Factors such as post-HCT immunosuppression, including sirolimus and cyclosporine, affect lipid homeostasis, and may contribute to the development of hypercholesterolemia²². Radiotherapy, including low-dose radiation, can also contribute to cardiovascular dysfunction and hypercholesterolemia^{22, 23}. Additionally, progesterone therapy for birth control is commonly used peri- and post-HCT and can cause dyslipidemia²⁴. This study is the first to evaluate the effects of HCT on SCD dyslipidemia. We hypothesized that the potentially protective hypocholesterolemia of SCD would end after transplant. We hypothesized that graft failure is associated with increased triglyceride: HDL-c ratio but decreased TC and HDL-c, given that these lipid findings are associated with a return of SCD.

Methods

This cohort study analyzed data from patients who received their first nonmyeloablative HLA-matched sibling or haploidentical HCT for SCD at the National Institutes of Health (NIH) from May 2009 to December 2021. Our study included four transplant protocols (ClinicalTrials.gov identifiers: NCT00061568, NCT00977691, NCT02105766, NCT03077542).^{25, 26} All study participants received alemtuzumab with total body irradiation (300 – 400 cGy). Two protocols included preconditioning with IV pentostatin and oral cyclophosphamide. All but 3 (8%) haploidentical patients received post-HCT cyclophosphamide 50–100 mg/kg. All patients received sirolimus as graft versus host disease (GVHD) prophylaxis after HCT. Sirolimus was continued for at least 1 year in patients with successful HCT.

Total cholesterol, HDL-c, LDL-c, and TG were collected at baseline pre-HCT, one-year post-HCT, and annually thereafter via a fasting serum-based lipid panel. Baseline data were collected within one year preceding HCT, prioritizing measurements closest to the start of conditioning. For patients who underwent a second HCT, data were censored at the time of the repeat HCT. Data were analyzed using linear generalized estimating equation regression models to measure the influence of factors such as years since HCT, sirolimus use, sex, age, transplant type, and graft failure on TC, LDL-c, HDL-c, and TG levels. Graft failure and sirolimus use were treated as time-varying covariates. P-values <0.05 were considered statistically significant. Estimated marginal means were obtained from the regression models to produce sample estimates of baseline and 5 year post-HCT lipid levels for comparison to National Health and Nutrition Examination Survey (NHANES) estimates ²⁷. These marginal mean estimates account for the model’s gender and transplant type effects on outcomes and correspond to an average age of 31.25 years (the sample average). The 5-year post-HCT estimates correspond to patient estimates of those without graft failure and off sirolimus at 5 years post-HCT. Post-HCT estimates were evaluated at 5 years as this corresponds to the median follow-up, and results are qualitatively similar if other years are chosen for evaluation.

We investigated whether changes in lipid levels (measured as change from baseline or change from the most recent previous lipid measurement) were associated with an increased likelihood of graft failure. This analysis employed the changes in lipid levels as a time-varying covariate in a time-to-graft failure model using Cox regression. Sex and age at HCT were included as covariates. For this analysis, all lipid measurements were included, not just those nearest to an annual follow-up date.

Results

This analysis included N=116 patients with a median age of 31 (range 10–64) years (Table 1). The cohort was predominantly HbSS genotype (90%), male (62%), and matched sibling donor HCT type (66%) Overall survival was 87%; 19% experienced graft failure with a median follow-up of 5 (range 0–10) years. No patients were on a statin in the pre-HCT setting; n=10 (9%) were started on a statin post-HCT. Similarly, n=4 patients (3%) were on pre-HCT progesterone therapy and n=10 (9%) were started post-HCT.

Table 1. Study Population Characteristics.

Sample Characteristic	Total (N = 116)
Median age (min, max)	31 (10, 64)
Sex
Male (%)	72 (62%)
Female (%)	44 (38%)
Ethnicity
Black/African American	108 (93%)
Multiple race	3 (3%)
Unknown	4 (3%)
White	1(1%)
Protocols
Haploidentical	77 (66%)
Matched sibling	39 (34%)
SCD genotype
SS	105 (90%)
Sβ0-thalassemia	6 (5%)
SC	3 (3%)
Sβ+-thalassemia	2 (2%)
Medication
Pre-HCT statin	0 (0%)
Post-HCT statin	10 (9%)
Pre-HCT progesterone therapy	4 (3%)
Post-HCT progesterone therapy	10 (9%)
Transplant Outcomes
Median follow-up (25th, 75th percentiles), years	5 (3, 7)
Graft failure	22 (19%)
No graft failure	94 (81%)
Survival status
Alive (%)	101 (87%)
Dead (%)	15 (13%)

Hematopoietic cell transplant (HCT)

Sickle cell disease (SCD)

The population’s mean TC, LDL-c, and HDL-c increased substantially within the first year of transplant (Figure 1). These early increases were sustained throughout the follow-up period. Figure 2 demonstrates that TC remains unchanged from the pre-HCT baseline in patients with graft failure. However, mean TC and LDL-c increased in patients with successful transplants within the first year after HCT. Mean HDL-c levels increased and mean TG levels decreased regardless of failure status (Figure 3). Beyond eight years post-HCT, there was a considerable decrease in the number of patients included in the analyses with a resulting increase in the confidence intervals (Figures 1 and 2).

Figure 1. Mean lipid levels and 95% confidence intervals after HCT for sickle cell disease.

Total Cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), and high-density lipoprotein cholesterol (HDL-c) increased, and Triglyceride (TG) decreased significantly after transplant. Lipid ranges for TC, LDL-c, HDL-c, and TG are based on the Adult Treatment Panel III classification for optimal, desirable, and/or normal lipid levels for the healthy adult population ²⁸.

Figure 2. Mean total cholesterol and low-density lipoprotein measures with 95% confidence intervals post-HCT stratified by presence or absence of graft failure.

Graft failure reverted total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-c) to pre-HCT baseline. N=22 for Failure at baseline counted those who ever failed; N for Failure at subsequent years counted those who contributed a yearly measurement and failed at or prior to that year’s measurement. N=94 for No-Failure at baseline counted those who never failed; N for No-Failure at subsequent years counted those who contributed a yearly measurement and had not failed by the time of the measurement.

Figure 3. Mean high-density lipoprotein cholesterol (HDL-c) and triglyceride measures with 95% confidence intervals post-HCT stratified by the presence or absence of graft failure.

The triglyceride decreased and HDL-c increased in the post-HCT setting. These changes were unaffected by graft status. N=22 for Failure at baseline counted those who ever failed; N for Failure represented those who contributed a yearly measurement and failed at or prior to that year’s measurement. N=94 for No-Failure at baseline counted those who never failed;-N for No-Failure represented those who contributed a yearly measurement and had not failed by the time of the measurement.

According to regression modeling, each year post-HCT is associated with a 43–50 mg/dL increase in TC from baseline (p < 0.0001 for years 1 through 9 and p = 0.02 for year 10, Table S1), while graft failure is associated with a 42 mg/dL decline in the TC (p<0.0001). Total cholesterol increased with every year of age (by +1.06 mg/dL, p = 0.006). Other factors, including transplant type (haplo vs. matched sibling), sex, and sirolimus use, had no statistically significant effect on TC levels.

The post-HCT LDL-c and HDL-c regression modeling trends mirrored TC. The LDL-c and HDL-c levels increased at year 1 (LDL-c: +36.18 mg/dL, p < 0.0001; HDL-c: +19.71, p < 0.0001) and remained similarly increased throughout the follow-up period (Tables S2 and S3). Graft failure, however, was associated with LDL-c reverting back toward the pre-HCT baseline (failed engraftment: −37.62 mg/dL, p< 0.0001). LDL-c increased with every year of age (age: +0.72 mg/dL, p = 0.03). Age had a marginally significant effect on HDL-c (age coefficient: 0.25 mg/dL, p = 0.05) while graft status had no significant impact on HDL-c levels post-HCT. Transplant type (haplo vs. matched sibling), sex, and sirolimus use did not affect LDL-c or HDL-c (Tables S2 and S3).

Unlike the other lipid parameters, TG decreased after HCT. The regression model demonstrated a 26 mg/dL decrease in the first year (p = 0.01) that was sustained annually at similar levels (Table S4). Unlike other lipid types, sirolimus use was associated with a significant increase in TG levels (+32.57 mg/dL, p = 0.0004). Patients were on sirolimus for GVHD prophylaxis for a median of one year. According to the model, after the one-year time point, when most patients stopped sirolimus, TG declined (32.57 mg/dL, p = 0.0004). Transplant type, age, sex, and engraftment status did not affect TG after HCT (Table S4).

Based on estimated marginal means obtained through regression modeling, baseline estimates for TC 132 mg/dL, 95% CI (126, 138), LDL-c 73 mg/dL, 95% CI (68, 79), and HDL-c 33 mg/dL, 95% CI (30, 35) were significantly lower than the non-Hispanic Black U.S. population mean TC 186 mg/dL, 95% CI [182, 190], LDL-c 111 mg/dL, 95% CI [107, 116], and HDL-c 57 mg/dL, 95% CI [56, 58] (Table 2)²⁷. However, the baseline mean TG level [128 mg/dL, 95% CI (117, 138) was substantially higher in this cohort than the non-Hispanic Black U.S. population mean TG [71 mg/dL, 95% CI (67, 74)]²⁷. At the 5-year median follow-up period the mean estimates increased to 180 mg/dL, 95% CI (171, 189) for TC, 111 mg/dL, 95% CI (102, 119) for LDL-c, and 49 mg/dL, 95% CI (45, 53) for HDL-c, but decreased to 108 mg/dL, 95% CI (92, 124) for TG (Table 2)²⁷. Thus, the lipid profiles at the 5-year post-HCT time point following successful HCT nearly normalize to that of the non-Hispanic Black U.S. population.

Table 2. Lipid profile mean estimates for Pre-HCT and 5-year post-HCT vs. the non-Hispanic Black U.S. population.

Pre-HCT and 5-year post-HCT lipid profile marginal mean estimates account for our population’s gender and protocol type (haploidentical vs. matched sibling donor) distributions and correspond to a sample average age of 31.25 years. The 5-year post-HCT mean estimates correspond to patient estimates of those without graft failure and off sirolimus at 5 years post-HCT. Non-Hispanic Black U.S. population means were according to the Center for Disease Control (CDC) 2017–2018 National Health and Nutrition Examination Survey (NHANES) that reports age-adjusted mean total cholesterol, low-density lipoprotein cholesterol (LDL-c), high-density lipoprotein cholesterol (HDL-c), and Triglyceride levels among U.S. adults by race and ethnicity²⁷.

Lipids	Pre-HCT marginal mean estimates (mg/dL) (95% CI)	5-year post-HCT marginal mean estimates (mg/dL) (95% CI)	NHANES mean Black non-Hispanic (mg/dL) (95% CI)
Total Cholesterol	132 (126, 138)	180 (171, 189)	186 (182, 190)
LDL-c	73 (68, 79)	111 (102, 119)	111 (107, 116)
HDL-c	33 (30, 35)	49 (45, 53)	57 (56, 58)
Triglyceride	128 (117, 138)	108 (92, 124)	71 (67, 74)

While we do not believe lipid changes cause graft failure, anecdotally, we have noted that changes in lipid levels may precede graft failure. Therefore, we investigated if any lipid parameters may be associated with graft status. We explored whether changes in TG: HDL-c ratio, TC, or HDL-c predicted the time to graft failure. We hypothesized that declines in TC and HDL-c and increases in the TG: HDL-c ratio might presage graft failure. Whether change was measured as change from baseline or change from the previously measured lipid level, no significant results were seen (see Supplementary Material S9).

Hemolysis is an integral part of the mechanism of hypolipidemia in SCD. Therefore, we examined the association of baseline ARC with pre-HCT lipid parameters. We found no correlation between pre-HCT ARC and baseline TC, LDL-c, HDL-c, or TG (Figures S5, S6, S7 and S8).

Discussion

Nonmyeloablative HCT with sustained engraftment in our patients with SCD led to significant increases in TC, LDL-c, and HDL-c and a decline in TG from baseline to one year post-HCT, normalizing the patients’ previously hypolipidemic profiles (Figure 2 and Table 2). Throughout the follow-up period, lipid levels increased to within age adjusted mean normal ranges and did not continue to increase to pathologically elevated levels^{27, 28}.

Hypercholesterolemia predisposes individuals to developing atherosclerosis and subsequent heart attacks and strokes^{25, 26}. Atherosclerosis typically takes decades to develop from this elevated cholesterol state^{25, 26}. Endothelial dysfunction, an important component in SCD pathophysiology, and elevated LDL-c are significant contributors to risk for atherosclerotic cardiovascular events (ASCVE) such as heart attacks and strokes²⁶. Meanwhile, HDL-c levels are inversely related to ASCVE²⁶.

Comparisons of pre-HCT and 5 year post-HCT estimated marginal means with the U.S. non-Hispanic Black population’s lipid profiles provide evidence that the SCD population is hypolipidemic at baseline and increases to a near normal lipid profile when off sirolimus in the absence of graft failure. At 5 years post-HCT, the TG mean estimate is slightly higher, and the HDL-c mean estimate is slightly lower than the U.S. non-Hispanic Black population. These discrepancies are minor and are unlikely to represent a clinically meaningful difference in atherosclerotic risk given that the 5-year post-HCT lipid mean estimates for TC, LDL-c, HDL-c, and TG all fall within normal, desirable, or optimal lipid panel classification according to Adult Treatment Panel III ²⁸.

These data suggest that the atheroprotective sickle hypolipidemia resolves and patients’ lipid-based risk factors for atherosclerosis become similar to that of the U.S. non-Hispanic Black population (Table 2)²⁷. On the other hand, this rise could be protective because HDL-c levels increase as well. An analysis of this cohort’s nuclear magnetic resonance lipid data, which is a better predictor of atherosclerotic risk, is ongoing^{31, 32}. Additionally, follow-up studies extending past 10 years with the inclusion of a non-HCT SCD cohort will be necessary to assess the impact of HCT on cardiovascular outcomes such as atherosclerosis and myocardial infarction.

Mean TC and LDL-c analyses stratified by graft failure (Figure 2) demonstrate that graft failure returns LDL-c and TC back to baseline. Graft failure is associated with disease return and resumption of the increased erythroid marrow activity of SCD; thus, we suspect there is a resultant increase in cholesterol absorption and an absence of normalization of the low TC and LDL-c levels.

According to linear regression analysis, TC, HDL-c, and LDL-c increased one year post-HCT, and the changes were sustained annually thereafter in the absence of failure. This finding was consistent with our hypothesis that the correction of SCD and sustained engraftment are associated with decreased erythroid marrow activity and increased cholesterol levels. The pathophysiology is complex and may also involve changes in liver lipid metabolism in responses to hemolysis, or lack thereof ⁷. Transplant type (haplo vs. matched sibling), sex, and sirolimus use did not affect TC, LDL-c, and HDL-c. The absence of a detectable effect of transplant type, sex, and sirolimus use further supports our hypothesis that reversal of SCD is more important than these variables when it comes to lipid trends post-HCT.

While the relationship between liver function and dyslipidemia has not been fully elucidated, some have suggested that dysregulation of liver protein production may be contributory⁷. We have experience transplanting patients with liver dysfunction and cirrhosis; indeed, sickle hepatopathy is an inclusion criteria on many NIH protocols. We have seen both amelioration and progression of liver disease after HCT for SCD. However, due to the limited numbers, we have not studied the impact of liver disease systematically, although this may be an important factor which should be explored.

TG decreased one-year post-HCT, and this change was sustained throughout the follow-up period. Sirolimus use is associated with higher TG. Sirolimus affects TG metabolism more than other lipids as it affects the insulin signaling pathway, leading to increased hepatic synthesis of TG³³. The TG decrease and HDL-c increase in the post-HCT setting were unaffected by graft status suggesting that the HCT conditioning regimen may influence HDL-c and TG homeostasis long term. Regardless of graft status, HCT is associated with an increase in HDL-c in our population, which is protective against ASCVE.

Lipid parameters, such as the TG: HDL-c ratio, may be predictive of SCD severity but did not predict the time to graft failure or engraftment status. The lack of significant findings may arise here because there is no systematic change in lipid levels prior to failure, or perhaps because the lipids were not measured regularly enough prior to failure to detect an association. The lack of association between pre-HCT ARC and baseline lipid levels may be due to inadequate ARC data given prior work that has demonstrated the association of hemolysis with lipid levels⁹.

The main limitations of our study include the retrospective nature of the data, possible confounding due to medications used during the HCT course that affected lipid homeostasis, data were not consistently collected more often than annually post-HCT, and the fact that data were from a single center. Confounding medications such as statins and progesterone therapies were taken by approximately 9% of the study population. Duration, doses, and compliance to lipid altering medications was not systematically tracked. Thus, implications of medication effects beyond the use of sirolimus in this population were not analyzed.

Missing data were also a limitation. Figure 1 shows the number of people contributing data at each yearly time point and indicates that the original cohort of 116 decreased to only 8 individuals providing data by year 10. The majority of the missing data occurred because of follow-up time censoring, i.e. 79 of the 116 did not provide data because not enough time had elapsed between transplant and data extraction. Additional reasons for missing 10-year data included death (N=14), censoring for a second transplant (N=3), and unknown/other reasons (e.g. COVID, lost to follow-up, unwillingness to provide samples, N=12). Follow-up time censoring, death, and second transplants are unavoidable reasons for missing data and as such should not be interpreted as poor follow-up. The relatively small number of individuals (10% of cohort) missing 10-year data for unknown/other avoidable reasons is laudable though the results for the later years of follow-up should be interpreted cautiously given the smaller numbers of participants providing data.

While 10 years of follow-up post-HCT is impressive, decades of follow-up and a larger patient population are critical to evaluate the impact of nonmyeloablative HCT on cardiovascular outcomes in individuals with SCD. Further, this single-institution study includes patients with compromised organ function who are often excluded from other HCT protocols. Larger longitudinal trials that use more standard conditioning regimens and typical inclusion criteria should be prioritized.

In conclusion, successful HCT results in significant rises in TC, LDL-c, and HDL-c and a decline in TG from baseline to one-year post-HCT. As a result, sustained engraftment led to normalization of the lipid profiles. These findings reveal that SCD dyslipidemia is corrected with the reversal of SCD. Long-term follow-up studies are indicated to assess the risk of atherosclerosis and adverse cardiac outcomes in individuals with SCD following non-myeloablative allogeneic HCT.

Highlights

• Non-myeloablative HCT in SCD is associated with normalization of lipid profiles

• Graft failure reverts TC and LDL-c levels back to pre-HCT baseline

• Changes in TG:HDL-c ratio, TC, and HDL-c didn’t predict time to graft failure

• Sirolimus effects on lipids were clinically significant on TG levels alone

• Pre-HCT mean TC, LDL-c, and HDL-c were lower in SCD vs. the healthy U.S. population