Changes in Glomerular Filtration Rate and Impact on Long-Term Survival among Adults after Hematopoietic Cell Transplantation

Abstract

Background and objectives

Kidney injury is a significant complication for patients undergoing hematopoietic cell transplantation (HCT), but few studies have prospectively examined changes in GFR in long-term survivors of HCT. We described the association between changes in GFR and all-cause mortality in patients up to 10 years after HCT.

Design, setting, participants, & measurements

We conducted a prospective, observational cohort study of adult patients undergoing HCT at the Fred Hutchinson Cancer Center in Seattle, Washington from 2003 to 2015. Patients were followed from baseline, before conditioning therapy, until a maximum of 10 years after transplant. We used Cox proportional hazard models to examine the association between creatinine eGFR and all-cause mortality. We used time-dependent generalized estimating equations to examine risk factors for decreases in eGFR.

Results

A total of 434 patients (median age, 52 years; range, 18–76 years; 64% were men; 87% were white) were followed for a median 5.3 years after HCT. The largest decreases in eGFR occurred within the first year post-transplant, with the eGFR decreasing from a median of 98 ml/min per 1.73 m² at baseline to 78 ml/min per 1.73 m² by 1 year post-HCT. Two thirds of patients had an eGFR<90 ml/min per 1.73 m² at 1 year after transplant. When modeled as a continuous variable, as eGFR declined from approximately 60 ml/min per 1.73 m², the hazard of mortality progressively increased relative to a normal eGFR of 90 ml/min per 1.73 m² (P<0.001). For example, when compared with an eGFR of 90 ml/min per 1.73 m², the hazard ratios for eGFR of 60, 50, and 40 ml/min per 1.73 m² are 1.15 (95% confidence interval, 0.87 to 1.53), 1.68 (95% confidence interval, 1.26 to 2.24), and 2.67 (95% confidence interval, 1.99 to 3.60), respectively. Diabetes, hypertension, acute graft versus host disease, and cytomegalovirus infection were independently associated with a decline in GFR, whereas calcineurin inhibitor levels, chronic graft versus host disease, and albuminuria were not.

Conclusions

Adult HCT recipients have a high risk of decreased eGFR by 1 year after HCT. Although eGFR remains fairly stable thereafter, a decreased eGFR is significantly associated with higher risk of mortality, with a progressively increased risk as eGFR declines.

Introduction

Hematopoietic cell transplantation (HCT) offers prolonged survival for patients diagnosed with many malignant and nonmalignant diseases. Despite an overall improvement in outcomes after HCT, kidney injury remains a frequent post-transplant complication and likely contributes to morbidity and mortality in this population (1,2). Multiple risk factors contribute to kidney injury after HCT, including chemotherapy, radiation, infections, antimicrobials, calcineurin inhibitors, thrombotic microangiopathy, and graft versus host disease (GVHD) (3–6). Survivors of an episode of AKI after HCT have an almost two-fold increased risk of developing CKD by 1 year post-transplant (7).

The cumulative incidence of CKD among HCT survivors varies from 7% to 48%, and CKD has been diagnosed anywhere from 6 months to 10 years post-transplant (8,9), leading to ESKD in up to 4% of patients with CKD after HCT (10). Approximately 20,000 patients receive an HCT annually in the United States, and with the expanding indications for transplant and the increasing numbers of survivors, there will be an increase in the public health burden faced by a growing population of patients with CKD.

Few studies have prospectively assessed for changes in kidney function in long-term survivors of HCT as the existing literature is limited by mostly retrospective observations with relatively short follow-up (8). We examined a prospectively followed, single center cohort of adult patients from before transplant to up to 10 years post-HCT, with three main objectives: (1) to describe the natural history of kidney injury after transplant, (2) to determine the effect of changes in eGFR after transplant on overall all-cause mortality, and (3) to identify potential risk factors for declines in eGFR in patients followed long-term after HCT.

Materials and Methods

Study Population

From 2003 to 2015, adult patients undergoing their first HCT at the Fred Hutchinson Cancer Center were prospectively enrolled if at baseline, before the start of conditioning therapy, they had the following: (1) a baseline serum creatinine ≤1.3 mg/dl in women and ≤1.5 mg/dl in men (to ensure enrollment of patients with normal kidney function at baseline); (2) were not taking angiotensin receptor blockers or angiotensin-converting enzyme inhibitors at the time of transplant; (3) no history of diabetes at the time of transplant; and (4) consented to a protocol approved by the Institutional Review Board of Seattle Children’s Hospital and the Fred Hutchinson Cancer Research Center. Participants were followed until death or loss to follow-up. The cohort was followed for a median of 5.3 years (interquartile range [IQR], 30 days to 11.2 years) after HCT.

Technique of HCT

Patients received a conditioning regimen followed by infusion of donor hematopoietic cells on “day zero,” by convention. Myeloablative conditioning regimens for allogeneic transplants were cyclophosphamide-based (with total body irradiation [TBI] or busulfan) or busulfan with fludarabine; autologous graft recipients received a number of different regimens. Reduced-intensity conditioning regimens consisted of fludarabine and TBI of 2–4 Gy (11). The kidneys were not shielded during TBI. Allograft recipients received prophylaxis against GVHD with immunosuppressive drugs, usually cyclosporine or tacrolimus plus methotrexate or mycophenolate mofetil (12). Standard prophylaxis for infections included acyclovir, trimethoprim-sulfamethoxazole, oral fluconazole or itraconazole, and pre-emptive ganciclovir for patients with cytomegalovirus (CMV) viremia (13–17). Prophylactic oral ursodiol was given routinely after July 1, 2003.

Specimen Collection and Clinical Parameters in Patients with HCT

Patients were prospectively followed from baseline, before any conditioning therapy, until a maximum of 10 years after transplant. Clinical data and urine samples were collected from patients at baseline, weekly through day 100, monthly through the first year after transplant, and annually for 5 years in patients with a creatinine eGFR <90 ml/min per 1.73 m² at 1 year. After year 5, our center’s long-term follow-up (LTFU) database was used for clinical and laboratory data. Urine was collected between the hours of 8 and 10 in the morning, placed on ice, separated into 2 ml aliquots, and frozen at −80°C until time of analysis for urine creatinine and urine albumin. Blood samples were collected in a citrated tube between the hours of 8 and 10 in the morning at baseline, and then weekly through day 100 post-HCT and again at 1 year.

Clinical data included age, sex, race/ethnicity, indication for HCT, preparative regimen, type of HCT, donor status, and conditioning regimen components. Each patient’s BP, medications, and temperature were recorded weekly on the morning of sample collection. Hypertension was defined as a systolic or diastolic BP >140/90 mm Hg or the prescription of antihypertensive medication. The development of diabetes, the need for insulin, and the presence of viral infections were noted during the first 100 days post-transplant and annually thereafter. Acute GVHD was graded as 0–1 and 2–4. Baseline serum creatinine was drawn at the time of enrollment in the study and before starting any conditioning therapy. GFR was estimated using the CKD Epidemiology Collaboration creatinine-only equation (18). In March of 2007, the calibration methodology for serum creatinine was changed; however, the test method and instrument did not change, and the results were deemed commutable. The degree of albuminuria was expressed as a urinary albumin-to-creatinine ratio (ACR) on a spot urine sample, collected as described above. A normal ACR was <30 mg/g creatinine, microalbuminuria was defined as an ACR of 30–299 mg/g creatinine, and macroalbuminuria was ACR≥300 mg/g creatinine. The average ACR over the first 100 days after HCT was grouped into the above categories. The ACR value at 1 year was also grouped into the above categories and used in the analysis for changes in eGFR after year 1. Diabetes was defined as a diagnosis of diabetes and/or insulin use listed as a medication. CMV infection was defined as blood positive for CMV DNA by PCR without organ involvement, and CMV disease was defined as detection of CMV from biopsy or autopsy tissue or bronchoalveolar lavage in patients with organ dysfunction. Severity of disease was defined on the basis of primary diagnosis and indication for transplant and categorized as low, intermediate, or high.

Statistical Analyses

To describe the natural history of kidney injury after HCT over time, we chose to model eGFR as a continuous variable and use all creatinine values that were available to estimate GFR. Annual GFR estimates among surviving patients were graphically displayed using box-plots. Cox proportional hazards models were fit to assess the association between eGFR and all-cause mortality after HCT. In the models, eGFR was treated as a time-dependent, continuous nonlinear covariate, specifically using a restricted cubic spline. We examined splines with three, four, or five knots, ultimately deciding on four knots. All Cox models were adjusted for age, severity of disease, type of donor, and year of transplant.

To investigate risk factors for eGFR decline over the 10-year observation period, generalized estimating equations (GEE) were created using an unstructured correlation matrix. Variables were treated as time-dependent in the GEE models. Variables included in this model were chosen a priori and included diabetes, hypertension, micro- and macroalbuminuria, acute GVHD grades 2–4 within the first 80 days post-HCT and chronic GVHD at day 200, calcineurin inhibitor use and 12-hour trough levels (cyclosporine and tacrolimus), CMV infection, and CMV disease. For the GEE model, hypertension was defined as above, and the presence or absence of hypertension was maintained throughout time until another measurement was provided that would change the hypertension status. Acute GVHD was defined as a grade of 2 or above at a given time. Average cyclosporine levels and average tacrolimus levels for the first 100 days were used. ACRs were calculated using the average ACR value over the first 100 days and then divided into ACR groups. CMV infection and disease were defined as the status of the patient at a given time during the first 100 days post-HCT. Analyses were conducted with R version 3.4.3 software and a two-sided P value <0.05 was considered statistically significant.

Results

Patient Demographics

Of the 446 patients who met eligibility criteria, 434 were enrolled in the study (Figure 1). Baseline demographic characteristics and baseline serum creatinine did not differ between those who were eligible and consented to the study and those who declined to participate (data not shown). Approximately 134 patients did not meet eligibility criteria and were excluded on that basis. Demographic data are presented in Table 1. The median age of the cohort was 52 years, with a range of 18–76 years and an IQR of 19 years. Fifty-three percent of the population were men, and 87% identified as white. The majority of patients were transplanted for a hematologic malignancy and 74% of patients received an allogeneic transplant. The cohort had a baseline median serum creatinine of 0.8 mg/dl (IQR, 0.2 mg/dl) and a median creatinine eGFR of 98 ml/min per 1.73 m² (IQR, 29 ml/min per 1.73 m²). The median ACR was 31.1 mg/g creatinine at baseline (IQR, 93 mg/g) and the median systolic BP was 122 mm Hg (IQR, 20 mm Hg) with a median diastolic BP of 78 mm Hg (IQR, 14 mm Hg).

Figure 1.

Flow diagram of patients enrolled in the study.

Table 1.

Patient demographic data and clinical characteristics at baseline, n=434

Patient Characteristic	Frequency (%)
Age at transplantation, yr
18–19	8 (2)
20–39	90 (21)
40–59	220 (51)
≥60	116 (27)
Sex
Women	155 (36)
Men	279 (64)
Race
Black	10 (2)
White	380 (87)
Hispanic	24 (6)
Other	20 (5)
Diagnosis
Acute Myelogenous Leukemia	133 (31)
Myelodysplastic Syndrome	73 (17)
Chronic Myelogenous Leukemia	27 (6)
Non-Hodgkin Lymphoma	65 (15)
Acute Lymphocytic Leukemia	30 (7)
Multiple Myeloma	38 (9)
Chronic Lymphocytic Leukemia	22 (5)
Aplastic Anemia	8 (2)
Other	38 (8)
Donor type
Allogeneic related	133 (31)
Autologous	114 (26)
Allogeneic unrelated	185 (43)
Conditioning regimen
Reduced intensity regimens	82 (19)
Myeloablative: CY/TBI 12–13.5 cGy	79 (18)
Myeloablative: BU, CY only	100 (23)
Stem cell source
Bone marrow	54 (12)
Peripheral blood	356 (82)
Cord blood	22 (5)
Other myeloablative regimens	173 (40)
Baseline serum albumin, g/dl, median (range, IQR)	3.3 (2.1–4.3, 0.8)
Baseline serum creatinine, mg/dl, median (range, IQR)	0.8 (0.4–1.5, 0.2)
Baseline eGFR, ml/min per 1.73 m2, median (range, IQR)	98 (52–283, 29)
Baseline ACR, mg/g, median (range, IQR)	31.13 (0.6–4397.0, 93)
Baseline systolic BP, mm Hg, median (range, IQR)	122.0 (82.0–178.0, 20)
Baseline diastolic BP, mm Hg, median (range, IQR)	78.0 (45.0–512.0, 14)
Recipient CMV serostatus
Positive	247 (57)
Negative	185 (43)
Missing	2 (0)

CY, cyclophosphamide; TBI, total body radiation; BU, busulfan; IQR, interquartile range; ACR, albumin-to-creatinine ratio; CMV, cytomegalovirus.

Natural History of Kidney Injury

Figure 2 illustrates changes in annual eGFR over time after HCT. The largest decreases in GFR occurred within the first year post-transplant. The median eGFR at baseline was 98 ml/min per 1.73 m² and decreased to 78 ml/min per 1.73 m² at 1 year post-HCT. Approximately 42% of patients (n=126) had an eGFR of 60–90 ml/min per 1.73 m² and 20% (n=59) had an eGFR<60 ml/min per 1.73 m² by 1 year post-HCT. After year 1, in patients who survived and had a serum creatinine available, the eGFR remained relatively stable over the subsequent years.

Figure 2.

Median eGFR of the cohort among patients with measurements near yearly time after HCT. Box plots show 25th and 75th percentiles. The numbers refer to the number of patients available at each year.

Association between GFR and Overall Mortality

Figures 3 and 4 summarize the association between eGFR and all-cause mortality; there was a total of 201 deaths by last contact among the patients analyzed. When modeled as a continuous variable, as eGFR declined from approximately 60 ml/min per 1.73 m² the hazard of mortality progressively increased (P<0.001). For example, when compared with an eGFR of 90 ml/min per 1.73 m², the hazard ratio for eGFRs of 60, 50, and 40 ml/min per 1.73 m² was 1.15 (95% confidence interval [95% CI], 0.87 to 1.53), 1.68 (95% CI, 1.26 to 2.24), and 2.67 (95% CI, 1.99 to 3.60) respectively (P<0.001). A test of linearity suggested a nonlinear association between eGFR and mortality (P<0.001). Although the number of surviving patients decreased as the time since HCT increased, the association between eGFR and mortality remained consistent over time (interaction P=0.48). The qualitative association between eGFR and the hazard of mortality was similar between allogeneic and autologous transplant recipients. After modeling change in eGFR (the slope) instead of the actual eGFR, the overall shape of this association was similar to that depicted in Figure 3. This association did not qualitatively change after the model was adjusted for diabetes, hypertension, albuminuria, and acute GVHD as time-dependent covariates (data not shown).

Figure 3.

Cox proportional hazards model of the association between creatinine eGFR and overall mortality, showing that mortality increases as eGFR decreases. Values on the y-axis represent the logarithm of the hazard ratio for eGFR value relative to an eGFR of 90 ml/min per 1.73 m². All eGFR values were used for this model, regardless of time measured. A horizontal line at 0 indicates a hazard ratio of 1, i.e., where the hazard of mortality associated with an eGFR at the relevant point on the x-axis is the same as the hazard of mortality associated with an eGFR of 90 ml/min per 1.73 m². Model is adjusted for type of transplant, donor type, and disease status. eGFR is modeled as a cubic spline with four knots. Gray areas represent pointwise 95% confidence limits for the logarithm of the estimated hazard ratio. Plots specific to type of transplant (autologous versus allogeneic) have the same qualitative shape as the plot for all patients.

Figure 4.

Cox proportional hazards model of the association between creatinine eGFR before, after 1 year, and any time point after HCT and overall mortality, showing that mortality increases as eGFR decreases. Values on the y-axis represent the logarithm of the hazard ratio for eGFR value on the x-axis relative to an eGFR of 90 ml/min per 1.73 m². All eGFR values (regardless of time of measurement) were used for the model depicted by the solid curve; eGFR values within 1 year were used for the model depicted by the curve with long dashes; and eGFR values >1 year were used for the model depicted by the curve with short dashes. A horizontal line at 0 indicates a hazard ratio of 1, i.e., where the hazard of mortality associated with an eGFR at the relevant point on the x-axis is the same as the hazard of mortality associated with an eGFR of 90 ml/min per 1.73 m². Model is adjusted for type of transplant, donor type, and disease status.

Risk Factors for Decreased GFR

Table 2 illustrates univariable analyses of risk factors associated with changes in eGFR over the 10-year observation period. Diabetes, hypertension, acute GVHD, micro- and macroalbuminuria, and CMV infection were significantly associated with a decline in eGFR over time. However, on multivariable analysis, only diabetes, hypertension, acute GVHD, macroalbuminuria in the first 100 days, and CMV infection (Table 3) at any time post-HCT were associated with a statistically significant decline in eGFR. There was no suggestion that these associations were dependent on time of eGFR measurement (interactions with time ranging in P values from 0.31 to 0.92). On the basis of our model, patients who developed diabetes or acute GVHD had an eGFR that was approximately 9 ml/min per 1.73 m² lower compared with patients who do not develop either of these complications. There was no significant association between average calcineurin inhibitor levels during the first 100 days after transplant or a diagnosis of chronic GVHD and changes in eGFR over time. Neither micro- nor macroalbuminuria at around 1 year were associated with further declines in GFR after 1 year.

Table 2.

Univariable GEE model estimates, 95% confidence intervals, and P values of clinical variables associated with a change in average GFR over the entire time period

Variable	Estimate	95% CI	P Value
Diabetes	−18.45	(−24.87 to −12.02)	<0.001
Hypertension	−5.96	(−8.34 to −3.57)	<0.001
aGVHD 2–4	−14.26	(−26.96 to −11.56)	<0.001
cGVHD	0.25	(−4.12 to 4.16)	0.91
Cyclosporine	−0.03	(−0.06 to 0.02)	0.22
FK506+tacrolimus	−0.15	(−0.86 to 1.16)	0.77
ACR mean to day 100 (low versus middle)	−7.44	(−12.58 to −2.40)	0.003
ACR mean to day 100 (low versus high)	−16.21	(−22.61 to −8.81)	<0.001
ACR at 1 yr (low versus middle)	−8.55	(−17.32 to 0.23)	0.06
ACR at 1 yr (low versus high)	−11.34	(−42.96 to 20.28)	0.48
CMV infection	−13.87	(−16.99 to −10.75)	<0.001
CMV disease	−7.40	(−16.71 to 1.91)	0.12

The listed estimate represents the estimated difference in GFR units, on average, between those with and those without the listed complication. Cyclosporine and FK506+tacrolimus were modeled as continuous variables, and the listed estimate represents the average change in GFR units for each increase of one unit of cyclosporine (or FK506/tacrolimus). Reference was “not having” the variable, i.e., patients without diabetes have a GFR of 19.15 ml/min per 1.73 m² less, on average, than someone who does have diabetes. GEE, generalized estimating equations; 95% CI, 95% confidence interval; aGVHD, acute graft versus host disease; cGVHD, chronic graft versus host disease; FK506, another name for tacrolimus; ACR, albumin-to-creatinine ratio; CMV, cytomegalovirus.

Table 3.

Multivariable GEE model estimates, 95% confidence intervals, and P values of clinical variables associated with a change in GFR over the entire time period

Variable	Estimate	95% CI	P Value
Diabetes	−9.65	(−14.97 to −4.32)	0.003
Hypertension	−2.92	(−5.13 to −0.71)	<0.01
aGVHD 2–4	−9.31	(−12.04 to −6.57)	<0.001
ACR mean to day 100 (low versus middle)	−2.92	(−9.18 to 3.33)	0.35
ACR mean to day 100 (low versus high)	−8.17	(−16.04 to −0.30)	0.04
CMV infection	−6.87	(−10.10 to −3.64)	<0.001

The listed estimate is to be interpreted as described above for Table 2. GEE, generalized estimating equations; 95% CI, 95% confidence interval; aGVHD, acute graft versus host disease; ACR, albumin-to-creatinine ratio; CMV, cytomegalovirus.

Discussion

We describe changes in kidney function among a large cohort of adults prospectively followed from baseline, before transplant, up to 10 years post-HCT. We observed that the largest decline in creatinine eGFR occurred during the first year after HCT, with almost two thirds of all patients developing an eGFR<90 ml/min per 1.73 m². Approximately 20% of the cohort developed an eGFR<60 ml/min per 1.73 m² by 1 year post-HCT. We identified a nonlinear association between eGFR and all-cause mortality, with the hazard of death increasing significantly as the eGFR approached 60 ml/min per 1.73 m². The risk of death continued to increase as the eGFR declined further. Risk factors independently associated with a decline in eGFR after HCT included diabetes, hypertension, acute GVHD grades 2–4, macroalbuminuria within the first 100 days, and CMV infection. Neither chronic GVHD, calcineurin inhibitor drug levels within the first 100 days, nor microalbuminuria were significantly associated with eGFR.

Progression to ESKD after HCT occurred in approximately 4% of patients with CKD in one study (19), and more recently 0.8% at one center (19,20). Risk factors that have been identified in the development of CKD include AKI, chronic GVHD, older age at transplant, hypertension, longer survival time post-HCT, and TBI (9,10,20–24). In earlier studies, both acute and chronic GVHD have been identified as risk factors for post-transplant CKD (7,25). We found an association between eGFR and acute GVHD but not chronic GVHD in this cohort of patients. This may be because of varying definitions of chronic GVHD and the timing of that diagnosis post-HCT. We also did not specify a stage of CKD and were looking more broadly at changes in eGFR over time whereas prior studies have defined CKD as a GFR≤60 ml/min per 1.73 m² at 1 year post-HCT.

Although AKI has been associated with an increased risk of mortality post-HCT (10,26), the associations with CKD and mortality have been mixed (22,27).We demonstrated an increased risk in mortality associated with a decrease in eGFR starting at approximately 60 ml/min per 1.73 m² at any time post-HCT. It is possible that our prospective design, comprehensive capture of clinical variables (including all creatinine values), and longer length of follow-up allowed us to observe a significant association between eGFR decline and mortality.

Microalbuminuria in the first 100 days was not associated with a subsequent change in eGFR as in previous studies, although earlier studies used only the day 80–100 time point (28) to look for associations with development of CKD (defined as a GFR<60 ml/min per 1.73 m² at 1 year post-HCT). Given the differing time points and outcomes, it is difficult to compare the results from the two studies.

One limitation of our study is the number of patients who died or were lost to follow-up over the time period of the study. In addition, we relied on patient and nurses’ responses on the annual follow-up questionnaires to get additional clinical information and laboratory values on patients when they were drawn by their local physicians. There may have been variations in BP measurement techniques and creatinine measurements at the various sites of patient care. We did not have consistent data on additional comorbidities that may have influenced the relationship between eGFR and mortality, such as cardiovascular disease. In addition, this is a single center study and our patient population is mostly white, which limits the generalizability of the findings. However, many of the protocols used at our center are standardized and in use at other centers as are the definitions of the post-transplant complications. Because we did not have measured GFR available for this study, we used the creatinine estimating equation to calculate eGFR, which may be less precise in this patient population as eGFR has tended to underestimate measured GFR in adults after HCT (29). We did not account for hyperfiltration status separately in the analyses as we were looking at changes in eGFR over time; however, it may be considered as a risk in a patient population with a high prevalence of albuminuria and hypertension.

We observed that the largest decline in creatinine eGFR occurred during the first year after HCT, with 64% of patients having an eGFR<90 ml/min per 1.73 m², offering a time frame within which to target potential interventions. Interventional trials, including potentially the use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, should be designed for use during this time period. The aim would be to prevent albuminuria, to treat hypertension, and to slow the progression of kidney disease. The association of declining eGFR with acute GVHD and lack of association with calcineurin inhibitor levels suggests perhaps aggressive management of GVHD, rather than a change or reduction in calcineurin inhibitor use or levels, would do more to offset the declines in eGFR post-HCT. Additionally, the high risk of low eGFR in this patient population supports the active screening for hypertension, albuminuria, and the need for long-term nephrology follow-up in those with decreased kidney function and/or other markers of CKD.

In summary, we report a dramatic decline in eGFR over the first year post-HCT, suggesting that early intervention may be the key to improve future outcomes. The increasing number of transplants performed both in the United States and worldwide will lead to an increase in the burden of kidney disease for the health care system. Future studies should be designed to identify patients early with kidney injury and to intervene before 1 year after HCT.

Disclosures

None.