Kidney Transplant Outcomes in Recipients with Cognitive Impairment: A National Registry and Prospective Cohort Study

Abstract

Background.

Cognitive impairment is common in patients with end-stage renal disease and is associated with poor outcomes on dialysis. We hypothesized that cognitive impairment might be associated with an increased risk of all-cause graft loss (ACGL) in kidney transplant (KT) recipients.

Methods.

Using the Modified Mini-Mental State (3MS) examination, we measured global cognitive function at KT hospital admission in a prospective, two-center cohort of 864 KT candidates (8/2009–7/2016). We estimated the association between pre-KT cognitive impairment and ACGL using Cox regression, adjusting for recipient, donor, and transplant factors.

Results.

In living donor KT (LDKT) recipients, the prevalence was 3.3% for mild impairment (60≤3MS<80) and 3.3% for severe impairment (3MS<60). In deceased donor KT (DDKT) recipients, the prevalence was 9.8% for mild impairment and 2.6% for severe impairment. LDKT recipients with cognitive impairment had substantially higher ACGL risk than unimpaired recipients (5-year ACGL: 45.5% vs 10.6%, p<0.01; aHR any impairment: 5.40 (95% CI: 1.78–16.34), p<0.01; aHR severe impairment: 5.57 (95% CI: 1.29–24.00), p=0.02). Similarly, DDKT recipients with severe impairment had higher ACGL risk than recipients without severe impairment (5-year ACGL: 53.0% vs 24.2%, p=0.04; aHR severe impairment: 2.92 (95% CI: 1.13–7.50), p=0.03).

Conclusions.

Given the elevated risk of ACGL among KT recipients with cognitive impairment observed in this two-center cohort, research efforts should explore the mechanisms of graft loss and mortality associated with cognitive impairment and identify potential interventions to improve posttransplant survival.

INTRODUCTION

Dementia is a relative contraindication to transplantation; however, there are no guidelines for evaluating cognitive impairment in transplant candidates when the degree of cognitive impairment does not fulfill the diagnostic criteria for dementia. Cognitive impairment is increasingly recognized as a risk factor for adverse outcomes in older adults, including adults reaching end-stage renal disease (ESRD) (1). In the community-dwelling geriatric population, cognitive impairment is associated with an increased risk of mortality (2, 3), hospitalization (4), admission to an intensive care unit (5), and discharge to a nursing home following hospitalization (4, 6). We previously reported that incident dementia and Alzheimer’s disease are associated with an elevated risk of graft loss and mortality in older adults following kidney transplantation (KT) (7). Cognitive impairment, which often proceeds dementia, detected prior to KT has not been well characterized and might also be independently associated with graft loss and mortality in KT recipients.

Studies of chronic kidney disease (CKD) patients have shown that cognitive impairment increases in prevalence and severity as kidney function declines (8–13). Cognitive changes across multiple domains, including executive function and memory, begin early in CKD progression and continue after reaching ESRD (13). As a result, patients with ESRD have twice the prevalence of moderate to severe cognitive impairment as the general population (14, 15). Additionally, dialysis initiation has been associated with a decrease in executive function in a multi-center cohort study (16). Cognitive impairment in dialysis patients has been independently associated with an elevated risk of mortality (1, 17, 18), emphasizing the importance of transplantation in this population.

Since the prevalence of cognitive impairment in ESRD patients is high, cognitive impairment is likely also prevalent but underdiagnosed in KT recipients. We hypothesized that baseline cognitive impairment might be independently associated with graft loss and mortality after KT. Using a prospective, longitudinal, two-center cohort of KT recipients, we measured global cognitive function and estimated the association between cognitive impairment and all-cause graft loss, adjusting for recipient, donor, and transplant characteristics.

MATERIALS AND METHODS

Prospective Cohort Data Source

This study used data from a prospective, longitudinal two-center cohort study at the Johns Hopkins Hospital (N=798), Baltimore, Maryland and the University of Michigan Hospital (N=66), Ann Arbor, Michigan, which has been described elsewhere (19–22). Briefly, study participants were enrolled prior to KT and consented to medical record abstraction to allow for the identification of demographics and co-morbidities. All study participants underwent cognitive testing including the Modified Mini-Mental State (3MS) exam upon admission for KT. The clinical and research activities being reported are consistent with the Declaration of Helsinki and Declaration of Istanbul. The Institutional Review Boards of Johns Hopkins Hospital and the University of Michigan approved this study, and all participants provided written informed consent.

National Registry Data Source

This study also used data from the Scientific Registry of Transplant Recipients (SRTR) external release made available in September 2017. The SRTR data system includes data on all donors, waitlist candidates, and transplant recipients in the United States (US), submitted by members of the Organ Procurement and Transplantation Network (OPTN), and has been previously described (23). The Health Resources and Services Administration (HRSA), U.S. Department of Health and Human Services provides oversight to the activities of the OPTN and SRTR contractors. Using SRTR, we identified 101 718 adults (age≥18) recipients who underwent KT between August 2009 and July 2016. We cross-validated key recipient, donor, and transplant factors in the linkage between our prospective cohort data source and the SRTR. All recipients were successfully linked with SRTR based on age, sex, unique transplant ID, and date of transplant. We rely on the national registry’s capture of data for all analytic variables and outcomes with the exception of cognitive impairment and additional comorbidity statuses (described below).

Global Cognitive Function

The 3MS examination, a validated assessment of global cognitive function (24), was administered to study participants at admission for KT, which was a median (interquartile range [IQR]) 1 (0–1) days prior to KT. Scores for the 3MS examination range between 0–100 (lower scores indicate worse cognition) based on responses to 15 exam components including temporal and spatial orientation, multi-stage commands, and recall. While there are no standardized thresholds for cognitive impairment in the KT candidate population, it is common in the literature to use standard deviation (SD) thresholds to define cognitive impairment in novel populations (24, 25). Consistent with this convention, we defined any cognitive impairment as a 3MS score<80 (−1 SD) and severe cognitive impairment as a 3MS score<60 (−2 SD). By definition, participants with severe cognitive impairment were a subset of those with cognitive impairment and were therefore included in both groups for the purposes of analyses. In sensitivity analyses, we considered cut-offs for cognitive impairment stratified by age and educational attainment based on normative data from an external population of community-dwelling adults (26) (SDC, Materials and Methods).

Model Variable Selection

We used Cox proportional hazards regression to assess the independent association between cognitive impairment and all-cause graft loss (ACGL), defined as graft loss or mortality. The independent association between cognitive impairment and death-censored graft failure (DCGF) and mortality, evaluated separately, are presented in supplemental materials using Cox proportional hazards regression (Tables S1, S2). All analyses were stratified by living or deceased donor status, given the differences in the relevant risk factors. For example, cold ischemia time (CIT) is known to be an important risk factor for graft failure among deceased donor recipients, but CIT is not associated with graft failure in living donor recipients (27). Stratification is also consistent with the SRTR risk-adjustment models (28). In addition to traditional Cox regression, we also present results from hybrid registry-augmented Cox regression models, a statistically efficient method that brings precisely estimated coefficients from the national registry model into the prospective cohort model (29, 30).

Potential confounders were identified using the SRTR risk-adjustment models (28). Covariates included in the final multivariable models were selected to optimize goodness-of-fit as assessed by the log-likelihood test. For living donor kidney transplant (LDKT) recipients, we adjusted for recipient characteristics (age, Black race, Hispanic ethnicity, years on dialysis, diabetes status, panel reactive antibody [PRA] at transplant, college education, employment status, public insurance status, Hepatitis C Virus [HCV] infection, body mass index [BMI], hypertension status, history of transplantation, and Charlson Comorbidity Index [CCI]), donor characteristics (age, BMI), and transplant characteristics (recipient and donor both male, zero HLA mismatches, blood type incompatibility, and transplant date) (31). For deceased donor kidney transplant (DDKT) recipients, we adjusted for recipient factors (age, sex, Black race, Hispanic ethnicity, years on dialysis, diabetes status, PRA at transplant, college education, BMI, hypertension status, history of transplantation, and CCI), donor factors (kidney donor profile index [KDPI]) (32), and transplant factors (CIT and transplant date).

Handling of Missingness

Variable missingness in the national registry data was quite low: CIT (5.2%), education (4.6%), PRA at transplant (2.8%), recipient BMI (1.6%), donor BMI (1.1%), time on dialysis (0.6%), HLA mismatch (0.2%), all other variables <0.1%. We handled missing covariate values using multiple imputation by chained equations (MICE) (SDC, Materials and Methods).

Statistical Analysis

For participants in the prospective cohort, differences in recipient, donor, and transplant characteristics by cognitive impairment were assessed using the χ² (categorical variables) and Mann-Whitney rank-sum (continuous variables) tests. We report frailty as measured by Fried (33); recipients were classified as frail if they had at least 3 of the 5 frailty components as we have previously published (34–37). Low functional status captures recipients unable to perform normal activities. Functional status is reported to the OPTN on a percent scale; we classified low functional status as 70% or lower, which is the point at which a patient is unable to perform normal activity. We classified induction agents as antibody depleting (muromonab-CD3, equine anti-lymphocyte globulin, lymphocyte immune globulin, thymoglobulin, rabbit anti-thymocyte globulin, alemtuzumab) or nonantibody depleting (daclizumab, basiliximab, and rituximab). Differences in the survivor function were assessed using the log-rank test. Functional forms of continuous variables were empirically derived using Martingale residuals. Proportional hazards were confirmed visually by graphing the log-log plot of survival and statistically using Schoenfeld residuals. We used a two-sided α of 0.05 to indicate a statistically significant difference. We report adjusted hazard ratio (aHR) 95% confidence intervals for as per the method of Louis and Zeger (38). This method shows the lower 95% confidence interval first as a subscript, then the point estimate, and finally the upper 95% confidence interval as a subscript. All analyses were performed using Stata 15/MP for Linux (College Station, Texas).

RESULTS

Prospective Cohort Study Population

Participants in our prospective, longitudinal two-center cohort were followed for a median (IQR) 3.3 (2.0–5.2) years and contributed a total of 2640 person-years at risk. Among the 864 KT recipients in our cohort, 362 underwent LDKT and 502 underwent DDKT. There were a total of 134 ACGL events (40 among LDKT recipients and 94 among DDKT recipients). Median (IQR) recipient age was 53 (42–63) years, median (IQR) BMI was 28.0 (24.0–32.0), median (IQR) time on dialysis prior to KT was 1.9 (0.2–4.2) years, 39.4% were female, 39.1% were Black, 2.3% were Hispanic, 65.8% were college educated, 46.6% were employed, 50.7% had public insurance, 7.1% were positive for HCV, 17.4% had a history of diabetes, 29.8% had a history of hypertension, 21.4% had a history of previous transplant, and 10.5% had a PRA > 80 at the time of transplant. Median (IQR) donor age was 41 (29–51) years, median (IQR) BMI was 26.9 (23.8–30.6), 50.3% were female, 18.1% were Black, and 4.1% were Hispanic. Deceased KT donors had a median (IQR) KDPI of 43.5 (27.2–64.5) and a median (IQR) CIT of 11.8 (2.0–26.4) hours. There were 5.7% recipient-donor pairs with zero HLA mismatches. Characteristics of the registry study population can be found in supplemental materials (SDC, Materials and Methods).

Prevalence of Cognitive Impairment in Prospective Cohort

LDKT recipients in our prospective cohort had a median (IQR) 3MS score of 96 (92–99) and DDKT recipients had a median (IQR) 3MS score of 94 (87–97) (Figure 1). The overall prevalence of cognitive impairment was 10.0% (6.6% in LDKT recipients and 12.4% in DDKT recipients), and the prevalence of severe cognitive impairment was 2.9% (3.3% in LDKT recipients and 2.6% in DDKT recipients). LDKT recipients with cognitive impairment were younger (p=0.03) and had lower BMI (p<0.01) than those without cognitive impairment (Table 1). DDKT recipients with cognitive impairment were more likely to be Black (p=0.04), older (p=0.03), and have diabetes (p=0.04) and less likely to be college educated (p<0.001) than those without cognitive impairment. DDKT recipients with cognitive impairment received kidneys with a higher median KDPI (p=0.049) (Table 2).

Figure 1. Distribution of Modified Mini-Mental State Exam (3MS) scores by donor type.

Lower 3MS scores indicate worse global cognition and a 3MS score<80 indicates cognitive impairment. The distribution of 3MS scores were left-skewed for both living donor kidney transplant (LDKT) and deceased donor kidney transplant (DDKT) recipients. The median (IQR) 3MS score was 96 (92–99) for LDKT recipients and 94 (87–97) for DDKT recipients (p<0.001). A larger portion of DDKT recipients were below the threshold for cognitive impairment than LDKT recipients (p<0.001).

Table 1.

Characteristics of 362 living donor kidney transplant (KT) candidates by pre-KT cognitive impairment status.

	No Impairment	Mild Impairment	Severe Impairment
N	338 (93.4%)	12 (3.3%)	12 (3.3%)
Recipient Characteristics1
Female	150 (44.4%)	4 (33.3%)	9 (75.0%)
Black	65 (19.2%)	1 (8.3%)	2 (16.7%)
Hispanic	7 (2.1%)	0 (0.0%)	1 (8.3%)
Age	49.0 (36.0–59.0)	50.0 (30.0–60.0)	33.5 (23.5–43.0)
Years on Dialysis	0.7 (0.0–2.4)	0.4 (0.0–2.0)	0.5 (0.0–1.9)
BMI	27.1 (23.0–30.9)	24.3 (22.2–28.3)	22.6 (22.0–24.4)
HCV	3 (0.9%)	0 (0.0%)	1 (8.3%)
College Educated	251 (76.1%)	7 (58.3%)	8 (72.7%)
Employed	204 (60.7%)	4 (33.3%)	6 (54.5%)
Public Insurance	123 (36.4%)	4 (33.3%)	3 (25.0%)
Diabetes	43 (12.7%)	2 (16.7%)	0 (0.0%)
Hypertension	63 (18.6%)	3 (25.0%)	1 (8.3%)
Previous Transplant	100 (29.6%)	3 (25.0%)	0 (0.0%)
PRA>80 at Transplant	26 (7.7%)	1 (8.3%)	1 (8.3%)
Fried Frailty	48 (14.2%)	2 (16.7%)	1 (8.3%)
Low Functional Status	9 (2.7%)	0 (0.0%)	0 (0.0%)
EPTS Score	34.2 (13.5–57.3)	10.5 (1.2–35.3)	5.8 (1.1–17.0)
Serum Albumin (g/dL)	4.2 (3.9–4.6)	4.6 (4.3–4.7)	4.5 (4.0–4.9)
Donor Characteristics
Female	211 (62.4%)	7 (58.3%)	8 (66.7%
Black	46 (13.6%)	1 (8.3%)	1 (8.3%
Hispanic	10 (3.0%)	1 (8.3%)	0 (0.0%
Age	46.0 (37.0–54.0)	45.5 (39.5–51.0)	49.0 (38.5–53.0)
BMI	26.6 (23.7–29.2)	27.7 (25.5–29.1)	24.0 (21.9–30.7)
eGFR	100.1 (85.2–113.4)	106.0 (90.2–125.5)	97.3 (82.8–108.1)
Biologically Related	122 (36.1%)	7 (58.3%)	5 (41.7%
Transplant Characteristics
ABO Incompatible	43 (12.7%)	1 (8.3%)	1 (8.3%)
Zero HLA mismatch	21 (6.3%)	1 (8.3%)	0 (0.0%)
Cold Ischemia Time	1.5 (1.0–3.0)	1.5 (1.0–2.0)	1.5 (0.8–7.0)
Antibody Depleting Induction	283 (83.7%)	9 (75.0%)	11 (91.7%)
Nonantibody Depleting Induction	65 (19.2%)	2 (16.7%)	1 (8.3%)

¹Characteristics are presented as percentages for binary variables and median (interquartile range) for continuous variables.

Table 2.

Characteristics of 502 deceased donor kidney transplant (KT) candidates by pre-KT cognitive impairment status.

	No Impairment	Mild Impairment	Severe Impairment
N	440 (87.6%)	49 (9.8%)	13 (2.6%)
Recipient Characteristics1
Female	153 (34.8%)	17 (34.7%)	7 (53.9%)
Black	229 (52.1%)	33 (67.4%)	8 (61.5%)
Hispanic	10 (2.3%)	2 (4.1%)	0 (0.0%)
Age	56.0 (46.0–64.5)	62.0 (54.0–70.0)	57.0 (52.0–67.0)
Years on Dialysis	2.9 (1.0–5.1)	2.8 (1.0–4.5)	2.8 (0.0–4.3)
BMI	28.7 (24.9–32.9)	30.3 (25.4–33.6)	29.4 (24.4–30.9)
HCV	53 (12.1%)	4 (8.2%)	0 (0.0%)
College Educated	267 (61.8%)	17 (35.4%)	7 (53.9%)
Employed	165 (37.8%)	15 (31.3%)	5 (38.5%)
Public Insurance	269 (61.1%)	31 (63.3%)	8 (61.5%)
Diabetes	86 (19.6%)	15 (30.6%)	4 (30.8%)
Hypertension	166 (37.7%)	22 (44.9%)	2 (15.4%)
Previous Transplant	76 (17.3%)	5 (10.2%)	1 (7.7%)
PRA>80 at Transplant	54 (12.3%)	3 (6.2%)	0 (0.0%)
Fried Frailty	78 (17.7%)	13 (26.5%)	2 (15.4%)
Low Functional Status	12 (2.7%)	3 (6.1%)	0 (0.0%)
EPTS Score	55.0 (31.4–76.0)	70.1 (34.8–85.4)	69.5 (29.4–90.0)
Serum Albumin (g/dL)	4.3 (4.0–4.6)	4.1 (3.9–4.3)	4.2 (4.0–4.6)
Donor Characteristics
Female	187 (42.5%)	14 (28.6%)	8 (61.5%)
Black	91 (20.7%)	14 (28.6%)	3 (23.1%)
Hispanic	21 (4.8%)	2 (4.1%)	1 (7.7%)
Age	35.0 (26.0–48.0)	41.0 (31.0–50.0)	37.0 (22.0–48.0)
BMI	27.7 (23.8–32.3)	27.5 (24.5–32.7)	27.2 (21.5–34.0)
KDPI	43.0 (26.5–64.1)	49.9 (34.1–69.7)	46.0 (29.1–66.5)
Transplant Characteristics
Zero HLA mismatch	25 (5.7%)	1 (2.0%)	1 (7.7%)
Cold Ischemia Time	25.0 (17.1–32.0)	25.5 (15.4–33.4)	22.5 (17.0–30.1)
Antibody Depleting Induction	395 (89.8%)	42 (85.7%)	11 (84.6%)
Nonantibody Depleting Induction	43 (9.7%)	4 (8.1%)	1 (7.7%)

¹Characteristics are presented as percentages for binary variables and median (interquartile range) for continuous variables.

Post-KT Outcomes, LDKT Recipients

LDKT recipients with any cognitive impairment had higher unadjusted rates of graft loss compared to those without cognitive impairment (p<0.01); however, there were no detectable differences by severe cognitive impairment status in LDKT recipients in unadjusted analyses (p=0.1) (Figure 2). ACGL for LDKT recipients with any cognitive impairment was 45.5% at 5 years versus 10.6% at 5 years for LDKT recipients without cognitive impairment. ACGL for LDKT recipients with severe cognitive impairment was 37.5% at 5 years versus 11.6% at 5 years for LDKT recipients without severe cognitive impairment (Table 3). After adjusting for recipient, donor, and transplant factors using Cox regression, any cognitive impairment in LDKT recipients was associated with a 5.40-fold increased risk of ACGL (aHR: _1.785.40_16.34, p<0.01) and severe cognitive impairment was associated with a 5.57-fold increased risk of ACGL (aHR: _1.295.57_24.00, p=0.02). After adjusting for recipient, donor, and transplant factors using hybrid registry-augmented regression (Table S6), any cognitive impairment in LDKT recipients was associated with a 3.22-fold increased risk of ACGL (aHR: _1.223.22_8.50, p=0.02); however, there was no statistically significant difference by severe cognitive impairment (aHR: _0.843.22_12.36, p=0.09) (Table 4). In sensitivity analyses where cognitive impairment was defined based on an external population, the magnitudes of the associations between cognitive impairment and ACGL increased in LDKT recipients, but our inferences were unchanged (Table S3).

Figure 2. Cumulative incidence of all-cause graft loss (ACGL) in living donor kidney transplant (LDKT) recipients.

LDKT recipients with (A) cognitive impairment (CI) and (B) severe CI.

Table 3.

Unadjusted Risk of all-cause graft loss (ACGL) for kidney transplant (KT) recipients with cognitive impairment.

	% 1-Year ACGL1	% 3-Year ACGL	% 5-Year ACGL
Living Donor KT
No Impairment	2.4 (1.2–4.7)	7.6 (5.0–11.5)	10.6 (7.1–15.7)
Cognitive Impairment	8.7 (2.3–30.5)	18.3 (7.3–41.8)	45.5 (20.3–80.4)
No Severe Impairment	2.6 (1.4–4.9)	8.0 (5.4–11.9)	11.6 (7.9–16.9)
Severe Impairment	8.3 (1.2–46.1)	16.7 (4.5–51.8)	37.5 (11.8–82.8)
Deceased Donor KT
No Impairment	6.3 (4.3–9.0)	17.7 (13.4–22.6)	24.9 (19.6–31.3)
Cognitive Impairment	9.7 (4.5–20.3)	18.6 (10.3–32.3)	26.8 (15.3–44.5)
No Severe Impairment	6.5 (4.6–9.0)	17.3 (13.6–21.8)	24.2 (19.3–30.2)
Severe Impairment	15.4 (4.1–48.8)	37.3 (15.3–73.3)	53.0 (24.5–86.8)

¹ACGL estimated by the Kaplan-Meier method with 95% confidence intervals shown in parentheses.

Table 4.

Risk of all-cause graft loss (ACGL) for kidney transplant (KT) recipients with cognitive impairment. The risk of ACGL was assessed using hybrid registry-augmented Cox proportional hazards (PH) regression. The exposure of interest, cognitive impairment, and the Charleston Comorbidity Index (CCI) were estimated in the prospective cohort model. All other confounders were estimated using the SRTR study population. Recipients with cognitive impairment were compared to recipients without cognitive impairment and recipients with severe cognitive impairment were compared to recipients without severe cognitive impairment.

	Cox PH		Hybrid Registry-Augmented Cox PH
Exposure	aHR ACGL	P value	aHR ACGL	P value
LDKT1
Any Cognitive Impariment2	1.785.4016.34	<0.01	1.223.228.50	0.02
Severe Cognitive Impairment3	1.295.5724.00	0.02	0.843.2212.36	0.09
DDKT4
Any Cognitive Impairment	0.581.051.89	0.9	0.591.041.84	0.9
Severe Cognitive Impairment	1.132.927.50	0.03	1.352.936.35	<0.01

¹LDKT Models adjusted for CCI and confounders from SRTR data: recipient characteristics (continuous age with knots at 35 and 65, Black race, Hispanic ethnicity, years on dialysis, diabetes status, PRA at transplant, college education, employment, public insurance status, HCV status, BMI, hypertension status, and history of transplantation), donor characteristics (age, BMI), and transplant characteristics (recipient and donor both male sex, zero HLA mismatches, blood type incompatibility, and date of transplant).

²3MS Score<80.

³3MS Score<60.

⁴DDKT Models adjusted for CCI and confounders from SRTR data: recipient factors (continuous age with knots at 35 and 65, sex, Black race, Hispanic ethnicity, years on dialysis, diabetes status, PRA at transplant, college education, employment, public insurance status, HCV status, BMI, hypertension status, and history of transplantation), donor factors (KDPI), and transplant factors (CIT and date of transplant).

In sensitivity analyses, any cognitive impairment was associated with a 3.01-fold increased risk of mortality among LDKT recipients (aHR: _1.333.01_6.80, p<0.01) after adjusting for recipient, donor and transplant factors using hybrid registry augmented regression. We did not detect an association between severe impairment and mortality among LDKT recipients (Table S1). When cognitive impairment was defined based on an external population of community dwelling older adults, the association between any cognitive impairment and mortality was slightly attenuated (aHR: _1.062.75_7.11, p=0.04) (Table S4). We did not detect an association between cognitive impairment and death-censored graft failure among LDKT recipients using hybrid registry augmented regression with either internally or externally defined cognitive impairment (Tables S2, S5).

Post-KT Outcomes, DDKT Recipients

DDKT recipients with any cognitive impairment had similar rates of graft loss as individuals without cognitive impairment (p=0.6); however, DDKT recipients with severe cognitive impairment had higher rates of graft loss, compared to DDKT recipients without severe cognitive impairment (p=0.046) (Figure 3). ACGL for DDKT recipients with any cognitive impairment was 26.8% at 5 years versus 24.9% at 5 years for DDKT recipients without cognitive impairment. ACGL for DDKT recipients with severe cognitive impairment was 53.0% at 5 years versus 24.2% at 5 years for DDKT recipients without severe cognitive impairment (Table 3). After adjusting for recipient, donor, and transplant factors using Cox regression, severe cognitive impairment in DDKT recipients was associated with a 2.92-fold increased risk of ACGL (aHR: _1.132.92_7.50, p=0.03); however, there was no statistically significant difference by any cognitive impairment (aHR: _0.581.05_1.89, p=0.9). After adjusting for recipient, donor, and transplant factors in DDKT recipients using hybrid registry-augmented regression (Table S7), severe cognitive impairment in DDKT recipients was associated with a 2.93-fold increased risk of ACGL (aHR: _1.352.93_6.35, p<0.01); however, there was no statistically significant difference by any cognitive impairment (aHR: _0.591.04_1.84, p=0.9) (Table 4). In sensitivity analyses where cognitive impairment was defined based on an external population, the magnitude of the associations between cognitive impairment and ACGL increased in DDKT recipients, but our inferences were unchanged (Table S3).

Figure 3. Cumulative incidence of all-cause graft loss (ACGL) in deceased donor kidney transplant (DDKT) recipients.

DDKT recipients with (A) cognitive impairment (CI) and (B) severe CI.

In sensitivity analyses, we did not detect an association between cognitive impairment and mortality or death-censored graft failure among DDKT recipients (Tables S1, S2). However, in sensitivity analyses where cognitive impairment was defined based on an external population, there was an association between severe cognitive impairment and mortality (aHR: _1.0032.26_5.09, p=0.049) (Table S4). We did not detect an association between cognitive impairment and death-censored graft failure using cognitive impairment definitions defined in the external population (Table S5).

DISCUSSION

In this prospective, longitudinal two-center cohort study of 864 KT recipients, we found that pretransplant cognitive impairment was common and was associated with an elevated risk of all-cause graft loss (a composite of graft loss and mortality). The prevalence of cognitive impairment in our study population was 10.0%. Any cognitive impairment was associated with a 5.40-fold higher risk of ACGL in LDKT recipients (aHR: _1.785.40_16.34). Severe cognitive impairment was associated with a 5.57-fold higher risk of ACGL in LDKT recipients (aHR: _1.295.57_24.00) and with a 2.92-fold higher risk of ACGL in DDKT recipients (aHR: _1.132.92_7.50).

A nationally representative survey in the US that estimated that the prevalence of cognitive impairment was 16% among adults without dementia age 71–79 (39). In our two-center cohort, the prevalence of cognitive impairment was 22.9% among adults aged 71–79 (N=61), possibly pointing to the increased risk of cognitive impairment among patients with chronic kidney disease. While older age was associated with cognitive impairment in our study, the prevalence of cognitive impairment across the study population (median age 53; IQR: 42–63) was noteworthy. Previous studies have identified an association between dialysis initiation and duration with progressively worsening cognitive function (15, 16, 18, 40) likely through the buildup of uremic toxins, inflammation, and cerebral hypotension and hypoxia during dialysis sessions (41, 42). These mechanisms lead to a higher risk of cognitive impairment in kidney transplant candidates of all ages, as we observed in our study. This underscores the need to consider screening for cognitive impairment in KT candidates, even those who would not otherwise be considered at risk due to age alone.

Our observed association of higher graft loss in kidney transplant recipients with cognitive impairment is consistent with prior studies that have shown cognitive impairment to be associated with inferior medical outcomes. One possible mechanism that might explain their elevated risk of all-cause graft loss is poorer medication adherence among recipients with cognitive impairment. Since adherence to the immunosuppressive medication regimen affects the longevity and function of the transplanted allograft, cognitive impairment might indirectly cause inferior posttransplant outcomes. The level of cognitive impairment observed in these studies is severe enough to interfere with adherence to medication and treatment regimens as seen in studies of dialysis patients (14) and nontransplant surgical patients (43–45).This potential pathway warrants further study, as it may represent a target for intervention to improve KT outcomes in this vulnerable population (46).

A limitation of this study was the relatively small sample size of our two-center cohort compared to national transplant recipient datasets. In order to address this limitation, we were able to adjust for important donor, recipient, and transplant characteristics using hybrid registry-augmented regression; however, future multi-center cohort studies will be necessary to improve the generalizability of our findings. Our study used 3MS thresholds based on algorithmically defined cut-offs for cognitive impairment. Given the lack of normative data on 3MS scores in KT candidates, cognitive impairment status might have been misclassified. In sensitivity analyses, we used 3MS thresholds based on an external population with detailed normative data and also stratified by educational attainment and age and found that our inferences did not change (Table S3). Another notable limitation of this study is that we were only able to identify an association between cognitive impairment and all-cause graft loss, not causation. However, we believe that the relationship between cognitive impairment and inferior outcomes is plausible and warrants further study. While we present results for the association of cognitive impairment and mortality and death-censored graft failure (evaluated separately) in the supplemental material, these analyses were likely underpowered based on the sample size and number of events. We were not able to account for social support or other important elements of social context. Strengths of this study include the prospective measurement of global cognitive impairment using a validated instrument (3MS) and reliable ascertainment of posttransplant outcomes using the national registry. Using prospectively collected data, we were also able to adjust for a wider range of comorbidities than is available in the national registry.

In summary, we found that cognitive impairment was common in KT recipients and was associated with an elevated risk of a composite outcome of graft loss and mortality. These findings underscore the importance of elucidating potential mechanisms underlying the relationship between cognitive impairment and inferior posttransplant outcomes, such as medication nonadherence, and emphasize the need for designing interventions to improve outcomes. Transplant centers may consider screening for cognitive impairment to identify higher-risk KT recipients and to inform pre and posttransplant clinical management of these patients. Screening processes might combine a test of global function, like 3MS, with other validated tools specific to the ESRD population, such as the Kidney Disease Quality of Life Cognitive Function scale (47). Further study of the effect of pretransplant interventions to preserve or improve cognitive function, such as cognitive or exercise training, is also warranted.

ABBREVIATIONS

3MS

modified mini-mental status

ACGL

all-cause graft loss

aHR

adjusted hazard ratio

CCI

Charlson Comorbidity Index

CIT

cold ischemia time

CKD

chronic kidney disease

DDKT

deceased donor kidney transplant(ation)

ESRD

end-stage renal disease

HCV

Hepatitis C virus

HRSA

Health Resources and Services Administration

IQR

interquartile range

LDKT

living donor kidney transplant(ation)

KDPI

kidney donor profile index

KT

kidney transplant(ation)

MICE

multiple imputation by chained equations

OPTN

Organ Procurement and Transplantation Network

PPM

predictive mean matching

SRTR

Scientific Registry of Transplant Recipients