https://orcid.org/0000-0001-9456-8243
Figures
Abstract
Autosomal dominant polycystic kidney disease (ADPKD) is a genetic kidney disease with high phenotypic variability. Furthering insights into patients’ ADPKD progression could lead to earlier detection, management, and alter the course to end stage kidney disease (ESKD). We sought to identify patients with rapid decline (RD) in kidney function and to determine clinical factors associated with RD using a data-driven approach. A retrospective cohort study was performed among patients with incident ADPKD (1/1/2002-12/31/2018). Latent class mixed models were used to identify RD patients using differences in eGFR trajectories over time. Predictors of RD were selected based on agreements among feature selection methods, including logistic, regularized, and random forest modeling. The final model was built on the selected predictors and clinically relevant covariates. Among 1,744 patients with incident ADPKD, 125 (7%) were identified as RD. Feature selection included 42 clinical measurements for adaptation with multiple imputations; mean (SD) eGFR was 85.2 (47.3) and 72.9 (34.4) in the RD and non-RD groups, respectively. Multiple imputed datasets identified variables as important features to distinguish RD and non-RD groups with the final prediction model determined as a balance between area under the curve (AUC) and clinical relevance which included 6 predictors: age, sex, hypertension, cerebrovascular disease, hemoglobin, and proteinuria. Results showed 72%-sensitivity, 70%-specificity, 70%-accuracy, and 0.77-AUC in identifying RD. 5-year ESKD rates were 38% and 7% among RD and non-RD groups, respectively. Using real-world routine clinical data among patients with incident ADPKD, we observed that six variables highly predicted RD in kidney function.
Citation: Sim JJ, Shu Y-H, Bhandari SK, Chen Q, Harrison TN, Lee MY, et al. (2024) Data driven approach to characterize rapid decline in autosomal dominant polycystic kidney disease. PLoS ONE 19(6): e0298484. https://doi.org/10.1371/journal.pone.0298484
Editor: Mohamed E. Elrggal, Kidney and Urology Center, EGYPT
Received: January 23, 2024; Accepted: May 21, 2024; Published: June 5, 2024
Copyright: © 2024 Sim et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Individual-level data reported in this study involving human research participants are not publicly shared due to potentially identifying or sensitive patient information. Upon request, and subject to review, KPSC may provide the deidentified aggregate-level data that support the findings of this study. Anonymized data (deidentified data including participant data as applicable) that support the findings of this study may be made available from the investigative team in the following conditions: (1) agreement to collaborate with the study team on all publications, (2) provision of external funding for administrative and investigator time necessary for this collaboration, (3) demonstration that the external investigative team is qualified and has documented evidence of training for human subjects protections, and (4) agreement to abide by the terms outlined in data use agreements between institutions. Interested researchers can submit their request via the “Contact Us” form (see link: https://www.kpscalresearch.org/aboutus/contact-us/).
Funding: This study was funded by a research grant from Otsuka Pharmaceutical Development & Commercialization, Inc., Princeton, NJ. (PI: JJS). The funder provided support in the form of salaries for authors SS, KP, and AN. The funder was involved in the study design and preparation of the manuscript. The funder did not have any additional roles in the data collection and analysis, or decision to publish.
Competing interests: Shirin Sundar, Kristin Pareja, and Ali Nourbakhsh are current or former employees of Otsuka Pharmaceutical Development & Commercialization, Inc., Princeton, NJ. Cynthia J. Willey has worked as a paid consultant for Otsuka Pharmaceuticals Development & Commercialization. John J. Sim currently has or has had research grant support from Otsuka Pharmaceutical Development & Commercialization, Vera Pharmaceuticals, Astra Zeneca Pharmaceuticals, NIH/Kidney Nutrition Obesity and Diabetes Study Section (co-investigator w Manjula Tamura PI), and Kaiser Permanente Southern California clinician investigator award. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Introduction
Autosomal dominant polycystic kidney disease (ADPKD) is a leading cause of genetic kidney disease with an estimated prevalence of 30–50 per 100,000 persons [1–4]. It is the fourth leading cause of end stage kidney disease (ESKD) accounting for 5% of ESKD in the United States [2,5,6]. The natural progression of the disease is characterized by pathogenic genetic mutations that lead to fluid filled cysts resulting in irreversible damage to kidney parenchyma and loss of kidney function [7,8]. ADPKD is genetically heterogenous and is caused by pathogenic mutations most commonly in the PKD1 (85% of cases) and PKD2 (15% of cases) genes. There is interfamilial and intrafamilial variability in the natural course of ADPKD [9–13]. In addition, we previously described differences in prevalence of and also progression to ESKD among different race/ethnic patients with ADPKD by race and ethnicity [4,14]. While 50% of patients progress to ESKD by their sixth decade of life, the decline in eGFR typically occurs rapidly later in adulthood due to initial compensatory glomerular hyperfiltration. While eGFR is an established marker of kidney function, it alone has not shown to reliably assess ADPKD burden nor prognosticate outcomes [15–17].
Treatment of ADPKD was previously limited to the management of symptoms and complications of the disease [18,19]. However, several disease modifying therapies are being studied to evaluate their benefit in patients with ADPKD who are at risk of rapid progression to ESKD. Tolvaptan, a vasopressin V2 receptor-specific antagonist, has shown efficacy in slowing the rate of kidney growth and the decrease in eGFR in ADPKD. It has been FDA approved for those with rapid progression and shown to improve outcomes when started earlier in the disease course [20–22]. As targeted therapies are developed, there remains an increased need for methods to accurately predict the natural course of ADPKD and distinguish rapidly progressing patients most likely to benefit from disease modifying therapy.
Current methods to identify rapid progression in ADPKD have relied on resource intensive prognostic approaches and tools. One of the first methods was developed using data from the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) cohort and Mayo database. Mayo Imaging Classification (MIC) necessitates imaging to determine height adjusted total kidney volume (htTKV), along with age, race, and sex to predict time to decline in eGFR [7,23]. While race is used to calculate eGFR, it is not factored into determining risk of progression to ESKD. The Predicting Renal Outcome in PKD (PROPKD) score uses genotype, sex, and the presence of a urologic event and/or hypertension before the age of 35 to predict onset of ESKD [24,25]. The PROPKD score relies on genotype results limiting its utility. Both the MIC and the PROPKD score were derived from primarily Caucasian populations and thus may be limited in generalizability [7,8,19]. We previously characterized and described an ADPKD cohort and observed racial/ethnic differences in the proportion of patients with kidney failure, age of kidney failure onset, and likelihood of having had kidney transplantation [14]. Thus, approaches to identify rapid progressors early in the disease course with readily available clinical data and considering population diversity are warranted.
Using a large diverse population of patients with ADPKD and longitudinal eGFR data collected within a routine clinical care environment, we sought to identify rapid decline (RD) in kidney function and determine clinical factors associated with RD using a data-driven approach.
Materials and methods
Study population
A retrospective cohort study of Kaiser Permanente Southern California (KPSC) members between January 1, 2002 to December 31, 2018 was performed. KPSC is an integrated health system comprised of 15 medical centers and over 230 satellite clinics providing care to over 4.8 million members. The membership population is racially, ethnically, and socioeconomically diverse, reflecting the general population of Southern California [26]. Complete healthcare encounters are tracked using a comprehensive electronic health record (EHR) from which all study information were extracted.
The study population from which this cohort was identified has been previously described [4,14,27]. In brief, the study population included patients of any age with a minimum of 1-year continuous membership in the health plan. This time requirement was used to reliably capture incident ADPKD diagnoses and comorbidities. Inpatient and outpatient International Classifications of Diseases, Ninth and Tenth Revision (ICD-9, ICD-10) ADPKD diagnoses codes (ICD-9: 753.12, 753.13; ICD-10: Q61.2, Q61.3) were used to identify patients.
Incident ADPKD was defined as newly diagnosed and not having a prior diagnosis of ADPKD. Patients were required to have ≥ 2 diagnosis codes on 2 separate encounter dates (which may have been consecutive days) from inpatient, emergency department (ED), or ambulatory care settings. Patients were excluded if they had a prior diagnosis of ADPKD, ≥ 2 diagnosis codes for autosomal recessive polycystic kidney disease (ARPKD) or did not have 1-year of continuous KPSC membership. Patients with ESKD (defined as treatment with dialysis or kidney transplant), eGFR <15mL/min/1.73m2, or no eGFR information at baseline and during follow-up were also excluded (S1 Fig in S1 File). Patients were followed until they experienced ESKD, death, disenrollment from the healthcare plan, or until the end of the observation period (January 31, 2020).
Data collection
Information on demographics, clinical characteristics, and medications were obtained for patients with incident ADPKD in the 1 year prior to the index date. All laboratory data, vital sign assessments (including blood pressure measurements and body mass index), and diagnostic and procedure codes were extracted from the EHR. Kidney function was expressed as eGFR calculated from serum creatinine levels using the 2009 Chronic Kidney Disease Epidemiology Collaboration equation [28]. Proteinuria was defined as urinalysis positive for protein, urine protein/creatinine ratio >0.2 g/g, urine albumin/creatinine ratio >30mg/g, or a 24-hour urine collection with >200mg total protein or >30mg of albumin. ESKD was defined as treatment with hemodialysis, peritoneal dialysis, or kidney transplant.
Medication use was retrieved from internal pharmacy dispensing records. Health care utilization and an ever history of comorbidities (diabetes mellitus (DM), hypertension (HTN), hyperlipidemia, ischemic heart disease (ICH), congestive heart failure (CHF), cerebrovascular disease, urologic diseases, abdominal pain, and liver disease) were extracted from the EHR. Cerebrovascular disease included any history of ischemic stroke, hemorrhagic stroke, subarachnoid hemorrhage, and cerebral aneurysms. The Elixhauser Comorbidity Index was also extracted from the EHR categorizing 31 comorbidities using diagnosis codes with each comorbidity being assigned a value of ‘1’ [29]. Data on hospitalizations and diagnoses that occurred outside of the KPSC healthcare system were available through administrative claims records. KPSC death records were obtained by identifying death that occurred within KPSC-owned facilities [30].
Analysis
Descriptive statistics stratified by the RD vs. non-rapid decliner groups were used to report demographic and clinical characteristics among patients with ADPKD. The standardized mean difference (SMD) was used to test for distance between group means (i.e., effect size). An SMD > 0.1 indicates the distributions between the two groups are unbalanced with meaningful differences [31].
Serial eGFRs were evaluated for each patient to determine the pattern of eGFR change over time. Group-based trajectory analysis was applied to identify possible patterns in potential distinct trajectory groups based on the longitudinal eGFR data using a latent class mixed model (LCMM). LCMM was performed by regressing eGFR on the measurement time assuming random effects of time between individuals. The appropriate assumptions of trajectories were explored with linear, beta, and spline distributions. For each distribution assumption, the number of distinct trajectory groups were evaluated from 2 to 5 in consideration of the sufficient sample size in each group. The best fitted results were determined based on the smallest model Akaike Information Criterion (AIC). After distinct trajectory groups were identified in the analysis, the trajectory curves of each group were visualized using a locally weighted scatterplot smoothing (LOWESS) approach. Patients with ADPKD showing the steepest decline in kidney function were categorized in one (or more) trajectory curve(s) and were defined as the RD group. Accordingly, the other group(s) was defined as the non-rapid decliner (non-RD).
After RD patients were identified based on the longitudinal eGFR data using LCMM, the next steps explored potential associations included baseline demographics, vital signs, comorbidities, and laboratory data. Given the numerous baseline variables that were considered as potential predictors for RD, several feature selection steps were performed to help identify the variables with higher importance. Variables with missingness above 30% were initially included, and the missing data were addressed using multiple imputation by chained equations (MICE). To achieve better reliability, three feature selection methods were applied: logistic regression, Least Absolute Shrinkage and Selection Operator (LASSO) regression, and random forest. The important features were determined based on mutual agreements between the three methods. In the logistic model, variables were selected if the p-values were less than 0.05 using likelihood ratio test. In LASSO, lambda parameter was tuned to find the top 80th percentile of important features as the variables with non-zero coefficients. In random forest, variables at top 80th percentile importance were selected. After performing the three methods in ten sets of the imputed data, pooled statistics were calculated to determine the selected variables in each method (pooled p-values in logistic model, grouped adaptive LASSO, and averaged-scaled importance in random forest). Finally, variables with mutual agreements of being selected between the methods were included as the predictors in building the prediction model based on the non-missing data.
The logistic model was used to build the prediction model for RD due to its accessibility and simplicity. Variables selected from the feature selection were included in the model, and the prediction performance was evaluated using area under the curve (AUC). A few additional clinical variables of interest were included in the model to see if the prediction performance could be improved, even if they were not initially selected in the feature selection steps. The final prediction model was determined based on the balance between model simplicity, AUC, and clinical relevance.
Time to ESKD and/or mortality were the primary outcomes. ESKD rates overall and before age 53 (50th percentile), 60 (based in the PROPKD score), 62 (based on Mayo Clinic Research), and 63 (75th percentile) were evaluated. ESKD rates within 5 years of incident ADPKD were evaluated for risk assessment among incident ADPKD patients.
Descriptive statistics and cumulative incidence plots were used to describe time to mortality and ESKD stratified by RD status for each of the outcomes of interest. The incidence rate (IR) per 1,000 person-years and the 95% confidence intervals (CI) were estimated using Poisson regression with robust standard error. Annual eGFR change were also calculated for RD and non-RD groups. The annualized rate of change of eGFR was calculated using an ordinary least squares (OLS) regression method fitted to all eGFR measurements for each patient.
All statistical analyses were generated using the SAS Enterprise Guide (version 9.4; SAS Institute, Cary, North Carolina, USA). The KPSC Institutional Review Board (IRB) reviewed and approved the protocol of this study (#11823). A waiver of informed consent was obtained due to the retrospective nature of this study. Data were accessed for research purposes for this study during the period 01/06/2019 through 01/12/2022.
Results
Study population
A total of 1,744 patients with incident ADPKD were included in the study. The mean (SD) age of the study population was 50.6 (19.1) years, 52.4% males, 41.6% White, 13.8% Black, 31.5% Hispanic, and 9.9% Asian/Pacific Islander (Table 1). Mean blood pressure was 129/76 mm Hg and mean (SD) eGFR was 73.8 (35.6) mL/min/1.73m2. At baseline, 60.0% of patients had a history of HTN, 33.0% had a history of urologic disease, and 12.6% had a history of DM.
RD vs non-RD
Based on the best AIC model, we identified two groups in the longitudinal eGFR data: 125 (7%) patients in the RD group and 1,619 (93%) patients in the non-RD group (Fig 1). Patients in the RD group had a mean (SD) age of 41.5 (18.0) years vs. 51.3 (19.0) years in the non-RD group. Compared to non-RD, the RD group had a higher proportion of patients who were male (55.2% vs 52.2%), Hispanic (39.2% vs 30.9%), had higher baseline BP (131/79 mm Hg vs 129/76 mm Hg), and were more likely to have HTN (64.0% vs 59.7%) and cerebrovascular disease (7.2% vs 3.1%) (Table 1). History of urologic disease and abdominal pain were lower in the RD group (28.0% vs. 33.4%, 40.8% vs. 51.1%, respectively). Mean (SD) baseline eGFR was 85.2 (47.3) and 72.9 (34.4) mL/min/1.73m2 in the RD and non-RD groups, respectively. Baseline sodium, bicarbonate, BUN, vitamin D, HDL, CRP, PTH and albumin were lower in RD compared to non-RD (S1 Table in S1 File). Baseline hemoglobin was 13.1 (1.6) g/dL and 13.6 (1.7) g/dL the in RD and non-RD groups, respectively. The RD group was found to have higher urinary protein, lower urinary phosphorus, and lower urinary calcium compared to non-RD group.
Predictors of rapid decline
Of the 42 variables assessed, 7 were found to have significant agreement in identifying RD including: age, hemoglobin, creatinine, proteinuria, hypertension, cerebrovascular disease, and liver disease (Table 2). In addition, 4 pre-selected clinically relevant variables were identified and included in the model: sex, race/ethnicity, diabetes, and history of urologic events. Based on a balance between clinical relevance and AUC assessed using non-missing data, creatinine and history of liver disease were excluded and sex was included in the final model. The 6 final predictors of RD were younger age at onset [Odds Ratio (OR) 0.941 (95% CI, 0.927, 0.955)], male sex [OR 1.795 (95% CI, 1.129, 2.879)], hypertension [OR 4.402 (95% CI, 2.469, 8.059)], cerebrovascular disease [OR 3.612 (95% CI, 1.378, 8.399)], a 1g/dL increase of hemoglobin [OR 0.820 (95% CI, 0.719, 0.936)], and proteinuria [OR 2.887 (95% CI, 1.861, 4.536)] (Table 3). Results showed 72% sensitivity, 70% specificity, 70% accuracy, and 0.77 AUC in identifying the RD group. AUC results were considered fair for values between 0.7 and 0.8 (Fig 2).
ESKD and mortality outcomes
Among the patients in the RD group, 59 (47.2%) progressed to ESKD with an incidence rate of 91.7 (95% CI, 75.2–111.8) per 1,000 person-years with a medium follow-up time of 4.8 years (S2 Table in S1 File). Among patients in the non-RD group, 169 (10.4%) progressed to ESKD with an incidence rate of 17.0 (95% CI, 14.7–19.6) per 1,000 person-years and median follow up time of 6.2 years. The RD group had higher incidence ESKD rates across ages compared to non-RD. After excluding patients >60 years of age, 1,169 patients progressed to ESKD before age 60 with incidence rates 87.2 (95% CI, 70.4–107.9) for RD and 13.3 (95% CI, 10.7–16.6) for non-RD. A total of 19 (15.2%) and 232 (14.3%) patients died in the RD and non-RD groups, respectively. Similar mortality rates were observed between the RD and non-RD (20.5 per 1000-patient years and 21.8 per 1000-person years, respectively). The median (IQR) annual eGFR decline for the RD group was 7.4 (6.2, 10.2) mL/min/1.73m2 compared to 1.7 (0, 3.8) mL/min/1.73m2 for the non-RD group.
After excluding those who disenrolled or died within 5-years of incident ADPKD, 976 were identified for analysis of ESKD within 5-years. A total of 31 (38.3%) developed ESKD within 5-years among 81 in the RD group compared with 63 (7.0%) among non-RD (S3 Table in S1 File). Cumulative incidence plots for ESKD, ESKD before age 60, and mortality were calculated. RD was highly associated with progression to ESKD during all follow-up years whereas cumulative index plots for mortality were similar between both RD and non-RD (S2 Fig in S1 File).
Discussion
Among patients with ADPKD, there remains a need to identify aggressive phenotypes in a practical manner that can help guide preventative and therapeutic considerations along with resource allocation. Using 42 baseline measurements from routine clinical practice along with longitudinal eGFR data in a large population with ADPKD, we performed multiple data set analyses to identify factors associated with RD and determine a model to further predict patients at risk for RD. Our study found 6 variables that highly predicted RD among patients with ADPKD. We observed younger age of onset, male sex, lower hemoglobin, the presence of proteinuria, hypertension, and cerebrovascular disease to be clinical predictors of RD. We demonstrated 72% sensitivity, 70% specificity, and 70% accuracy in identifying RD.
The European Renal Association-European Dialysis and Transplant Association (ERA-EDTA) algorithm for RD identification amongst ADPKD relies on historical eGFR decline to identify RD. However, using historical eGFR decline alone to identify RD is thought to be of limited value in predicting disease progression in early stages given eGFR decline in ADPKD is nonlinear and often occurs rapidly late in the disease [15]. In our study, baseline eGFR was higher in the RD group compared to the non-RD group (85.2 vs 72.9 respectively). These findings are consistent with increased hyperfiltration thought to happen first in patients with ADPKD and RD prior to rapid eGFR decline. Similarly, lower hemoglobin may represent an earlier and more sensitive indicator of kidney function decline. Baseline creatinine was initially identified as one of the 7 significant variables in identifying RD. However, after including only non-missing data and balancing the clinical relevance and AUC, creatinine and history of liver disease were both excluded in our final modeling tool. The absence of creatinine as a risk factor in our final model likely relates to our paradoxical finding of lower creatinine in RD at baseline and the fact that we are using a single creatinine measurement rather than the change over time.
We found age at diagnosis to be significant factors in identifying RD. The CRISP observational study used a large cohort of patients and assessed several prognostic indicators among patients with ADPKD over a six year follow up-period. They similarly found younger age at diagnosis along with total kidney volume (TKV) to be an indicator of early disease progression. Several smaller cohort studies found younger age to be a predictor of RD [17,32]. We found presence of proteinuria to also be a significant in identifying RD vs non-RD. This is consistent with prior studies which have shown that urinary protein excretion is correlated with higher mean arterial pressure, larger renal volumes, and increased filtration fraction amongst ADPKD [32–35].
Total kidney volume (TKV) typically increases continuously from the early stages of the disease and is associated with decline in kidney function. Kidney enlargement progresses significantly in the early stages of the disease making TKV a significant early indicator of disease progression [17,36–38]. One of the most commonly used tools for identifying progression amongst ADPKD is the Mayo Imaging Classification (MIC) [23]. The MIC uses height adjusted total kidney volume (htTKV) indexed for age to predict future decline in eGFR with Mayo Classification Class 1E having the most aggressive predicted decline in eGFR of 4.58 to 4.78 ml/min per 1.73 m2 per year [23]. The RD group in our ADPKD cohort had a decline in eGFR of 7.4 ml/min per 1.73 m2 per year compared to 1.7 ml/min per 1.73 m2 per year for the non-RD group.
All cerebrovascular disease which included intracranial aneurysms and subarachnoid hemorrhage was a predictor of RD in our study. It has recently been reported that progression of renal function and kidney volume in patients with ADPKD is associated with intracranial aneurysms/subarachnoid hemorrhage [39]. It has also been reported that the similar genetic mutations of PKD such as PKD1 splicing mutations or PKD1 frameshift mutations are involved in the progression of renal function, increased kidney volume, and intracranial aneurysms/subarachnoid hemorrhage. Interestingly, these genetic mutations seem to be involved in early-onset renal function decline and early-onset intracranial aneurysms/subarachnoid hemorrhage [40].
We observed that male sex and history of hypertension to be 2 of the 6 significant variables in identifying RD in our final modeling tool. This is similar to the PROPKD score which uses 4 variables to predict RD: male sex, history of HTN before 35 years of age, first urologic event before 35 years, and PKD mutation type. PROPKD further relies on identifying PKD mutation type. Genetic factors play an important role in determining severity of ADPKD. PKD1 truncating mutations are associated with the most severe disease with average age of ESKD onset being 56 years whereas PKD1 non-truncating has an average age of ESKD onset of 68 years. PKD2 mutations are associated with the least severe disease with average of ESKD onset being 79 years [8–13]. Although genetic testing provides prognostic information for ADPKD, it is not often used in routine clinical practice and not widely available on all patients. In addition, ADPKD shows significant phenotypic variability among family members which suggests that disease modifiers including clinical and environmental factors should be considered in evaluating disease progression and prognosis [8,41].
The PROPKD score uses both clinical and genetic data to predict likelihood of reaching kidney failure before the age of 60. By assigning points to patients with ADPKD, a score > 6 indicates risk of rapid progression with a 92% chance of reaching kidney failure before age 60. Use of genetic mutation in the scoring system does not take into consideration the intrafamilial variability. A recent long-term follow-up of the CRISP cohort found that while ADPKD genotype was associated with CKD outcomes, it was not considered an independent prognostic factor after adjusting for htTKV.13 Similarly, a retrospective study of 164 patients with ADPKD in Spain found the PROPKD score very specific but had low sensitivity in identifying patients with high risk for progression [16]. Finally, the PROPKD and MIC classification tools were both developed in primarily Caucasian populations. Our clinical risk prediction model is not only an additional tool to help determine RD across a diverse population but also relies on data from routine clinical practice which is often readily available.
Previously, treatment options for ADPKD were limited to management of symptoms and complications.17 Management was focused on hypertension, hydration, dietary changes, treatment of pain, urinary tract infections, and ephrolithiasis.18 Most patients with ADPKD are diagnosed more than two decades before they reach ESKD [42,43]. As novel therapies are developed, there is an increased need for tools that accurately identify higher risk populations which are most likely to benefit from treatment [44]. While there is no cure for ADPKD, tolvaptan, a selective vasopressin V2 receptor blocker, has been FDA approved as the first treatment to slow kidney function in ADPKD. Tolvaptan reduced kidney growth by 45%, reduced eGFR decline by 26% in early ADPKD, and reduced eGFR decline by 35% in advanced ADPKD19,18. Newer therapeutic developments further illustrate the need for clinical tools and resources that reliably identify patients eligible for treatment to and guide clinical decision making to help improve outcomes.
Limitations
There are several potential limitations that may confound the interpretation of our findings. We relied on diagnosis codes from the EHR to identify patients with ADPKD, which could introduce selection bias if some patients were misdiagnosed or undiagnosed. The reliance solely on ICD codes without genetic information or imaging data could lead to over or under-capturing. However, use of EHRs to identify to rare diseases within KPSC has been described to have modestly high positive predictive values [45]. Studies evaluating the accuracy of ADPKD diagnosis by ICD codes have demonstrated high sensitivity and positive predictive values exceeding 85% [46,47].
Patients with ESKD, eGFR <15 mL/min/1.73m2, or no eGFR information at baseline and during follow-up were excluded, which could affect the generalizability of the results to the entire ADPKD population. In our study, we did not have genetic information nor abdominal imaging with age adjusted TKV total kidney volume for the entire study population as these diagnostic studies were not 100% adopted into practice over this observation window. But we also felt our approach utilizing routine clinical information that was available was a strength with the potential to be more readily utilized in the real world. Finally, our study was conducted within a single integrated healthcare system in Southern California where the study’s findings may not be generalizable to all ADPKD patients, especially those outside this geographic region or within different healthcare systems. The population’s specific demographic and healthcare system characteristics may influence the results. While our study population was racially and ethnically diverse, we did perform a stratified analysis by race/ethnicity, which could mask potential differences in the predictors and outcomes of RD among different groups. Despite these potential limitations, our ADPKD cohort remains one of the largest to date with detailed clinical information capturing laboratory results, medication use, and health care utilizations.
Conclusion
Among a large, diverse population using real-world data from routine clinical practice, we observed that 6 variables (age, hemoglobin, proteinuria, hypertension, cerebrovascular disease, and sex) highly predict RD among patients with incident ADPKD. Clinical prediction tools may serve as a practical screening tool to capture and manage high-risk patients with ADPKD who may need earlier and more intensive management strategies.
Acknowledgments
The authors would like to thank the KPSC Renal Business Group for their assistance with identifying and providing comprehensive information for the ESKD population. Preliminary results of this manuscript were presented at the National Kidney Foundation Spring Clinical Meetings, April 2022.
References
- 1. Levy M, Feingold J. Estimating prevalence in single-gene kidney diseases progressing to renal failure. Kidney Int. 2000;58(3):925–43. pmid:10972657.
- 2. Willey C, Kamat S, Stellhorn R, Blais J. Analysis of Nationwide Data to Determine the Incidence and Diagnosed Prevalence of Autosomal Dominant Polycystic Kidney Disease in the USA: 2013–2015. Kidney Dis (Basel). 2019;5(2):107–17. Epub 20190109. pmid:31019924; PubMed Central PMCID: PMC6465773.
- 3. Davies F, Coles GA, Harper PS, Williams AJ, Evans C, Cochlin D. Polycystic kidney disease re-evaluated: a population-based study. Q J Med. 1991;79(290):477–85. pmid:1946928.
- 4. Aung TT, Bhandari SK, Chen Q, Malik FT, Willey CJ, Reynolds K, et al. Autosomal Dominant Polycystic Kidney Disease Prevalence among a Racially Diverse United States Population, 2002 through 2018. Kidney360. 2021;2(12):2010–5. Epub 20210922. pmid:35419536; PubMed Central PMCID: PMC8986058.
- 5. Saran R, Robinson B, Abbott KC, Bragg-Gresham J, Chen X, Gipson D, et al. US Renal Data System 2019 Annual Data Report: Epidemiology of Kidney Disease in the United States. Am J Kidney Dis. 2020;75(1 Suppl 1):A6–A7. Epub 20191105. pmid:31704083.
- 6. Suwabe T, Chamberlain AM, Killian JM, King BF, Gregory AV, Madsen CD, et al. Epidemiology of autosomal-dominant polycystic liver disease in Olmsted county. JHEP Rep. 2020;2(6):100166. Epub 20200804. pmid:33145487; PubMed Central PMCID: PMC7593615.
- 7. Cornec-Le Gall E, Alam A, Perrone RD. Autosomal dominant polycystic kidney disease. Lancet. 2019;393(10174):919–35. Epub 20190225. pmid:30819518.
- 8. Chebib FT, Torres VE. Assessing Risk of Rapid Progression in Autosomal Dominant Polycystic Kidney Disease and Special Considerations for Disease-Modifying Therapy. Am J Kidney Dis. 2021;78(2):282–92. Epub 20210308. pmid:33705818.
- 9. Lavu S, Vaughan LE, Senum SR, Kline TL, Chapman AB, Perrone RD, et al. The value of genotypic and imaging information to predict functional and structural outcomes in ADPKD. JCI Insight. 2020;5(15). Epub 20200806. pmid:32634120; PubMed Central PMCID: PMC7455088.
- 10. Hwang YH, Conklin J, Chan W, Roslin NM, Liu J, He N, et al. Refining Genotype-Phenotype Correlation in Autosomal Dominant Polycystic Kidney Disease. J Am Soc Nephrol. 2016;27(6):1861–8. Epub 20151009. pmid:26453610; PubMed Central PMCID: PMC4884120.
- 11. Cornec-Le Gall E, Audrezet MP, Chen JM, Hourmant M, Morin MP, Perrichot R, et al. Type of PKD1 mutation influences renal outcome in ADPKD. J Am Soc Nephrol. 2013;24(6):1006–13. Epub 20130221. pmid:23431072; PubMed Central PMCID: PMC3665389.
- 12. Cornec-Le Gall E, Audrezet MP, Renaudineau E, Hourmant M, Charasse C, Michez E, et al. PKD2-Related Autosomal Dominant Polycystic Kidney Disease: Prevalence, Clinical Presentation, Mutation Spectrum, and Prognosis. Am J Kidney Dis. 2017;70(4):476–85. Epub 20170327. pmid:28356211; PubMed Central PMCID: PMC5610929.
- 13. Hateboer N, v Dijk MA, Bogdanova N, Coto E, Saggar-Malik AK, San Millan JL, et al. Comparison of phenotypes of polycystic kidney disease types 1 and 2. European PKD1-PKD2 Study Group. Lancet. 1999;353(9147):103–7. pmid:10023895.
- 14. Harrison TN, Chen Q, Lee MY, Munis MA, Morrissette K, Sundar S, et al. Health Disparities in Kidney Failure Among Patients With Autosomal Dominant Polycystic Kidney Disease: A Cross-Sectional Study. Kidney Med. 2023;5(2):100577. Epub 20221205. pmid:36718187; PubMed Central PMCID: PMC9883284.
- 15. Yu ASL, Shen C, Landsittel DP, Grantham JJ, Cook LT, Torres VE, et al. Long-term trajectory of kidney function in autosomal-dominant polycystic kidney disease. Kidney Int. 2019;95(5):1253–61. Epub 20190304. pmid:30922668; PubMed Central PMCID: PMC6478515.
- 16. Naranjo J, Furlano M, Torres F, Hernandez J, Pybus M, Ejarque L, et al. Comparative analysis of tools to predict rapid progression in autosomal dominant polycystic kidney disease. Clin Kidney J. 2022;15(5):912–21. Epub 20211228. pmid:35498884; PubMed Central PMCID: PMC9050526.
- 17. Woon C, Bielinski-Bradbury A, O’Reilly K, Robinson P. A systematic review of the predictors of disease progression in patients with autosomal dominant polycystic kidney disease. BMC Nephrol. 2015;16:140. Epub 20150815. pmid:26275819; PubMed Central PMCID: PMC4536696.
- 18. Chebib FT, Perrone RD, Chapman AB, Dahl NK, Harris PC, Mrug M, et al. A Practical Guide for Treatment of Rapidly Progressive ADPKD with Tolvaptan. J Am Soc Nephrol. 2018;29(10):2458–70. Epub 20180918. pmid:30228150; PubMed Central PMCID: PMC6171265.
- 19. Chebib FT, Torres VE. Recent Advances in the Management of Autosomal Dominant Polycystic Kidney Disease. Clin J Am Soc Nephrol. 2018;13(11):1765–76. Epub 20180726. pmid:30049849; PubMed Central PMCID: PMC6237066.
- 20. Torres VE, Chapman AB, Devuyst O, Gansevoort RT, Grantham JJ, Higashihara E, et al. Tolvaptan in patients with autosomal dominant polycystic kidney disease. N Engl J Med. 2012;367(25):2407–18. Epub 20121103. pmid:23121377; PubMed Central PMCID: PMC3760207.
- 21. Torres VE, Chapman AB, Devuyst O, Gansevoort RT, Perrone RD, Koch G, et al. Tolvaptan in Later-Stage Autosomal Dominant Polycystic Kidney Disease. N Engl J Med. 2017;377(20):1930–42. Epub 20171104. pmid:29105594.
- 22. Torres VE, Higashihara E, Devuyst O, Chapman AB, Gansevoort RT, Grantham JJ, et al. Effect of Tolvaptan in Autosomal Dominant Polycystic Kidney Disease by CKD Stage: Results from the TEMPO 3:4 Trial. Clin J Am Soc Nephrol. 2016;11(5):803–11. Epub 20160223. pmid:26912543; PubMed Central PMCID: PMC4858477.
- 23. Irazabal MV, Rangel LJ, Bergstralh EJ, Osborn SL, Harmon AJ, Sundsbak JL, et al. Imaging Classification of Autosomal Dominant Polycystic Kidney Disease: A Simple Model for Selecting Patients for Clinical Trials. Journal of the American Society of Nephrology. 2015;26(1):160–72. pmid:24904092-201501000-00019.
- 24. Cornec-Le Gall E, Audrezet MP, Rousseau A, Hourmant M, Renaudineau E, Charasse C, et al. The PROPKD Score: A New Algorithm to Predict Renal Survival in Autosomal Dominant Polycystic Kidney Disease. J Am Soc Nephrol. 2016;27(3):942–51. Epub 20150706. pmid:26150605; PubMed Central PMCID: PMC4769200.
- 25. Cornec-Le Gall E, Blais JD, Irazabal MV, Devuyst O, Gansevoort RT, Perrone RD, et al. Can we further enrich autosomal dominant polycystic kidney disease clinical trials for rapidly progressive patients? Application of the PROPKD score in the TEMPO trial. Nephrol Dial Transplant. 2018;33(4):645–52. pmid:28992127; PubMed Central PMCID: PMC5888998.
- 26. Koebnick C, Langer-Gould AM, Gould MK, Chao CR, Iyer RL, Smith N, et al. Sociodemographic characteristics of members of a large, integrated health care system: comparison with US Census Bureau data. Perm J. 2012;16(3):37–41. pmid:23012597; PubMed Central PMCID: PMC3442759.
- 27. Sim J, Shu Y-H, Harrison T, Chen Q, Munis M, Morrissette K, et al.,. Predictors Associated with Rapid Decline of eGFR to End-Stage Kidney Disease (ESKD) among a Diverse Autosomal Dominant Polycystic Kidney Disease (ADPKD) Population. American Journal of Kidney Diseases. 2022:79(4) Supp2:S7.
- 28. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–12. pmid:19414839; PubMed Central PMCID: PMC2763564.
- 29. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8–27. pmid:9431328.
- 30. Bhandari SK, Zhou H, Shaw SF, Shi J, Tilluckdharry NS, Rhee CM, et al. Causes of Death in End-Stage Kidney Disease: Comparison between the United States Renal Data System and a Large Integrated Health Care System. Am J Nephrol. 2022;53(1):32–40. Epub 20220111. pmid:35016183.
- 31. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083–107. pmid:19757444; PubMed Central PMCID: PMC3472075.
- 32. Panizo N, Goicoechea M, Garcia de Vinuesa S, Arroyo D, Yuste C, Rincon A, et al. Chronic kidney disease progression in patients with autosomal dominant polycystic kidney disease. Nefrologia. 2012;32(2):197–205. Epub 20120127. pmid:22425799.
- 33. Chapman AB, Johnson AM, Gabow PA, Schrier RW. Overt proteinuria and microalbuminuria in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 1994;5(6):1349–54. pmid:7894001.
- 34. Schrier RW, Brosnahan G, Cadnapaphornchai MA, Chonchol M, Friend K, Gitomer B, et al. Predictors of autosomal dominant polycystic kidney disease progression. J Am Soc Nephrol. 2014;25(11):2399–418. Epub 20140612. pmid:24925719; PubMed Central PMCID: PMC4214531.
- 35. Ozkok A, Akpinar TS, Tufan F, Kanitez NA, Uysal M, Guzel M, et al. Clinical characteristics and predictors of progression of chronic kidney disease in autosomal dominant polycystic kidney disease: a single center experience. Clin Exp Nephrol. 2013;17(3):345–51. Epub 20121020. pmid:23085781.
- 36. Chapman AB. Approaches to testing new treatments in autosomal dominant polycystic kidney disease: insights from the CRISP and HALT-PKD studies. Clin J Am Soc Nephrol. 2008;3(4):1197–204. Epub 20080625. pmid:18579674.
- 37. Grantham JJ, Chapman AB, Torres VE. Volume progression in autosomal dominant polycystic kidney disease: the major factor determining clinical outcomes. Clin J Am Soc Nephrol. 2006;1(1):148–57. Epub 20051019. pmid:17699202.
- 38. Xue C, Zhou C, Mei C. Total kidney volume: the most valuable predictor of autosomal dominant polycystic kidney disease progression. Kidney Int. 2018;93(3):540–2. pmid:29475545.
- 39. Kataoka H, Akagawa H, Yoshida R, Iwasa N, Ushio Y, Akihisa T, et al. Impact of kidney function and kidney volume on intracranial aneurysms in patients with autosomal dominant polycystic kidney disease. Sci Rep. 2022;12(1):18056. Epub 20221027. pmid:36302803; PubMed Central PMCID: PMC9613770.
- 40. Ushio Y, Kataoka H, Akagawa H, Sato M, Manabe S, Kawachi K, et al. Factors associated with early-onset intracranial aneurysms in patients with autosomal dominant polycystic kidney disease. J Nephrol. 2024. Epub 20240205. pmid:38315279.
- 41. Lanktree MB, Guiard E, Li W, Akbari P, Haghighi A, Iliuta IA, et al. Intrafamilial Variability of ADPKD. Kidney Int Rep. 2019;4(7):995–1003. Epub 20190507. pmid:31317121; PubMed Central PMCID: PMC6611955.
- 42. Dicks E, Ravani P, Langman D, Davidson WS, Pei Y, Parfrey PS. Incident renal events and risk factors in autosomal dominant polycystic kidney disease: a population and family-based cohort followed for 22 years. Clin J Am Soc Nephrol. 2006;1(4):710–7. Epub 20060608. pmid:17699277.
- 43. McEwan P, Bennett Wilton H, Ong ACM, Orskov B, Sandford R, Scolari F, et al. A model to predict disease progression in patients with autosomal dominant polycystic kidney disease (ADPKD): the ADPKD Outcomes Model. BMC Nephrol. 2018;19(1):37. Epub 20180213. pmid:29439650; PubMed Central PMCID: PMC5810027.
- 44. McGill RL, Saunders MR, Hayward AL, Chapman AB. Health Disparities in Autosomal Dominant Polycystic Kidney Disease (ADPKD) in the United States. Clin J Am Soc Nephrol. 2022;17(7):976–85. Epub 20220620. pmid:35725555; PubMed Central PMCID: PMC9269641.
- 45. Sun AZ, Shu YH, Harrison TN, Hever A, Jacobsen SJ, O’Shaughnessy MM, et al. Identifying Patients with Rare Disease Using Electronic Health Record Data: The Kaiser Permanente Southern California Membranous Nephropathy Cohort. Perm J. 2020;24. Epub 20200207. pmid:32069207; PubMed Central PMCID: PMC7021143.
- 46. Kalatharan V, Pei Y, Clemens KK, McTavish RK, Dixon SN, Rochon M, et al. Positive Predictive Values of International Classification of Diseases, 10th Revision Coding Algorithms to Identify Patients With Autosomal Dominant Polycystic Kidney Disease. Can J Kidney Health Dis. 2016;3:2054358116679130. Epub 20161214. pmid:28781884; PubMed Central PMCID: PMC5518965.
- 47. Kalot MA, El Alayli A, Al Khatib M, Husainat N, McGreal K, Jalal DI, et al. A Computable Phenotype for Autosomal Dominant Polycystic Kidney Disease. Kidney360. 2021;2(11):1728–33. Epub 20210916. pmid:35372997; PubMed Central PMCID: PMC8785841.