Familial Variability of Disease Severity in Adult Patients With ADPKD

Abstract

Introduction

Autosomal dominant polycystic kidney disease (ADPKD) is the most common monogenic nephropathy and has striking familial variability of disease severity.

Methods

To better comprehend familial phenotypic variability, we analyzed clinical and pedigree data on 92 unrelated ADPKD kindreds with ≥2 affected individuals (N = 292) from an Irish population. All probands underwent genetic sequencing. Age at onset of kidney failure (KF), decline in estimated glomerular filtration rate (eGFR), predicting renal outcome in polycystic kidney disease (PROPKD) score, and imaging criteria were used to assess and grade disease severity as mild, intermediate, or severe. One mild and 1 severe case per family defined marked intrafamilial variability of disease severity.

Results

Marked intrafamilial variability was observed in at least 13% of the 92 families, with a higher proportion of families carrying PKD1-nontruncating (PKD1-NT) variants. In families with ≥2 members affected by KF, the average intrafamilial age difference was 7 years, and there was no observed difference in intrafamilial variability of age at KF between allelic groups. The prespecified criteria showed marked familial variability in 7.7%, 8.4%, and 24% for age at KF, the PROPKD score, and imaging criteria, respectively. In our multivariate mixed-effects model, the intrafamilial variability in kidney survival was independent of the measured genotypic factors associated with prognosis and survival (P = <0.001).

Conclusion

Using objective measures, we quantified marked intrafamilial variability in ADPKD disease phenotype in at least 13% of families. Our findings indicate that intrafamilial phenotypic variability remains incompletely understood and necessitates a more thorough identification of relevant clinical and genotypic factors.

Graphical abstract

See Commentary on Page 523

ADPKD (OMIM 173900) is the most prevalent monogenic nephropathy, with an estimated prevalence of 1:1000 in the general population.^1,2 In affected individuals, kidney cyst formation and gradual enlargement exert pressure and damage to adjacent kidney parenchyma, resulting in chronic kidney disease (CKD) and ultimately leading to KF. 70% of patients may develop KF by their 70s.¹ Affected individuals are also at increased risk of liver cysts, hypertension, urinary tract infections, and intracranial artery aneurysms.³ Clinical genetic testing is usually reserved for atypical situations, such as disease progression variability, especially within families, which has become a clinical necessity since the approval of tolvaptan.⁴ However, tolvaptan eligibility is restricted to patients with ADPKD with rapid disease progression, leaving numerous patients with ADPKD, including young patients with early nonprogressive disease and preserved kidney function, in need of additional disease-modifying therapies.⁵ Therefore, it is necessary to investigate a broad spectrum of patients with ADPKD, not just those with severe disease.

ADPKD is genetically heterogenous, and pathogenic variants in PKD1 or PKD2, which code for polycystin-1 and polycystin-2, are the most common causes of ADPKD.⁴ ADPKD manifests as a recessive disorder at the cellular level, and the likelihood of cyst formation depends on the level of functional polycystin-1 or polycystin-2. Numerous studies have found loci and allelic effects that can modulate disease severity outcomes.^3,6,7 Although ADPKD is well-known for its familial variability, it remains incompletely understood,^8,9 though it can be explained for a small subset of cases.10, 11, 12, 13 Also, in a large ADPKD cohort, the prevalence estimate of the marked intrafamilial variability in disease severity was reported as at least 12%.¹⁴ Based on clinical progression and total kidney volume (TKV), the same group found 18% of patients with PKD1 protein-truncating (PT) variant, who had a poor prognosis, had a less severe kidney phenotype, suggesting modifying factors that may accelerate or slow disease progression.¹⁵

Here, we set out to describe the prevalence of familial phenotypic variability of ADPKD depending on disease severity in an Irish population, stratified by pathogenic variant. We had 3 research questions as follows: (i) How prevalent is the marked intrafamilial variability in disease severity? (ii) What is the frequency of familial variability of ADPKD for each metric? and (iii) Does the marked intrafamilial variation have any identifiable characteristics?

Methods

Ethics

The study was approved by the ethics committee of Beaumont Hospital, Dublin, Ireland (REC 12/75). Written informed consent for participation in clinical and genetic studies was obtained before enrollment.

The Study Cohort

Phenotype information was retrieved from the Irish Kidney Gene Project at Beaumont Hospital, Dublin.¹⁶ Individuals were referred from the nephrology clinics and dialysis units all over Ireland. With a family history consistent with dominant inheritance, ADPKD was diagnosed by conventional kidney imaging using age-dependent diagnostic criteria.¹⁷ We included families with at least 2 confirmed affected individuals. Individuals younger than 18 years, cases with no apparent family history, singletons with a family history of ADPKD, and those with definite genetic or clinical evidence of nephropathies other than ADPKD were excluded.

Clinical Phenotype

Cohort characteristics were reported as of the dataset available in June 2022. Clinical baseline data was collected during recruitment, outpatient clinic visits, medical records, and phone calls to collect additional information. A web-based platform, Progeny version 9.5.5, was used to store and analyze clinical data and family pedigrees of Irish patients with ADPKD. Year of birth, sex at birth, age at diagnosis, age at study recruitment, and kidney survival at the last follow-up were collected for all patients. Patients were considered to have reached KF at the commencement of dialysis or received a preemptive kidney transplant. Creatinine measurement was performed in clinical laboratories, and the eGFR was calculated using the CKD Epidemiology Collaboration (eGFR CKD-EPI, ml/min per 1.73 m²/year) equation.¹⁸ eGFR measurements were collected for all patients with CKD. Linear regression was used to calculate annual eGFR decline from at least 2 creatinine measurements over 6 months apart. However, for those patients who had progressed to KF for >10 years, data of eGFR decline were unavailable.

TKV was estimated using computed tomography or magnetic resonance imaging. Kidney imaging was acquired for routine clinical indications prior to KF. A trained physician and experienced radiologist measured TKV by manual segmentation.¹⁹ Integrating age and height, the Mayo Clinic imaging classification (MCIC) of ADPKD was used to stratify patients into 5 risk classes (1A–1E).²⁰ In the absence of computed tomography or magnetic resonance imaging data, kidney ultrasound (US) was used to measure the kidney length of patients. US scans were collected by reviewing medical records, and kidney length was measured in a sagittal plane.

For all patients with sufficient clinical and genetic data, we calculated the PROPKD score to stratify KF risk into low-risk, intermediate-risk, and high-risk subgroups.²¹

Genetic Analysis

Next-generation sequencing technologies were utilized to identify pathogenic variants in PKD1 and PKD2 in all probands; gene panels, and exome-sequencing as previously described.^22,23 Gap-filling long-range polymerase chain reaction, Sanger sequencing, and multiplex ligation-dependent probe amplification were applied in a minority of patients with no pathogenic variant detected on targeted next-generation sequencing (n = 13), as described previously.²² The pathogenicity of ADPKD variants was evaluated using the American College of Medical Genetics and Genomics guidelines.²⁴ Pathogenic and likely pathogenic variants were categorized based on their functional impact as PT variants (frameshift, canonical splice, nonsense, and, in a small percentage of cases, copy number variants) and NT variants (missense and nonframeshift indel variants). Likely-causative PKD variants of uncertain significance, likely upgradable to pathogenicity upon familial segregation, were included.

Evaluation of Disease Severity of ADPKD

Age at onset of KF, the annual decline in eGFR, the PROPKD score, and imaging classification according to MCIC and US criteria were used to quantify disease severity. We stratified patients into 3 categories of disease severity as adopted by Lanktree et al..¹⁴ These categories were as follows:

• •Severe disease: Defined as patients who reached KF before the age of 55 years, or with eGFR annual decline >5 ml/min/annum, or with PROPKD score >6, or with MCIC class 1D or 1E, or with a kidney length on US >16.5 cm at age <45 years.
• •Mild disease: Defined as patients who developed KF later than the age of 70 years, or with PROPKD score ≤3 at an age later than 35 years, or with MCIC risk class 1A or 1B.
• •Intermediate Disease: Failed to meet the criteria for either mild or severe disease.

The primary objective was to assess familial variability in ADPKD by evaluating disease severity; familial concordance or discordance of onset of KF; the PROPKD score; and MCIC risk class for each participant with severe, intermediate, or mild disease, whenever available. Concordance was defined as the agreement of the same severity threshold within all family members, whereas discordance was defined as having varying severity in families across all 3 categories, including the intermediate. We defined marked intrafamilial variability in disease severity as the presence of at least 1 severe and 1 mild case per family, excluding the intermediate category.

Statistical Analysis

Standard descriptive statistics were expressed as frequencies, mean (± SD), and median (interquartile range). Chi-squared tests were used to compare cohort genotype (disease-causing genes and allelic type [PT and NT variants]) and degree of intrafamilial variability, whereas the t-test or Mann–Whitney test was used for continuous variables, as appropriate. As a measure of variability dispersion, the familial coefficient of variation was calculated as the ratio of the SD to the mean of age at KF (coefficient of variation % = SD/mean).

Using Cox proportional-hazards models, the factors associated with disease progression were examined. A multivariate mixed-effects model with random effects for family identification was used to account for intrafamilial variation in kidney survival by clustering within families, where multiple members are interrelated. This model included fixed and random effects, often referred to as frailty effects in a Cox Model. If significant, this model evaluates the independence of the shared random effect within families (i.e., the variability), which were assumed to be randomly distributed across family groups. The probability of a type 1 error (P value) of less than 0.05 was deemed significant. All tests were performed using STATA SE software version 16.0 (Stata Corp, College Station, TX) and R (https://www.r-project.org/).

Results

Population

Five hundred twenty-two consecutive individuals with ADPKD from 313 unrelated families were recruited to the Irish Kidney Gene Project registry. Given that the prevalence of ADPKD is estimated at approximately 1:1000,² and because Ireland's population was reported to be approximately 5 million in 2021, we estimate that we recruited a minimum of 10% of ADPKD-affected individuals in Ireland. Clinical data was available for 292 individuals from 92 unrelated (to the best of our knowledge) ADPKD families with at least 2 confirmed affected individuals (Figure 1a). There was a slight nonsignificant excess of females (162/292; 55.5%) over males (130/292; 44.5%). Half the families (n = 45) contributed 3 or more individuals to this analysis. The demographics of the entire ADPKD cohort, stratified by the underlying genotype, are detailed in Table 1.

Figure 1.

ADPKD families in the Irish Kidney Gene Project. (a) Study flow chart displaying the number of ADPKD families included in this cohort. Seventy-five of 522 (14.4%) recruited ADPKD cases reported a family history of PKD; however, no additional family members were recruited; 58 cases of PKD1 and 17 cases of PKD2. The non-ADPKD phenotypes of 8 probands (PKHD1 (n = 3), MAP2K2 (n = 2), PRKCSH (n = 1), and COL4A1 (n = 1)) were excluded. Sporadic cases were defined as those with no apparent familial history of PKD or it being unknown or unreported at the time of analysis. (b) The sunburst chart depicts the proportion of participants per family with their sequencing status. The percentages next to the variants represent the proportion of affected families. CNV, copy number variants; LP, likely-pathogenic variants; NFS, nonframeshift variants; NMD, no mutation detected; NT, nontruncating; P, pathogenic; PT, protein-truncating variants; VUS, variant of uncertain significance

Table 1.

Clinical characteristics of the 292 individuals (92 families) with autosomal dominant polycystic kidney disease stratified by the underlying disease-causing variants

Patient characteristics	Total	PKD1		PKD2	Non-PKD1/PKD2, NMDa
		PKD1-PT	PKD1-NT
Families/Individuals, n	92/292	54/170	26/91	5/12	7/19
Male, (%)	44.5	46.5	41.8	41.7	42.2
Age at initial diagnosis, yr	27 (19–39.2)	21 (17–32)	29 (22–45)	48 (40.5–55.7)	29 (19.25–35)
Age at last follow-up, median (IQR), yr	55.5 (44–65)	54 (44–62)	60 (45.5–67)	63 (58.5–74)	47 (28–62)
CKD, n (%)	117 (40.1)	57 (33.5)	38 (41.8)	8 (66.7)	14 (73.7)
KDIGO Stage 1 and 2	65 (55.6)	30 (52.6)	20 (52.6)	5 (62.5)	10 (71.4)
KDIGO Stage 3	30 (25.6)	14 (24.6)	11 (28.9)	2 (25)	3 (21.4)
KDIGO Stage 4 and 5	22 (18.8)	13 (22.8)	7 (18.5)	1 (12.5)	1 (7.2)
eGFR decline (ml/year), (n = 91)	−3.4 (−5.6 to −1.5)	−3.5 (−7 to −1.5)	−3.5 (−5.5 to −1.9)	−1 (−2 to −0.5)	−2.8 (−3.2 to −1.1)
KF, n (%)	175 (59.9)	113 (66.5)	53 (58.2)	4 (33.3)	5 (26.3)
Age at start of KRT, median (IQR), yr	48 (42–55)	48 (42–54)	48 (41–58)	61.5 (56.5–69.5)	37 (28–38)
Mean difference in familial age at KF, yr	7.2 ± 6.9	6 ± 4	9 ± 6	5 ± 2	10 ± 5.1
Liver cysts, n (%)	188 (64.4)	108 (63.5)	67 (73.6)	8 (66.7)	5 (26.3)
Diagnosis of HTN, n/N (%)	236/270 (87.4)	135/152 (88.8)	83/89 (93.3)	10/11 (90.9)	8/18 (44.4)
Age at diagnosis of hypertension, yr	33 (25–42)	33 (25–42)	32 (27–44)	48.5 (39–58)	26 (20–35.5)
HTN at age < 35 year, n (%)	131 (44.9)	74 (43.5)	50 (54.9)	1 (8.3)	6 (31.6)
Urological events, n/N (%)	194/261 (74.6)	112/149 (74.8)	66/85 (78.1)	7/11 (60)	9/16 (72.7)
Age at onset of urological events, year	32 (21.5–42.5)	31 (20–42)	31.5 (21–40)	55 (48.5–63)	30 (21–44)
Urological events at age <35 year, n (%)	124 (44%)	71 (41.8)	45 (66.7)	1 (11.1)	7 (62.5)
PROPKD Score, n (%)	263 (90.1)	161 (94.7)	91 (100)	11 (68.7)	-
Low-risk score (0–3)	33 (12.5)	0 (0)	22 (24.2)	11 (100)	-
Intermediate-risk score (4–6)	139 (52.9)	85 (52.8)	54 (59.3)	0 (0)	-
High-risk score (7–9)	91 (34.6)	76 (47.2)	15 (16.5)	0 (0)	-
MCIC class, n (%)	89 (30.5)	52 (30.6)	29 (35.8)	4 (25)	4 (10)
1A and 1B	20 (22.5)	9 (17.3)	7 (24.9)	1 (25)	3 (75)
1C	28 (31.4)	16 (30.8)	11 (37.9)	1 (25)	0 (0)
1D and 1E	41 (46.1)	27 (51.9)	11 (37.9)	2 (50)	1 (25)
Genetic sequencing performed (Individuals, n)	280	159	90	12	19

CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; HTN, hypertension; IQR, interquartile range; KDIGO, Kidney Disease: Improving Global Outcome criteria; KF, kidney failure; MCIC, Mayo Clinic Imaging Class; NMD, no mutation detected; NT, nontruncating variant; PROPKD, Predicting Renal Outcome in Polycystic Kidney Disease scoring system; PT, protein-truncating variant.

^aIncludes 1 family with monoallelic IFT140 variant.

Genetic Characteristics

Data on genetic sequencing was available for 280 (95.9%) individuals, comprising 92 families in whom at least 2 family members had been sequenced. A genetic diagnosis was made in 86 (93.5%) families and 263 (93.9%) individuals. Within the subgroup where a genetic diagnosis was made, 93% of families had a genetic cause attributable to PKD1 and 5.8% PKD2. Two-thirds (67.5%; 54/80) of PKD1 families harbored PKD1-PT variants: 23 (42.6%) families had frameshift variants, 20 (37%) had nonsense, 10 (18.5%) had splice variants, and 1 (1.9%) had copy number variants; whereas PKD1-NT variants accounted for 32.5% (26/80) of families: 15 [65.2%] families have American College of Medical Genetics and Genomics pathogenic or likely pathogenic missense or nonframeshift indel variants and 11 [34.8%] families have variants of uncertain significance according to the American College of Medical Genetics and Genomics criteria. All sequenced affected family members with a genetic diagnosis carried the same disease-causing ADPKD variant. For individuals in whom sequence data were unavailable (n = 12), the genetic diagnosis was inferred to be the same as their family members with a molecular diagnosis.

Eight percent of the sequenced families were negative for PKD1/PKD2: 1 family (n = 2) harbored a pathogenic IFT140 variant, whereas 6 (6.5%) families (n = 17) had no pathogenic alternation detected. Diagnostic variants are illustrated and summarized in Figure 1b and Supplementary Table S1.

Kidney Disease Severity and Familial Variability per Metric

In order to evaluate the intrafamilial and interfamilial variability of disease severity in ADPKD, we assessed kidney disease severity and the frequency of variability for each metric. We observed a large interfamilial and intrafamilial variability in age of kidney survival according to the underlying ADPKD variants (Figure 2). At the most recent assessment, 175 individuals (59.9% of the cohort) had progressed to KF. The average age of KF was 48.5 ± 10.8 years; 73.1% (128/175) displayed severe disease, that is, KF before 55 years (Table 1). The intrafamilial and interfamilial age at onset of KF was determined for 52 families with ≥2 members diagnosed with KF. The intrafamilial mean difference in age of KF was 7 ± 4.8 years, with 15 (26.9%) families exhibiting more than a 10-year difference in the age at KF between members (Figure 3a). Families with PKD1-PT did not differ statistically from those with PKD1-NT regarding the difference in KF age among family members [6.2 ± 4 vs. 8.7 ± 6 years; P = 0.08]. The median intrafamilial variability in age at onset of KF was modest (coefficient of variation 13.3%; range: 1.1%–49.3%). In families sharing PKD1-PT variants, the variability of onset of KF was not statistically significant to families sharing PKD1-NT variants (intrafamilial coefficient of variation: 12.2%, interquartile range: 5.9%–17.8% vs. 17%, interquartile range: 10.5%–24.1%; Mann-Whitney test P = 0.164) (Figure 3b). Of the families with ≥2 members diagnosed with KF (n = 52; 143 individuals), 37 (71.2%) families (95 individuals) were concordant, that is, had "same-severity" in terms of progressing to KF less than or greater than 55 years. In contrast, 15 (28.8%) families (48 individuals) were discordant in severity, that is, having ‘different-severity’ as some family members progressing before and after 55 years. Of these, 4 families (7.7%; 13 individuals) demonstrated marked intrafamilial variability of the onset of KF, defined as having at least 1 member progressing before 55 years and 1 member at ≥70 years.

Figure 2.

Dot plot displaying the age of kidney survival within 92 families, stratified by the underlying primary ADPKD variant. The affected members of each family (vertical lines) were plotted in ascending order of families based on the earliest age of onset of kidney failure (KF) (solid red circles). At most recent follow-up, hollow green circles denote subjects who remained chronic kidney disease. For relatives with chronic kidney disease (at the most recent follow-up) who are younger than relatives with KF, the family vertical line is colored green. Horizontal reference lines represent the severity criteria of KF; solid red line denotes individuals who developed KF before age 55 years old, and were considered to have severe disease, and the green dash-line represents those with mild disease, who developed KF at ≥70 years. Families were stratified by the underlying primary ADPKD variant; PKD1-protein-truncating variants (PKD1-PT), PKD1-nontruncating (PKD1-NT); PKD2, and those with no-mutation-detected (NMD). (∗) indicates a family with a monoallelic IFT140 variant.

Figure 3.

The intrafamilial variability in mean difference in age at kidney failure and coefficient of variation. (a). Approximately one-quarter (26.9%) of families exhibited more than a 10-year difference in age at kidney failure (KF), with an average mean difference in age at KF between family members being 7 years. The inset table summarizes the number of families with a cumulative proportional difference in familial average age of KF. (b) Intrafamilial variability in coefficient of variation (CV). For each family, the familial CV was calculated as the ratio of the SD to the mean of age at KF (CV% = SD/mean). Blue to red, which was depicted on the heat scale to the right, corresponded to the lowest CV to the highest CV (i.e., variability). Families were plotted in ascending order of families based on the earliest onset of KF, where each family connected by a vertical line, and grouped by the underlying primary ADPKD variant. The median intrafamilial variation was 13.3%, with a range of 1.1% to 49.3%; however, there was no difference in the age of KF onset based on allelic type. ∗One family with no molecular diagnosis was excluded from this analysis.

At the last assessment, 40% (117/292) of individuals were free of KF and at various stages of CKD. We studied the rate of decline in eGFR and variability between family members. A total of 91 individuals had more than 2 blood tests (available to us for analysis) within a period of 7.8 ± 2.3 years, resulting in mean creatinine and eGFR of 143.5 ± 117 and 67.1 ± 35.1 respectively, and a median annual decline of 3.4 (interquartile range: 1.5–5.6 ml/year). Twenty-eight (30.8%) individuals experienced severe disease, indicating eGFR annual loss >5 ml/year. The variation in decline in eGFR is illustrated in Supplementary Figure S1. Nineteen individuals remained free of KF at their last assessment at an age beyond their proband developing KF (Supplementary Figure S2). There was a significant difference in KF-free survival by genotype as shown in Figure 4.

Figure 4.

Kaplan-Meier survival estimates of kidney survival. The Kaplan-Meier survival curve and the number of individuals at risk for each ADPKD variant group indicate overall kidney survival. The Log-rank test for comparing survival curves and a P-value of less than 0.05 indicated a statistically significant difference in the probability of KF at last follow-up. Compared to other variants, PKD1-PT is associated with a significant reduction in kidney survival (log-rank test, P < 0.0001), although the small sample size for PKD2 and no-mutation-detected results in a wide confidence interval.

The PROPKD score was calculated in 263 individuals from 85 families; 91 (34.6%) individuals had a score of >6 and were considered to have severe disease, whereas all the 33 (12.5%) individuals in the low-risk subgroup were considered to have mild disease and were older than 35 years old. In Supplementary Figure S3, we display the distribution of 83 families with 2 or more individuals with calculated PROPKD. Discordance was observed among individuals of 52 (62.7%) families, that is, having different risk subgroups to their relatives. Of these, 7 (8.4%) families demonstrated marked variability in PROPKD score, defined by 1 family member in an high-risk subgroup (score >6 points) and another family member in a low-risk group (score ≤3 points) (Supplementary Figure S4).

Overall, 236 (80.8%) individuals were treated for hypertension at age 35.1 ± 11.9 years. Of all the individuals, 131 (55.5%) were treated with antihypertensive medications at age <35 years at a mean age of 26.3 ± 4.9 years, whereas 105 (35.9%) were hypertensive at ≥35 years (mean age of 46 ± 8.6 years). Thirty-eight families had at least 2 individuals treated before 35 years, 34 (11.6%) were normotensive at the last follow-up, and 22 (7.6%) had missing data. Urological events were observed in 194 (74.6%) individuals at a mean age of 32.4 ± 14.3 years. Of the individuals, 124 (44.7%) encountered urological events before 35 years of age, whereas 70 (23.9%) experienced at least 1 urological event after 35 years (mean of 23.5 ± 7.5 and 48 ± 9.3 years, respectively). Thirty-eight families had ≥2 individuals who experienced urological events at age <35 years. Sixty-seven (22.9%) individuals had never encountered prior urological complications; data were not recorded in 31 (10.6%) individuals. The number of documented hypertension and urological events is presented in Table 1.

We assessed TKV and MCIC classes in 89 of 292 (only 30.4%) individuals. Severe disease (MCIC 1D and 1E) had a median height-adjusted TKV of 2546 (1672–3391) ml, whereas mild disease had 438 (266–750) ml (MCIC 1A and 1B). Utilizing US imaging, 16 individuals had kidney diameters greater than 16.5 cm before the age of 45 years. To explore familial variability using imaging criteria, according to MCIC and US data, 19 (76%) of the 25 families have at least 2 members with imaging assessment had concordance in severity criteria. Six (24%) of the 25 families displayed marked variability based on imaging criteria (1 severe and 1 mild case per family). The familial distribution of MCIC classes is illustrated in Figure 5.

Figure 5.

Dot plot displaying distribution of familial individuals with ADPKD according to Mayo Clinic Imaging Class (MCIC). Total number of families (x-axis) plotted in order of wide discrepancy in familial MCIC risk class (y-axis); each vertical line represents a unique family. Each symbol represents a unique patient. The first 6 families represented with marked intrafamilial variability based on imaging criteria (1 severe and 1 mild case per family). The probands of the first 3 families harbored PKD1 protein-truncating variants of PKD1, whereas the probands of the subsequent families carried protein-nontruncating variants of PKD1. All patients with high-risk MCIC class (1D or 1E) developed KF before their relatives, with the exception of the fifth family, in which the proband (with massive kidneys and MCIC class 1E) developed KF at a similar age of 39 years, despite the relative having nonenlarged polycystic kidneys with MCIC class 1B.

Familial Variability of Kidney Disease Progression

Sixty-two percent (181/292) of patients were deemed to have severe disease, 9.6% (28/292) to have mild disease, and 28.4% (83/292) to have intermediate disease. Most of our patients (90.1%) have been assessed by more than 1 metric (Supplementary Table S2). Twelve of 92 (13%) families had marked intrafamilial variability with at least 1 severe and 1 mild case (Figure 6a). Notably, compared to concordant families, discordant families had a higher proportion of families with PKD1-NT variants (8 [67%] vs. 18 [22.5%]; P = 0.004) (Table 2 and Figure 6b). In order to assess the consistency of this result in families with marked intrafamilial variability, we excluded families with PKD1-NT variants of uncertain significance in a sensitivity analysis (Supplementary Table S3). We show that sensitivity analysis yields results similar to primary analysis. In addition, discordant families had a higher proportion of larger families (91.6% vs. 43.7%; P = 0.002) and lower PROPKD scores (4.1 vs. 5.8; P = <0.0001). On the other hand, concordant families were less variable and characterized by severe disease course; hypertension at a younger age (34.3 vs. 38.8 years; P = 0.028), and earlier KF (47.3 vs. 56.2 years; P = 0.0008).

Figure 6.

The prevalence of marked intrafamilial variability of kidney disease severity across ADPKD families. (a). Variability is positively associated with family size, where large families display marked variability, defined as presence of 1 severe case and 1 mild case per family, as high as 50%. Count of families is illustrated with bar columns. Nonparametric (Cochrane-Armitage test) test for trend: Z = 4.07; P ≤ 0.001. Participants with PKD1-PT and PKD1-NT variants were younger (52.4 ± 14.2 and 56.5 ± 13.3 years) compared to patients with PKD2 (64.4 ± 12.4 years) (1-way analysis of variance; F, 5.93; P = 0.003). (b) Based on the underlying allelic type of ADPKD, families with marked variability of disease severity were more likely to harbor PKD1-NT variants compared to other variants; the Pearson chi-square test is 15.6; P = 0.001. Error bars indicate SD across each group. PT, protein-truncating variant, NT, nontruncating variant, NMD, no mutation detected

Table 2.

Characterization of ADPKD familial variability of disease severity

Variables	Marked intrafamilial variability	Concordant families	P-value
Number of families (number of individuals)	12 (54)	80 (238)	-
2 members per family, N (%)	1 (8.3)	45 (56.25)	0.002
More than 2 members per family, N (%)	11 (91.6)	35 (43.75)
Individuals reach KF, N (%)	24 (44.4)	151 (63.5)	0.013
Age at KF	56.20 ± 11.61	47.28 ± 10.26	0.0008
Mean difference of age at KF	14.12 ± 5.67	6.39 ± 3.94	<0.0001
Age at onset of hypertension	38.80 ± 12.65	34.31 ± 11.65	0.028
Age at onset of urological events	36.11 ± 15.23	31.58 ± 14.06	0.094
PROPKD Score	4.13 ± 1.9	5.8 ± 1.9	<0.0001
eGFR, mean ± SD	72.2 ± 6.3	65.3 ± 3.77	0.363
eGFR Decline, mean ± SD	2.9 ± 2.4	3.7 ± 3.1	0.237
TKV	2217.25 ± 3194	1685.88 ± 1069	0.242
PKD1-PT, N (%)	2 (16.6)	52 (65)	0.004
PKD1-NT, N (%)	8 (66.6)	18 (22.5)
PKD2, N (%)	2 (16.6)	3 (3.75)
NMD, nonPKD1/PKD2. N (%)	0 (0)	7 (8.75)

eGFR, estimated glomerular filtration rate; KF, kidney failure; MCIC, Mayo Clinic Imaging Class; NMD, no mutation detected; NT, nontruncating variant; PROPKD, Predicting Renal Outcome in Polycystic Kidney Disease scoring system; PT, protein-truncating variant

In this analysis, marked familial variability was defined as families where 1 or more individuals within that family are discrepant in terms of disease severity as defined by the following: (i) KF at age <55, (ii) MCIC 1 D and 1 E, (iii) eGFR decline of >5 ml/year, and (iv) PROPKD score >6 points. Compared to concordant families, defined as the agreement of the same severity threshold within all family members, the intrafamilial difference in the age at KF was wider in the 12 discordant families (14.12 ± 5.67 vs. 6.39 ± 3.94 years). Discordant families were more likely to have PKD1 nontruncating variants.

Impact of Relatedness on Variability of Disease Severity

We used Cox proportional hazard regression models to predict disease progression to KF across our cohort (Table 3). Univariate analysis showed that among the prognostic variables, male individuals, early-onset hypertension, early-onset urological events, and MCIC class were significant risk factors for KF. Compared to PKD1-PT variants, PKD1-NT, PKD2, and non-PKD1/PKD2/no-mutation-detected conferred reduced risk of progression into KF. These observations were held for the multivariate Cox model (MCIC class was not included due to limited available data). In the multivariate mixed-effects model that accounts for intrafamilial variability (the shared random effect frailty model), we confirmed that intrafamilial variability in kidney survival was independent of the measured genotypic factors associated with the kidney survival (P = <0.001, Table 3).

Table 3.

Cox proportional hazard model (shared frailty model) for familial variability of ADPKD disease severity

	Univariate analysis			Multivariate analysis
Variables	HR (95% CI)	P-value (Cox regression)	P-value (shared frailty)	HR (95% CI)	P-value (Cox regression)	P-value (shared frailty)
Male	1.57 (1.10–2.24)	0.013	<0.001	1.60 (1.10–2.33)	0.01	<0.001
Urological events at age <35 year	2.97 (2.04–4.35)	<0.001	<0.001	2.43 (1.61–3.36)	<0.001
HTN at age <35 year	2.42 (1.19–4.90)	0.014	<0.001	2.20 (1.45–3.34)	<0.001
MCIC Risk Class (1D and 1E)	2.69 (1.69–4.29)	<0.001	<0.001	-
PKD1-PT	Reference		<0.001	Reference
PKD1-NT	0.56 (0.39–0.79)	0.001		0.53 (0.31–0.90)	0.02
PKD2	0.23 (0.11–0.48)	<0.001		0.13 (0.04–0.45)	1.30 × 10−3
NMD	0.41 (0.24–0.72)	0.002		0.57(0.16 –1.98)	0.37

CI, confidence interval; HR, hazard ratio; HTN, hypertension; MCIC, Mayo Clinic imaging class; NMD, no mutation detected; NT, nontruncating variant; PT, protein-truncating variant

Cox proportional hazards univariate and multivariate models (nonfrailty models) were employed to investigate the association between predictor variables and the kidney survival time. We also employed a multivariate mixed-effects model with random effects for family identification in order to account for intrafamilial variation in kidney survival. We employed a mixed-effect model to account for clustering within families, that is, multiple members within the same families have an interrelatedness. This model included fixed effects and random effects often referred to as frailty effects in a Cox model. If significant, this model evaluates the independence of the shared random effect within families (i.e., the variability), which were assumed to be randomly distributed across family groups. MCIC class was not included in the multivariate model due to limited available data. In essence, we modelled the variability of the survival function across family groups. The assumption of proportional hazard was also tested and confirmed.

Discussion

In day-to-day clinical practice, we have observed significant variation in the severity of kidney disease between and within ADPKD families; therefore, we conducted this large observational study in an Irish population to better understand the phenotype- and genotype-derived familial variability in the severity of the kidney phenotype. Our study corroborates and extends previous findings trifold. First, using modified 4-metric criteria to assess disease severity,¹⁴ familial discordance (having different severity criterium) varied according to each metric: 28.8% for age at KF, 62.7% for the PROPKD score, and 24% for MCIC metric. We also demonstrated marked intrafamilial variability per each metric (at least 1 family member with severe disease and 1 member with mild disease): 7.7% for age at KF, 8.4% of the PROPKD score, and 24% of the MCIC risk class. Only typical kidney imaging patterns of PKD (class 1) were identified from the available TKV data in our cohort. This is likely attributed to the positive family history and detected variants of PKD1 and PKD2.²⁵ Second, our cohort's average age at KF was significantly younger (48.9 ± 10.9 years) compared to other large cohorts.^6,21,26 As a measure of disease severity, 27% of families had a difference in KF age of >10 years. Thirdly, 13% of ADPKD families had marked intrafamilial variability, with a higher proportion in families with PKD1-NT variants.

Until now, predictors of familial variability in ADPKD come primarily from 2 studies. In a study of pairs with ADPKD, which included 9 monozygotic twins, the variability of KF was reported to be 6.9 years, compared to our results of 7 years.⁸ Of importance, Persu et al. found much reduced intrafamilial variability in KF age between monozygotic twins of 2.1 years, suggesting a greater influence of background genomic factors, including modifier genes, on kidney disease progression. Using a modified 4-metric criterion, our cohort has a comparable proportion of marked intrafamilial variability to the large eTGESP cohort (13% vs. 12%).¹⁴ However, we observed a more significant proportion of PKD1-NT variants in families with marked intrafamilial variability than in concordant families (66.6%, 8/12 vs. 22.5%, 18/80). This result remained significant in a sensitivity test after excluding families with PKD1 variants of uncertain significance. These observations may be explained by cohort differences in phenotypic severity; combined with the clustering of PKD1-enriched families in our cohort, with an increased PKD1:PKD2 ratio (18:1), no missense PKD2 families included, and 60% of patients progressed to KF, which may have contributed to this disparity; suggesting that patients with more severe phenotypes were overrepresented in our cohort. A more descriptive cohort comparison is presented in Supplementary Table S4. Because of the complexity of variant interpretation in ADPKD, especially missense variants, the ClinGen cystic kidney disease variant curation expert panel is developing a version of optimized American College of Medical Genetics and Genomics guidelines for ADPKD variants pathogenicity.²⁷

One might hypothesize that disease onset is a monogenic trait whereas disease progression is multifactorial influenced by both genes and the environment to produce either severe or mild phenotypes. The genetic background, somatic variants, and other stochastic factors, such as acute kidney injury, have been reported to influence cystogenesis and hypothetical disease severity.28, 29, 30 Our shared random effect frailty model confirms a strong familial effect on known progression markers. Studying family members "protected" from KF compared to their siblings may reveal "protective" pathways that could be used as therapeutic targets.²⁶

Although studies have shown that tens or hundreds of loci can affect CKD progression and other phenotypes,31, 32, 33, 34, 35 polygenic scores enable the quantification and integration of effects across multiple common variants to a single metric for risk stratification of relevant phenotypes, such as lower eGFR, hypertension, and cardiovascular diseases.36, 37, 38 Recent research in coronary artery disease demonstrates that polygenic scores can affect treatment response.³⁹ This may also be applicable to ADPKD.

We have used 4 clinically-relevant disease severity estimates to understand intrafamilial variability better, assessing a large proportion of the Irish ADPKD population (at least 10% of ADPKD-affected individuals). We do, however, acknowledge several limitations. It is important to note that each metric has its limitations, and the performance of disease-severity determinants requires additional work.⁵ Considering that data were collected for routine clinical care, we lacked eGFR decline data on subjects who developed KF more than 10 years ago and TKV data on only a third of patients. It makes sense to predict familial variability using disease progression markers before KF; a more individualized approach to assess ADPKD severity is imperative.⁴⁰ We have not performed focused analyses of somatic or mosaic variants to account for within-family variation Supplementary Table S5.

Conclusions

We have observed marked intrafamilial variation in the ADPKD disease phenotype in at least 13% of studied families, with a higher frequency in families with PKD1-NT variants. Our shared random effect frailty model shows the intrafamilial impact of ADPKD disease severity. Genomic analysis of secondary genetic factors in the ADPKD genes or gene-gene interactions might explain this familial variation.

Disclosure

EAEE reports funds from the Royal College of Surgeons in Ireland StAR PhD. GLC reports receipt of research funding from BioMarin. KAB and PJC report funds in part by the Health Research Board and Irish Nephrology Society under the HRCI-HRB Joint Funding Scheme HRCI-HRB-2020-032. KEC is supported by a research grant from Science Foundation Ireland (SFI) under Grant number 18/CRT/6214. All the other authors declared no competing interests.

Supplementary Material

Figure S1. Proportion of average estimated glomerular filtration rate decline across 98 individuals with ADPKD with chronic kidney disease.

Figure S2. Familial distribution of individuals with ADPKD at various chronic kidney disease status at age beyond index patients with kidney failure.

Figure S3. Dot plot displaying the familial distribution of PROPKD score (83 families).

Figure S4. Distribution of ADPKD families with marked variability in the PROPKD score.

Table S1. Molecular genetic diagnosis in the 92 families in whom disease-causing ADPKD variants were identified.

Table S2. Proportion of patients with severe, intermediate, or mild kidney disease according to the criteria for defining ADPKD disease severity.

Table S3. Sensitivity analysis of the characteristics of ADPKD familial variability of disease severity.

Table S4. Characteristics of included studies of familial variability of ADPKD utilizing prespecified criteria for assessing disease severity.

Table S5. List of genes sequenced included in the in silico gene panel using whole exome sequencing.

STROBE Statement (PDF)