Timing of Kidney Replacement Therapy among Children and Young Adults

Visual Abstract

Abstract

Background

No randomized trials exist to guide the timing of the initiation of KRT in children. We sought to define trends and predictors of the eGFR at initiation of KRT, center-related clinical practice variation, and any association with patient survival.

Methods

Children and young adults (1–25 years) commencing KRT (dialysis or kidney transplantation) between 1995 and 2018 were included using data from the Australia and New Zealand Dialysis and Transplant Registry. The associations between eGFR on commencing KRT and covariates were estimated using quantile regression. Cox regression was used to estimate the association between eGFR and patient survival. Logistic regression, categorizing eGFR about a value of 10 ml/min per 1.73 m², was used in conjunction with a random effect by center to quantify clinical practice variation.

Results

Overall, 2274 participants were included. The median eGFR at KRT initiation increased from 7 to 9 ml/min per 1.73 m² over the study period and the 90th centile from 11 to 17 ml/min per 1.73 m². The effect of era on median eGFR was modified by modality, with a greater increase among those receiving a preemptive kidney transplant (1.0 ml/min per 1.73 m² per 5 years; 95% confidence interval [CI], 0.6 to 1.5) or peritoneal dialysis (0.7 ml/min per 1.73 m² per 5 years; 95% CI, 0.4 to 0.9) compared with hemodialysis (0.1 ml/min per 1.73 m² per 5 years; 95% CI, −0.1 to 0.3). There were 252 deaths (median follow-up 8.5 years, interquartile range 3.7–14.2) and no association between eGFR and survival (hazard ratio, 1.01 per ml/min per 1.73 m²; 95% CI, 0.98 to 1.04). Center variation explained 6% of the total variance in the odds of initiating KRT earlier. This rose to over 10% when comparing pediatric centers alone.

Conclusions

Children and young adults progressively commenced KRT earlier. This change was more pronounced for children starting peritoneal dialysis or receiving a preemptive kidney transplant. Earlier initiation of KRT was not associated with any difference in patient survival. A substantial proportion of clinical practice variation was due to center variation alone.

Podcast

This article contains a podcast at https://dts.podtrac.com/redirect.mp3/www.asn-online.org/media/podcast/CJASN/2023_08_08_CJN0000000000000204.mp3

Introduction

There is no consensus regarding the optimal timing to commence KRT in children with kidney failure. The Initiating Dialysis Early and Late (IDEAL) trial, the only randomized trial conducted to date addressing this question, included 828 adults and found no difference in survival or other clinical outcomes between the two groups.¹ After the publication of the IDEAL trial, there has been a reversal in the previously observed trend toward earlier initiation of dialysis among adults.² While indications for dialysis remain similar throughout life, there are technical differences, along with additional growth and neurodevelopmental considerations, that make it important to establish the benefits and risks specifically among children and young adults.

The observational data that have been published suggest no benefit or a higher risk among children with a higher eGFR at initiation of KRT. Two retrospective cohort studies using the United States Renal Data System showed that a higher eGFR at dialysis initiation was associated with lower survival.^3,4 Data from the European Society of Pediatric Nephrology/European Renal Association and European Dialysis and Transplant Association Registry also found no improvement in health outcomes with earlier initiation of KRT, including growth, over the first year of therapy.⁵ Despite these findings, a trend to earlier dialysis has been observed in the United States and Canada.⁶ There are little data on whether trends observed among older adults apply to younger patients treated at adult centers or whether they extend to children receiving a kidney transplant as their initial KRT modality.

The Australia and New Zealand Dialysis and Transplant (ANZDATA) Registry has the benefit of complete capture of all people treated for kidney failure across both countries.⁷ Using this binational registry, we aimed to examine temporal trends and practice variation in the timing of the initiation of KRT, including preemptive transplantation, and any associations with patient survival among children and young adults.

Methods

Population

All children and young adults from 1 to 25 years commencing KRT in Australia and New Zealand between 1998 and 2018 were included. All pediatric and tertiary adult nephrology units in Australia and New Zealand contribute data to the ANZDATA Registry, which collects participant data from the initiation of maintenance KRT and operates on an opt-out consent policy.⁷ Children younger than 1 year were excluded because estimates of eGFR are unreliable in this age group.⁸ We chose to include patients up to 25 years because young adults are rarely examined in studies of adult populations, and they served as a reference when comparing practice between pediatric and adult centers. The clinical and research activities being reported are consistent with the principles of the Declaration of Istanbul as outlined in the Declaration of Istanbul on Organ Trafficking and Transplant Tourism. This study was approved by the University of Western Australia Human Research Ethics Committee.

Outcome

eGFR, expressed as ml/min per 1.73 m², at dialysis initiation was calculated using the u25 Chronic Kidney Disease in Children formula.^9,10 Where eGFR was categorized, we defined an early start as eGFR ≥10 ml/min per 1.73 m², consistent with thresholds used in other studies. Children who initiated KRT with eGFR ≥30 ml/min per 1.73 m² were removed from the dataset on the assumption that the timing of KRT was driven by factors aside from eGFR, such as tumor resection or nephrosis.

Explanatory Variables

Body mass index (BMI) was calculated and normalized for each child's age and sex using the International Obesity Task Force (2–18 years old) centile curves, the 95th centile of which passes through a BMI of 30 kg/m² at 18 years of age, which was used to define obesity thereafter. A weight for length ≥95th centile was used to define obesity among children younger than 2 years. A period of <3 months between referral to a nephrologist and dialysis initiation was considered late referral. The dialysis start date was recorded as the commencement of dialysis rather than creation of access. When grouping etiology, a primary cause that included reflux or renal dysplasia or posterior urethral values or bladder malformations was considered congenital, polycystic kidney disease and nephronophthisis were considered cystic, and other causes that were not glomerulonephritis were categorized as such. Acute kidney injury, including hemolytic uremic syndrome and congenital nephrotic syndrome, were categorized as other. The rural or remote variable was coded using the Australian Bureau of Statistics Remoteness Area Classification and postcode data (available for Australian participants only).

Statistical Analyses

Categorical variables were presented as n (%) and continuous variables as mean (SD) or median (interquartile range [IQR]), depending on skewness. Differences in proportions were compared using the chi-squared test. Quantile regression was used to determine whether potential explanatory variables were associated with eGFR at initiation of KRT. To adjust for important confounding, we adjusted for era and modality, followed by age and era and modality, comparing estimates of effect and variance about the estimate between models (a change in the effect size >25% being considered important). Sensitivity analyses included constructing a logistic regression model with covariate selection by least absolute shrinkage and a selection operator (lasso), the exclusion of children who received a preemptive kidney transplant, and testing for significant interaction between explanatory variables. Quantile regression models with splines (polynomial, b-spline with three internal knots) were included to test for linearity for the continuous variables era and age. Cox regression was used to determine the association between eGFR and patient survival, with simple and multivariable models constructed. Age and era were included in the survival model using nonlinear terms a priori (natural cubic spline with three internal knots), on the basis of previous data.¹¹ A Fine–Gray model was then constructed for the survival of children on dialysis considering transplantation as a competing risk. We used logistic regression to quantify the contribution of center to eGFR at initiation of KRT by including center as a random intercept term. A two-sided P-value of 0.05 was considered to denote statistical significance for explanatory variables of interest and a priori hypotheses. Exploratory and sensitivity analyses were tested at an α threshold of 0.01. All analyses were performed using R 4.0 (R Core Team, Vienna, Austria); packages used are listed in the Supplemental Material.

Results

Participant Characteristics

During the study period, 2354 children and young adults started KRT. Of these, 2296 (98%) had sufficient data included to calculate an eGFR. A further 22 (1%) participants had an eGFR at initiation of KRT ≥30 ml/min per 1.73 m² and were excluded. The median serum creatinine at initiation of KRT of the 2274 participants was 8.0 mg/dl (IQR, 5.8–11.3 mg/dl), corresponding to a median eGFR of 8 ml/min per 1.73 m² (IQR, 6–11 ml/min per 1.73 m²).

Overall, there was a slight male predominance, with congenital anomalies the cause of kidney failure in 660 participants (29%) (Table 1). The median age at initiation of KRT was 20 years (IQR, 14–23 years). The median height, weight, and BMI z-scores at initiation of KRT among participants younger than 18 years were −1.1 (IQR, −2.1 to −0.1), −0.6 (IQR, −1.5 to 0.4), and 0.4 (IQR, −0.5 to 1.2), respectively. Underweight (defined by BMI z-score < −2 among participants younger than 18 years and BMI <17 kg/m² among participants 18 years and older) was present in 98 (4%) participants at initiation of KRT (3% of children younger than 5 years). First Nations and Pacific People were over-represented, consistent with a known excess of kidney disease among these groups.

Table 1.

Characteristics of children and young adults at initiation of KRT

Characteristic	n (%) N=2274
eGFR (ml/min per 1.73 m2)a	8 (6–11)
Sex (male)	1314 (58)
Age (yr)
1–5	189 (8)
6–9	165 (7)
10–13	203 (9)
14–18	495 (22)
19–22	654 (29)
23–25	568 (25)
Etiology
Congenital	660 (29)
Cystic	146 (6)
Glomerulonephritis	996 (44)
Other	464 (20)
Ethnicity b
Aboriginal	130 (6)
Māori	147 (6)
Pacific	138 (6)
Other	1930 (82)
Obesity	373 (17)
Pediatric center	744 (33)
New Zealand center	463 (20)
Rural or remotec	540 (30)
Late referrald	629 (28)
Modality
HD	1227 (54)
PD	730 (32)
Transplant	317 (14)
Era
1998–2002	509 (22)
2003–2007	513 (23)
2008–2012	570 (25)
2013–2018	682 (30)

Missing data <5% for all variables. Aboriginal, Aboriginal and/or Torres Strait Islander People; HD, hemodialysis; PD, peritoneal dialysis.

^aMedian (interquartile range).

^bWhen restricted to New Zealand participants, there were 131 (28%) Māori and 79 (17%) Pacific People included in the cohort.

^cValid for Australian participants only.

^dLate referral <3 months between referral and dialysis initiation.

Temporal Trends and Predictors of eGFR at Initiation of KRT

In 1998 and 1999, 31 participants (16%) started KRT with an eGFR ≥10 ml/min per 1.73 m² compared with 108 (43%) in 2017 and 2018. In these same periods, four (2%) and 40 (16%) participants, respectively, started KRT with an eGFR ≥15 ml/min per 1.73 m². The effect of era on timing eGFR at initiation of KRT is further illustrated in Figure 1 for the cohort as a whole and by relevant subgroups (age, sex, modality, and etiology) in Figure 2. During the study period, the median eGFR at KRT initiation increased from 7 to 9 ml/min per 1.73 m² (28% increase), and the 90th centile of eGFR at initiation of KRT increased from 11 to 17 ml/min per 1.73 m² (56% increase). This apparent discordance in rate of change between the median and 90th quantile of eGFR was reflected in a significant difference between slopes in the regression model (P < 0.001). As illustrated in Figure 2, the change in median eGFR by era was greater for transplant and peritoneal dialysis (PD) recipients than in children receiving hemodialysis (HD) as their initial KRT modality. The increase per 5 years for HD was 0.1 ml/min per 1.73 m² (95% confidence interval [CI], −0.1 to 0.3), compared with 0.7 ml/min per 1.73 m² (95% CI, 0.4 to 0.9) for PD and 1.0 ml/min per 1.73 m² (95% CI, 0.6 to 1.5) for transplantation (interaction P < 0.001). There was no difference in the 90th centile by era stratified by modality (interaction P = 0.7).

Figure 1.

Trend in eGFR at initiation of KRT by year. The higher line is the 90th quantile trend and the lower line is the 50th quantile trend (median), both estimated using a b-spline (polynomial) with three internal knots. A slight variation about the x-axis values has been introduced to minimize the number of overlapping points. Figure 1 can be viewed in color online at www.cjasn.org.

Figure 2.

Trend in eGFR at initiation of KRT by year and subgroup. Bars represent the 10th to 90th centiles for that group, with a dot for the median value. Figure 2 can be viewed in color online at www.cjasn.org.

Most of the potential explanatory variables were associated with timing of KRT, except for obesity and geography (Tables 1 and 2). The strongest unadjusted predictors were late referral to a nephrologist (lower eGFR), transplant as the initial modality (higher eGFR), age (higher eGFR with increasing age), and year (higher in more recent era) (Tables 2 and 3, column 1). The remaining predictors were associated with only a small difference in median eGFR (<2 ml/min per 1.73 m²). Adjustment for era and modality had the greatest effect on the effect sizes for etiology and initial treatment at a pediatric center (Tables 2 and 3, column 2). Further adjustment for age at initiation of KRT (Tables 2 and 3, column 3) led to substantial changes in the estimates for PD as an initial modality and initial treatment at a pediatric center, the latter being associated with a higher eGFR after adjustment. Although both these variables were correlated with age, the variances about estimates of effect remained stable, consistent with a stable model and lack of significant collinearity. Restricting the analysis to participants between 14 and 20 years (removing the more collinear sections of the distribution) made no material difference to the results. There was little difference in the results when restricting analyses to those who received PD or HD only (Table 3).

Table 2.

Predictors of median eGFR at initiation of KRT

Predictor	β ml/min per 1.73 m2 (95% CI)	Adjusted for Era and Modality	Adjusted for Age, Era, and Modality
Sex (male)	0.8 (0.5 to 1.1)	0.5 (0.2 to 0.9)	0.5 (0.2 to 0.8)
Age (per 5 yr)	0.4 (0.3 to 0.5)	0.6 (0.5 to 0.7)	0.6 (0.5 to 0.7)
Etiology
Congenital	Ref	Ref	Ref
Cystic	−0.7 (−1.6 to 0.0)	−0.5 (−0.9 to 0.1)	0.0 (−0.8 to 0.9)
Glomerulonephritis	−1.0 (−1.2 to −0.6)	−0.3 (−0.6 to 0.0)	−0.7 (−1.3 to −0.4)
Other	−0.5 (−0.9 to 0.0)	0.0 (−0.5 to 0.4)	−0.2 (−0.7 to 0.5)
Ethnicity
Other	Ref	Ref	Ref
Aboriginal	−1.2 (−1.8 to −0.5)	−0.7 (−1.4 to −0.2)	−0.8 (−1.7 to −0.1)
Māori	−1.0 (−1.5 to −0.6)	−0.6 (−1.1 to 0.0)	−0.7 (−1.2 to −0.1)
Pacific	−1.6 (−2.4 to −1.2)	−1.5 (−2.0 to −1.0)	−1.7 (−2.1 to −1.0)
Obesity	−0.1 (−0.5 to 0.5)	0.0 (−0.5 to 0.2)	−0.1 (−0.5 to 0.2)
Pediatric center	−0.6 (−1.0 to −0.2)	−1.1 (−1.4 to −0.8)	0.4 (0.1 to 1.0)
Rural or remote	−0.2 (−0.6 to 0.2)	−0.2 (−0.5 to 0.1)	−0.2 (−0.5 to 0.1)
Late referrala	−2.0 (−2.2 to −1.7)	−1.7 (−2.0 to −1.2)	−1.8 (−2.1 to −1.4)
Modality
HD	Ref	Ref	Ref
PD	0.6 (0.3 to 0.9)	0.5 (0.2 to 0.8)	1.0 (0.7 to 1.3)
Transplant	2.5 (2.2 to 3.1)	2.6 (2.0 to 3.1)	3.0 (2.4 to 3.6)
Era (per 5 yr) b	0.4 (0.3 to 0.5)	—	—
HD (per 5 yr)	—	0.1 (−0.1 to 0.3)	0.2 (−0.1 to 0.4)
PD (per 5 yr)	—	0.7 (0.4 to 0.9)	0.7 (0.4 to 0.9)
Transplant (per 5 yr)	—	1.0 (0.6 to 1.5)	0.9 (0.5 to 1.4)

CI, confidence interval; Aboriginal, Aboriginal and/or Torres Strait Islander People; HD, hemodialysis; PD, peritoneal dialysis.

^aLate referral <3 months between referral and dialysis initiation.

^bInteraction between era and modality expressed as combination of slopes with CI by bootstrapping.

Table 3.

Predictors of median eGFR at initiation of dialysis

Predictor	β per ml/min per 1.73 m2 (95% CI)	Adjusted for Era and Modality	Adjusted for Age, Era, and Modality
Sex (male)	0.6 (0.3 to 1.1)	0.5 (0.2 to 0.9)	0.4 (0.1 to 0.8)
Age (per 5 yr)	0.4 (0.3 to 0.5)	0.5 (0.4 to 0.6)	0.5 (0.4 to 0.6)
Etiology
Congenital	Ref	Ref	Ref
Cystic	−0.6 (−1.5 to 0.2)	−0.5 (−1.0 to 0.3)	−0.2 (−0.8 to 1.0)
Glomerulonephritis	−0.5 (−1.0 to 0.0)	−0.4 (−0.7 to 0.0)	−0.7 (−1.1 to −0.4)
Other	−0.3 (−0.7 to 0.4)	−0.1 (−0.6 to 0.2)	−0.2 (−0.8 to 0.4)
Ethnicity
Other	Ref	Ref	Ref
Aboriginal	−0.8 (−1.5 to −0.1)	−0.8 (−1.3 to −0.2)	−0.6 (−1.5 to −0.2)
Māori	−0.6 (−1.2 to −0.2)	−0.6 (−1.1 to 0.0)	−0.9 (−1.1 to −0.1)
Pacific	−1.4 (−2.1 to −0.9)	−1.5 (−2.0 to −0.9)	−1.6 (−2.1 to −0.9)
Obesity	0.1 (−0.3 to 0.8)	0.1 (−0.3 to 0.5)	0.0 (−0.3 to 0.3)
Pediatric center	−0.6 (−1.0 to −0.1)	−0.8 (−1.2 to −0.5)	0.5 (0.0 to 1.2)
Rural or remote	−0.2 (−0.6 to 0.2)	−0.2 (−0.6 to 0.1)	−0.2 (−0.5 to 0.1)
Late referrala	−1.7 (−2.1 to −1.4)	−1.7 (−2.0 to −1.2)	−1.8 (−2.2 to −1.5)
Modality
HD	Ref	Ref	Ref
PD	0.6 (0.3 to 0.9)	0.5 (0.2 to 0.8)	0.8 (0.5 to 1.1)
Era (per 5 yr)	0.3 (0.2 to 0.4)	—	—
HD (per 5 yr)	—	0.1 (−0.1 to 0.3)	0.1 (−0.1 to 0.4)
PD (per 5 yr)	—	0.7 (0.4 to 0.9)	0.7 (0.5 to 0.9)

CI, confidence interval; Aboriginal, Aboriginal and/or Torres Strait Islander People; HD, hemodialysis; PD, peritoneal dialysis.

^aLate referral <3 months between referral and dialysis initiation.

KRT Initiation as a Predictor of Survival

There were 252 deaths (11%) among the cohort during a median follow-up of 8.5 years (IQR, 3.7–14.2 years). The actuarial 1- and 5-year patient survivals for the cohort were 98% (95% CI, 98 to 99) and 93% (95% CI, 92 to 94), respectively. Participants whose initial modality was transplantation had a 5-year survival of 98% (95% CI, 97 to 100), compared with 95% (95% CI, 94 to 97) among those who started PD and 90% (95% CI, 88 to 92) among those who started HD. Of those who started PD or HD, 1432 (73%) progressed to transplantation after a median of 2.0 years (95% CI, 1.9 to 2.2).

eGFR at commencement was not associated with survival (hazard ratio 1.01 per ml/min per 1.73 m²; 95% CI, 0.98 to 1.04) (Figure 3, Supplemental Figure 2). After adjustment for age, year of initiation of KRT, modality, ethnicity, and etiology, there was a higher risk of death associated with earlier initiation of KRT (adjusted hazard ratio [aHR] 1.04 per ml/min per 1.73 m²; 95% CI, 1.01 to 1.07). The results were similar considering only those receiving dialysis as their initial modality, with transplantation as a competing risk and adjusting for the same covariates (aHR 1.06 per ml/min per 1.73 m²; 95% CI, 1.02 to 1.09).

Figure 3.

Survival by eGFR at initiation of KRT. eGFR about a threshold of 10 ml/min per 1.73 m². When retained as a continuous covariate, the relationship was linear (i.e., not threshold-dependent). Shaded area represents 95% confidence interval about the Kaplan–Meier estimator. Figure 3 can be viewed in color online at www.cjasn.org.

Center Variation

There was substantial center variation in the timing of KRT (Figure 4). To quantify this further, we constructed a logistic regression model with a random intercept term to allow clustering by center for the outcome of initiation of KRT at a lower eGFR (eGFR <10 ml/min per 1.73 m²). In this, the proportion of total variance attributable to center variation alone (intraclass correlation coefficient [ICC]) was 6%. Accounting for participant characteristics made little difference, with a conditional ICC of 7% after accounting for multiple covariates (all the significant variables in Tables 2 and 3). The amount of center variation comparing pediatric centers alone was greater, with an ICC of 10% before adjustment and a conditional ICC of 16% when potential confounders were included. These variances were higher again when an eGFR threshold of <15 ml/min per 1.73 m² was used (ICC and conditional ICC for the total cohort were 9% and 10%, respectively, and 13% and 22% comparing pediatric centers alone).

Figure 4.

Center variation in eGFR at initiation of KRT. Each bar represents the deviation from the cohort median (8 ml/min per 1.73 m²) for a single pediatric center, encompassing the 10th to 90th centiles with a dot for the median value. The long-dash and short-dash bars are all adult and pediatric centers combined, respectively. The x-axis is the proportion of children at that center who received a kidney transplant as their initial modality. Figure 4 can be viewed in color online at www.cjasn.org.

Sensitivity Analyses

The effect modification of modality on era was clearly demonstrated during inspection of exploratory plots, as illustrated in Figure 2. We went on with the test for interactions between era and age (P = 0.6), sex (P = 0.7), etiology (P = 0.02), and late referral (P = 0.02). There was a difference in the proportion of participants referred late for dialysis stratified by modality, 471 (39%) of those starting HD compared with 153 (21%) starting PD and five (2%) receiving a kidney transplant (Supplemental Table 1). Adjusting for this, but not age or era, reduced the difference in eGFR between PD and HD (compared with HD, the adjusted median eGFR for PD was 0.3 ml/min per 1.73 m² higher, and transplantation 1.9 ml/min per 1.73 m² higher). However, adjusting for late referral did not affect the trend in eGFR with era or interaction between era and modality described. When building the quantile regression model, nonlinear terms for age and era were examined, but did not improve model accuracy (illustrated in Figure 1 for era; age in Supplemental Figure 1). The use of a group lasso to select variables made no material difference to the results presented (Supplemental Material; ethnicity and pediatric status were discarded from the model, consistent with having smaller effect sizes). The results presented were similar when the modified Schwartz formula was used to calculate the eGFR.⁸ Given the results of the center variation analysis, we tested the addition of a random effect by center (frailty term) to the survival model and found the estimates of effect reported to be stable (aHR for eGFR within 1%). Splitting the survival dataset into children (1–17 years) and young adults (18–25 years) made no difference to the results presented (Supplemental Figures 3 and 4). Although the fraction of missing data was low (Table 1), we confirmed consistency of the results when missing data were multiply imputed using chained equations (Supplemental Material).

Discussion

There has been a persistent and substantial increase in eGFR at initiation of KRT among children and young adults in Australia and New Zealand over the past 20 years. This trend was observed among all participant groups, but greatest in those receiving a kidney transplant or PD and at the earliest quantiles of the distribution. Aside from era and modality, age, ethnicity, etiology, late referral, treatment at a pediatric center, and sex all predicted eGFR at initiation of KRT. Earlier KRT was not associated with any difference in patient survival. Clinical practice variation accounted for a substantial amount of the observed variability in eGFR at initiation of KRT.

The proportion of participants starting KRT with an eGFR ≥10 ml/min per 1.73 m² doubled over the 20-year period examined, from 16% in 1998–1999 to more than 43% in 2017–2018, consistent with observations in the United States over the same period.³ However, when eGFR was maintained as a continuous variable, the trend was less marked, with the median increasing by 1 ml/min per 1.73 m² per decade. This reflects the median eGFR at initiation of KRT lying close to the threshold value of 10 ml/min per 1.73 m². For this reason, we mostly maintained eGFR as a continuous variable in analyses, using quantile regression to account for the typical right skew observed and to allow for comparison between changes in median and 90th centile (earliest 10% of starters). This demonstrated that the increase in eGFR was relatively greater at the 90th centile, 3 ml/min per 1.73 m² per decade. The trend to earlier initiation with time was observed among all subgroups, including young adults, a population either not included or not specifically considered in previous studies. Young adults have causes of kidney disease and similar effects of kidney failure on quality of life as children, rather than older adults with whom they are often combined for analytic purposes.¹² There was a difference in the rate of increase in median (but not 90th centile) eGFR by era according to modality. The increase was greatest among those receiving a kidney transplant or PD as their initial modality compared with HD. This may reflect an increasing preference for preemptive kidney transplantation, which is associated with better long-term outcomes, including life participation.^13,14 While there are benefits to preemptive transplantation, observational studies suggesting a difference in patient survival over any period of dialysis need to be interpreted in the context of lead-time and indication biases favoring transplantation.^15,16

The trend to earlier initiation of KRT over time described may seem to be small, but equates to a substantially longer exposure to the complications and burden of care associated with treatment. Among children participating in the Chronic Kidney Disease in Children study, the rate of eGFR decline in the 18 months before dialysis was 32% per year.¹⁷ This would equate to the average participant in our study starting KRT 10 months earlier in 2018 compared with 1998. Among those starting PD, the average difference in initiation of KRT between 1998 and 2018 would be 12 months. Among the earliest 10% of starters, this difference would be longer again.

Keeping era aside, multiple participant factors influenced the timing of KRT. Younger children tended to start KRT later. This might be expected, given most centers avoid transplantation before children reaching 10 kg in size and that dialysis is technically difficult in small children, affecting families and clinicians' assessment of net benefit when initiating treatment. However, the relationship between age and eGFR persisted among older children and young adults. After adjusting for age, pediatric centers tended to initiate KRT slightly earlier than adult centers. Late referral to a nephrologist predicted a lower eGFR at initiation of KRT, as did belonging to a First Nations population, for whom cultural and structural barriers to health care access persist.¹⁸ Participants with glomerulonephritis as a cause of kidney failure also started dialysis with a lower eGFR, which might be explained by a substantially faster progression to kidney failure on average among young people with glomerular disease and additional risk factors (e.g., proteinuria or hypertension).^19,20 Male participants started dialysis slightly earlier than their female counterparts. A recent study also used the ANZDATA Registry for sex-related differences in access to preemptive transplantation among children in Australia and New Zealand and, contrary to international results, found no difference between male and female participants.²¹

There was no benefit to earlier KRT for patient survival. The higher mortality observed with earlier initiation of KRT in the multivariable model is consistent with data from the United States Renal Data System and the European Society of Pediatric Nephrology/European Renal Association and European Dialysis and Transplant Association.^3–5 However, these findings should be interpreted with caution given the potential effect of indication bias whereby sicker patients receive more aggressive therapy, in this case starting dialysis earlier.²² Lead-time bias also has direct and tangible effects when attempting to answer questions of survival using registries of participants with kidney failure, but because the direction of the bias is toward better survival among early starters, it will have favored the null hypothesis in this instance.²³

There was substantial clinical practice variation between centers. There are plausible reasons why centers serving different populations or with different resources might differ, although we did not find any evidence that the variation observed was explainable by differences in participant characteristics or the mix of modalities used. In lieu of evidence indicating benefit to earlier initiation of KRT, it might be assumed that physician preference and logistics are the main driver of the variation observed. Our findings are consistent with contemporary results from Canada.⁶

The inclusion of all pediatric centers in Australia and New Zealand treating children over a long period is a strength of this study. Despite the breadth of statistical methods used, there will remain unmeasured and residual confounding. The effect of initiation of KRT on growth and development was not considered. Lead-time and indication biases are important considerations, as discussed. Despite these limitations, the trends and associations described were consistent in sensitivity and subgroup analyses. As for other eGFR equations, there are limited data validating the accuracy of the u25 Chronic Kidney Disease in Children formula among those with an eGFR <15 ml/min per 1.73 m².²⁴

It is clear that GFR should not be the primary factor determining the timing of KRT.²⁵ However, given the effect of KRT on patients and families, persistent temporal trends toward earlier dialysis initiation need to be questioned in the absence of any demonstrable benefit.^26,27 The large amount of practice variation observed is consistent with a lack of trial data and uncertainty. There may be benefits to openly reporting the timing of KRT, ensuring that health care providers and consumers have the opportunity to reflect on where their practice lies within the broader spectrum. Ideally, a similar trial to IDEAL would be performed to answer the questions raised by these data; acknowledging this would be difficult given the relative rarity of kidney failure in children and adverse mortality outcomes.

Supplementary Material

cjasn-18-1041-s001.pdf ^{(852KB, pdf)}

Disclosures

J.C. Craig reports serving as Vice President of Flinders University and Coordinating Editor of Cochrane Kidney and Transplant and serving on the Editorial Boards of Clinical Epidemiology and Prognosis and Diagnosis Research. S. Kim reports serving as IPNA Juniors Committee ANZPNA Representative and TSANZ SPEC Committee Member. W. Lim reports research funding from AstraZeneca (education grant), honoraria from Alexion and Astellas, and advisory or leadership roles for Alexion and Astellas. H. McCarthy reports ownership interest in Australian Foundation Investment Company Limited, BHP Group Limited, Sonic Healthcare Limited, and Woodside Energy Group Limited. All remaining authors have nothing to disclose.

Funding

None.

Author Contributions

Conceptualization: Jonathan C. Craig, Anna Francis, Carrie Goh, Siah Kim, Nicholas G. Larkins, Hugh McCarthy.

Data curation: Nicholas G. Larkins.

Formal analysis: Nicholas G. Larkins.

Investigation: Nicholas G. Larkins, Wai Lim.

Methodology: Jonathan C. Craig, Anna Francis, Siah Kim, Nicholas G. Larkins, Wai Lim, Hugh McCarthy, Germaine Wong.

Project administration: Carrie Goh, Nicholas G. Larkins.

Software: Nicholas G. Larkins.

Supervision: Jonathan C. Craig, Wai Lim, Germaine Wong.

Visualization: Nicholas G. Larkins.

Writing – original draft: Carrie Goh, Nicholas G. Larkins.

Writing – review & editing: Jonathan C. Craig, Anna Francis, Siah Kim, Nicholas G. Larkins, Wai Lim, Hugh McCarthy, Germaine Wong.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/B777.

Supplemental Material.

Supplemental Table 1. Characteristics of children at initiation of KRT by modality.

Supplemental Figure 1. Examining linearity of age in a quantile regression model.

Supplemental Figure 2. Survival by eGFR at initiation of KRT (15 ml/min per 1.73 m² threshold).

Supplemental Figure 3. Survival by eGFR at initiation of KRT (participants younger than 18 years).

Supplemental Figure 4. Survival by eGFR at initiation of KRT (participants older than 18 years).