This study was designed and reported according to the PRISMA guidelines (13). The protocol was prospectively registered with PROSPERO (CRD42014010565) (14).

Type of studies: Any study design, including registry analyses, cohort studies, case-control studies and case series comparing PKT with nPKT, were eligible for inclusion. Case reports, and narrative reviews, editorials without primary data and non-English studies were excluded. We included both full articles and congress abstracts, and also checked for overlap in case abstracts were later published as full texts.

Type of participants and intervention: Eligible studies included those that compared paediatric recipients of a first or subsequent, living donor (LD) or deceased donor (DD) PKT versus nPKT. We included studies that described their population as paediatric or reported an age range of up to 18 years. PKT was defined as transplantation prior to any initiation of peritoneal dialysis (PD) or haemodialysis (HD). nPKT refers to transplantation after any given period of PD or HD. No restrictions were imposed on pre-transplant dialysis duration (dialysis vintage). Studies reporting on recipients with either a history of a previous organ transplant other than kidney or recipients of multi-organ transplants were excluded.

Type of outcomes: The outcomes of interest were overall graft loss (non-censored for death), death-censored graft loss, patient death (from all causes), delayed graft function (DGF), incidence of acute rejection (any definition, including clinically suspected and biopsy-proven acute rejection), renal function [serum creatinine or estimated glomerular filtration rate (eGFR)], primary non-function, quality of life, return to school after transplantation, height/growth measures, and incidence of cardiovascular morbidity, infections and malignancy.

As this review was part of a larger study that reviewed the available evidence for both paediatric and adult KT patients, a broad bibliographic search was carried out up to 31 July 2020 using a mixture of free text and controlled vocabulary terms (Supplementary Table S1), which retrieved references for both paediatric and adult studies. Five electronic databases including EMBASE, MEDLINE (OvidSP), Cochrane Central Register of Controlled Trials (CENTRAL), Web-of-science and Google Scholar were searched. No limits for date of publication or language were applied. The references of identified studies or review articles were scanned to find potentially eligible studies that may have been missed during the literature search. Attempts were made to contact the study authors in case of missing data or unclear study information.

The study selection was carried out in two stages by independent reviewers (RRM, LP, ST, and JL). Initially, titles and abstracts of the retrieved studies were screened against the inclusion/exclusion criteria, followed by full-text review of potentially eligible papers and final selection of the studies to be included in the review. Discrepancies between reviewers were resolved by consensus.

Two reviewers (RRM and LP) independently extracted the data using a standardized data extraction sheet. Discrepancies between reviewers were solved by discussion. Where there was more than one publication of the same study, data were only extracted from the publication that had the most complete data or the largest sample size. We extracted data on general study information and demographics, and primary and secondary outcomes. Where possible, data for LD and DD were extracted separately.

The methodological quality of the included studies, published as full text papers, was assessed by two independent authors (RRM and LP) using the Downs and Black checklist (15). Two out of the 27 items from the checklist were removed, i.e., the items relating to intervention compliance and the power of the study, as these were considered irrelevant or could not be calculated.

Where at least three studies reported on an outcome, meta-analysis was performed using the statistical software R version 3.6.3. Data were pooled using the random-effects model to calculate the relative risk (RR) with a 95% confidence interval (CI). We planned to analyze data according to LD vs. DD, however, this was not always feasible as most studies combined LD and DD in their analyses. Hence, data were pooled regardless of whether they were LD and/or DD. Patient or graft survival rates were converted to the number of deaths and graft losses. Data on graft loss were categorized as either overall graft loss or death-censored graft loss. If a study neither defined graft loss nor specified whether the graft loss data was death-censored or non-censored for death, we categorized graft loss as being non-censored for death. We calculated a pooled estimate for the nPKT group if the study reported the results for nPKT according to different dialysis durations or separately for PD or HD. If a single study reported an outcome at more than one time point, the most recent follow-up data was used. Data were pooled for any duration of follow up. In order to account for the role of confounders in the analysis of the overall graft loss, we also calculated a pooled ratio consisting of adjusted ratios either calculated or directly extracted from the studies. Secondary analyses were conducted excluding smaller studies with overlapping countries and study periods to avoid duplicate use of data. If less than three studies reported on an outcome we summarized the results in a narrative review.

Heterogeneity was analyzed using the I² statistic (16). Where heterogeneity was significant (I² ≥ 50%), a mixed effect analysis was performed to explore its potential causes.

The literature search retrieved 8,583 references. Following full-text analysis of 332 studies, 216 studies were excluded (Figure 1). Of the remaining 116 studies that met the inclusion criteria, 22 were identified as paediatric studies reporting on a total of 22,622 patients (Table 1). Cransberg (17) and Cransberg (18) were considered as one study due to insufficient data on the extent of overlap between the studies. Only the estimate for adjusted graft survival was extracted from Cransberg (18).

The methodological quality of the included studies varied with quality scores ranging from 10 to 19 out of a maximum possible score of 26 (Supplementary Table S2). Eleven studies adjusted for confounders in their analysis.

Ten studies (5–8, 17, 19–23) reported data on patient deaths. The pooled analysis showed no significant difference in the risk of patient death for PKT vs. nPKT (n = 13,490; RR: .77; CI: .53–1.11; p = .16; Figure 2). Heterogeneity was not significant (I² = 35.13%). The difference in the risk remained nonsignificant after excluding four studies (8, 17, 20, 22) with overlapping countries and study periods (n = 11,988; RR: .86; CI: .53–1.39; p = .53; I² = 57.94%; Supplementary Figure S1).

Patient death for LD transplants was reported in three studies (5, 6, 17). The pooled analysis revealed a significantly lower risk of patient death in PKT patients (n = 3,617; RR: .53; CI: .34–.83; p = .0054; I² = 0%; Supplementary Figure S2).

Two studies (5, 17) reported data on patient survival for DD. Amaral et al (5) reported a significantly higher 5-year patient survival in the PKT versus nPKT group (97.5% vs. 95.0%; p = .004). However, in the Cransberg et al (17) study, patient survival at 6 years following transplantation was similar between these groups.

Sixteen studies (5–10, 17, 20, 22–29) reported on overall graft loss. The meta-analysis revealed that the risk of graft loss following PKT was significantly lower than that of nPKT (n = 20,212; RR: .57; CI: .49–.66; p < .0001; I² = 51.24%; Figure 3). Results were similar after excluding four (8, 24–26) studies with overlapping countries and study periods (n = 16,314; RR: .54; CI: .47–.62; p < .0001; I² = 32.22%; Supplementary Figure S3). Eight of the 16 studies reported ratios adjusted for various confounders, using multivariate analyses or by matching the PKT and nPKT group (5, 6, 8, 9, 18, 22, 25, 29). Pooling of these adjusted ratios showed a similar result (n = 16,715; RR: .61; CI: .40–.92; p = .018; I² = 60.7%; Supplementary Figure S4). The adjusted ratios and confounders are presented in Supplementary Table S3.

In an attempt to explain the heterogeneity between studies for overall graft loss, a mixed-effect analysis was performed which looked at the role of four moderator variables: the percentage of HD patients in the nPKT group, length of follow-up, percentage of LD, and the year of publication (Supplementary Figures S5–S8). None of these variables were found to significantly influence the relative risk of graft loss. It may be worth noting that on visual inspection of the forest plot, the heterogeneity is in the size of effect rather than the direction of effect.

Five studies (5, 6, 23, 26, 27) reported on overall graft loss for LD, and the pooled analysis showed that PKT significantly reduced the risk of graft loss (n = 4,973; RR: .57; CI: .46–.69; p < .0001; I² = 0%; Supplementary Figure S9).

Two studies (5, 23) reported on overall graft survival in DD recipients. Amaral et al (5) reported a significantly higher 5-year graft survival rate in PKT patients compared to nPKT patients (85.4% vs. 76.4%; p < .001). However, Vats et al (23) reported similar 3-year graft survival in PKT versus nPKT (PD and HD) patients.

Death-censored graft loss was reported in two studies (9, 30) for LD and DD data combined. Garcia et al (9) reported a higher 12-, 36-, 60- and 90-month death-censored graft survival rate, adjusted by donor type, in PKT patients compared with nPKT patients (97% vs. 87%; 92% vs. 79%; 86% vs. 72%; 76% vs. 65%, respectively). The difference was significant at 90 months (p < .05); however, the study did not clearly report if the differences were significant at the other time points. The study by Fitzwater et al (30), found no significant difference in the 2-year death-censored graft loss between PKT and nPKT.

DGF was reported in three studies (17, 25, 27). The RR for the incidence of DGF was .57 (n = 4,871; CI: .22–1.50; p = .26; Supplementary Figure S10). Heterogeneity was high (I² = 81.51%). We could not explore heterogeneity as the number of studies was too small.

DGF for LD was reported in two studies (17,27). Cransberg et al (17) showed a slightly higher incidence of DGF in PKT patients (3.5% vs. 2.4%), but did not report if this difference was significant. No significant difference was observed between PKT vs. nPKT in terms of DGF in the study by Kim et al (27).

The only study that reported on DGF in DD patients was Cransberg et al (17), which observed no difference in the DGF rate between PKT and nPKT.

Incidence of acute rejection was reported in seven studies (6, 17, 22, 25–27, 30). The pooled analysis revealed that the risk of acute rejection in PKT patients was significantly lower than that of nPKT patients (n = 4,897; RR: .81; CI: .75–.88; p < .0001; Figure 4). Heterogeneity was low (I² = 0%). Similar results were observed after excluding Fitzwater et al (30) from the analysis due to overlapping country and study period (n = 4,851; RR: .81; CI: .74–.87; p < .0001; I² = 0%; Supplementary Figure S11). Of the seven studies, only two (6, 22) adjusted for confounders; hence, a pooled estimate of the adjusted acute rejection rate could not be calculated.

Three studies (6, 26, 27) reported on the rate of acute rejection for LD. Although the effect size was similar to the overall analysis, it did not reach statistical significance (n = 247; RR: .79; CI: .55–1.15; p = .22; I² = 0%; Supplementary Figure S12).

Cransberg et al (17) was the only study that included data on acute rejection for DD. In the study, a significantly higher percentage of patients remained acute rejection-free following PKT than after nPKT (52% vs. 37%; p = .039) at 3 years.

Two studies reported cardiovascular morbidity outcomes (17, 31). Cransberg et al (17) measured the incidence of severe hypertension between PKT vs. nPKT at one, three and 5 years post-transplant, and found significantly lower incidence of severe hypertension in the PKT group in the third year (40% vs. 64%; p = .016), among patients with DD transplants. The study by Sinha and Marks (31) also showed a significantly lower incidence of hypertension in the PKT versus nPKT group (31% vs. 53%; p = .02) for combined LD and DD data. No studies reported on infections and malignancy.

Renal function was reported in six studies as either eGFR or serum creatinine, with four studies (20, 22, 30, 32) reporting on LD and DD data combined. Mahmoud et al (22) evaluated the mean GFR at one and 4 years post-transplant, and found no statistical differences in the GFR values between the PKT and nPKT group at both follow-ups. The study by Kaya et al (20) also showed no significant difference in the mean GFR between these groups within a median follow-up of 7.23 years. Duzova et al (32) measured the mean GFR values at one, two, three and 5 years after transplantation, and reported a significantly lower mean GFR in the PKT group only in the third year (mean ± standard deviation (SD): 86 ± 31 ml/min/m² vs. 101 ± 31 ml/min/m²; p < .05). Likewise, Fitzwater et al (30) reported no statistical differences in the serum creatinine levels between PKT and nPKT at 1 month, 3 months, 6 months, 1 year and 2 years post-transplant.

Two studies (26, 27) reported renal function for LD only. Kim et al (27) reported no differences between PKT and nPKT in the median GFR at 1 month and 1 year. Flom et al (26) reported a higher mean (±SD) GFR for PKT (68 ± 28 ml/min/1.73 m²) versus nPKT (HD and PD) (both 60 ± 26 ml/min/1.73 m²), calculated over a median follow-up of 3.5, 3.6 and 5.1 years for PKT, PD and HD respectively. However, the study did not report whether this difference was significant.

No studies reported on primary-non function.

Quality of life was reported in only two studies (6, 33). Splinter et al (33) assessed the health-related quality of life (HRQoL) of patients who spent at least 6 months on their treatment modality, using the PedsQL™ questionnaire. The PedsQL™ consisted of five major domains, including physical health, emotional functioning, social functioning, school functioning, and psychosocial health. The mean ± SD HRQoL scores for physical health was significantly higher in the PKT vs. nPKT group (78.6 ± 18.0 vs. 70.4 ± 20.5; p < .05), but showed no differences between the groups for the other domains. Harada et al (6) asked patients about the benefits and disadvantages of renal transplantation. The percentage of patients that reported feeling satisfied with the improvement in their physical condition was significantly higher in the PKT vs. the nPKT group (p < .01). On the other hand, a significantly higher percentage of patients in the nPKT group reported satisfaction related to the freedom from restrictions of liquid intake, daily diet and time spent on dialysis, following renal transplantation (p < .01). No significant differences were observed between the two groups regarding disadvantages felt due to renal transplantation, which included anxiety about the fate of the renal graft and annoyance resulting from frequent hospital visits and daily medications.

No studies reported data on return to school.

Three studies (6, 8, 31) reported findings on the height/growth of patients. Harada et al (6) assessed the mean ± SD heights of the patients at transplantation and at one and 3 years post-transplant, using the national cross-sectional standard growth chart for boys and girls. The study showed significantly better mean ± SD height in the PKT vs. nPKT group at transplantation (−.84 ± 0.73 vs. −2.86 ± 1.93; p < .05) and at 3 years post-transplant (−.53 ± 1.65 vs. −3.22 ± 1.94; p < 0.05), only for patients less than 15 years old. Sinha and Marks (31), who measured the height of the patients at the last clinical visit (range 1–15 years) using the median standard deviation score (SDS), found no significant differences in the scores between the two groups. Similar results were reported by Offner et al (8), who also used the median SDS to measure the height of the patients at 1 year post-transplantation.

Secondary analyses comparing PKT versus nPKT patients with primary KT are presented in Supplementary Figures S13–S15.