Edited by: Patrick Mark, University of Glasgow, United Kingdom
Reviewed by: Giulia Capitoli, University of Milano Bicocca, Italy; Edwin Castillo Velarde, Hospital Base Guillermo Almenara Irigoyen, Peru
*Correspondence: Paramit Chowdhury,
This article was submitted to Clinical Research in Nephrology, a section of the journal Frontiers in Nephrology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Post-transplant glomerulonephritis (PTGN) has been associated with inferior long-term allograft survival, and its incidence varies widely in the literature.
This is a cohort study of 7,623 patients transplanted between 2005 and 2016 at four major transplant UK centres. The diagnosis of glomerulonephritis (GN) in the allograft was extracted from histology reports aided by the use of text-mining software. The incidence of the four most common GN post-transplantation was calculated, and the risk factors for disease and allograft outcomes were analyzed.
In total, 214 patients (2.8%) presented with PTGN. IgA nephropathy (IgAN), focal segmental glomerulosclerosis (FSGS), membranous nephropathy (MN), and membranoproliferative/mesangiocapillary GN (MPGN/MCGN) were the four most common forms of post-transplant GN. Living donation, HLA DR match, mixed race, and other ethnic minority groups were associated with an increased risk of developing a PTGN. Patients with PTGN showed a similar allograft survival to those without in the first 8 years of post-transplantation, but the results suggest that they do less well after that timepoint. IgAN was associated with the best allograft survival and FSGS with the worst allograft survival.
PTGN has an important impact on long-term allograft survival. Significant challenges can be encountered when attempting to analyze large-scale data involving unstructured or complex data points, and the use of computational analysis can assist.
Glomerulonephritidies are an important cause of end-stage kidney disease (ESKD), and their tendency to affect the younger age groups results in them being the most prevalent cause of ESKD in patients undergoing renal transplantation in the UK and the second most common in the USA (
The incidence of recurrent glomerulonephritis (GN) post-transplant varies widely in studies (
Recurrent GN is an important cause of allograft loss (
When conducting epidemiological studies using large-scale data, care needs to be taken regarding the accuracy of the clinical data points being considered. The use of computational analysis, such as text mining and natural language processing (NLP), allows the extraction of informative features from many types of raw and unstructured data, thus allowing greater accurately and standardisation. Such techniques may lead to more accurate results and conclusions (
This multicenter study is the first in nephrology, utilizing computational analysis to aid in the diagnosis of GN and aiming to accurately estimate the incidence of GN post-transplantation in a United Kingdom cohort, analyse epidemiological factors influencing their emergence in the post-transplant period and comment on allograft outcomes.
This is a multicenter cohort study conducted in four UK transplant centres—Guy’s and St. Thomas’ NHS Foundation Trust, Cambridge University Hospitals NHS Foundation Trust, Oxford University Hospitals NHS Trust, and Imperial Healthcare NHS Trust.
The study was approved by the East Midlands Nottingham 2 Research Ethics Committee, with reference number 15/EM/0449.
The inclusion criteria for the patients during the study period were any individual (>16 years old) who has undergone kidney or simultaneous pancreas and kidney transplantation between January 2005 and October 2016 in the participating centres.
Only the first transplant episode per patient was considered during the time of study. There were no exclusion criteria.
Routinely collected clinical data were extracted from electronic health records (EHRs) at each center, combined with data from NHS Blood & Transplant and aggregated in a data warehouse on a per transplant and personal level.
Centers 1, 2, and 4 used mostly basiliximab as induction agent and the maintenance immunosuppression regimen of tacrolimus, mycophenolate (MMF), and prednisolone. Center 3 used a steroid-sparing protocol. Alemtuzumab was used as an induction agent in combination with tacrolimus monotherapy and a 7-day prednisolone regime. MMF was added in cases of
Regarding biopsy practices, center 3 had criteria-driven protocol biopsies at 1 year post-transplant. The other 3 centers only performed biopsies when clinically indicated.
Software to extract the diagnoses of various glomerular diseases from unstructured biopsy text reports was developed using the open-source General Architecture for Text Engineering (
Frequencies and percentages were used to present categorical factors and missing data. Means and standard deviations were used for normally distributed continuous data, while medians and interquartile ranges were used if their distribution was skewed. The distribution of continuous data was explored using histograms.
For the evaluation of possible risk factors for the diagnosis of GN post-transplantation, recipient-related characteristics were explored.
Univariate Cox regression analysis was used to explore the influence of each factor on time from transplant to the development of GN and time from transplant to graft failure. The category with the highest frequency was used as reference category in each exploratory factor. Factors with
Kaplan–Meier plots were used for the survival analysis. For patients not presenting a graft failure, the date when their creatinine value was measured last was considered. A
All statistical analyses were performed with Stata/IC version 15.1.
During the period of the study, 7,623 patients had their first renal transplant. The population characteristics between centers were similar, except for an expected variation in ethnicity, with two of the four centres serving large multicultural urban areas. Only 45.9% of the patients had information regarding the cause of ESKD. Of these, in 38.1% of cases, the cause of ESKD was unknown, while 11.1% had a diagnosis of a specific GN and 3.2% had a diagnosis of chronic glomerulonephritis of an unspecified nature.
The demographics of our population and patients’ transplant characteristics are described in
Demographic and transplant characteristics of patients who had their first kidney transplant during the period 2005–2016 by center.
Baseline and transplant characteristics | Centers | |||||
---|---|---|---|---|---|---|
Center 1 | Center 2 | Center 3 | Center 4 | Total | ||
|
|
|
|
|
||
Gender | Female | 620 (36.6) | 897 (39.5) | 638 (36.6) | 732 (38.2) | 2,887 (37.9) |
Male | 1,074 (63.4) | 1,373 (60.5) | 1,102 (63.3) | 1,186 (61.8) | 4,735 (62.1) | |
Unknown | – | – | 1 (0.1) | – | 1 (0.0) | |
Age at date of transplant | Years, mean, SD, (range), |
48.8 (13.5) (17.0, 76.0) |
47.1 (13.7) (16.0, 79.0) |
49.5 (12.9) (18.0, 78.0) |
48.2 (12.6) (17.0, 81.0) |
48.3 (13.2) (16.0, 81.0) |
Age groups at date of transplant | 16–30 | 185 (10.9) | 294 (13.0) | 168 (9.6) | 163 (8.5) | 810 (10.6) |
31–45 | 487 (28.7) | 722 (31.8) | 478 (27.5) | 662 (34.5) | 2,349 (30.8) | |
46–55 | 434 (25.6) | 583 (25.7) | 483 (27.7) | 532 (27.7) | 2,032 (26.7) | |
≥56 | 588 (34.7) | 671 (29.6) | 612 (35.2) | 561 (29.2) | 2,432 (31.9) | |
Ethnicity |
White | 1,415 (83.5) | 1,587 (69.9) | 782 (44.9) | 1,558 (81.2) | 5,342 (70.1) |
Mixed | 7 (0.4) | 32 (1.4) | 21 (1.2) | – | 60 (0.8) | |
Asian or Asian British | 60 (3.5) | 135 (5.9) | 465 (26.7) | 252 (13.1) | 912 (12.0) | |
Black or Black British | 171 (10.1) | 371 (16.3) | 266 (15.3) | 77 (4.0) | 885 (11.6) | |
Other ethnic groups | 12 (0.7) | 58 (2.6) | 142 (8.2) | 25 (1.3) | 237 (3.1) | |
Not stated | 29 (1.7) | 87 (3.8) | 65 (3.7) | 6 (0.3) | 187 (2.5) | |
Blood group | Missing | – | – | 11 (0.6) | – | 11 (0.1) |
A | 711 (42.0) | 858 (37.8) | 587 (33.7) | 817 (42.6) | 2,973 (39.0) | |
AB | 62 (3.7) | 107 (4.7) | 97 (5.6) | 104 (5.4) | 370 (4.9) | |
B | 168 (9.9) | 293 (12.9) | 331 (19.0) | 241 (12.6) | 1,033 (13.6) | |
0 | 753 (44.5) | 1,012 (44.6) | 715 (41.1) | 756 (39.4) | 3,236 (42.5) | |
Comorbidities | Cardiovascular disease | 28 (1.7) | 183 (8.1) | 5 (0.3) | – | 216 (2.8) |
Diabetes | 72 (4.3) | 485 (21.4) | 19 (1.1) | – | 576 (7.6) | |
Current smoker | 3 (0.2) | 89 (3.9) | – | – | 92 (1.2) | |
Ex-smoker | – | 55 (2.4) | – | – | 55 (0.7) | |
Hypertension | 1,316 (77.7) | 1,406 (61.9) | 4 (0.2) | – | 2,726 (35.8) | |
Donor type | DBD | 530 (31.3) | 887 (39.1) | 791 (45.4) | 982 (51.2) | 3,190 (41.8) |
DCD | 740 (43.7) | 428 (18.9) | 197 (11.3) | 467 (24.3) | 1,832 (24.0) | |
Living—related | 243 (14.3) | 571 (25.2) | 408 (23.4) | 268 (14.0) | 1,490 (19.5) | |
Living—unrelated | 181 (10.7) | 384 (16.9) | 345 (19.8) | 201 (10.5) | 1,111 (14.6) | |
Donor’s age |
Years, mean, SD (range), |
– | 46.4 (15.4) (0.0, 83.0) |
54.8 (14.8) (11.0, 90.0) |
45.5 (16.0) (1.0, 85.0) |
48.5 (16.0) (0.0, 90.0) |
Donor’s gender | Missing | – | 1 (0.0) | – | – | 1 (0.0) |
Female | 786 (46.4) | 1,150 (50.7) | 936 (53.8) | 923 (48.1) | 3,795 (49.8) | |
Male | 908 (53.6) | 1,119 (49.3) | 805 (46.2) | 995 (51.9) | 3,827 (50.2) | |
Years of transplantation | 2005–2010 | 787 (46.5) | 966 (42.6) | 904 (51.9) | 858 (44.7) | 3,515 (46.1) |
2011–2016 | 907 (53.5) | 1,304 (57.4) | 837 (48.1) | 1,060 (55.3) | 4,108 (53.9) | |
Cold ischemia time | Hours, median (IQR), |
12.7 (7.0, 16.2) |
10.8 (4.0, 15.1) |
12.0 (3.0, 19.0) |
12.0 (9.0, 15.0) |
12.0 (4.0, 16.0) |
Missing | 36 (2.1) | 261 (11.5) | 31 (1.8) | 262 (13.7) | 590 (7.7) | |
HLA mismatch A | 0 | 336 (19.8) | 491 (21.6) | 354 (20.3) | 401 (20.9) | 1,582 (20.8) |
1 | 876 (51.7) | 1,192 (52.5) | 867 (49.8) | 1,025 (53.4) | 3,960 (51.9) | |
2 | 482 (28.5) | 559 (24.6) | 519 (29.8) | 491 (25.6) | 2,051 (26.9) | |
Missing | – | 28 (1.2) | 1 (0.1) | 1 (0.1) | 30 (0.4) | |
HLA mismatch B | 0 | 249 (14.7) | 353 (15.6) | 264 (15.2) | 256 (13.3) | 1,122 (14.7) |
1 | 1,087 (64.2) | 1,342 (59.1) | 1,091 (62.7) | 1,079 (56.3) | 4,599 (60.3) | |
2 | 358 (21.1) | 547 (24.1) | 385 (22.1) | 582 (30.3) | 1,872 (24.6) | |
Missing | – | 28 (1.2) | 1 (0.1) | 1 (0.1) | 30 (0.4) | |
HLA mismatch DR | 0 | 639 (37.7) | 771 (34.0) | 644 (37.0) | 627 (32.7) | 2,681 (35.2) |
1 | 840 (49.6) | 1,118 (49.3) | 869 (49.9) | 972 (50.7) | 3,799 (49.8) | |
2 | 215 (12.7) | 353 (15.6) | 227 (13.0) | 318 (16.6) | 1,113 (14.6) | |
Missing | – | 28 (1.2) | 1 (0.1) | 1 (0.1) | 30 (0.4) | |
HLA mismatch A/B/DR | 0–2 | 591 (34.9) | 772 (34.0) | 587 (33.7) | 547 (28.5) | 2,497 (32.8) |
3–4 | 904 (53.4) | 1,188 (52.3) | 934 (53.6) | 1,142 (59.5) | 4,168 (54.7) | |
5–6 | 199 (11.7) | 282 (12.4) | 219 (12.6) | 228 (11.9) | 928 (12.2) | |
Missing | – | 28 (1.2) | 1 (0.1) | 1 (0.1) | 30 (0.4) | |
ESKD cause captured | No | 802 (47.3) | 960 (42.3) | 447 (25.7) | 1,918 (100.0) | 4,127 (54.1) |
Yes | 892 (52.7) | 1,310 (57.7) | 1,294 (74.3) | – | 3,496 (45.9) |
IQR, interquartile range (25 to 75% quartiles).
Ethnicity was coded according to the “Ethnic Category Code” of NHS; for more details, please see the link below:
Results from CUH available for 30 patients. Missing data from CUH: 1,664 (98.2%).
During the period of the study, 277 (3.6%) patients presented with evidence of glomerular disease in their allograft biopsy.
The histological findings in the renal allograft were IgA nephropathy (IgAN) in 48.7% of the biopsies, focal segmental glomerulosclerosis (FSGS) in 18.1%, membranous nephropathy (MN) in 7.2%, membranoproliferative GN and mesangiocapillary GN (MPGN/MCGN) in 2.9%, minimal change disease (MCD) in 0.4%, and thrombotic microangiopathy (TMA) in 22.7%.
A statistical analysis was carried out using the four most common GNs, namely, IgAN, FSGS, MN, and MPGN/MCGN. TMA represents a heterogeneous group of conditions and often cannot be unequivocally attributed to a single underlying etiology (
Between 24 and 32 patients per 1,000 transplanted patients (95% CI) are estimated to develop a GN post-transplantation (
Incidence estimates by post-transplant glomerulonephritis (GN) (N = 214).
Post-transplant GD | Incidence estimate (per 1,000 patients) (95% CI) |
---|---|
Total |
|
IgA nephropathy | 17.71 (14.87 to 20.93) |
FSGS | 6.56 (4.87 to 8.64) |
Membranous nephropathy | 2.62 (1.60 to 4.05) |
MPGN | 1.05 (0.45 to 2.07) |
MCD | 0.13 (0.00 to 0.73) |
All patients who developed any GN |
28.07 (24.48 to 32.03) |
FSGS, focal segmental glomerulosclerosis; aHUS, atypical hemolytic uremic syndrome; MPGN, membranoproliferative glomerulonephritis; MCD, minimal change disease.
Any glomerulonephritis disease: IgAN, FSGS, MPGN, MN, or MCD.
The median time from transplant to histopathological diagnosis of GN was 701.5 days (IQR: 168–1,742).
Younger age groups, 16–30 years old and 31–45 years old, were found to have a 1.7 (95% CI: 1.113–2.635;
Living related donation was identified to increase the risk for the development of a GN (HR: 1.74; 95% CI: 1.238–2.445;
Protective factors against the development of GN in the allograft were identified as female gender (HR: 0.70; 95% CI: 0.521–0.933;
Risk factors for the development of post-transplant glomerulonephritis (IgAN, FSGS, MPGN, MN, and MCD) (N = 7,560).
Characteristics | UnivariateHR (95% CI) |
|
MultivariableHR (95% CI) |
|
|
---|---|---|---|---|---|
Gender | Female | 0.697 (0.521–0.933) | 0.015 | 0.754 (0.556–1.023) | 0.069 |
Male | 1 | 1 | |||
Age groups at date of transplant | 16–30 | 1.712 (1.113–2.635) | 0.014 | ||
31–45 | 1.461 (1.027–2.078) | 0.035 | |||
46–55 | 1.069 (0.720–1.586) | 0.741 | |||
≥56 | 1 | ||||
Age at date of transplant | 0.985 (0.975–0.995) | 0.004 | 0.988 (0.977–0.999) | 0.036 | |
Ethnic groups | White | 1 | 1 | ||
Black | 0.542 (0.318–0.923) | 0.024 | 0.556 (0.319–0.971) | 0.039 | |
Asian | 1.204 (0.838–1.729) | 0.316 | 1.288 (0.888–1.867) | 0.182 | |
Other | 1.584 (0.957–2.621) | 0.073 | 1.714 (1.027–2.861) | 0.039 | |
Blood group | A | 0.889 (0.654–1.207) | 0.45 | – | |
AB | 1.288 (0.747–2.22) | 0.362 | – | ||
B | 0.914 (0.606–1.377) | 0.666 | – | ||
O | 1 | – | |||
Donor type | DBD | 1 | 1 | ||
DCD | 1.244 (0.835–1.854) | 0.283 | 1.351 (0.892–2.046) | 0.155 | |
Living—related | 1.74 (1.238–2.445) | 0.001 | 1.777 (0.990–3.192) | 0.054 | |
Living—unrelated | 1.536 (1.042–2.264) | 0.03 | 2.136 (1.165–3.915) | 0.014 | |
Donor’s gender | Female | 1.208 (0.923–1.581) | 0.169 | 1.195 (0.897–1.593) | 0.224 |
Male | 1 | 1 | |||
Donor’s age | 1.005 (0.996–1.015) | 0.281 | – | ||
Year of transplantation | 2005–2010 | 0.637 (0.47–0.864) | 0.004 | 0.596 (0.430–0.824) | 0.002 |
2011–2016 | 1 | 1 | |||
Cold ischemia time | 0.983 (0.966–1) | 0.05 | 1.010 (0.980–1.042) | 0.504 | |
HLA mismatch A | 0 | 1.047 (0.747–1.467) | 0.79 | – | |
1 | 1 | – | |||
2 | 0.977 (0.708–1.348) | 0.887 | – | ||
HLA mismatch B | 0 | 1.268 (0.894–1.799) | 0.182 | 1.132 (0.757–1.692) | 0.545 |
1 | 1 | 1 | |||
2 | 0.959 (0.684–1.344) | 0.806 | 0.941 (0.638–1.388) | 0.759 | |
HLA mismatch DR | 0 | 1.316 (0.987–1.754) | 0.062 | 1.399 (1.007–1.942) | 0.045 |
1 | 1 | 1 | |||
2 | 1.078 (0.707–1.643) | 0.728 | 1.054 (0.653–1.701) | 0.829 | |
HLA mismatch A/B/DR | 0–2 | 1.194 (0.893–1.597) | 0.232 | – | |
3–4 | 1 | – | |||
5–6 | 1.21 (0.805–1.819) | 0.359 | – |
HR, hazard ratio.
From the adjusted analysis, living donation was shown to be a risk factor for developing a GN post-transplantation, with living unrelated donation associated with a risk increase of 2.1 times (95% CI: 1.171–3.928;
The protective factors against the development of GN in the allograft were also identified, specifically, black race (HR: 0.56; 95% CI: 0.319–0.972;
Patients who developed GN post-transplantation presented a median allograft lifespan of 1,207 days (IQR: 365.0–2,208.0). Allograft failure was observed in 25% (54 patients) with a median time from the histopathological diagnosis of GN to failure of 224 days (IQR: 17–414). Death occurred in 8.9% (19 patients) at 587 days (IQR: 204.0–1,223.0) following the diagnosis of GN in the allograft.
From the adjusted analysis, the year of transplantation between 2005 and 2010 was shown to be the only risk factor for graft failure, with an increased risk of 1.4 times (95% CI: 1.14–1.61;
Risk factors for graft loss (N = 7,560).
Characteristics | UnivariateHR (95% CI) |
|
MultivariableHR (95% CI) |
|
|
---|---|---|---|---|---|
Gender | Female | 0.942 (0.833–1.065) | 0.342 | - | |
Male | 1 | - | |||
Age groups at date of transplant | 16–30 | 0.630 (0.508–0.780) | <0.001 | – | – |
31–45 | 0.685 (0.592–0.794) | <0.001 | – | ||
46–55 | 0.711 (0.609–0.830) | <0.001 | – | ||
≥56 | 1 | – | |||
Age at date of transplant | 1.014 (1.01–1.019) | <0.001 | 1.013 (1.007–1.020) | 0.000 | |
Ethnic groups | White | 1 | 1 | ||
Black | 0.920 (0.764–1.109) | 0.384 | 0.914 (0.740–1.130) | 0.401 | |
Asian | 0.696 (0.573–0.847) | <0.001 | 0.659 (0.530–0.819) | 0.000 | |
Other | 0.704 (0.512–0.968) | 0.031 | 0.760 (0.545–1.060) | 0.106 | |
Blood group | A | 1.034 (0.909–1.177) | 0.605 | 1.010 (0.866–1.178) | 0.899 |
AB | 0.741 (0.547–1.003) | 0.052 | 0.716 (0.504–1.017) | 0.062 | |
B | 0.760 (0.625–0.923) | 0.006 | 0.887 (0.708–1.110) | 0.294 | |
O | 1 | 1 | |||
Donor type | DBD | 1 | 1 | ||
DCD | 0.975 (0.835-1.136) | 0.745 | 0.805 (0.655–0.989) | 0.039 | |
Living—related | 0.583 (0.493–0.689) | <0.001 | 0.408 (0.302–0.525) | 0.000 | |
Living—unrelated | 0.609 (0.506–0.734) | <0.001 | 0.338 (0.251–0.456) | 0.000 | |
Donor’s gender | Female | 1.036 (0.920–1.166) | 0.552 | – | |
Male | 1 | – | |||
Donor’s age | 1.007 (1.004–1.012) | <0.001 | 1.007 (1.002–1.011) | 0.009 | |
Year of transplantation | 2005–2010 | 1.282 (1.110–1.479) | 0.001 | 1.355 (1.141–1.61) | 0.001 |
2011–2016 | 1 | 1 | |||
Cold ischemia time | 1.016 (1.009–1.023) | <0.001 | 0.974 (0.961–0.988) | 0 | |
HLA mismatch A | 0 | 0.920 (0.788–1.075) | 0.297 | – | |
1 | 1 | – | |||
2 | 1.041 (0.906–1.196) | 0.566 | – | ||
HLA mismatch B | 0 | 0.898 (0.756–1.068) | 0.226 | – | |
1 | 1 | – | |||
2 | 0.966 (0.837–1.114) | 0.636 | – | ||
HLA mismatch DR | 0 | 0.988 (0.868–1.125) | 0.862 | – | |
1 | 1 | – | |||
2 | 0.967 (0.807–1.157) | 0.715 | – | ||
HLA mismatch A/B/DR | 0–2 | 0.917 (0.805–1.044) | 0.194 | 0.911 (0.778–1.067) | 0.246 |
3–4 | 1 | 1 | |||
5–6 | 0.872 (0.722–1.053) | 0.157 | 1.132 (0.900–1.424) | 0.291 |
HR, hazard ratio.
When comparing the survival estimate in patients who developed any type of GN post-transplantation to patients who did not develop a disease, the former group of patients was shown to have a similar allograft survival until after approximately 7.5–8 years post-transplantation, at which point there is a suggestion that they did less well (
Graft survival for patients with and without a post-transplant glomerulonephritis (
The probability of graft survival for patients who developed a diagnosis of GN after transplantation, regardless of its histopathological pattern, was approximately 95% at 1 year (95% CI: 0.91–0.97) and 83% at 5 years (95% CI: 0.76–0.88).
Kaplan–Meier survival estimates of post-transplant glomerulonephritis by histopathological type.
IgAN demonstrated the most favorable outcome showing a small and stable decline in allograft survival until 8 years post-transplant, with an allograft survival of approximately 75% at this time point. In contrast, FSGS experienced the worst outcome with a progressive decline post-transplant, showing an approximate allograft survival of 30% at 8 years post-transplant. MN presented good allograft survival in the first year post-transplant, with an allograft survival of approximately 55–60% at 7 years post-transplant.
This is the first large-scale UK study investigating the incidence and effects of post-transplant GN. It is one of the first reports detailing the use of computational techniques to aid in the analysis of renal biopsy reports in a large patient cohort across multiple centers. We found the overall incidence of GN after renal transplantation to be 2.8%. A large study using registry data focused on recurrent disease and found that the recurrence of the four most common GN subtypes was around 10% (
A major limitation of this study was the inability to accurately define the cause of ESRD in a large proportion of patients and prevented us commenting specifically on recurrent disease. As well as a significant amount of missing data (46%), even where data was available, a large proportion of patients with ESKD had no determined cause (41%). This is particularly true in black patients where a biopsy is less likely to occur and ESKD ascribed to hypertension (
Another factor why our study has highlighted a low rate of post-transplant GN might relate to biopsy practice and the low frequency of protocol biopsies carried out. Only one center carried out a form of protocol biopsy at 1 year post-transplant based on specific criteria such as low eGFR. However, the practice of early protocol biopsy is likely to have a little effect as most post-transplant GN develops after the first year, at least histologically. The criteria for performing a biopsy in the longer term vary widely not just between centers but also between individual clinicians and has also changed over time, with the understanding that there are multiple causes of long-term allograft loss.
In common with other studies, we found IgAN as the most common GN found in post-transplant biopsies (
Living donation has been identified as a risk factor for developing a post-transplant GN, however with conflicting results (
Black race was identified as a protective variable for the development of post-transplant GN. This finding could be explained by the fact that this group is more greatly affected by other diseases involving the kidneys (hypertension and sickle cell disease) or because some forms of GN related to APOL1 genotype have a very low risk of recurrence (
With regards to immunosuppression and its role in the development or reduction of GN disease in the renal allograft (
Cold-ischemia time and HLA mismatches have been implied in the development of GN in the allograft (
In our study, allograft failure was observed in 25% of the patients who developed a GN, regardless of histopathologic type. This percentage is in line with previous studies (
In our cohort, patients with any type of post-transplant GN presented a similar allograft survival to patients who did not develop the disease until after approximately 7.5-8 years post-transplantation, at which point there is a suggestion that they did less well. Allograft survival at 1 year was 95% and at 5 years was 83%. Recipients with allograft GN also presented a lower allograft survival when compared with the allograft survival estimate for all UK centers (
From all the variables studied, year of transplantation between 2005 and 2010 was shown to be the only risk factor for graft failure. Whether this is due to the immunosuppression used in the early era (
The different types of GN, both in native and transplanted kidneys, behave in distinct ways. As in other studies (
A major strength of our study is the direct analysis of biopsy reports to determine post-transplantation GN rather than relying on registry data, thus reducing potential inaccuracy in data gathering or disease coding.
The increasing use of EHRs provides an opportunity for the automated collection and analysis of large patient cohorts and “big data” which is of particular importance in nephrology, where the incidence of specific diseases is rare. Our study highlights the difficulties in standardization and accurate collection of complex data points, such as the diagnosis of renal disease, which rely on the analysis of unstructured data. We demonstrate that computational analysis has potential for use in this area, with more advanced techniques involving NLP and artificial intelligence offering the potential of fully automated extraction of complex data points, such as the diagnosis of renal disease. Other limitations to our study are the type of study, retrospective, as well as the abovementioned absence of cause of ESKD, which did not allow us to differentiate recurrent from
In conclusion, in this large-scale UK study, we found that GN post-renal transplantation has an important impact on long-term allograft survival, and significant challenges can be encountered when attempting to analyze large-scale data. Nonetheless, machine learning can aid in the study of complex data points.
The original contributions presented in the study are included in the article/
The studies involving human participants were reviewed and approved by the East Midlands Nottingham 2 Research Ethics Committee with reference number 15/EM/0449. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
The research idea and study design were contributed by PC, RA, CB, MC, and RM. EB, BC, FD, LJ, GO, SH, AMu, BG, LM, JD, KV, and KW contributed to the analysis and interpretation of data. JS and BC contributed to the development of the text-mining biopsy software. Statistical analysis was done by EB and CB. AW, FH, and AS contributed to the management of the project. Drafting the article and providing intellectual content were done by RA, PC, EB, CB, CR, NS, CB, and RP. Approval of the submitted version of the manuscript was done by RA, EB, CB, PC, RP, CB, MH-F, CR, NS, AMc, RP, and GL. Supervision and mentorship was provided by PC. All authors contributed to the article and approved the submitted version.
This research was funded by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St. Thomas’ NHS Foundation Trust and King’s College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health.
MH-F is currently an employee of UCB Celltech, a pharmaceutical company. Her involvement in the conduct of this research was solely in her capacity as academic at King’s College London.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
This research has been conducted using National Institute for Health Research (NIHR)- Health Informatics Collaborative (HIC) data resources. The NIHR-HIC is a joint initiative between the NIHR Biomedical Research Centres at Guy’s and St. Thomas’ NHS Foundation Trust, Imperial Healthcare NHS Trust, Cambridge University Hospitals NHS Foundation Trust, and Oxford University Hospitals NHS Trust, which has provided data services, infrastructure, and expertise.
The Supplementary Material for this article can be found online at: