Risk factors for graft loss and death among kidney transplant recipients: A competing risk analysis.

INTRODUCTION: Kidney transplantation is the best therapeutical option for CKD patients. Graft loss risk factors are usually estimated with the cox method. Competing risk analysis could be useful to determine the impact of different events affecting graft survival, the occurrence of an outcome of interest can be precluded by another. We aimed to determine the risk factors for graft loss in the presence of mortality as a competing event.
METHODS: A retrospective cohort of 1454 kidney transplant recipients who were transplanted between July 1, 2008, to May 31, 2019, in Colombiana de Trasplantes, were analyzed to determine risk factors of graft loss and mortality at 5 years post-transplantation. Kidney and patient survival probabilities were estimated by the competing risk analysis. The Fine and Gray method was used to fit a multivariable model for each outcome. Three variable selection methods were compared, and the bootstrapping technique was used for internal validation as split method for resample. The performance of the final model was assessed calculating the prediction error, brier score, c-index and calibration plot.
RESULTS: Graft loss occurred in 169 patients (11.6%) and death in 137 (9.4%). Cumulative incidence for graft loss and death was 15.8% and 13.8% respectively. In a multivariable analysis, we found that BKV nephropathy, serum creatinine and increased number of renal biopsies were significant risk factors for graft loss. On the other hand, recipient age, acute cellular rejection, CMV disease were risk factors for death, and recipients with living donor had better survival compared to deceased-donor transplant and coronary stent. The c-index were 0.6 and 0.72 for graft loss and death model respectively.
CONCLUSION: We developed two prediction models for graft loss and death 5 years post-transplantation by a unique transplant program in Colombia. Using a competing risk multivariable analysis, we were able to identify 3 significant risk factors for graft loss and 5 significant risk factors for death. This contributes to have a better understanding of risk factors for graft loss in a Latin-American population. The predictive performance of the models was mild.

Introduction

Kidney transplantation is the optimal renal replacement therapy for suitable patients with end stage renal failure [1]. Identifying risk factors for graft failure in kidney transplant recipients is useful for recognizing those patients at high risk and anticipating potential therapeutic interventions to improve graft survival [2, 3]. Risk factors in the field of organ transplantation are typically assessed using time-to-event outcomes, for instance, when recording time-to-death or time-to-graft loss. Survival analyses are key statistical tools in transplantation research [4]. This analyses are the most used methods to estimate the incidence of an outcome of interest, often censoring for a competing event [4]. For example, death is competing event to graft loss because a patient may die before losing the graft, as such no opportunity for graft loss in that case. Thus, competing events are present when another event precludes the event of interest. In this condition, the Kaplan-Meier (KM) approach is not suitable because this method assumes that censored patients are at the same risk as patients who remain in the study. In general, this leads to an overestimation of the cumulative incidence of the event of interest [5-7]. To solve this limitation, Kalbfleisch and Prentice introduced the Cumulative Incidence Function (CIF) [8]. The CIF calculates all events’ probability as the sum of the event of interest’s probabilities and those of the competing risks. The competing risk analysis (CRA) allows using a modified chi-squared test to compare CIF curves between groups and the Fine and Gray model with subdistribution hazards (sdHR) [9, 10]. Thus, patients are followed until observing the first of multiple event types in the CRA. Adjusting this fact, the inferred incidence of the event of interest is lower, and the sum of calculated incidences across all event types sums up to 100% [4, 10].

The CRA may provide further insights into the effect of interventions on the separate endpoints, comparing CIF curves to explore the association between covariates and the absolute risk. Indeed, CRA may be essential for clinical decision-making and prognostic research questions [11]. Despite this advantage, competing risk models have not been used frequently by researchers [2, 10]. In particular, the advantage of using CRA method was highlighted in a study evaluating race, age, and survival among patients undergoing dialysis, where accounting for transplant as competing risk brought to light a greater disparity in death on dialysis among younger black patients (related to disparity in access to transplantation) [12]. Other studies used the CRA method to evaluate risk of mortality and subsequent graft survival in older recipients after sustaining fracture [13], as well as the risk of graft loss and mortality in older recipients (age ≥60 y) receiving older kidneys (age ≥80 y) versus remaining on dialysis [14]. So far, there are no Latin-American studies related to competing risks analysis in transplantation. In this context and given the substantial variability of the identified risk factors for graft loss across different transplant populations, our transplant program (Colombiana de Trasplantes—CT) aims to use a well characterized Latin-American cohort of kidney transplant recipients with long term follow-up to determine the risk factors for graft loss in the presence of mortality as a competing risk event.

Materials and methods

Study design and population

We conducted a retrospective cohort study at Colombiana de Trasplantes. To give a context it must be said that Colombiana de Trasplantes is a transplant network in Colombia with 4 centers with around 21% of the annual national kidney transplant activity. We included first time, kidney transplant recipients aged ≥ 18 years who were transplanted between July 1, 2008, to May 31, 2019. Patients with primary renal graft thrombosis (arterial or venous) were excluded. Recipients were followed up to graft failure, death, or end of follow up at 5 years post transplantation, whichever was earliest.

Kidney donors

Informed consent was obtained from organ/tissue deceased donors (DD) by a family interview where both family and the donation team go through information related to answer inquiries about encephalic death, emotional support, and the possibility of organ donation. The main causes of death in our DD were cerebrovascular/stroke, followed by head trauma, anoxia, CNS tumor, or others. Organ donation and tissue consent form is provided in S1 Table in S1 File.

In the case of the live kidney donors, our transplant team provides kidney donation and nephrectomy informed consent and the affidavit from the live kidney donors. Overall, less than 1% of our kidney donors were previously registered as organ donors. According to Colombian law, the total of donor medical costs is covered including organ donor maintenance and procurement with an average of 5000 USD and without any economic contribution to the family donor.

Immunosuppression and follow-up protocol

All patients received standard induction therapy with alemtuzumab, basiliximab or antihuman thymocyte immunoglobulin according to immunologic risk or transplant clinical guidelines. All patients received a fixed-dose of methylprednisolone perioperatively for 3 days with a transition to fixed-dose oral prednisone from day 4 to day 7 in the postoperative period. One-week post transplantation steroids were withdrawn. Chronic immunosuppression consists of Calcineurin inhibitors-based therapy and mycophenolate mofetil. Patients are monitored closely in the first 4 weeks post transplantation, and they return for follow-up monthly thereafter.

The acute rejection was classified under parameters described by Banff (2015) [15]. Biopsy was performed on those patients with increase of serum creatinine by >20% from baseline. Our center does not perform biopsies per protocol.

Treatment for acute cellular rejection was started once the histological diagnosis was confirmed as follows: Methylprednisolone: 500 mg. IV / day in infusion for three days. Oral prednisolone from the fourth day at a dose of 0.5 mg / Kg / day divided into two doses and for two weeks. After completing the two weeks, a weekly decrease of 10 mg / day was made until reaching the previous dose that was received before the rejection episode. Serum creatinine was done 5 days after finishing the boluses. A response to corticosteroid treatment was defined with a decrease in serum creatinine greater than and equal to 20% of the patient’s baseline creatinine.

Histocompatibility tests performed in our center correspond to Human Leukocyte Antigens (HLA), Panel Reactive Antibodies (PRA) I and II, flow cytometric crossmatch and anti-HLA antibodies (the latter only in living donors when the crossmatch is positive). HLA matching is when the recipient and the donor shared the same HLA antigens (HLA-A,-B,-DR antigens) [16].

We do not perform routine preimplant biopsies. The decision to take or not the organs from expanded criteria donors is made by macroscopic evaluation of the graft and, if required, it is sent for histological evaluation. Indices like KDPI / KDRI are not considered for taking or allocating organs [17]. Organ allocation is made according to the allocation criteria for kidney transplantation in Colombia [18].

Outcomes

Our primary outcome of interest was graft loss, not including death with function. Graft loss was defined by center reported as permanent return to dialysis or retransplantation. Death was defined as mortality from any cause and was ascertained by review of the Colombiana de Trasplantes database which records patient’s death and supplemented with the National register Master File. Patients were censored at 5 years of follow-up since the last follow-up date if they were transferred to another transplant center or lost to follow-up. Thus, survival analyses were performed using a competing risk approach, where graft loss and mortality were treated as competing events.

Statistical analysis

Descriptive analysis

Descriptive statistics were used to report the population characteristics. Frequencies and percentages were used for categorical variables. The numerical variables were reported according to its distribution using mean and standard deviation for normally distributed variables, and median and inter-quartile range (IQR) for non-normally distributed variables. Multiple imputation was not considered as there were few missing values (5.9% of the total number of cases), and those values were at random. According to this, we performed a complete case analysis in the univariable and multivariable models.

Predictors

Prespecified variables based on published literature and those available in our data, were collected as potential risk factors for graft loss. Data collected included demographics, medical history and clinic characteristics of kidney transplant recipient and donor. Definition of predictor measurement is provided in Supplementary material (S1 Table in S1 File).

Incidence estimates

The overall incidence of graft loss and/or death at 5 years post transplantation was calculated by Competing risk analysis method (CRA) using cumulative incidence function (CIF) where mortality was treated as a competing risk with graft loss. Log Rank test for graft loss and death were compared in the entire population and in specific patient population including living and deceased donor.

Comparisons between the two groups (graft loss yes/no and death yes/no) were performed using modified χ2 test. The subdistribution Hazard Ratio (sdHR) also known as Fine and Gray model was calculated for each independent variable and the two outcomes.

Variable selection and prediction

Variables with p value <0.25 in an univariable analysis and those with clinical importance were selected to perform further analysis. Variable selection to build the final model for graft loss was performed comparing three methods:

Full model: contains all the predictors selected in previous analysis and no variable selection was done.

Backward selection based on the Akaike information criterion (AIC).

Backward selection based on the Bayesian Information Criterion (BIC).

The model was selected on the model´s better performance.

Bootstrapping technique was used for internal validation as split method for resample a single dataset to create many simulated samples. The prediction models were trained on B bootstrap samples with replacement. The models were assessed in the observations that were not included in the bootstrap sample. This allowed us to calculate the prediction error, brier score and c-index. Calibration plot was used to compare the predicted probability with the observed probability.

The Fine and Gray model directly models the covariate effect on CIF, and it reports the sdHR. To model the impact of covariates on graft loss, we used the Fine and Gray method [9] for performing competing risk regression. The association between the primary outcome and the independent variables were assessed by the sdHR.

The model development and report was based on The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) [19]. More details of modeling process can be found in Supplementary material (Modelling process in S1 File).

Analysis was performed using the Software R version 3.6.3. Library used to perform competing risk analysis was cmprsk [20].

Ethics considerations

This study was approved by the ethics research committee of the institution, acting in concordance with local and national regulations, as well as with the Helsinki declaration. Confidentiality of all patients was secured all the time during the execution of the research. None of the transplant donors was from a vulnerable population and all donors or next of kin provided written informed consent that was freely given.

Results

Patient characteristics

A total of 1454 out of 1621 recipients met selection criteria. Exclusions took place in 167 (113 pediatric transplants and 54 kidney transplants with graft thrombosis). In gender distribution most of patients were male, the overall mean age was 43.58 ± 13 years. Table 1 summarizes the clinical and demographic characteristics.

Table 1

Clinical and demographic characteristics of kidney transplant recipients.

Variable	Total	Graft loss		P-value	Death		P-value
		No	Yes		No	Yes
	(N = 1454)	(N = 1285)	(N = 169)		(N = 1317)	(N = 137)
Sex, n (%)				0.495			0.323
Female	586 (40.3)	525 (40.9)	61 (36.1)		539 (40.9)	47 (34.3)
Male	868 (59.7)	760 (59.1)	108 (63.9)		778 (59.1)	90 (65.7)
Age, mean (SD)	43.6 (13.2)	43.6 (13.1)	43.2 (14.3)	0.926	42.8 (13.2)	50.9 (11.3)	<0.001
BMI, mean (SD)	23.3 (3.82)	23.3 (3.79)	23.2 (4.04)	0.988	23.2 (3.81)	24.1 (3.78)	0.030
Missing	49 (3.4)	39 (3.0)	10 (5.9)		47 (3.6)	2 (1.5)
Cause of CKD, n (%)				0.946			0.076
Congenital	96 (6.6)	84 (6.5)	12 (7.1)		90 (6.8)	6 (4.4)
Unknown	638 (43.9)	569 (44.3)	69 (40.8)		583 (44.3)	55 (40.1)
Diabetic	200 (13.8)	174 (13.5)	26 (15.4)		167 (12.7)	33 (24.1)
Glomerular	272 (18.7)	241 (18.8)	31 (18.3)		256 (19.4)	16 (11.7)
Hypertensive	163 (11.2)	145 (11.3)	18 (10.7)		144 (10.9)	19 (13.9)
Obstructive	37 (2.5)	34 (2.6)	3 (1.8)		35 (2.7)	2 (1.5)
Other	48 (3.3)	38 (3.0)	10 (5.9)		42 (3.2)	6 (4.4)
RRT, n (%)				0.266			0.213
Hemodialysis	618 (42.5)	531 (41.3)	87 (51.5)		554 (42.1)	64 (46.7)
Peritoneal	447 (30.7)	406 (31.6)	41 (24.3)		415 (31.5)	32 (23.4)
Pre-Dialysis	181 (12.4)	165 (12.8)	16 (9.5)		168 (12.8)	13 (9.5)
Unknown	208 (14.3)	183 (14.2)	25 (14.8)		180 (13.7)	28 (20.4)
Time on dialysis, months (SD)	27.2 (35.4)	26.7 (34.7)	30.8 (40.4)	0.421	26.6 (35.5)	33.5 (34.2)	0.149
Time on waiting list, months (SD)	554 (596)	561 (602)	484 (541)	0.423	547 (594)	615 (623)	0.510
Medical history n (%)
CVD	47 (3.2)	38 (3.0)	9 (5.3)	0.262	41 (3.1)	6 (4.4)	0.728
Stroke	13 (0.9)	8 (0.6)	5 (3.0)	0.010	12 (0.9)	1 (0.7)	0.977
Hypertension	1162 (79.9)	1033 (80.4)	129 (76.3)	0.465	1045 (79.3)	117 (85.4)	0.242
DM	205 (14.1)	179 (13.9)	26 (15.4)		171 (13.0)	34 (24.8)	<0.001
Smoking	210 (14.4)	184 (14.3)	26 (15.4)	0.934	186 (14.1)	24 (17.5)	0.561
Type of donor, n (%)				0.176			<0.001
DD	1002 (68.9)	875 (68.1)	127 (75.1)		881 (66.9)	121 (88.3)
LD	452 (31.1)	410 (31.9)	42 (24.9)		436 (33.1)	16 (11.7)
ECD, n (%)	189 (13.0)	158 (12.3)	31 (18.3)	0.157	155 (11.8)	34 (24.8)	<0.001
CIT, hours mean (SD)	18.3 (14.4)	17.8 (13.3)	21.3 (20.3)	0.039	18.3 (15.1)	18.3 (7.74)	1
CIT >14 hours, n (%)	675 (46.4)	573 (44.6)	102 (60.4)	<0.001	588 (44.6)	87 (63.5)	<0.001
Match, n (%)				0.808			0.628
0	117 (8.0)	105 (8.2)	12 (7.1)		110 (8.4)	7 (5.1)
1	202 (13.9)	175 (13.6)	27 (16.0)		183 (13.9)	19 (13.9)
2	298 (20.5)	258 (20.1)	40 (23.7)		267 (20.3)	31 (22.6)
3	498 (34.3)	447 (34.8)	51 (30.2)		457 (34.7)	41 (29.9)
4	220 (15.1)	194 (15.1)	26 (15.4)		189 (14.4)	31 (22.6)
5	69 (4.7)	60 (4.7)	9 (5.3)		62 (4.7)	7 (5.1)
6	40 (2.8)	39 (3.0)	1 (0.6)		39 (3.0)	1 (0.7)
Missing	10 (0.7)	7 (0.5)	3 (1.8)		10 (0.8)	0 (0)
Qualitative PRA I, n (%)				0.321			0.477
Negative	748 (51.4)	663 (51.6)	85 (50.3)		685 (52.0)	63 (46.0)
Positive	78 (5.4)	63 (4.9)	15 (8.9)		73 (5.5)	5 (3.6)
Unknown	628 (43.2)	559 (43.5)	69 (40.8)		559 (42.4)	69 (50.4)
Qualitative PRA II, n (%)				<0.001			0.859
Negative	95 (6.5)	69 (5.4)	26 (15.4)		87 (6.6)	8 (5.8)
Positive	736 (50.6)	662 (51.5)	74 (43.8)		672 (51.0)	64 (46.7)
Unknown	623 (42.8)	554 (43.1)	69 (40.8)		558 (42.4)	65 (47.4)
CMV Disease, n (%)	76 (5.2)	62 (4.8)	14 (8.3)	0.165	58 (4.4)	18 (13.1)	<0.001
BKV nephropathy, n (%)	36 (2.5)	21 (1.6)	15 (8.9)	<0.001	35 (2.7)	1 (0.7)	0.385
Number of renal allograft biopsies, n (%)	1.01 (1.30)	0.874 (1.20)	2.08 (1.54)	<0.001	0.995 (1.30)	1.20 (1.36)	0.201
Acute cellular rejection, n (%)	473 (32.5)	368 (28.6)	105 (62.1)	<0.001	413 (31.4)	60 (43.8)	0.012
Serum creatinine at 12 months, mean (SD)	1.51 (0.848)	1.42 (0.588)	2.73 (2.10)	<0.001	1.48 (0.851)	1.84 (0.748)	0.004
Coronary stent	13 (0.9)	12 (0.9)	1 (0.6)	0.906	9 (0.7)	4 (2.9)	0.030
Number of hospital readmissions, mean (SD)	1.30 (1.81)	1.20 (1.77)	2.07 (1.92)	<0.001	1.26 (1.79)	1.70 (2.02)	0.024

SD: standard deviation, BMI: Body Mass Index, RRT: Renal Replacement Therapy, CVD: Cardiovascular Disease, DM: Diabetes Mellitus. DD: Deceased donor, LD: Live donor, ECD: Expanded criteria donor, CIT: Cold isquemia time, PRA: Panel Reactive Antibody Test, BMI: Body Mass Index, CMV: citomegalovirus, BKV: BK virus. Negative PRA test is indicative of a lack of anti-HLA antibodies.

Overall cumulative incidence

During the follow-up period graft loss occurred in 169 patients (11.6%) and death occurred in 137 (9.4%). Cumulative incidence for graft loss and death was 15.8% and 13.8% respectively. Fig 1 displays the combined cumulative incidence for the entire cohort. Significant differences in estimates of both outcomes were found when analyzing live and deceased transplant separately, where deceased transplant (17.1% and 16.3% for death and graft loss respectively) had greater cumulative incidences (deceased transplant 17.1% and 16.3% for death and graft loss respectively vs live 5.4% and 15% for death and graft loss respectively). Fig 2 shows the difference between type of transplant in the cumulative incidence of graft loss and death.

Fig 1

Cumulative incidence of graft loss and death estimated by the method for competing risk.

Fig 2

Log Rank of cumulative incidence of risk of death by type of transplant, and Log Rank of cumulative incidence of risk graft loss by type of transplant.

Risk factors for cumulative incidence of graft loss and death with functioning graft

We fit the Fine and Gray competing risk survival regression model for identifying the potential determinants of graft loss using covariates with significant association and those with clinical importance. The covariates that had a significant impact on the graft loss were stroke, cold ischemia time, qualitative PRA II, BKV nephropathy, number of allograft biopsies, acute cellular rejection, serum creatinine at 12 months and number of hospital readmissions. The covariates that had a significant impact on death were recipient age, diabetes mellitus, type of donor, expanded criteria donor, CMV disease, cold ischemia time, coronary stent, and number of hospital readmissions.

Table 2 provides the sdHR of risk factors estimated by using the final multivariate Fine and Gray model. The risk of graft failure was noted to increase in the presence of nephropathy due BK virus, higher rates of serum creatinine at 12 months post transplantation, and greater number of kidney biopsies. Significant risk factors associated with cumulative incidence of death were recipient age, deceased donor, CMV disease, coronary stent, and acute cellular rejection.

Table 2

Factors associated with graft loss using death as a competing risk in kidney transplant recipients in a final Fine and Gray model.

Characteristic	Graft loss outcome		p-value	Death outcome		p-value
	sdHR	CI 95%		sdHR	CI 95%
BKV nephropathy	4.43	2.02–9.72	<0.001	-	-	-
Serum creatinine at 12 months	1.76	1.55–2.00	<0.001	-	-	-
Number of renal allograft biopsies	1.45	1.28–1.64	<0.001	-	-	-
Recipient age (years)	-	-	-	1.039	1.02–1.05	<0.001
Living donor (Vs deceased)	-	-	-	0.386	0.21–0.68	<0.001
CMV Disease	-	-	-	2.459	1.46–4.11	<0.001
Coronary stent	-	-	-	3.032	0.99–9.23	0.05
Acute cellular rejection, n (%)	-	-	-	1.336	0.93–1.90	0.11

sdHR: subhazard distribution; CI: Confidence interval

Variable selection to build the final model for graft loss was performed comparing three methods:

Full model: contains all the predictors selected in previous analysis and no variable selection was done. 2. Backward selection based on the AIC. 3. Backward selection based on the BIC.

The performance and prediction error of the three models were evaluated using Bootstrap cross-validation, showing similar results for the AIC and BIC models. The C-index for the full model was 0.57, for the AIC model was 0.6 and, for the BIC model was 0.6.

The same process for variable selection and performance assessment was performed for death model. Similar results were obtained for the AIC and BIC models. The C-index for the full model was 0.78, for the AIC model was 0.72 and, for the BIC model was 0.72. Fig 3 provides the prediction errors and calibration plot of the final Fine and Gray model for graft loss. Fig 4 provides the prediction errors and calibration plot of the final Fine and Gray model for death.

Fig 3

Prediction errors and calibration plot of the final Fine and Gray model for graft loss.

Fig 4

Prediction errors and calibration plot of the final Fine and Gray model for death.

Discussion

Kidney transplantation is the best therapy available for most patients with end- stage kidney disease [21]. We developed two predictive models of risk of graft loss and risk of death in kidney transplant patients. Risk prediction models are useful for identifying kidney recipients at high risk of graft failure, and optimize clinical care, decision-making and resource allocation; is a challenging issue in kidney transplantation [2]. Our objective was characterized Latin-American cohort of kidney transplant recipients with long term follow-up and to predict the risk factors for graft loss in the presence of mortality as a competing risk event. We were able to identify 3 significant risk factors for graft loss and 5 significant risk factors for death. This contributes to have a better understanding of risk factors for graft loss in a Latin-American population.

Graft survival is one of the most critical concerns in kidney transplant recipients, and our ability to accurately monitor the cumulative incidence of graft loss its importance. Risk prediction models are useful for identifying kidney recipients at high risk of graft failure, thus optimizing clinical care. Therefore, using competing risks methods that provide more accurate estimates, we sought to identify risk factors leading to graft loss, considering death as a competing risk in kidney transplant recipients [4].

Particularly, one study highlights the advantage of using CRA method assessing both the probabilities of death and graft loss.

Risk factors for graft loss

Late failure of kidney transplants remains an important clinical problem [2, 22]. Renal allograft loss is multifactorial [23]. In the United States, 5469 kidney transplants developed end-stage kidney failure in 2008 (data provided by Jon Snyder from USRDS), making kidney transplant failure the fourth leading cause of end-stage renal disease. The reasons for failure are not well understood. Some have postulated that late deterioration reflects dysregulated fibrosis, drug toxicity or progressive “chronic allograft nephropathy” [24-26]. In our study we found that BKV nephropathy, serum creatinine at 12 months and increased number of renal allograft biopsies were significant risk factors for graft loss. Sellarés et al., attributed causes of graft failure in the biopsy-for-cause population to antibody-mediated rejection (ABMR), probable ABMR or mixed rejection, with nonadherence recorded in nearly half. There was evidence of ABMR in 18 of 19 nonadherent patients who failed. There were also three groups of nonrejection causes of failures: glomerulonephritis, BKV nephropathy and failure in the context of an intercurrent illness. The results emphasize the burden of ABMR and mixed rejection and its interaction with nonadherence in observed failures, making these key targets for further progress. The results also illustrate the range of clinical courses leading to failure and the sometimes-complex relationships to the indication biopsy findings [22].

On the other hand, renal allograft biopsy (RAB) is still the best approach to diagnose renal transplant complications [27]. We found that kidney recipients with more significant requirements to perform RAB had greater risk of graft loss. According to our guidelines, biopsy was performed on those patients with increase of serum creatinine by >20% from baseline, generally when acute or chronic renal allograft rejection is suspected, antibody-mediated rejection, polyoma virus nephropathy, glomerular diseases, atrophy u other. Thus, greater number of RAB may be associated with renal allograft disfunction or detection of other lesions that may influence graft loss [27-29]. However, as a limitation, we do not have an electronic database of all the histological findings of dysregulated fibrosis and C4d of the renal biopsies of our patients. Of those factors related to graft loss, there is a high magnitude of association with BKV nephropathy [30, 31]. Previous studies have associated the BKV nephropathy with premature loss of kidney function [32-35], graft loss and alteration of renal histology [21–23, 36]. The reactivation of the virus may occur with the use of immunosuppression. Polyomavirus BK virus reactivation in kidney transplant recipients can lead to BK polyoma virus-associated nephropathy (BKPyVAN), which is associated with graft dysfunction in >90% and graft loss in over 50% of the affected individuals [37].

Our results also showed an association with higher serum creatinine level at 12 months. This factor has been widely described as predictor of graft loss [3, 38–45].

The identification of risk factors for graft loss in the long term has been provided by several studies, however, there is substantial variability in data collection, the methods used for model development and included predictors [2]. Among others, the most described predictors are: chronic dysfunction [38, 42, 46], episodes of acute rejection [3, 38–41], death with functional graft [38, 46], glomerulonephritis [38], donor age [47], hypertension [47, 48], diabetes [41, 47], type of immunosuppression [47], delayed graft function [47], recipient age [3], race [3], albumin [3], proteinuria [3, 42, 47], low-density lipoprotein (LDL) cholesterol levels [48] and higher BMI [49]. However, some of them included in the analysis but that finally were not significant.

Risk factors for death

Identification and quantification of the relevant factors for death can improve patients’ individual risk assessment and decision-making [50]. In this study we confirm risk factors for death like recipient age, deceased donor, CMV disease, CMV disease, coronary stent, acute cellular rejection. Our findings show that older recipients are more likely to die, which is consistent with several published studies that report youngest age groups demonstrated a clear trend toward lower mortality compared with those ≥60–65 years [50-53]. However, it must be said that long term patient survival in the elderly has been shown to be significantly better in transplant patients compared with remaining on the waitlist [54-57]. Similarly to what happens with large series (Collaborative Transplant Study [58] and UNOS Register [59]), it is observed that living-donor kidney transplantation provides better outcomes than deceased-donor transplantation. Besides, it is associated with shorter transplant waiting list period and better early outcome [60]. We have found that CMV disease represents a risk of death in our population. This is one of the most important infectious complications in transplant recipient leading to significant morbidity and mortality [61]. Various direct and indirect detrimental effects occur because of CMV infection on patients and grafts. Indirect effects may include rejection, immunosuppression resulting in infections by other microorganisms, graft dysfunction, and poor survival of the kidney graft [62].

Cardiovascular disease (CVD) is frequent after kidney transplantation, is a major cause of morbidity and of death with functioning graft in recipients [63, 64]. We found as a risk factor in the model that coronary disease, specifically coronary stent placement, as a risk factor for death. OPTN/SRTR 2017 Annual Data Report: Kidney, Death with a functioning graft is the leading cause of graft loss in kidney transplant recipients, and a major cause of death is cardiovascular disease, accounting for about one third of known causes [65]. Another of the factors related in the model with the death of kidney transplant patients that we found was the presence of Acute Rejection (AR). Clayton et al., proposed that AR and its treatment may directly or indirectly affect longer-term outcomes for kidney transplant recipients, they found AR was also associated with death with a functioning graft (HR, 1.22; 95% CI, 1.08 to 1.36), and with death due to cardiovascular disease (HR, 1.30; 95% CI, 1.11 to 1.53) and concluded AR is associated with increased risks of longer-term graft failure and death, particularly death from cardiovascular disease and cancer. The results suggest AR remains an important short-term outcome to monitor in kidney transplantation [66].

Previous studies have attempted to identify and integrate risk factors for death into predictive models, including the pre-transplant variables gender, race, body mass index (BMI), time on dialysis, cause of end-stage renal disease, panel reactive antibodies, HLA mismatches, comorbidities such as diabetes, and heart failure, and donor age. In some models, the post-transplant factors Delayed graft function (DGF), and graft function were included [50], however, in our study population these were not significant. We think that in the case of diabetes the sample size was not sufficient.

Strengths and limitations

Unlike most previous studies, the main strength of this study is that our analysis includes a competing risk model. Many papers have pointed out the important issue of competing events in kidney transplantation [2, 4, 5, 7]. This method allowed us to determine graft loss risk factors differentiating those who increase recipient mortality. We believe that this integral view is best suited to a rational and patient-centered risk assessment.

Further, our cohort is the largest reporting risk factors for graft loss and death by a unique transplant program in Colombia and contributes to have a better understanding of Latin-American population as most of previous studies have been reported by transplant programs that treat mainly Caucasian patients. Other strengths include consistent data collection with a high degree of completeness and several variables.

Potential limitations attendant with the nature of data collection. The retrospective nature of our study prohibited adjusting for unmeasured confounding factors that may explain the association between independent factors and adverse graft outcomes. Besides, donor age was no considered in our analysis due to no available information.

On the other hand, variable selection with backward regression is not ideal. A fundamental problem with stepwise regression is that some real explanatory variables that have causal effects on the dependent variable may happen to not be statistically significant, while nuisance variables may be coincidentally significant. As a result, the model may fit the data well in-sample but do poorly out-of-sample. Unfortunately, penalized methods for Fine-Gray models have some limitations and the output from the crrp () function does not include convenient parameters such as a p-value. In addition, this package has not been maintained since its first commit in 2015.

We did not perform external validation, and this could be useful to assess the generalizability to other similar populations.

Conclusion

In summary, we found that stroke, BKV nephropathy, serum creatinine at 12 months and increased number of renal allograft biopsies were significant risk factors for graft loss. On the other hand, recipient age, acute cellular rejection, CMV disease were risk factors for death, and recipients with living donor had better survival compared to deceased-donor transplant and coronary stent. This contributes to have a better understanding of Latin-American population. However, the predictive performance of the models was mild.

(DOCX)

Click here for additional data file.

28 Sep 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Justyna Gołębiewska

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that your study involved tissue/organ transplantation. Please provide the following information regarding tissue/organ donors for transplantation cases analyzed in your study.

1. Please state in your response letter and ethics statement whether the transplant cases for this study involved any vulnerable populations; for example, tissue/organs from prisoners, subjects with reduced mental capacity due to illness or age, or minors.

- If a vulnerable population was used, please describe the population, justify the decision to use tissue/organ donations from this group, and clearly describe what measures were taken in the informed consent procedure to assure protection of the vulnerable group and avoid coercion.

- If a vulnerable population was not used, please state in your ethics statement, “None of the transplant donors was from a vulnerable population and all donors or next of kin provided written informed consent that was freely given.”

2. In the Methods, please provide detailed information about the procedure by which informed consent was obtained from organ/tissue donors or their next of kin. In addition, please provide a blank example of the form used to obtain consent from donors, and an English translation if the original is in a different language.

3. Please indicate whether the donors were previously registered as organ donors. If tissues/organs were obtained from deceased donors or cadavers, please provide details as to the donors’ cause(s) of death.

4. Please provide the participant recruitment dates and the period during which transplant procedures were done (as month and year).

5. Please discuss whether medical costs were covered or other cash payments were provided to the family of the donor. If so, please specify the value of this support (in local currency and equivalent to U.S. dollars)

3. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Here are my comments/questions to the authors:

Abstract: the conclusion does not reflect the results provided in that section which may confuse a reader that has not yet read the paper.

Introduction:

First paragraph providing details on CKD epidemiology not needed.

Line 78-79: most helpful to say: death is competing event to graft loss because a patient may die before losing the graft, as such no opportunity for graft loss in that case.

Please clarify the difference between CFI and CRF: authors seem to use the terms interchangeably, best to limit to one term to avoid confusion.

Methods

152: AIC and BIC methods: please spell out the first time mentioned (I did not see that).

Results

Tables 1 and 2: would be helpful to either put the p values or indicate where the p values are significant

Table 2: post-op complications: the ureteral leak statistic under graft loss (section of no graft loss) is not clear where it came from: 170 (77.5%) is that the percentage of total of complications: does not fit with the remainder of statistics. Please clarify if typo.

Table 2: any data on histology such as TG or IFTA and effect on graft survival?

Tables 3 and 4:

I am surprised that stroke was found to be statistically significant association with graft loss independent of other factors: how do you explain this result? It is a reflection of uncontrolled cardiovascular risk factors such as HTN and HLD?

I am sceptical of the data related to number of biopsies as independent association with graft loss: unless the rate of biopsy complications at the centers studied is very high, the number of biopsies itself should not be a risk factor for graft loss, rather a reflection of some persistent allograft dysfunction requiring multiple biopsies (indication bias). This would typically be either high creatinine (authors included the one year creatinine but might be useful to look at earlier time points such as 3 or 6 months and later such as 2 years) and proteinuria (which is not included here and would be very useful to include in the analysis).

Expanded criteria donor is an older term best replaced with the KDPI if available.

I am surprised that DM did not make it as an independent association with graft loss and seems to have been supplanted by obesity (although the actual HR in the univariate analysis for DM were higher than for obesity). This is a very interesting result which seems to suggest that obesity rather than diabetes is responsible for the excess mortality. Please clarify: 1) history of DM: is that only pre-transplant DM or was new onset post-transplant DM included?. 2) what is the proportion of DM type 1 vs type 2?

Discussion

220: please include reference 9 in that sentence as well.

You have the same reference listed as 9 and 19: please consolidate.

Line 243: I would argue that there is a significant amount of data that links BK nephropathy with graft loss and unfavorable histology including the following that would be important to acknowledge:

Gago et al: Kidney allograft inflammation and fibrosis, causes and consequences. PMID: 22221836

Sellares et al: Understanding the causes of kidney transplant failure: the dominant role of antibody-mediated rejection and nonadherence. PMID: 22081892

El-Zoghby et al: Identifying specific causes of kidney allograft loss. PMID: 19191769

Naesens et al: The histology of kidney transplant failure: a long-term follow-up study. PMID: 25243513

Paragraph 260-266: I am not clear as to the point of this paragraph. Please clarify how does that link to your analysis and how your analysis is better and risk factors that you included are better than what you listed from the existing literature. I do agree as pointed above that proteinuria is an important risk factor that is not included in your analysis.

You also allude to unfavorable histology as possible explanation to the number of biopsies being a risk factor and stroke: as I asked above, would be interesting to have that data analyzed in your cohort.

In the paragraph of risk factors for death, it is important to address why DM did not remain an independent risk factor of death (contrary to other studies) and give possible hypotheses for that observation.

Reviewer #2: Since the KM model does not account for competing risk, the KM model should not be used at all. As the title clearly state “a competing risk analysis”, the KM model should be omitted. The author may try different competing risk models though.

The conclusion is not accurate and did not focus on the primary aim of this paper which is to identify risk factors. Even though the results of KM and competing risk analysis are different, you cannot claim that KM overestimates the incident based on a specific data. You need to perform more rigorous methodology analysis to compare and validate the two models.

A propensity score matching should be performed.

Were the final models in Table 4 selected by p value or AIC/BIC? Please clarify.

Minor concerns:

Use “univariable/multivariable” instead of “univariate/multivariate”.

Cite the reference to the competing risk analysis.

Add p values to Table 1 and 2.

Table 3 footnote: “IC” should be “CI”.

The “Model performance” section in the results seems to be more “methods” not “results”.

Reviewer #3: I thank the authors and editors for the opportunity to review this interesting manuscript.

Summary

Using a 10-year long retrospective cohort of first-time kidney transplant from Columbia, Pinto-Ramirez et al. applied competing risks analyses and used forward stepwise predictor selection to identify risk factors for graft failure and patient survival. Stroke, BKV nephropathy, acute cellular rejection, serum creatinine, and increased number of renal biopsies were significant predictors for graft loss, while PRA II negative was protective. Recipient age, obesity, and cytomegalovirus disease were associated with an increased risk of death, while living donor transplantation was protective.

Major

It is important for the authors to clarify the main message of the manuscript. Is it to (i) educate the readership on competing risks analysis or (ii) respond to the main aim of identifying risk factors for graft failure and patient survival among kidney transplant recipients from Columbia? While application of suitable analytical methods as outlined in the METHODS section is encouraged, the discrepancy between KM and CIF estimates (derived from competing risks analysis) in and of itself, is not novel. Thus, the main contribution of this manuscript relates to the development of prediction models in kidney transplant recipients form Columbia. This could be better emphasized in the INTRODUCTION and DISCUSSION sections of the article.

Consultation of the STROBE and TRIPOD statements on reporting of observational studies and prediction models, respectively, is advisable.

Additional points for consideration regarding predictor variable definition and handling, as well as for prediction model development are outlined below:

ABSTRACT:

- The METHODS section should clarify predictor variable definition and handling, as well as the variable selection process for each model depending on outcome

- The main conclusion should be aligned with the main objective(s) of the manuscript.

INTRODUCTION:

- Kindly review previously published prediction models for similar outcomes in kidney transplant recipients and the important predictors identified from those.

METHODS:

- Please mention protocols for treating rejection

- Please discuss HLA (in)compatibility assessment pre-transplant and how it affects allocation decisions

- Definition, handling, and timing of predictor measurement

o What period was considered for the “number of biopsies” variable definition?

o Why was donor sex not considered among the predictor variables?

- Multiple imputation for handling missing values is advisable

- Modelling technique and predictor selection

o Why were no prespecified variables (based on published literature) included in the prediction model?

o Please discuss how correlation between variables was considered prior to fitting the final prediction model

o Please consider LASSO (least absolute shrinkage and selection operator) for fitting parsimonious prediction models for the study endpoints.

o Given the consideration of post-transplant characteristics (e.g., 12-month creatinine and number of biopsies), was landmark analysis or time-varying analysis considered?

- Please specify the timeframe the models intend to prognosticate for

- Justification is needed for lack of internal validation of the final prediction models

RESULTS:

- Consider organizing variables presented in Table 3 based on recipient, donor, and transplant characteristics. For the latter, also consider timing in reference to transplantation

- Is it “biological sex” or “gender” that is considered among the predictor variables?

- FIGURES 1 to 3

o Please use a similar scale for the y-axis for all figures.

o Please modify Y-axis for Figure 1 to indicate CIF of graft failure and death

o Please clarify if Figure 3 provides CIF and Gray test or K-M estimates and log rank

- The model performance section in its current form is better placed in the METHODS section. Please report on model discrimination (e.g., C-statistic) with confidence intervals in the RESULTS

- For which models were AIC and BIC compared?

- For the final models, please provide equations including intercept and regression coefficients

DISCUSSION:

- The first paragraph should summarize the main novel observations. Currently it reads as the rationale for the study.

- Limitations

o Please discuss limitations of forward stepwise selection for prediction model specification

o Please mention and justify lack of internal validation (as well as external validation)

- Please explain the utility of the final prediction models and how they could be incorporated in clinical care

Minor

- Line 95: “competitive” should be replace with “competing”

- Line 134: “available case analysis” should be replaced with “complete case analysis”

- Define PRA II negative (i.e., no preformed antibodies against class II HLA)

- Does BMI in Table 2 relate to donor or post-transplant recipient BMI?

- Define “Match” (e.g., HLA-A,-B,-DR antigens)

- Which classification was used for rejection diagnoses?

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes: Ruth Sapir-Pichhadze

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

13 Apr 2022

Bogotá DC, Abril 1th 2022

Dear Doctor,

Emily Chenette, PhD

Editor-in-Chief

PLOS ONE

Response Letter

We summarized the corrections of the research manuscript “Risk factors for graft loss and death among kidney transplant recipients: A competing risk analysis”, please see below:

Journal requirements

1. We ensured our manuscript met PLOS ONE's style requirements.

2. Ethics statement: None of the transplant donors was from a vulnerable population and all donors or next of kin provided written informed consent that was freely given.

3. We provided the original and the English version of the “Organ and Tissue Donation Consent Form” from deceased donors.

4. The participant recruitment dates and the period during which transplant procedures were done are provided in the methods section.

5. In the methods section is specified thar “the total of donor medical costs is covered including organ donor maintenance and procurement with an average of 5000 USD and without any economic contribution to the family donor”.

6. We provided a repository of information for our data that is available with DOI: 10.17632/3j2m4ftn69.1

Reviewer 1

Abstract:

- We changed the conclusion

Introduction:

1. CKD epidemiology was removed from the manuscript.

2. We corrected the paragraph in lines 78-79 about death as a competing event to graft loss.

3. We did not use the Competing Random Forest (CRF) to analyze our data. Instead, we used the Cumulative Incidence Function (CIF).

Methods

4. Line 152: We spelled out The Akaike information criterion (AIC) and the Bayesian Information Criterion (BIC).

5. Tables 1 and 2: We added the p values that are significant in Table 1, and we removed further Tables that were all summarized in one table.

6. Table 2: Table 2 was removed

7. Table 2: We did not have any TG or IFTA data available to analyze the effect on graft survival. Table 2 was removed.

8. Tables 3 and 4: Our center does not perform protocol biopsies. We perform kidney biopsies only with clinical indications. Thus, this could be the reason for the correlation between the number of biopsies and graft loss (indication bias). Tables 3 and 4 were summarized in Table 2.

9. We do not have KDPI available for donors in our country.

10. 1)We included only pretransplant DM, 2) We did not have available data about the proportion of DM type 1 and 2.

Discussion:

11. Line 220: We added reference 9 and consolidate it.

12. Line 243: We included the articles suggested about BK nephropathy and graft loss.

13. Paragraph 260-266: in the methods section we specified how were collected the potential risk factors for graft loss, we included those published from the existing literature and also those available in our data. Those predictors not included are recognized as a limitation.

14. Why DM did not remain an independent risk factor of death (contrary to other studies)? We think that in the case of diabetes the sample size was not sufficient

Reviewer 2

1. Since the KM model does not account for competing risk, we removed the KM model.

2. We changed the conclusion and focused it on the aim of this paper which is to identify risk factors.

3. Propensity scores are used to balance observed covariates between subjects from the study groups to mimic the situation of a randomized trial (Joffe & Rosenbaum, 1999) and can be used for matching, stratification, or in a regression model as a covariate or weight (Rubin, 1997; D’Agostino, 1998). Because propensity scores are used to address potential confounding by indication, they would not be expected to improve pure prediction, which is not concerned with specific coefficient estimation. Additionally, propensity scores are estimated from regressions that comprise the same covariates included in the traditional prediction models, and only those covariates, thus it would seem mathematically impossible for the propensity scores to add anything – they are simply functions of the same variables already included in the traditional models. Despite this argument, requests for the addition of propensity scores to pure prediction models persist https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3740143/

4. We specified how the final models were selected in the methods section.

5. We used “univariable/multivariable” instead of “univariate/multivariate”.

6. We cited the reference to the competing risk analysis.

7. We Added p values to Table 1 and 2. (now summarized in table 1)

8. Table 3 footnote: “IC” should be “CI”. (now in table 2)

9. We removed The “Model performance” section from the results

Reviewer 3

Major

1. We emphasized the development of prediction models in the INTRODUCTION and DISCUSSION sections of the article.

2. We consulted and used STROBE and TRIPOD statements

ABSTRACT:

3. We clarified the predictor variable definition and handling, as well as the variable selection process for each model depending on outcome in the methods section

4. The main conclusion was improved to be aligned with the main objective(s) of the manuscript.

INTRODUCTION:

5. We reviewed previously published prediction models for similar outcomes in kidney transplant recipients and the important predictors identified from those.

METHODS:

6. We mentioned protocols for treating rejection

7. We discussed HLA (in)compatibility assessment pre-transplant and how it affects allocation decisions.

8. We provided the list of predictors and variables in a supplementary material. We also clarified predictors in the method section

9. Multiple imputation for handling missing values was no performed as there were few missing values

10. We specified the Modelling technique and predictor selection

11. We considered LASSO for fitting parsimonious prediction models. Unfortunately, for the package crrp:::crrp (this package is useful for penalized Fine-Gray models, using LASSO, SCAD, MCP, and their group versions) there is no predict function, and the output from the crrp() function does not include convenient parameters such as a P value. There are standard errors, so a P value can be manually calculated. Creating a tidy wrapper may be difficult. In addition, this package has not been maintained since its first commit in 2015.

12. Analysis of time-dependent covariates has some limitations in both the survival analysis and the competing risks setting. When different causes are acting simultaneously, the main interest is in estimating quantities such as cause-specific hazards, cumulative incidences or marginal survival probabilities. If random (internal) time-dependent covariates are to be included in the modeling process, then it is still possible to estimate cause-specific hazards but prediction of the cumulative incidences and survival probabilities based on these is no longer feasible (Andersenet al., 1993, Chapter VII; Kalbfleisch and Prentice, 2002, Chapter 6). This limitation is also encountered in the regression approach for cumulative incidences (or ‘‘subdistribution hazards’’) (Fine and Gray, 1999) and has been further discussed by Latouche, Porcher, and Chevret (2005) and Latoucheet al.(2007) and tackled by Beyersmann and Schumacher (2008)

13. We specified the timeframe the model’s intent to prognosticate for in the methods section (5 years postrasplant)

14. Internal validation of the models was performed.

RESULTS:

15. Table 3 was removed.

16. Biological sex was considered among the predictor variables, not gender.

17. Figure 3, is now figure 2. We change the scale of figures.

18. We indicated that figure 3 (now figure 2) shows CIF of graft failure and death

19. The model performance section is explained in the methods section.

20. The model discrimination and calibration were reported

21. Characteristics of modelling are provided in a supplementary material

DISCUSSION:

- We summarized the main novel observations in the first paragraph

-We discussed the limitations of stepwise regression selection.

- We performed internal validation using resampling techniques.

Minor

1. Line 95: We replaced “competitive” with “competing”.

2. Line 134: We replaced “available case analysis” with “complete case analysis”.

3. We defined PRA II negative in Table 1.

4. We corrected the observation about BMI in Table 2 which is related to the post-transplant recipient BMI (Now summarized in Table 1).

5. We gave a definition of a match.

6. We added the Banff classification for rejection diagnoses.

Submitted filename: Response Letter.docx

Click here for additional data file.

2 Jun 2022

Risk factors for graft loss and death among kidney transplant recipients: A competing risk analysis

PONE-D-21-25029R1

Dear Dr. Pinto-Ramirez,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Justyna Gołębiewska

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: (No Response)

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: (No Response)

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: (No Response)

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: (No Response)

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

**********

6 Jul 2022

PONE-D-21-25029R1

Risk factors for graft loss and death among kidney transplant recipients: A competing risk analysis

Dear Dr. Pinto-Ramirez:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Justyna Gołębiewska

Academic Editor

PLOS ONE