Objectives: This study was undertaken to derive a risk prediction model for skin cancer utilizing the United Network for Organ Sharing database population
Materials and Methods: Of the 24?734 adults (>18 years old) heart transplant recipients (2000-2015) in the United Network for Organ Sharing database, 2625 recipients developed skin cancer. Univariate and multivariate Cox regression analyses were performed; P values, hazard ratios, and confidence intervals were derived. The model was tested using receiver operating characteristics curves and area under the curves. MATLAB software (MathWorks) was used for analyses.
Results: Multivariate analysis showed that White patients had a hazard ratio of 31.7 compared with Black patients (P < .001). Male patients had a hazard ratio of 2.52 (P < .001) compared with female patients. Malignancy at listing showed a hazard ratio of 1.77 (P < .001). Thymoglobulin had a hazard ratio of 1.19 (P = .005) compared with other induction agents. The receiver operating characteristic curves generated for 5 years, 8 years, and 10 years after transplant showed area under the curve values of 0.78, 0.77, and 0.76, respectively, in the training set and 0.75, 0.75, and 0.74, respectively, in the validation set.
Conclusions: Male sex, White ethnicity, older age, malignancy at the time of listing or at time of transplant, and thymoglobulin induction are major risk factors for skin cancers after transplant. This risk prediction model has a C statistic of 0.75. To our knowledge, this is the first time such a model has been generated for skin cancers in this population.
Key words : Cardiac transplant, Latitude effects, Risk prediction model, Skin cancers, Thymoglobulin induction
Skin cancers are the most common malignancies among solid-organ transplant recipients,1 and the incidence of skin cancer in heart transplant patients is higher versus the general population.2 Analyses have been conducted to predict the risk of skin cancers after solid-organ transplant.3-7 A few studies have investigated risk factors of posttransplant skin cancer in heart transplant patients and the relevance of immunosuppressive therapy.2,8-13 However, there have been relatively few studies from large registry data sets on prediction of the risk of skin cancer after heart transplant. As treatment outcomes improve and advancements are made in immunosuppressive therapy, it is necessary to document the risk of skin cancer in heart transplant recipients in more recent years. We studied heart transplant recipients for the period 2000-2015 in the United Network for Organ Sharing (UNOS) database to investigate the risk for the development of skin cancers.
Materials and Methods
Data source and study population
The study was approved by the local institutional review board. We studied 24?374 adult patients (?18 years old) from the UNOS database who received a heart transplant between January 1, 2000, and October 31, 2015. Patients who received multiple organ transplants were excluded from the study. Both pretransplant and posttransplant data and baseline characteristics were analyzed. The data included 5995 male recipients, with mean age at transplant of 52.1 ± 12.6 years and mean post-transplant follow-up of 5.3 years (range, 1-188 months). For patients who did not develop any types of skin cancer, the time of last follow-up was used; for patients who developed skin cancers, time to the first diagnosis was extracted.
A Cox proportional hazards regression model was used to investigate the association of risk factors with posttransplant skin cancer events. Univariate analysis was performed to identify significant risk factors. Variables with P < .1 were further analyzed by multivariate analysis. Patient data were divided into derivation (80%) and validation (20%) sets. A multivariate model was developed from the derivation set and tested on the validation set. Stepwise forward selection was performed to determine the effect of each variable on the model predictions, and variables with significant effects were included in the final model. The proportional hazards assumption was tested, and variables that did not satisfy the assumption were stratified to generate different baseline hazard rate functions. The multivariate model was used to predict the probability of a patient to develop skin cancer 5 years, 8 years, and 10 years after transplant. Specifically, patients were separated into cancer and noncancer groups based on whether they had developed skin cancer after 5 years, 8 years, or 10 years. The multivariate model was then used to compute the probability of a patient to develop skin cancer, and the prediction accuracy was evaluated using area under the receiver operating characteristic curves. Furthermore, a risk score was developed from the multivariate model to stratify patients into different risk groups, and the predicted risk and observed risk to develop skin cancer were compared. Kaplan-Meier survival curves were generated to compare the risk to develop skin cancer between different groups, and the log-rank test was performed to quantitatively compare the intergroup difference. All the analyses were performed with MATLAB software (MathWorks).
The final study cohort had 24?374 recipients after excluding patients with unknown status at transplant and/or unknown skin cancer event. The chi-square test was done to compare patients’ characteristics between the cancer group and the no-cancer group for all variables (Table 1). Patients in the cancer group were older, had fewer human leukocyte antigen (HLA) mismatches, and were more likely to have coronary artery disease and less likely to have a congenital heart defect before transplant. Patients in the cancer group were more likely to be male. More patients in the skin cancer group were in status 1B and status 2. The study cohort was composed of 71.6% White, 17.5% Black, and 7.22% Hispanic patients, and 3.64% of patients were from other ethnic backgrounds. Compared with patients without skin cancer, patients who developed skin cancer were more likely to be White. In addition, patients who had skin cancer weremore likely to have malignancy at listing or malignancy at transplant. Patients in the cancer group were more likely receive induction with muromonab CD3 (OKT3) and daclizumab than with basiliximab.
Frequency and distribution of skin cancer
Of 2625 recipients who were diagnosed with skin cancer, there were 1880, 1211, and 200 patients who were diagnosed with basal cell carcinoma, squamous cell carcinoma, and melanoma cancer, respectively, and 638 had more than 1 type of cancer. The cumulative incidence rate of skin cancer in all heart transplant recipients is shown in (Figure 1). The risk to develop skin cancer for all recipients increased from 0.091 after 5 years to 0.191 after 10 years. Male recipients had a higher risk (P < .001) than female recipients, and the 10-year incidence rate of skin cancer for male recipients was 0.223. The incidence rates by ethnicity are plotted in (Figure 2A), where White patients had a significantly higher probability to develop skin cancer compared with the rest of the patients. The probability of a White patient to develop skin cancer 5 years after transplant was 0.121 and 10 years after transplant was 0.244. A log-rank test indicated significant differences (P < .001) between different groups, and Black patients had the lowest risk to develop skin cancer after transplant.
Risk prediction and stratification
The univariate Cox regression analysis indicated that age, sex, ethnic background, HLA mismatch level, panel reactive antibodies, malignancy at listing or at transplant, and induction with OKT3 were highly associated (P < .001) with posttransplant skin cancer development. In addition, induction with thymoglo-bulin was identified as a risk factor for posttransplant skin cancer (P < .05), and congenital heart defect showed protective effects (hazard ratio [HR] = 0.284; P < .001) against posttransplant skin cancer. The multivariate analysis selected 7 risk predictors that were highly correlated with posttransplant skin cancer events. Age was categorized into 3 groups: >60 years old, >40 and ?60 years old, and ?40 years old. A stratified Cox model was developed, and different baseline hazard rate functions were generated for patients who were up to 40 years old, over 40 years old but not older than 60 years, and older than 60 years. Scaled Schoenfeld residuals were used to test the proportional hazards assumption, and a global P value of .13 was obtained, indicating that the Cox model satisfied the assumption. The HR values of the predictors are reported in (Table 2). We excluded patients who had OKT3 and daclizumab from the multivariate analysis because these 2 drugs are not used presently. The final multivariate model predicted skin cancer 5 years, 8 years, and 10 years after transplant and provided values for the area under the receiver operating characteristic curves of 0.78, 0.77, and 0.76, respectively, for the derivation set and 0.75, 0.75, and 0.74, respectively, for the validation set. To determine the risk level of patients, a risk score was developed based on the HR value of each risk factor from the multivariate model (Table 2). Patients >60 years old were given 2 points, and patients over 40 years old but not older than 60 years were given 1 point. White patients were assigned 3 points. Patients with HLA mismatch ?2 were given 2 points, and patients with HLA mismatch of 3 or 4 were given 1 point. Male sex, latitude below the 32nd parallel north, malignancy at listing, and induction with thymoglobulin each increased the risk by 1 point. These scores separated all patients into 4 groups: very low risk, low risk, intermediate risk, and high risk. White ethnicity and male sex were identified as significant risk factors. The 10-year incidence rates of posttransplant skin cancer among White, Hispanic, Black, and other patients were 0.244, 0.058, 0.006, and 0.039, respectively, and among male and female patients were 0.223 and 0.087, respectively ((Figure 1) and (Figure 2A)). Age is another significant risk factor. Patients who were ?40 years old, between 40 years old and 60 years old, and >60 years old, respectively, had 10-year posttransplant skin cancer risk of 0.038, 0.179, and 0.311 (Figure 2B). These 3 factors were consistently identified as risk factors for the development of posttransplant skin cancer.1-3 At transplant, 101 patients had malignancy, among which 23 had skin cancer, 4 had genitourinary cancer, 2 had breast cancer, 4 had thyroid cancer, 2 had lung cancer, 5 had leukemia cancer, and the rest had other or unknown types of cancer. Patients with malignancy at transplant had significantly higher risk to develop skin cancer, and the 1-year, 3-year, 5-year, and 10-year incidence rates were 0.059, 0.153, 0.218, and 0.334, respectively (Figure 2C). Patients reported with malignancy at listing also had higher risk to develop skin cancer. At listing, 1412 patients had malignancy, and their cumulative risk to develop skin cancer 5 years and 10 years after transplant was 0.164 and 0.264, respectively (Figure 2C). Thymoglobulin was found to increase the risk of skin cancer (by 1.40% at 5 years after transplant) compared with the rest of the induction drugs (P = .008). Patients with an HLA mismatch level ?2 had a significantly higher risk for skin cancer compared with those with an HLA mismatch level >2 (Figure 2D). A possible explanation is that patients with a lower HLA may have a less activated tumor surveillance system, which protects them against skin cancer after heart transplant.2 The predicted and observed probability to develop skin cancer 5 years after transplant is given in (Figure 3). The Kaplan-Meier estimators of each risk group are shown in (Figure43). The log-rank test indicates a significant intergroup difference (P < .001).
Skin cancers are a common complication after cardiac transplant. Hence, development of a risk prediction score would help improve surveillance, diagnosis, and early treatment and thereby improve outcomes. The risk factors identified in this study that have not been reported in the past are thymoglobulin induction and HLA mismatches. White ethnicity has the highest risk versus intermediate risk for Hispanic patients and low risk for Black patients, which is suggestive of the pro-tective role of higher concentrations of skin melanin. More patients in the skin cancer group were diagnosed with coronary artery disease (P < .001), but the univariate analysis did not identify coronary artery disease as significant. In addition, congenital heart defect showed protective effects against posttransplant skin cancer. However, since this is correlated with age, the multivariate model did not include congenital heart defect as a risk predictor. Following the previous study,3 latitude was utilized as a surrogate indicator for the sun exposure level. Our analysis found that patients in the cancer group were more likely to live in the low latitude area, ie, area with greater sun exposure level, which is consistent with previous findings.3,7 The higher risk attributed to populations living in the lower latitudes is consistent with the increased exposure in those regions of the world. This suggests that appropriate sun protection may reduce the chance to develop skin cancers among heart transplant patients.
This study is a retrospective analysis of a single data source for both derivation and validation cohorts. The data may contain missing information, and the results are recommended to be replicated in a separate patient population and ideally prospectively. In addition, skin cancer is considered as the most frequent posttransplant malignancy and is often underreported in large registry databases. Our study only included patients who had a clear indication of posttransplant malignancy status. Furthermore, patients in the UNOS database had different lengths of follow-up. Our analysis extracted the time of the last follow-up as the censoring time.
This study investigated the risk factors of post-transplant skin cancers in heart transplant patients for the period 2000-2015 in the UNOS database. Older age, male sex, and White ethnicity were consistent risk factors for posttransplant skin cancer as in previous studies. Patients who were reported with malignancy at listing or at transplant had a significantly higher risk to develop skin cancer. Lower HLA mismatch and induction with thymog-lobulin greatly increased the risk of post-transplant skin cancer. In addition, a risk stratification model was presented to separate patients into different risk groups with significantly different incidence rates of skin cancer, which enables an early prediction of skin cancer after heart transplant.
Volume : 21
Issue : 1
Pages : 41 - 46
DOI : 10.6002/ect.2022.0252
From the 1Division of Cardiology, Department of Internal Medicine, Texas Tech Health Sciences Center; and the 2Department of Industrial, Manufacturing and Systems Engineering, Texas Tech University, Lubbock, Texas, USA
Acknowledgements: This work was supported by a medical practice grant from the Department of Internal Medicine, Texas Tech Health Sciences Center, to Nandini Nair and by a National Science Foundation grant to Dongping Du. Other than described, the authors have not received any funding or grants in support of the presented research or for the preparation of this work and have no declarations of potential conflicts of interest.
Corresponding author: Nandini Nair, 3601, 4th Street, Lubbock, TX 79430, USA
Table 1. Patient Characteristics in Cancer and No-Cancer Groups
Figure 1. Skin Cancer Incidence Rate in Heart Transplant Recipients
Figure 2. Skin Cancer Incidence Rate
Table 2. Risk Predictors in the Multivariate Model
Figure 3. Predicted and Observed Probabilities to Develop Skin Cancer 5 Years After Cardiac Transplant
Figure 4. Cancer-Free Survival Curves for Different Risk Groups