1. Problem overview

The ASTREG (Asian Organ Transplantation Registry) is a multinational, multicenter prospective registry collecting longitudinal data on organ transplantation across participating centers in Asia. This project uses ASTREG-H, the hospital-level clinical dataset, which includes detailed baseline and follow-up information on kidney transplantation from 2018 to 2024. While transplant outcomes have improved over time, less is known about how mortality rates and underlying causes of death have evolved in this population. Specific dataset details are not disclosed due to confidentiality.

The primary objective was to examine temporal changes in mortality and to identify factors contributing to shifts in cause-of-death patterns, including infection-related and cardiovascular events. Using registry data, I constructed a longitudinal cohort and first evaluated annual mortality trends and cause-specific distributions. I then applied Kaplan–Meier survival analysis and multivariable Cox proportional hazards models to assess differences in survival across transplant eras, providing a structured evaluation of how risk patterns evolved over time.

2. Data and cohort construction

The study population included all patients in the ASTREG-H registry with available baseline and follow-up data required to define survival outcomes. A longitudinal cohort was constructed by linking baseline transplant records with outcome data across participating centers, reflecting real-world clinical practice.

Follow-up time was defined from the date of transplantation to the earliest of death or last known follow-up, with censoring applied to individuals who remained alive at the end of observation.

The primary outcome was all-cause mortality. Cause-of-death information was extracted and recoded into clinically meaningful categories, including cardiovascular, infection, malignancy, sudden cardiac death, liver disease, accident, suicide, other, and unknown.

To examine temporal patterns, transplant activity was summarized annually from 2018 to 2024, with particular focus on changes observed during the COVID-19 period.

3. Data quality and preprocessing

Data quality checks were performed to ensure consistency across participating centers. Key variables, including transplant date, follow-up time, and mortality status, were reviewed for completeness and logical consistency.

Data preprocessing included standardizing date formats, resolving inconsistencies across centers, and excluding records with implausible or incomplete key variables.

Cause-of-death information was available for the majority of deaths, with only a small proportion of records having missing data. Records with missing or implausible dates were corrected when possible or excluded from time-to-event analyses.

Cause-of-death data were harmonized into standardized categories to reduce inter-center variability. A small proportion of cases remained classified as “unknown” or “other,” reflecting real-world clinical complexity rather than data absence.

Overall, the dataset showed sufficient completeness and consistency to support both descriptive analyses and survival modeling.

5. Cause of death patterns

5.1 Overview of the 2023 mortality peak

A total of deaths with available cause-of-death information were analyzed across participating centers from 2018 to 2024. Overall mortality demonstrated a gradual increase over time, reaching a peak in 2023, followed by a decline in 2024.

This peak was not attributable to a single cause but rather reflected a combined increase across multiple categories, most notably infection-related deaths, cardiovascular and sudden cardiac death events, and deaths classified as unknown or other.

5.2 Infection as the leading cause of death

Infection was the leading cause of death across most centers and years, showing a sustained increase over time. In particular, infection-related mortality remained consistently high in major contributing centers and peaked around 2023–2024.

This pattern suggests that infection represents a structural and ongoing risk rather than a transient phenomenon. The persistence of infection-related deaths beyond the peak period may reflect long-term vulnerabilities in immunosuppressed populations, as well as potential healthcare system strain during and after the COVID-19 era.

5.3 Cardiovascular and sudden cardiac death: an acute event cluster

Cardiovascular deaths and sudden cardiac death, when considered together, formed a notable cluster of acute events. These causes showed a relative increase around the peak period, particularly in 2023.

Although individually smaller in magnitude compared to infection, their concurrent rise suggests a shared underlying mechanism, potentially related to acute physiological stress, delayed care, or post-infectious complications. Grouping these causes provides a more clinically meaningful interpretation of acute mortality risk during this period.

5.4 Increasing proportion of unknown and other causes

An increase in deaths classified as “Unknown” or “Other” was observed in certain centers, particularly during the peak year. This trend may reflect increased clinical complexity, diagnostic uncertainty, or limitations in cause-of-death ascertainment.

Rather than representing random missingness, this pattern likely indicates a broader shift toward more heterogeneous and less clearly defined mortality profiles, especially under conditions of system-level stress.

6. Analytical extension

6-1. Kaplan-Meier survival analysis

Kaplan–Meier curves were used to compare patient survival across transplant eras (pre-COVID [2018–2019], COVID-era [2020–2023], and post-COVID [2024]).

Overall, survival probabilities were highly similar across all eras, with no clear separation between curves. This was supported by the log-rank test, which showed no statistically significant difference in survival (p = 0.38).

However, this result should be interpreted with caution. The post-COVID group (2024) has substantially shorter follow-up time compared to earlier eras, limiting the ability of the KM curves to capture long-term survival differences. Furthermore, with a relatively small number of total events, the analysis may lack sufficient statistical power to detect modest but clinically meaningful differences across eras. The absence of statistical significance does not preclude the possibility of a true difference that this study was underpowered to detect.

6-2. Cox regression analysis

Variables included in the model were age, transplant era, primary disease, donor type, and comorbidities. Gender was excluded after the proportional hazards assumption was assessed using Schoenfeld residuals (cox.zph), which indicated a significant time-varying effect for gender; it was therefore incorporated as a stratification variable rather than a covariate.

Age, the COVID-19 era (2020–2023), and comorbid conditions (diabetes and hypertension) were associated with higher mortality hazard. Comorbid hypertension showed the strongest association with increased risk (HR = 3.06, 95% CI 1.87–5, p = <0.001). The COVID-19 era was also independently associated with higher hazard (HR = 2.14, 95% CI 1.35–3.39), consistent with the pandemic period contributing to excess mortality beyond baseline patient characteristics. In contrast, donor type did not show a significant association with mortality in this analysis.

Compared to glomerulonephritis (GN) as the reference group for primary disease, other disease categories showed relatively lower or similar hazard ratios. This may reflect the relatively favorable prognosis of GN as a primary indication for transplant, though some estimates showed wide confidence intervals due to small subgroup sizes.

Some estimates showed wide confidence intervals, reflecting the limited number of events in certain subgroups. The concurrent inclusion of primary renal disease (HTN/DM as primary disease) and comorbidity status (HTN/DM as comorbidities) may introduce some conceptual overlap; this is discussed further in the limitations section.

Overall, these findings suggest that patient survival in this registry was more strongly influenced by the pandemic period and pre-existing comorbid conditions than by donor type or transplant era alone.

7. Limitation

Several limitations should be considered when interpreting these findings.

Short follow-up in recently transplanted patients. Patients transplanted in 2023–2024 have substantially shorter observation time compared to earlier cohorts. This introduces informative censoring and limits the comparability of survival estimates across transplant eras, particularly for the post-COVID group in the Kaplan–Meier analysis.

Variability in cause-of-death ascertainment. Cause-of-death classification relied on center-reported data, which may reflect differences in diagnostic practice, clinical documentation standards, and coding conventions across participating sites. The increase in deaths classified as “Unknown” or “Other,” particularly in certain centers, likely reflects this inter-center variability rather than a true clinical shift, which limits the interpretability of cause-specific mortality trends.

Conceptual overlap between primary disease and comorbidity variables. The Cox model includes both primary renal disease (with HTN and DM as categories) and comorbid conditions (HTN, DM as separate binary variables). These variables are not independent, and their concurrent inclusion may lead to redundant or partially collinear covariate adjustment. Future analyses should consider either consolidating these variables or formally assessing collinearity.

Observational design and residual confounding. As a registry-based observational study, this analysis is subject to unmeasured confounding. Some clinically relevant variables, such as immunosuppression regimen intensity, were not included in the model, while others (e.g., post-transplant infection surveillance practices and center-level care quality) were not captured in the dataset. These factors may explain part of the observed variation in mortality.

8. Discussion

This analysis identified a progressive increase in all-cause mortality among kidney transplant recipients in the ASTREG-H registry, culminating in a peak in 2023 before declining in 2024. The findings from descriptive, cause-specific, and survival analyses collectively point to a multifactorial explanation for this pattern, with infection, cardiovascular events, and the broader effects of the COVID-19 pandemic each contributing to the observed trend.

Reconciling the KM and Cox findings. The Kaplan–Meier analysis showed no statistically significant difference in survival across transplant eras (p = 0.38), whereas the Cox model identified the COVID-19 era (2020–2023) as associated with higher mortality hazard after adjustment for age, comorbidities, and other covariates. This apparent discrepancy may be explained by two factors. First, the KM analysis is unadjusted and does not account for differences in patient characteristics across eras. Second, the post-COVID group (2024) has relatively limited follow-up time, which may compress its survival curve and reduce the ability to detect divergence. Taken together, the Cox model provides complementary insight into adjusted risk patterns across eras.

The gap between cause-of-death patterns and the survival model. Infection emerged as a leading cause of death across years and centers, yet infection-related variables were not included in the Cox model. This reflects a structural limitation of survival analysis, as cause of death is a post-event outcome and cannot be used as a predictor. Therefore, the model captures overall risk factors but does not directly explain underlying mechanisms. The descriptive cause-of-death analysis provides complementary insight, suggesting that excess mortality may be partly mediated through infectious and cardiovascular deaths.

Infection as a structural risk in immunosuppressed populations. The sustained burden of infection-related mortality beyond the acute COVID-19 period is consistent with the known vulnerability of transplant recipients under chronic immunosuppression. This pattern may reflect a combination of factors, including immunosuppression-related susceptibility, changes in pathogen exposure during and after the pandemic, and potential healthcare system strain affecting timely diagnosis and treatment. The lack of a clear decline in infection-related deaths after 2023 suggests that this pattern may extend beyond a transient COVID-era effect and could represent an ongoing risk that warrants continued surveillance.

9. Conclusion

This project examined temporal patterns in all-cause and cause-specific mortality among kidney transplant recipients using the ASTREG-H multi-country registry from 2018 to 2024. Mortality rates increased over the study period, with a peak observed around 2023, reflecting a combined rise in infection-related, cardiovascular, and indeterminate deaths, followed by a decline in 2024.

Survival analysis using Cox proportional hazards models identified the COVID-19 era, older age, and comorbid hypertension and diabetes as factors associated with higher mortality hazard, while donor type was not significantly associated. The absence of a statistically significant difference in Kaplan–Meier curves across transplant eras may reflect limited statistical power and shorter follow-up in the most recent cohort.

Taken together, these findings suggest that infection remains a major contributor to mortality in this immunosuppressed population and highlight the potential influence of the pandemic period and pre-existing comorbidities on survival outcomes. Future analyses could benefit from incorporating a competing risks framework, extending follow-up for recent transplant cohorts, and improving standardization of cause-of-death reporting across centers.