GBD 2021-derived counts indicate that global incident cases of chronic kidney disease (CKD) rose from 8.24 million in 1992 to 19.94 million in 2021 — a 141.99% increase and roughly a 2.42-fold growth. The global age-standardized incidence rate (ASIR) increased from 194.74 to 233.56 per 100,000 over the same period. The APC-derived net drift quantified an average annual increase of 0.79% (95% CI: 0.76%–0.83%), signaling persistent growth in incidence over three decades. All five socio-demographic index (SDI) regions showed positive net drifts, though magnitudes varied by region.
High SDI regions carried the highest absolute ASIR in 2021 — 277.75 per 100,000 (95% UI: 260.70–295.01) — but the smallest net drift (0.34% per year, 95% CI: 0.27%–0.40%). Middle SDI regions demonstrated the fastest relative increase, with a net drift of 1.06% per year (95% CI: 1.02%–1.10%). Low SDI regions recorded the lowest ASIR in 2021 (155.00 per 100,000; 95% UI: 143.40–167.34) alongside a moderate net drift of 0.60% per year (95% CI: 0.57%–0.63%).
The study team applied an age-period-cohort (APC) log-linear Poisson regression to disentangle three overlapping temporal influences. Age effects, period effects and cohort effects were estimated as separate functions using estimable-function approaches to mitigate the collinearity inherent to APC analyses.
Investigators reported similar qualitative APC patterns across global and regional strata. Age effects increased within the same birth cohort: older age groups carried greater CKD risk. Period effects rose over time in most SDI strata, with disease risk lower in the early 1990s relative to the reference period (2002–2006) and higher in recent periods. Birth cohort RRs increased for later-born cohorts; cohorts born after the 1957–1961 reference showed relative incidence ratios significantly greater than one. The authors provided Wald χ2 test results confirming statistical significance for these APC components in their tables and figures.
Local drift — the annual percentage change for specific age groups — varied across the lifespan. Globally, trends from ages 25–29 through 55–59 were relatively stable. The investigators observed a modest deceleration in local drift after age 60: the local drift declined from 0.76% (95% CI: 0.72%–0.80%) at 60–64 years to 0.68% (95% CI: 0.62%–0.74%) at 75–79 years.
Regional differences were evident. High SDI regions recorded smaller increases across age groups compared with global averages; the largest increase in that stratum occurred in the 35–39 age group (local drift 0.58%, 95% CI: 0.44%–0.72%). Middle SDI regions showed higher-than-global local drifts across all age bands; even their lowest age-specific drift (0.94% at 30–34 years, 95% CI: 0.84%–1.05%) exceeded the global net drift.
A univariate autoregressive integrated moving average (ARIMA) model was fit to historic ASIR time series to generate ten-year forecasts. The authors selected optimal (p,d,q) parameters via automated AIC-based selection and checked residuals with the Ljung–Box test.
Using that model, the global ASIR was projected to reach 237.91 per 100,000 (95% CI: 227.54–248.28) by 2031. Regionally, the ARIMA forecasts suggested divergent near-term dynamics: the high SDI region was predicted to show an initial decline then partial recovery — falling from 277.75 per 100,000 in 2021 to a nadir of 255.90 (95% CI: 240.06–271.75) in 2027 and increasing to 275.92 (95% CI: 255.17–296.68) by 2031, producing an overall downward trajectory across the decade. The middle SDI region was projected to display the largest rise in ASIR, from 232.96 per 100,000 in 2021 to 252.36 (95% CI: 246.58–258.14) per 100,000 in 2031.
The source document included projected trajectories for low-middle and low SDI regions but the provided text was truncated and did not contain the complete numeric end-of-decade estimates for those strata. The authors reported that low-middle and low SDI regions were expected to increase at a slower pace, but the exact 2031 ASIR values for those regions were not available in the excerpt we received.
All incidence counts and ASIR were extracted from the GBD 2021 GHDx query tool and were standardized to the GBD global standard population. CKD was defined using ICD-11 code GB61. The study excluded individuals younger than 20 and those 80 and older; twelve five-year age groups from 20–24 to 75–79 years were analyzed. For APC decomposition, the 1992–2021 span was partitioned into six non-overlapping five-year periods, producing 17 overlapping ten-year birth cohorts for estimable-function calculation.
GBD supplies 1,000 posterior draws for each metric, but the APC and ARIMA analyses used only the mean point estimates as fixed inputs; consequently, confidence intervals reported from APC and ARIMA reflect model-based statistical uncertainty and do not incorporate draw-level uncertainty from GBD.
The ARIMA forecasts intentionally excluded external covariates; the authors stated the model was designed to capture intrinsic historical dynamics of incidence rather than to simulate effects of changing risk-factor prevalences or interventions.
The investigators noted, and the methods clarify, several analytic constraints that affect interpretation. First, APC estimates and ARIMA forecasts were generated from GBD mean point estimates; the uncertainty intervals from those models do not propagate GBD draw-level uncertainty, which limits the full characterization of overall uncertainty. Second, the ARIMA projection was univariate and omitted covariates such as diabetes, hypertension, or screening coverage that plausibly influence CKD incidence trends. Third, the analysis excluded people under 20 and those ≥80 years; the burden and dynamics in these excluded age groups therefore remain unaddressed by this work. Finally, the excerpted source text available for this rewrite was truncated before providing the complete numeric ARIMA forecasts for the low-middle and low SDI regions; those specific end-of-decade estimates were not available for reporting here.
These limitations define unanswered questions within the published analysis: how would incorporation of risk-factor trajectories alter projections; how sensitive are APC-derived net and local drifts to alternative reference-period choices; and what are full region-specific forecasts when draw-level uncertainty is propagated? The source article provides the empirical APC and ARIMA outputs and statistical test results; further exploration of covariate-informed forecasting and uncertainty propagation would be necessary to answer the remaining questions.