Methods
Data source
We collected data on annual cases of CKD between 1990 and 2021, categorised by sex, 17 age groups (from under 5 to 80 years old in 5 year intervals), country, the six regions of the WHO and five groups of countries based on the socio-demographic index (SDI) from the Global Burden of Disease (GBD) online query tool (https://vizhub.healthdata.org/gbd-results/).9 The SDI is a standardised, composite summary measure on a scale of zero to one that can be compared geographically and over time. It includes the following indices: (i) average income per person, (ii) educational attainment and (iii) the total fertility rate of the country. A higher SDI usually reflects a more advanced location with better health outcomes.10 Data on the aetiology of CKD, including type 1 and 2 diabetes mellitus, hypertension, glomerulonephritis and other causes, were also extracted from the GBD database.9 We obtained the relevant population data, which was stratified by sex, age group, nationality and year, from the United Nations Department of Economic and Social Affairs’s Population Division.11 In the GBD study, CKD was considered both a disease and a metabolic risk factor. For this study, however, we approached CKD solely as a disease. CKD is a long-term condition involving changes in the structure and function of the kidneys due to various factors. It is usually diagnosed when the estimated glomerular filtration rate is less than 60 mL/min/1.73 m² or when there are indicators of kidney injury, such as protein or blood in the urine. CKD may also be diagnosed when abnormalities are found through a kidney biopsy or imaging techniques. These indicators must persist for at least 3 months.12
Patient and public involvement
Patients or the public were not involved in the design, conduct, reporting or dissemination plans of our research.
Projections of CKD incidence rates and case numbers from 2022 to 2030
We employed several methods, including Poisson regression, regression discontinuity, smooth spline models, generalised additive models and Bayesian age-period-cohort (BAPC) models, using CKD incidence data at the global level. First, we performed model selection based on predictive performance. In other words, we divided the CKD case data into training and testing sets at the global level (training set: 1990–2015; testing set: 2016–2021). Using the 1990–2015 data, we trained the following five prediction models: BAPC, Poisson regression, regression discontinuity, smooth spline and GAM. We then predicted CKD incidence between 2016 and 2021 using the 1990–2015 data and compared the predicted values with the actual values from the same period. We calculated the mean absolute percentage error (MAPE) to assess model performance. The MAPE was calculated using the following formula:
where N is the number of data points and At and Ft represent the actual and forecast values at data point t, respectively.13 The results of the model selection are shown in figure 1. Considering that the MAPE value in the BAPC model was the least one among the five models, we used it for the projections of CKD incidence rates and case numbers through 2030. We projected CKD case numbers and age-specific incidence rates up to 2030 by conducting a BAPC analysis with integrated nested Laplace approximation. The details of this model, its assumptions and how it is run are well documented and validated elsewhere.14 It is important to note that the selection of 2030 as the endpoint of the forecast is related to the Sustainable Development Goals. In September 2015, leaders and representatives from various United Nations (UN) agencies and civil society organisations established these goals, which were approved by the UN General Assembly. The goals aim to be achieved by 2030. Since this document focuses on global poverty, many public health and medical research studies have chosen 2030 as the target date for their predictions. However, other studies with a similar goal and a similar dataset to this study have considered the years 2040 or even 2050 as the final prediction point. But the accuracy of forecasting models generally tends to decrease as the time horizon for the forecast increases. One reason for the decrease in accuracy as the number of forecast years increases is that forecasting models often rely on historical data and patterns. Over longer periods, underlying conditions may change due to economic shifts, medical technological advancements or policy changes, rendering historical data less relevant for predicting future outcomes. Additionally, new trends or disruptions, such as medical technological changes or social changes, can emerge unpredictably over the long term and are not accounted for in the model.
Results of model comparison based on chronic kidney disease data at the global level.
The Bayesian APC model consists of three factors: age, period and cohort. In this research, ‘age’ represents biological age, with the age effect reflecting how different age groups contribute to the onset of CKD. ‘Period’ refers to the specific year of CKD diagnosis, and the period effect indicates how factors like economic conditions, medical advancements and intervention policies during a particular timeframe influence the disease risk across all age groups. ‘Cohort’ represented birth year, and the cohort effect reflected how being born during a specific time period influenced CKD occurrence. For instance, a cohort that experienced a historical event, such as a war or economic depression, may have different health outcomes later in life. In the BAPC model, alpha (α), beta (β) and gamma (γ) parameters are used to represent the effects of age, period and cohort, respectively. In this study, we used the parameter values α=0.0005 and 1 for the age effect, β=0.0005 and 0.00005 for the period effect and γ=0.00005 for cohort effects.14 Another component of the BAPC model is the random walk (RW). RW priors are used to smooth the age, period and cohort effects, essentially acting as a form of regularisation. These priors constrain the parameters to be relatively similar to their neighbours in the time series. This helps to address the identification problem inherent in APC models, where it is difficult to disentangle the individual effects of age, period and cohort. In our analysis, since the expectation that effects adjacent in time might be similar, the second-order random walk (RW2) model with inverse-gamma prior distribution was used for age, period and cohort effects. RW2 assumes independent mean-zero normal distribution on the second differences of all time effects. This is a natural target for smoothing since the second differences in APC models are identifiable. Considering the age effects, the RW2 prior is given by:
where i denotes the age index running from 1 to I=4 in this study because we projected the CKD incidence rates and CKD incidence numbers in four age groups (0–19, 20–39, 40–59 and ≥60 years), 
denotes the variance parameter and Q is rank deficient. To complete the RW2 model specification, we used the usual conjugate hyperprior for the precision, 
Gamma (α, λ). This led to the full conditional 
Gamma (α+0.5 rank (Q), 
), which might be directly simulated.
We used the world population in 2000 to standardise the CKD incidence rates. All statistical analyses were performed using the R programme (version 4.1.2, R core team, Vienna, Austria).
Model validation and assumption checking
To validate the BAPC model and confirm the appropriateness of the assumptions, we performed several diagnostic checks. First, we assessed over-dispersion under the Poisson likelihood assumption by comparing the observed and expected variances of the counts across the age, period and cohort dimensions. The model did not exhibit significant over-dispersion, supporting the adequacy of the Poisson assumption. Second, we reported 95% Bayesian credible intervals (CrIs) for all incidence estimates to reflect model uncertainty. These CrIs were derived directly from the posterior distributions of the model parameters.
Quantifying the CKD incidence trends
To capture the temporal patterns of age-standardised incidence rates (ASRs) for CKD during the periods of 1990–2021 and 2022–2030, we used the average annual percentage change (AAPC), which reflects past and future trends, respectively. A regression model was applied to the natural logarithm of the incidence rates, expressed as 
, where y represents the log-transformed ASR, x stands for the calendar year and the AAPC was derived using the formula (exponential (β)−1) × 100. Generally, we described the trend by AAPC index in Joinpoint trend analysis software version 4.9.1.0.
Results
CKD case numbers and incidence, 1990–2021
From 1990 to 2021, the number of newly diagnosed CKD cases globally increased from 7758.60 thousand to 19 950.85 thousand. During this period, the ASR of CKD increased from 145.66 to 252.93 per 100 000 (AAPC=0.74, 95% CI: 0.73, 0.75) (table 1, figures 2 and 3). Both males and females experienced an increasing trend in case numbers and ASR. For men, the incidence increased from 3374.90 thousand cases in 1990 to 8818.39 thousand cases in 2021, and for women, it increased from 4383.70 thousand cases in 1990 to 11 132.47 thousand cases in 2021. The ASR increased from 125.78 to 222.80 per 100 000 people (AAPC=0.77, 95% CI: 0.76 to 0.78) for men, and from 165.84 to 283.28 per 100 000 people (AAPC=0.71, 95% CI: 0.70 to 0.72) for women, during this period. Case numbers and ASRs were higher in women than in men during the study period. CKD case numbers increased in all age groups (table 1 and figure 3). The most significant increase was observed among older individuals (≥60 years), with a case increase of over 8 million between 1990 and 2021. In terms of aetiology, after cases of CKD with an unknown cause, the largest number of new cases and ASRs were attributed to type two diabetes mellitus (T2DM) and hypertension, respectively. From 1990 to 2021, CKD cases due to T2DM increased from 751.93 thousand to 2012.15 thousand. Meanwhile, CKD cases caused by hypertension rose from 461.65 thousand in 1990 to 1282.33 thousand in 2021. The ASR of CKD due to T2DM increased from 14.12 per 100 000 in 1990 to 25.52 per 100 000 in 2021. Additionally, from 1990 to 2021, the ASR of CKD due to hypertension increased from 8.67 to 16.26 per 100 000 people. Regarding the SDI, the highest CKD ASR was found in countries with a high SDI. In these countries, the ASR increased from 316.19 per 100 000 people in 1990 to 520.84 per 100 000 people in 2021. This increasing trend was also observed for other SDI groupings. Although the highest ASR of CKD was seen in high-SDI countries, the most significant increase during the 32 year study period (1990 to 2021) was seen in middle-, high-middle- and low-middle SDI countries. The AAPC was 1.26% in middle-SDI countries, 1.06% in high-middle SDI countries and 0.67% in low-middle SDI countries. We also examined the CKD trend by country grouping based on the WHO classification. The regions with the highest increase slopes from 1990 to 2021 were the Western Pacific (AAPC=1.15%), the Eastern Mediterranean and Europe (AAPC=0.83%) and the Americas (AAPC=0.80), respectively. Further details on the number of new cases and the ASR of CKD are presented in table 1. Across the countries, the highest CKD ASR in 1990 was found in Costa Rica (436.29 per 100,000), followed by the United Arab Emirates (327.34 per 100,000), Mexico (304.82 per 100,000), Iran (304.82 per 100,000) and the USA (303.42 per 100 000) (see figures 4 and 5). During the study period, 201 countries or territories experienced increases in the CKD ASR, while three experienced decreases (see figures 4 and 5). Estonia had the most significant increase (AAPC=1.47, 95% CI: 1.46, 1.48), followed by Hungary, North Macedonia, Ecuador and Armenia with AAPCs of 1.46%, 1.37%, 1.34% and 1.28%, respectively (figure 4E). Between 1990 and 2030, only Greece (AAPC=−0.12%), Ireland (AAPC=−0.11%) and Sweden (AAPC=−0.05%) showed a decreasing trend.
(A–L) Temporal trends of age-standardised incidence rates (ASRs per 100 000) of chronic kidney disease between 1990 and 2021 and their projections up to 2030 at the global level. Incidence rates for the four age groups are crude, not age-standardised. The open dots represent the observed values, and the fan shape denotes the predictive distribution between the 2.5% and 97.5% quintiles. The predictive mean value is shown as a solid line. The vertical dashed line indicates where the prediction starts. ASRs, age-standardised incidence rates; CKD, chronic kidney disease.
Trends in the number of chronic kidney disease new cases between 1990 and 2030 at the global level according to sex and aetiology. The error bars denote the 95% CIs of the prediction values. The y-axes are on a scale of thousands. CKD, chronic kidney disease; DM, diabetes mellitus.
Global distribution and average annual percentage changes (AAPCs) in age-standardised incidence rates (ASRs per 100 000) of chronic kidney disease (CKD) at the national level. (A) ASR of CKD in 1990; (B) ASR of CKD in 2021; (C) ASR of CKD in 2022; (D) ASR of CKD in 2030; (E) AAPC of CKD ASR between 1990 and 2030. ASRs, age-standardised incidence rates; AAPCs, average annual percentage changes; CKD, chronic kidney disease.
Temporal trends of age-standardised incidence rates (ASRs) of chronic kidney disease (CKD) between 1990 and 2021 and their projections through 2030 in countries with the top 20 highest ASR due to CKD in 1990. The open dots represent the observed values, and the fan shape denotes the predictive distribution between the 2.5% and 97.5% quintiles. The predictive mean value is shown as a solid line. The vertical dashed line indicates where the prediction starts.
Case number and age-standardised rate (ASR) of chronic kidney disease in 1990 and 2021, by sex, aetiology, age groups and grouping countries based on the sociodemographic index (SDI)
CKD case numbers and incidence, 2022–2030
Between 2022 and 2030, the number of CKD cases is projected to increase by up to 25 057.70 thousand (table 2; figure 3). During the same period, the CKD ASR is predicted to increase from 257.83 to 297.62 per 100 000 people (AAPC=0.15%, 95% CrI: 0.14%, 0.62%) (table 2; figure 2). An increasing trend is expected for both sexes. Bayesian APC model prediction shows that between 2022 and 2030, the number of CKD cases is likely to increase to 11 069.64 thousand in males and 13 981.39 thousand in females. The CKD ASR is also predicted to increase to 261.26 per 100 000 people in men (95% CrI: 242.25 to 280.26) and to 332.52 per 100 000 people in women (95% CrI: 311.28 to 353.77) (table 2; figure 2). The case numbers and ASR are predicted to increase for all countries grouped by SDI and WHO region (table 2). The increasing trend in the ASR of CKD due to T2DM and hypertension is predicted to continue after 2021. A similar pattern was detected for CKD due to other causes, which experienced a significant increase in the ASR from 2022 to 2030. Figure 5 shows the 20 countries or territories with the highest CKD ASRs in 1990. These countries and territories are located in different regions and experienced various CKD incidence patterns between 1990 and 2030. Costa Rica’s CKD ASR is expected to rise from 455.96 to 460.82 per 100 000 people between 2022 and 2030. In the United Arab Emirates, it is expected to increase from 476.16 to 502.84 per 100 000 during the same period. Mexico’s CKD ASR increased from 304.15 in 1990 to 469.18 in 2021 and is predicted to reach 491.24 in 2030. Figure 4 depicts the temporal trend of CKD ASRs per 100 000 and their projection up to 2030 for each country.
Predictive number and age standardised rate (ASR) of chronic kidney disease (CKD) in 2022 and 2030, by sex, aetiology, age groups and grouping countries based on the socio-demographic index. The case numbers and ASR of CKD in 2020 and 2030 were estimated by the Bayesian age-period-cohort model
Discussion
CKD is a severe condition whose incidence varies widely worldwide. Although the burden attributed to this disease is higher than all common cancers; nevertheless, its importance has received much attention neither globally nor regionally. In the current study, we used GBD data and BAPC models to describe the temporal trends of CKD incidence over the last three decades and predict its future trends in the next decade at the global levels and across countries. Generally, the incidence rate and case numbers of CKD are expected to increase between 2022 and 2030, regardless of sex, age group, aetiology and the level of SDI. Regardless of CKD cases with unknown causes, more than 50% of the total CKD cases were attributable to T2DM. The most pronounced increase in incidence will be observed in CKD due to T2DM and hypertension. Between 2022 and 2030, 95% of countries are expected to experience an increase in CKD incidence, whereas 5% of countries are estimated to experience a decreasing or steady trend. Furthermore, among the six regions of WHO, it is predicted that in 2030, America, Europe and the Western Pacific regions will have the highest ASR of CKD. In terms of grouping countries based on SDI, high-, high-middle and middle-SDI regions are also projected to have the highest incidence of CKD by 2030.
The projected rise in CKD incidence by 2030 underscores the importance of analysing the contributing risk factors. The influence of both genetic and environmental elements on the development and progression of CKD has been well-documented in the literature.15 16 Smoking is known to elevate the risk of CKD through mechanisms such as inducing a pro-inflammatory state, increasing oxidative stress, causing a prothrombotic shift, impairing endothelial function and promoting glomerulosclerosis and tubular atrophy.17 Previous studies have shown that the likelihood of developing CKD is 30% higher in current smokers and 15% higher in former smokers compared with non-smokers.3 Furthermore, each additional five cigarettes smoked daily is linked to a 31% rise in serum creatinine levels, indicating a decreased ability of the kidneys to clear waste products and toxins from the body, thus elevating the risk of CKD.18 The WHO estimates that 20.2% of the world’s population aged ≥15 years were current smokers in 2015. The global prevalence of tobacco smoking based on the WHO estimates in 2015 was 20.2% for both sexes, 34.1% for males and 6.1% for females. In other words, about one-fifth of the world’s population is current smokers. Another point is that the prevalence of tobacco smoking appears to be increasing in the African and East Mediterranean regions, where the incidence of CKD is high.19 Also, obesity and being overweight are the most substantial modifiable risk factors for CKD in the 21st century, increasing strikingly over the past four decades.3 20 Obesity leads to kidney damage through inflammation, endothelial dysfunction, oxidative stress, hypervolemia and a prothrombotic state.21 Globally, 39% of adults were overweight in 2016, and 13% were obese. Also, rates of overweight and obesity continue to grow among adults and children, especially in developing countries, where the rate of CKD is high.22 Moreover, alcohol consumption has been associated with CKD progression; global consumption in adults increased from 5.9 L to 6.5 L and is forecasted to reach 7.6 L by 2030.23 Experimental evidence supports a direct, acute nephrotoxic effect of alcohol on the kidney; alternatively, chronic use in humans may result in alcohol-induced hypertension, indirectly increasing the risk of CKD. In specific circumstances, alcohol abuse or dependence has been linked to particular renal pathologies, including renal papillary necrosis, infection-associated glomerulonephritis and acute renal failure due to non-traumatic rhabdomyolysis. This alarming increase in smoking, alcohol consumption per capita and overweight and obesity might drive an unexpected increase in CKD incidence rates worldwide.24 25
DMT2 is a common chronic disease and a severe public health condition that is one of the most critical risk factors for CKD. According to our findings, CKD due to DM is predicted to continue to increase by 2030. Based on previous research, the worldwide incidence of DM has increased by 102.9% from 11 303 084 cases in 1990 to 22 935 630 cases in 2017, while the age-specific incidence rate increased from 234 per 100 000 persons to 285 per 100 000 persons in this period.26 In other words, over the past few decades, the point prevalence, case incidence numbers and the age-specific incidence rate of DM increased significantly between 1990 and 2017 worldwide in nearly all countries. It may be considered a growing epidemic, and patients with T2DM have an increased risk of diabetic complications, including CKD.27 Mechanisms that lead to CKD in DM include hyperfiltration injury, advanced glycosylation end products and reactive oxygen species. At the molecular level, numerous cytokines, growth factors and hormones, such as transforming growth factor-beta and angiotensin II, cause pathologic changes associated with diabetic nephropathy.28 With these explanations, in the absence of new and effective preventive interventions, the increasing global prevalence of T2DM will inevitably be associated with increasing CKD prevalence.29
The increasing global trend of CKD incidence and new cases might also be attributed to the increasing prevalence of hypertension. Hypertension is a severe medical condition and can increase the risk of adverse health outcomes such as CKD. Hypertension is a definite risk factor for CKD that leads to glomerulosclerosis and loss of kidney function.3 Based on previous studies, hypertension can damage the blood vessels in the kidneys and reduce their ability to work correctly. When the force of the blood flow is high, the blood vessels are stretched, so the blood flows more easily. Eventually, this stretching causes the blood vessels, including the kidneys, to thin and weaken. If the kidney’s blood vessels are damaged, it may be difficult or stopped to eliminate waste and excess fluid from the body. In this situation, the background is provided for kidney diseases.30 Based on the latest WHO report in 2022, about 1.13 billion people worldwide have hypertension, and its prevalence is predicted to increase in the future.31 Therefore, appropriate and timely public health interventional measures regarding hypertension are essential for CKD control and prevention.
In summary, the increasing trend of CKD over the last three decades and the next decade in most countries of the world might be partly explained by the following causes: (I) the increasing use of diagnostic methods and increased use of imaging techniques that lead to the increasing CKD detection rate and reporting rates;32 (II) the growing population, particularly the elderly population;33 and (III) shifting trend towards the adoption of Western diets, change in occupational patterns, increased high-risk behaviours (eg, excessive calorie intake and physical inactivity) and changes in established CKD risk factors (eg, smoking, obesity, DM and hypertension).34 35
While our overall findings indicate an increasing trend in CKD incidence across most high-SDI countries and developed regions such as Europe, a few countries—namely Ireland, Greece and Sweden—emerged as outliers showing a decreasing AAPC in incidence. These trends warrant closer examination to determine whether they reflect true epidemiological shifts or are artefacts resulting from data limitations and modelling assumptions within the GBD framework. In the case of Ireland and Greece, it is important to acknowledge the historical lack of comprehensive, nationally representative epidemiological studies on CKD prevalence or incidence prior to 2020. This is well documented in previous sources and studies, which highlight significant gaps in CKD surveillance in both countries.36–38 Consequently, GBD estimates for these regions rely heavily on mathematical modelling and data extrapolation from comparable geographies rather than direct local data, which may limit the accuracy of the reported trends. More recent evidence, however, contradicts the notion of a declining trend—particularly in Ireland. For instance, a prospective cohort study published in Clinical Kidney Journal reports a rising prevalence of CKD in Ireland,39 indicating that the GBD projections may not fully capture the current epidemiological reality. This discrepancy underscores the importance of continuously updating global burden estimates with local data as they become available. Data from Sweden are more nuanced. While some GBD estimates suggest a potential decrease in CKD incidence, this trend must be interpreted with caution given the country’s ageing population, increasing prevalence of DM and historically low levels of CKD awareness.39 40 These factors suggest that CKD burden could in fact be increasing or, at the very least, not decreasing as significantly as the modelled data imply. A deeper dive into national datasets and literature reveals uncertainties and potential inconsistencies that challenge the validity of the decreasing trend. These outlier cases highlight the inherent limitations of global modelling efforts, particularly in settings where primary data are sparse or outdated. When direct data are unavailable, GBD employs statistical models that integrate regional patterns and covariates to generate estimates. While this methodology is robust in many respects, it remains vulnerable to misrepresentation in countries where local data inputs are limited. Therefore, any interpretation of decreasing CKD incidence in such countries should be made cautiously and supplemented with local literature and expert input where possible. Importantly, we stress that a reported decline in incidence does not imply that CKD is no longer a public health concern in these countries. On the contrary, prevention, early detection and interventions to slow progression remain critical priorities. Continuous data collection, surveillance and validation of global estimates with national and subnational data are essential for accurate policymaking and public health planning.
Finally, this study used data from the GBD Study from 1990 to 2021. The results may differ slightly from those of other CKD studies that used data from 1990 to 2019. These differences are due to several key methodological updates. First, although the GBD 2019 and GBD 2021 datasets cover 204 countries and territories, improvements in data completeness, regional weighting and estimation methods have likely led to discrepancies in calculated ASRs and AAPCs.41 Second, incorporating two additional years of data (2020 and 2021) accounts for the potential impact of the COVID-19 pandemic on access to healthcare, diagnoses and reporting, especially for chronic diseases like CKD.42 43 These additional years may have altered trends, especially in countries with a high SDI and more complete registries. Third, updated demographic inputs and revised model priors in the BAPC framework may have shifted the distribution of cases across age groups while keeping total case numbers relatively stable. This results in lower ASRs for certain age groups.
Our study had a few limitations, which are as follows:
-
This is a macro-level assessment of global, regional and national epidemiological trends and CKD projections for more than a quarter century, and as such, it may have yet to capture micro-level trends. This is especially relevant in large countries such as China, India and the USA, which may have significant subnational variations in CKD incidence that are not reflected in our analyses.
-
We used GBD data to produce the estimates in this report. Although GBD methodologies and results are considered advanced, robust and reliable, they are necessarily limited by the quality of available data. For countries with limited data, GBD estimates were derived through mathematical modelling. On the other hand, in regions where no direct data on CKD are available, the GBD study employs statistical modelling techniques such as Bayesian meta-regression using the DisMod-MR tool. This approach allows for the integration of indirect data by borrowing information from neighbouring or demographically similar countries, based on the assumption that disease patterns and progression can be inferred from shared regional and socio-economic characteristics. Additionally, the model incorporates predictive covariates such as the prevalence of hypertension and DM—major risk factors for CKD—to improve the accuracy of estimates in data-sparse settings. Despite the explanations given about the areas of limited data, sparse data and no data, the findings related to these areas should be interpreted with caution.
-
The incidence of CKD in the past and future may be influenced by diagnostic approaches and reporting rates. Underreporting and detection biases are even higher in low- and middle-income countries.
-
Another limitation of long-term forecasts is that they do not consider changes in characteristics and variables whose changes can affect the predicted values. These characteristics include environmental and sociocultural changes; medical advances; demographic changes; and the spread of chronic diseases that increase the risk of CKD, economic fluctuations and access to healthcare.
A key strength is leveraging the availability of GBD data from 1990 to 2021. The GBD is the most comprehensive compilation and analysis of global health information available. We conducted separate analyses for aetiological and demographic variables as drivers of change in the burden of CKD over the past 32 years and the next 9 years.
