Twenty years of real-world data to estimate chronic kidney disease prevalence and staging in an unselected population

ABSTRACT

Chronic kidney disease (CKD) represents a global public health burden, but its true prevalence is not fully characterized in the majority of countries. We studied the CKD prevalence in adult users of the primary, secondary and tertiary healthcare units of an integrated health region in northern Portugal (n = 136 993; representing ∼90% of the region’s adult population). Of these, 45 983 (33.6%) had at least two estimated glomerular filtration rate (eGFR) assessments and 30 534 (22.2%) had at least two urinary albumin:creatinine ratio (UACR) assessments separated by at least 3 months. CKD was defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines as a persistent decrease in eGFR (<60 ml/min/1.73 m²) and/or an increase in UACR (≥30 mg/g). The estimated overall prevalence of CKD was 9.8% and was higher in females (5.5%) than males (4.2%). From these, it was possible to stratify 4.7% according to KDIGO guidelines. The prevalence of CKD was higher in older patients (especially in patients >70 years old) and in patients with comorbidities.

This is the first real-world-based study to characterize CKD prevalence in a large, unselected Portuguese population. It probably provides the nearest estimate of the true CKD prevalence and may help healthcare providers to guide CKD-related policies and strategies focused on prevention and on the improvement of cardiovascular disease and other outcomes.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

Chronic kidney disease (CKD) is a general term for heterogeneous disorders that irreversibly affect kidney structure and function for >3 months and is implicated in cardiovascular, metabolic, endocrine and xenobiotic toxicity-related complications and in premature mortality [1–3]. CKD is typically defined as a decreased glomerular filtration rate (GFR) and/or increased albuminuria. The worldwide prevalence of CKD was estimated to be ∼11–13% [4] and globally in 2017 it was estimated that nearly 700 million persons had CKD and 1.2 million people died from CKD-related disorders [5]. Moreover, the burden of CKD is expected to increase in the future, especially due to the increase in global aging and the increasing prevalence of hypertension, obesity and type 2 diabetes mellitus (T2DM) [6, 7].

Fortunately, the development of CKD comorbidities can be delayed or prevented if they are rapidly detected [8]. To achieve this, CKD epidemiology needs to be carefully assessed. However, data regarding CKD prevalence and staging in the early stages and morbidity and mortality are scarce or non-existent in many countries [5]. Moreover, even where data are available, a significant heterogeneity of CKD prevalence between regions exists, probably due to disparities in clinical risk factors, methodologies used for creatinine determination, formulas for calculation of estimated GFR (eGFR) and statistical approaches [4, 9, 10]. For instance, across the European population there were considerable differences in the prevalence of both CKD stages 1–5 and CKD stages 3–5 [11]. In the adult general population of the USA, the adjusted prevalence of CKD stages 3–5 ranged from 4.8% to 11.8% in the Northeast and Midwest, respectively [12]. In a very recent publication, the CaReMe CKD study was designed to estimate the prevalence of CKD, key clinical adverse outcomes and costs of CKD across 11 countries [13]. Relevant individual-level data for a cohort of 2.4 million CKD patients was obtained from digital healthcare systems and revealed a pooled prevalence of possible CKD of 10% and a confirmed prevalence of CKD ranging from 5.6 to 9.8% [13].

Specifically in Portugal, the PREVADIAB study showed a prevalence of CKD stages 3–5 of 6.1% [14, 15]. Although this study was a very relevant starting point, some limitations can be highlighted, such as the absence of data on the estimation of prevalence of CKD stages 1 and 2, inclusion of subjects only 20–79 years of age and non-compliance with the criterion for verifying kidney disease chronicity by 3 months after diagnosis, as recommended by the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines [8]. To overcome the reported limitations, more recently the RENA study aimed to estimate the prevalence of CKD and characterize patients on a national level [16]. This cross-sectional study included users of primary health care units (PCHUs) ≥18 years of age and the sociodemographic and clinical data were recorded through a structured questionnaire. Results showed a higher CKD prevalence compared with the global and European previously reported average. Nevertheless, the applied methodology based on voluntary participation of PCHU users presenting in the waiting room, offers some constraints. Indeed, as highlighted by the authors, despite all efforts, this approach may have compromised the global picture by unbiasing results, as the PCHU attendees are not representative of the real population since attendees typically possess multiple comorbidities [17, 18].

Taking this into consideration, the current study aimed to fully characterize the prevalence of CKD in a non-selected population of a group of PCHUs supported by a unique secondary and tertiary care health unit (STCHU) in northern Portugal and simultaneously compare the variation in CKD prevalence and staging by demographic, clinical, analytical and echocardiographic data for the population.

MATERIALS AND METHODS

Study design

This is an observational cohort and cross-sectional study performed in the Health Local Unit of Matosinhos (Unidade Local de Saúde de Matosinhos; ULSM), a regional health system in the district of Matosinhos in northern Portugal, including 14 PCHUs assisted by the same STCHU, the Pedro Hispano Hospital. We selected all persons ≥18 years of age who were seen at least once in the healthcare units in the 3 years before the index date (31 May 2022). A 22-year period of data analyses (since 1 January 2000) was applied. A total of 136 993 users matching the inclusion criteria were enrolled, representing ∼90% of the adult population of the geographic region of Matosinhos, according to the 2021 Portuguese census (the eighth most inhabited municipality in the country and the fourth in the northern region). In other words, ∼90% of the adult population of Matosinhos was attended at a healthcare unit at least once in the 3 years before data access. Data access for analysis was granted after approval by the Ethical Committee and Data Protection Officer of the ULSM [approval 34/CE/JAS of 23 April 2020 (original) and 64/CE/JAS of 10 July 2020 (addenda)]. Following the Health Insurance Portability and Accountability Act Safe Harbor Standard, de-identified data regarding age, gender, body mass index (BMI), waist circumference, systolic blood pressure, diastolic blood pressure, echocardiography and laboratory measurements (including general, iron, diabetes, lipid, liver, heart, thyroid and kidney panels) and general, cardiovascular and bone comorbidities classified by the International Classification of Diseases, Ninth and Tenth Revisions (ICD-9 and ICD-10) codes and current cardiovascular, diabetes and bone disease medications registered according to the Anatomical Therapeutic Chemical Classification System were extracted from electronic health records. Plasmatic creatinine determination was performed in the same laboratory and by the same method for all samples and was used for eGFR calculation. Urinary albumin:creatinine ratio (UACR) was used for albuminuria detection, as defined by the KDIGO guidelines. Only those patients with two or more tests for serum creatinine and/or albuminuria, at least 3 months apart, were included in this study. Patients with only one CKD test were not included in the prevalence calculations.

Nutritional status was classified in the following categories according to BMI: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), pre-obesity (25.0–29.9 kg/m²) and obesity class 1 (30–34.9 kg/m²), class 2 (35–39.9 kg/m²) and class 3 (≥40 kg/m²).

CKD definitions and calculations

CKD stages 1–5 were defined and classified based on the KDIGO guidelines [8] as either decreased eGFR (<60 ml/min/1.73 m²) or the presence of albuminuria assessed as UACR ≥30 mg/g or 3 mg/mmol for >3 months. eGFR was estimated by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, considering gender and serum creatinine [19]: eGFR (ml/min/1.73 m²) = 141 × min(SCr/κ, 1)^α × max(SCr/κ, 1)^−1.209 × 0.993^Age × [1.018 (if female)] or [1.159 (if Black)], where SCr is serum creatinine (mg/dl), α is 0.7 for females and 0.9 for males, κ is −0.329 for females and −0.411 for males, min indicates the minimum of SCr/κ or 1 and max indicates the maximum of SCr/κ or 1. As an additional analysis, the creatinine clearance was estimated by the Cockcroft–Gault (CG) equation as previously described [10]. eGFR categories were defined as follows: G₁, normal or high (≥90 ml/min/1.73 m²); G₂, mildly decreased (60–89 ml/min/1.73 m²); G_3a, mildly to moderately decreased (45–59 ml/min/1.73 m²); G_3b, moderately to severely decreased (30–44 ml/min/1.73 m²); G₄, severely decreased (15–29 ml/min/1.73 m²); and G₅, kidney failure (<15 ml/min/1.73 m²). Albuminuria was defined in three categories: A₁, normal to mildly increased (UACR <30 mg/g); A₂, moderately increased (UACR 30–300 mg/g); and A₃, severely increased (UACR ≥300 mg/g). Patients in stages G1/A1 and G2/A1 were not characterized for CKD since other data regarding renal lesions, such as echography, urinary sediment and renal biopsy reports, were not available.

Statistical analysis

Statistical analyses were performed using Spark 3.1.0 (Apache Software Foundation, Wilmington, DE, USA). Normally distributed variables were presented using mean and respective percentages and non-normally distributed data as medians with interquartile ranges (IQRs). Overall population and CKD staging were stratified into the 50–60, 60–70, 70–80 and >80-year age groups. The prevalence of CKD was estimated as the number with confirmed CKD divided by the number of all individuals registered in the health units enrolled at the time of data access.

RESULTS

Characterization of population

A population of 136 993 individuals, 59 867 (43.7%) males and 77 126 (56.3%) females, were enrolled in this study. A median age of 52.0 years (IQR 30.0) was recorded. Of note, hypertension and T2DM had a prevalence of 42.5% (n = 58 200) and 23.0% (n = 31 494), respectively, while obesity was present in 20.3% of the population studied (n = 27 835).

Prevalence and characterization of the CKD population

To reduce the possibility of CKD false-positive results, we evaluated and confirmed CKD by assessing eGFR and UACR at least twice at least 3 months apart. In total, 45 983 (33.6%) persons had at least two eGFR assessments (Table 1, Supplementary Tables S1 and S2) and 30 534 (22.3%) had at least two UACR assessments separated by at least 3 months (Table 2, Supplementary Tables S3 and S4). Tables 1 and 2 present a detailed characterization of CKD in the population according to the KDIGO guidelines, using CKD-EPI and UACR, respectively. Individual characterizations for female and male populations are provided in Supplementary Figures S1 and S2, respectively. According to the KDIGO guidelines, which define CKD as two eGFR values <60 ml/min/1.73 m² (G3–G5) and/or two UACR values ≥30 mg/g (A2–A3) persistent for at least 3 months, the estimated overall prevalence of CKD was 9.8% and was higher in females (5.5%) than males (4.2%). From these, 4.7% could be stratified according to the KDIGO guidelines (Figure 1). The prevalence of CKD was higher in older patients (especially in patients >70 years old) and in patients with comorbidities. We were also able to identify a significant percentage of patients [27.2% (n = 37 292)] with an eGFR of 60–89 ml/min/1.73 m². The prevalence of CKD, using two measurements of creatinine clearance calculated by the CG equation was 11.3% (detailed data not shown).

A significant increase in the prevalence of CKD was seen in the older age groups (Table 3). Of note, we also observed an increase in the prevalence of comorbidities with CKD stage, namely T2DM, structural heart disease, microvascular disease, familial hypercholesterolemia, cardiovascular disease, hypertension, atrial fibrillation, stable and unstable angina, atherosclerotic disease, myocardial infarction, ischaemic and haemorrhagic stroke, peripheral artery disease and heart failure. Renin–angiotensin system agents, diuretics, antiplatelet agents, calcium channel blockers, beta blockers, glucose-lowering drugs (excluding insulins) and magnesium were among the most prescribed drugs in CKD patients, a finding observed throughout all CKD stages.

DISCUSSION

CKD is a major worldwide public health problem and a relevant cost burden to healthcare systems. It is currently defined by abnormalities of kidney structure or function assessed by the eGFR, thresholds of albuminuria and duration of injury. Estimates of CKD prevalence vary widely, both within and between countries, due to effective differences in CKD regional prevalence, different understandings regarding the use of eGFR for identifying CKD, eGFR thresholds considered to define CKD in elderly populations, analytical methodologies applied for creatinine measurement, formulas for calculation of the eGFR and statistical approaches to estimate CKD prevalence in large-scale epidemiological studies. In an interesting review, solutions to overcome discrepancies were proposed [7]. The KDIGO guidelines [8] are also critical to design epidemiological studies to characterize the global burden of CKD in the general population and subgroups at increased risk with certain comorbidities. In the present work we aimed to study the prevalence of CKD in a population in northern Portugal.

According to the KDIGO guidelines using CKD-EPI and UACR calculations, our results suggest a CKD prevalence of 9.8% for patients in stages ≥G3a/A1. The prevalence of CKD, using two measurements of creatinine clearance calculated by the CG equation and two measurements of UACR, revealed a higher prevalence of 11.3%. The RENA study found a prevalence of CKD stages 1–5 of 20.9% (10.7% for stages ≥G3a) in the Portuguese population that attends the PCHU, while the PREVADIAB study found a prevalence of CKD stages 3–5 of 6.1%, without estimating CKD in stages G1 and G2. As the samples were very similar regarding sex (i.e. 65% of women in the RENA study versus 60% in the PREVADIAB study), age class distribution (i.e. 48% ≥60 years of age in RENA and 46% between 60 and 79 years in PREVADIAB) and comorbidities (i.e. self-reported hypertension, T2DM and obesity in 38%, 16% and 31%, respectively, in the RENA study versus 45%, 12% and 34% in the PREVADIAB study), the different recruitment strategy may partially explain the discrepancies. In fact, in the RENA study the participants were not recruited from the general population, but from primary care attendees, who are possibly less healthy, while in the PREVADIAB study, primarily designed to estimate the prevalence of DM in the Portuguese population, analysed data from a nationally representative sample of 5167 subjects. Therefore the CKD estimation in each study might have introduced some bias compromising full characterization. In our study, the real prevalence was determined from a large and unselected population of 136 993 individuals (121 643 ages 20–79 years), representing 59 867 (43.7%) men and 77 126 (56.3%) women. Comorbidities, such as hypertension and T2DM, with a prevalence of 42.9% (58 698) and 22.9% (31 494), respectively, were highly comparable to the PREVADIAB study, while obesity was less prevalent [n = 27 835 (20.3%)] in our study. It is also important to underline that comparisons of CKD prevalence between publications should be made with the due care. Indeed, different equations for eGFR may impact in the estimation of CKD. While the PREVADIAB eGFR was calculated using the simplified (the four-variable formula) Modification of Diet in Renal Disease study equation, in the RENA the CKD-EPI equation was used. In our study, besides the CKD-EPI equation, the CG equation was also used to increase the robustness of our studies.

Compared with other countries, the CKD burden shows marked variations in the prevalence: 3.3% in Norway, 17.3% in northeast Germany [11, 20] and 15.1% in Spain according to the ENRICA study when considering CKD stages 1–5 [21] and more recently estimated to be 4.91% for CKD stages 3–5 [22]. A meta-analysis performed to determine the global prevalence of CKD in 100 studies from all over the world presented a global mean prevalence of CKD stages 1–5 of 13.4% (range 11.7–15.1%) and stages 3–5 of 10.6% (range 9.2–12.2%) [4]. Interestingly, our study also uncovered a higher prevalence of CKD for stages ≥G3a compared with other countries. Indeed, Portugal has one of the highest prevalences in Europe of patients undergoing renal replacement therapy [23]. Nevertheless, another possible explanation is that in Portugal, primary care programs frequently remind patients to visit their family doctor by letter at least once every 3 years, suggesting that early detection may help to diagnosis CKD in the earliest stages [13]. We also emphasize that we were able to identify a significant percentage of patients [27.2% (n = 37 292)] with an eGFR of 60–89 ml/min/1.73 m². Although these cases were formally excluded according to the KDIGO guidelines (as they do not have albuminuria as an additional criteria), some of them may represent true cases of CKD. Therefore, in a global screening perspective, revision of these criteria may be useful to reduce underdiagnosis in the earliest stages. Compared with most of the studies in this meta-analysis [4], we also observed a higher prevalence of CKD in females (5.5%) than males (4.2%), a fact that does not corroborate RENA results.

As shown in Tables 1–3, prevalence increased with age, as demonstrated in several previous studies [4, 11, 20, 21]. Nevertheless, these results may be overestimated since eGFR naturally declines with age and the increased prevalence of CKD described in older groups might be due not only to real CKD, but also to normal biological variations in kidney function. Our results also show a higher prevalence of T2DM, structural heart disease, microvascular disease, familial hypercholesterolemia, cardiovascular disease, hypertension, atrial fibrillation, stable and unstable angina, atherosclerotic disease, myocardial infarction, ischaemic and haemorrhagic stroke and heart failure in the CKD population when compared with the population without CKD. In fact, CKD is an increasingly recognized cardiovascular risk factor, associated with greater therapeutic burden, high healthcare costs and reduced life expectancy, as up to half of individuals with CKD die from cardiovascular disease [24–26]. Recently we demonstrated that the coexistence of heart failure and CKD is associated with increased premature mortality, as well as non-fatal cardiovascular events in T2DM patients <65 years old [27, 28]. Moreover, in 2016 the European Guidelines on Cardiovascular Disease Prevention incorporated CKD as a non-traditional cardiovascular disease risk factor, readily identifiable from the analytical measurements of eGFR and UACR, and whose early identification and management may have a significant positive impact on cardiovascular disease prevention [29, 30]. Specifically, they classified individuals with an eGFR <30 ml/min/1.73 m² and diabetic patients with proteinuria as ‘very high risk’ (equivalent to a 10-year predicted risk of cardiovascular mortality ≥10%) and those with an eGFR of 30–59 ml/min/1.73 m² as ‘high risk’ (equivalent to a 10-year predicted risk of cardiovascular mortality of 5–10%).

Of note, 33.8% (n = 46 329) of patients were taking renin–angiotensin system blockers (i.e. angiotensin-converting enzyme inhibitors and angiotensin-receptor blockers and angiotensin receptor–neprilysin inhibitors), a fact that is not in line with a recent Spanish study that demonstrate that almost 70% patients were taking these drugs [22, 31].

Our study has some key strengths compared with previous Portuguese CKD prevalence studies, namely, it was not based on an estimation of CKD as in the RENA study [16] and it includes patients >18 years old and without an upper age limit. Moreover, the inclusion of a large and unselected sample offers more robustness to our results, since it is less likely to suffer from non-response bias. As recommended by the KDIGO guidelines [8], evaluation and confirmation of CKD was performed at two different time points at least 3 months apart, in order to fulfil the chronicity criterion and therefore to reduce the possibility of false-positive results, with consequent overestimation of CKD prevalence [32]. Moreover, to increase the accuracy of CKD prevalence estimation and staging, measurements were performed in the same laboratory and by the same method and ICD diagnostic codes were not included. Indeed, it has been demonstrated that ICD diagnostic codes display poor sensitivity and specificity in rapidly identifying progressing CKD patients when compared with the gold standard of eGFR measures, especially due to different practices among health units [33, 34].

There are, however, some limitations to our study. Specifically, our study population was predominantly Caucasian. Therefore the lack of ethnic diversity may restrict the translation of our results to other populations, especially those with substantial genetic differences [35–37]. Indeed, studies have shown that the development of CKD is largely influenced by multiple genetic loci [38]. As >95% of patients were Caucasian, the expected impact of other ethnicities on the overall prevalence estimation is negligible. In addition, since our population is representative of northern Portugal, this may hamper the interpretation and external validity of the results to the rest of the country. Moreover, this is a retrospective study that used secondary data from electronic health records, meaning that measurements such as UACR, fundamental for CKD staging, could not be defined for 77.7% of patients. On the other hand, the higher prevalence of CKD among patients who did have two eGFR or two UACR estimates should be considered an overestimation, as these characteristics may select patients at higher risk of CKD. Missing data may be easily explained since albuminuria, as a biomarker of kidney disease, is usually measured in the primary care setting in patients with comorbidities such as T2DM or hypertension. Therefore this analytical measurement is not widely available in the population free of or at low risk of developing the reported comorbidities and thus may underestimate the prevalence of the first stages of CKD. Finally, our laboratory results may also be influenced by the heterogeneity of techniques and storage conditions used to measure creatinine [39] and albuminuria by immunoassays [40].

CONCLUSIONS

Estimation of the prevalence of CKD is a key factor guiding healthcare system policies and strategies [41, 42]. In our population, CKD prevalence is estimated to be 9.8%, which is in accordance with the global prevalence of CKD across Europe. Further studies are needed to evaluate if there was a real change in CKD prevalence over a 13-year period between the PREVADIAB study (conducted in 2008), the RENA study (conducted in 2018) and our 2021 study. In a very similar population, recent data suggest that the prevalence of CKD could have changed in the last few years in Spain [22]. The frequency used to screen for UACR presents a considerable variation between high-risk population groups, resulting in a low awareness of CKD as a modifiable risk factor in the no-T2DM population. It is clear that CKD patients must be identified earlier and to develop awareness and educational programs to prevent CKD and its associated diseases, such as T2DM, cardiovascular disease and obesity, to reduce the CKD burden for patients, caregivers and society.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the editorial support, namely the constructive review of the manuscript and raised comments. The authors also would like to acknowledge Pedro Hispano Hospital for granted permissions for this study and to Daniel Santos, Tiago Morais and José Castanheira from the Department of Information Technologies for conducting data extraction.

FUNDING

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the article. Funding for this study was provided by AstraZeneca without influence in the preparation of data or in the writing of the article. The authors take final responsibility for the decision to submit for publication.

AUTHORS’ CONTRIBUTIONS

All authors contributed for the study conception and design, selection of the bibliography and revision and approval of the final version for submission. R.J.D.-O. prepared the first draft. T.T.-G. developed the analytic code used in the study and takes responsibility for the assessment of data integrity, for the accuracy of the data analysis and for preparation of the first draft of the tables. C.A. and L.M. were leads in selection of the bibliography regarding prevalence studies. All authors contributed to the interpretation of data and critical revision of the manuscript for important intellectual content. All attest that the listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

DATA AVAILABILITY STATEMENT

All data are incorporated in the article and its online supplementary material.

CONFLICT OF INTEREST STATEMENT

C.S.-A. and L.M. declare speaker and consulting fees from AstraZeneca. C.G. declares speaker and consulting fees from AstraZeneca, Bayer, BIAL, Boehringer-Ingelheim, Daiichi-Sankyo, Eli Lilly, MSD, Novartis and Novo Nordisk. D.S. declares speaker fees from Daiichi-Sankyo. M.P., F.B., J.C. and H.M. are employees of AstraZeneca and Produtos Farmacêuticos SA. T.T.-G. declares speaker and consulting fees from AstraZeneca, BIAL, Daiichi-Sankyo, MSD, Novartis and Medinfar and holds shares in MTG. R.J.D.-O. declares no conflicts of interest. The results presented in this article have not been published previously in whole or part.

REFERENCES

Author notes