Introduction

Chronic kidney disease (CKD) has emerged as a global public health issue [1] and has become an important cause of morbidity and mortality [2]. Patients with CKD constitute a significant economic burden, both directly in terms of resource utilization, and indirectly through loss of productivity and impaired quality of life [3]. In the past decade, 1.1 trillion dollars have been spent on dialysis worldwide [4]. Therefore, early detection of asymptomatic CKD patients is important because early intervention has a reasonable chance of making a positive impact on outcome [5]. Similarly, it is important to prevent or delay disease progression from mild to severe stages [4]. The global increase in the prevalence of CKD and its disproportionate burden on economically developing countries is being driven by an increase in the prevalence of the main risk factors for CKD; namely, diabetes, hypertension, obesity, and aging of the population [6].

However, in the North Central Province (NCP) of Sri Lanka, the dramatically increasing prevalence of CKD during the past two decades has not been attributable to any of these factors [7]. Thus, it has been named CKD of uncertain etiology (CKDu). The majority of CKDu patients in the NCP are from the low socioeconomic farming community [8]. As a result of the insidious onset and progression of the disease, these patients present in the late stages, which requires renal replacement therapy such as dialysis and transplantation. In 2005, 4.6% of the national health budget (annual) was being spent on the management of renal disease patients [9].

In this CKDu endemic region, an epidemiological survey to evaluate the possible risk factors has revealed that a family history of CKDu, taking ayurvedic treatment, and a history of snake bite are significant predictors for CKDu occurrence [8]. As a result of the clinically observed heterogeneity in disease progression, we hypothesized that the rate of disease progression and death could be modified by different demographic, clinical, and lifestyle-related factors. To evaluate this, we followed up a CKDu cohort that was identified in a high-risk population screening program, which is believed to be the first-ever cohort study of CKDu in the NCP of Sri Lanka. It is expected that risk factor identification will be crucial in the prevention of CKDu progression from mild to severe stages.

Methods

Cohort

Target population screening programs using dipstick proteinuria were started in 2003 when there was a surge in new registrations with CKDu in the local renal clinics in the NCP. Early-morning urine samples from people aged >5 years (except for menstruating women) were checked on three occasions, with an interval of 1 week between consecutive samples. Subjects with proteinuria on two or more occasions were referred to renal clinics for further confirmatory investigations. Patients with CKDu were diagnosed after excluding known etiologies. Diabetes mellitus was excluded by the absence of a history, no current treatment, and hemoglobin A1C (HbA1C) <6.5%. Malignant hypertension was excluded by the absence of a history of chronic and/or severe hypertension. If blood pressure was <160/100 mmHg untreated or <140/90 mmHg with up to two antihypertensive agents, those people were included in the CKDu group. Systemic lupus erythematosus was excluded by the absence of anti-nuclear factor and double-stranded DNA antibody. In all cases, biopsy was conducted for pathological diagnosis. CKDu was pathologically defined as a region of focal-to-diffuse interstitial fibrosis and tubular atrophy. IgA nephropathy was excluded by immunostaining for IgG, IgM, IgA, and complement component C3. Urological diseases of known etiology were excluded by clinical symptoms and routine investigations.

All patients who were diagnosed with CKDu were registered in local clinics. We recruited 143 CKDu patients at the time of registration at the Medawachchiya renal clinic after screening in 2003. All the patients recruited for the cohort were followed up for 72 months from registration until 2009 or death.

Baseline risk factor assessment

After registering the patients in the clinic, demographic and lifestyle-related information was collected by physicians through face-to-face interviews using a structured questionnaire. If available, medical records were used to collect the relevant clinical information. The risk factors of interest were: source of drinking water (agro-well, tube well, or garden well); habits (alcohol consumption, smoking, or betel chewing); family history of CKD, and history of snake bites; frequently used medication; and parental consanguinity. A history of hypertension, malaria, or bed-wetting, and ultrasonographic findings of shrunken kidney were collected from medical records. The presence of dental fluorosis was recorded after examination. Patients with mild hypertension (blood pressure <160/100 mmHg untreated, or <140/90 mmHg with up to two antihypertensive agents) were classified as hypertensive patients. If patients developed diabetes mellitus during the follow-up period, they were classified as diabetics.

Covariates

Body mass index

The height and weight of the patients were measured in increments of 0.1 cm and 0.1 kg, respectively, while they were wearing light clothing. Body mass index (BMI) was calculated as body weight (kg) divided by the square of height (m). BMI was categorized according to the World Health Organization guidelines: <18.5 kg/m2 (underweight), 18.5–24.9 kg/m2 (normal), 25–29.9 kg/m2 (overweight), and ≥30 kg/m2 (obese).

Serum creatinine and glomerular filtration rate

The latest serum creatinine level was used to calculate the estimated glomerular filtration rate (eGFR) using the Modified Diet in Renal Disease (MDRD) formula [10]:

$$ \begin{aligned}{\text{GFR }}\left( {{\text{ml}}/{ \min }/ 1. 7 3 {\text{m}}^{ 2} } \right)& = 1 8 6 \times \left( {\text{serum creatinine}} \right)^{ - 1. 1 5 4} \,\times \,\left( {\text{Age}} \right)^{ - 0. 20 3}\\ &\quad\times \left( {0. 7 4 2 {\text{ if female}}} \right) \times \left( { 1. 2 3 3} \right). \end{aligned} $$

According to one of the most widely adopted classifications, proposed by the Kidney Disease Outcomes Quality Initiative, five stages were defined for CKDu using the eGFR value [11]. If eGFR was ≥90 ml/min/1.73 m2, but there was evidence of kidney damage such as proteinuria, patients were classified as CKDu stage 1. Stages 2–5 were classified according to eGFR values of 60–89, 30–59, 15–29, and <15 ml/min/1.73 m2 in stages 2, 3, 4, and 5, respectively.

Kidney disease follow-up

For each member of the cohort, the person-years of follow-up were calculated from 2003 to the date of death or to the final year of the study period (2009). The entire cohort was closely monitored for 2 years (cohort established in 2003 and followed up in 2004 and 2005) for deaths and disease progression, and then the patients were invited to participate in the annual clinic to determine disease progression. It was observed that 34 patients were missing in 2005 without any records, and 81 were reported in 2006 and 76 in 2007. Forty-four patients were reported in 2008, and in 2009 only 25 were reported to the annual follow-up program.

Statistical analysis

Values for demographic variables, BMI, and history of hypertension in the study participants are presented as means with standard deviation (SD). MDRD-based eGFR was categorized by sex and various age groups, using the SAS Statistical Package version 9.1 (SAS Institute, Cary, NC, USA). After stratification by age and eGFR, sex/age adjusted hazard ratios (HRs) for CKDu were obtained using a Cox proportional hazards model. The risk factors considered in the analysis were: source of drinking water (agro-well, tube well, or garden well); habits (alcohol consumption, smoking, or betel chewing); family history of CKD; history of hypertension, malaria, bed-wetting, or snake bites; frequently used medication; parental consanguinity; and clinical findings (dental fluorosis, ultrasonographic findings of shrunken kidney, BMI, serum creatinine levels, and CKDu stage at follow-up). In the first trial of the Cox model, each factor was individually considered (univariate Cox model), and age-corrected, sex-stratified significant factors for disease progression were identified for CKDu stage 1–4 patients, using SAS (PROC PHREG—outcome of the model was disease progression). Disease progression was determined according to the advance of CKDu stage over the monitoring period. In the second trial of the Cox model, all significant factors identified in the first analysis were combined (multivariate Cox model). Similar analysis was carried out for CKDu stage 1 and stage 2–3 patients at the third and fourth Cox model analyses for disease progression. Univariate and multivariate Cox model analyses (fifth and sixth Cox model analyses) were carried out to determine individual and combined mortality factors. A Kaplan-Meier curve was generated with SAS soft ware. Log-rank and Wilcoxon tests were carried out to check the homogeneity of the survival curves. All analyses were performed with the SAS Statistical Package version 9.1 (SAS Institute). This was a fact-finding study; therefore, we placed an emphasis on false-negative results. Thus, we considered p < 0.10 as significant.

Results

Study cohort

The study cohort consisted of 143 patients with a mean age (SD) of 43 (6.5) years. They were predominantly Sinhalese in ethnicity (86%) and belonged to rural communities. The number of participants and their demographic characteristics, BMI, and history of hypertension are shown in Table 1.

Table 1 Distribution of demographic variables, BMI, and history of hypertension among study participants in 2003
Full size table

Fifty-eight percent of the participants were men. Importantly, the majority of the participants had low BMI (42%) and only 8% had high BMI (overweight and obese). The prevalence of hypertension was 24%, and 3% developed diabetes during the follow-up period. The distribution of MDRD-estimated GFR categories by sex and age is presented in Table 2. The proportion of early-stage CKDu patients was markedly reduced in the older age groups. For example, in the youngest population (age <20 years), 92% were CKDu stage 1 (GFR >90 ml/min/1.73 m2), but in the older population (aged 60–69 years), the CKDu stage-1 population was only 9%. In contrast, very few or no late-stage patients were found in the younger age group. For example, no patients were found in CKDu stage 4 (GFR 15–29 ml/min/1.73 m2) in the younger age group (<20 years), but almost 40% of the patients in the older age group (60–69 years) were CKDu stage 4.

Table 2 MDRD-based GFR according to sex and age in 2003
Full size table

The cohort was closely monitored for 2 years for disease progression and death, and they were invited to join the annual monitoring program for another 4 years. The mean period of observation was 39.9 months. It was found that 11 patients died during the monitoring period of 2003–2005, and the numbers of patients who came to the follow-up clinics in 2005, 2006, 2007, 2008, and 2009 were 90, 81, 76, 44, and 25, respectively. Table 3 shows the number of deaths recorded during the first 2 years, categorized by CKDu stage and age (above and below 65 years). The mortality rates in the present study were compared with those reported in similar CKD cohort studies from different parts of the world.

Table 3 CKDu cohort deaths during first 2 years follow-up (2004 and 2005)
Full size table

Predictors of disease progression

A sex-stratified, age-adjusted, Cox proportional hazards model was used to identify the individual parameters for disease progression: occupation, drinking water source, alcohol consumption, smoking, betel chewing, family history, hypertension, diabetes mellitus, malaria, bed-wetting, snake bites, medication used, parental consanguinity, dental fluorosis, shrunken kidney, and BMI. Table 4 shows the significant factors obtained for disease progression for patients with stage 1–4 CKDu. Alcohol consumption, betel chewing, nonsteroidal anti-inflammatory drugs (NSAIDs), and hypertension were identified as major individual factors (age-adjusted and sex-stratified) for disease progression, with HRs of 3.39, 3.61, 2.33, and 2.02, respectively (p < 0.1). In the multivariate analysis with all significant factors, alcohol consumption and hypertension were the main factors for disease progression, with HRs of 3.64 (p = 0.076) and 3.38 (p = 0.011), respectively. We repeated the same analysis for stage 1 and stage 2–3 patients, and the significant results are shown in Table 5. Hypertension was the only factor identified for disease progression in stage 1 patients, with an HR of 7.29 (p = 0.34). Even though it was statistically not significant (p = 0.12), it is worth highlighting the factor of family history, with an HR of 21.63. Hypertension, BMI, and malaria were identified as the significant factors for disease progression in stage 2 and 3 patients (sex-stratified and age-adjusted). Hypertension showed the highest HR, of 3.07 (p = 0.060), in this analysis. The dominance of hypertension was further emphasized by multivariate analysis, with an HR of 26.14 (p = 0.003); BMI and malaria showed lower HRs, of 0.64 (p = 0.011) and 0.27 (0.083), respectively.

Table 4 Sex-stratified, age-adjusted univariate and multivariate Cox proportional models for CKDu disease progression and death
Full size table
Table 5 Sex-stratified, age-adjusted univariate and multivariate Cox proportional models of disease progression for stage-specific analysis
Full size table

It was observed that hypertension repeatedly emerged as an important factor for disease progression, and we developed Kaplan–Meier survival curves to illustrate this further (Fig. 1). Different rank statistics tests were used to check the homogeneity of the survival curves. Log-rank and Wilcoxon tests showed χ 2 values of 11.4 (p = 0.0007) and 10.1 (p = 0.0014), respectively, which suggested a comparatively rapid disease progression if the patients had a history of hypertension in the early stages of the disease.

Fig. 1
figure 1

Kaplan–Meier curves showing the probability of chronic kidney disease of uncertain etiology (CKDu) disease progression over the monitoring period with and without self-reported hypertension for CKDu stage 1 (glomerular filtration rate [GFR] >90 ml/min/1.73 m2) patients (please refer to the “Results” section in the text for Log-rank test and Wilcoxon test results)

Full size image

In addition to hypertension, alcohol consumption and betel chewing were also identified as significant factors for disease progression by univariate Cox model analysis, but in the multivariate analysis, their significance disappeared. However, it is noteworthy that in a previous study, by Chou et al. [12], it was found that betel-nut use was associated with CKD in men and this association was independent of age, BMI, smoking, alcohol consumption, hypertension, diabetes, and hyperlipidemia.

Predictors of mortality

Univariate Cox survival analysis (age-adjusted and sex-stratified) suggested that hypertension, CKDu stage, and shrunken kidney were the significant factors associated with mortality rate, with HRs of 2.29 (p = 0.1), 3.00 (p = 0.00), and 2.74 (p = 0.09), respectively (Table 4). In multivariate Cox survival analysis, the dominance of CKDu stage was observed over other significant factors, with an HR of 2.94 (p = 0.001).

Discussion

This study is believed to be the first to identify possible risk factors of CKDu disease progression and death in a prospective cohort in Madawachchiya, Sri Lanka, where CKDu is endemic.

The age-structure pyramid of this CKDu cohort was broad-based, with a relatively large proportion of younger patients. For example, 17% of the CKDu patients were <20 years old. In contrast, the proportion of CKD patients aged <20 years in the United States is <2% [13]. Almost all the younger patients (<20 years) in this cohort had CKDu stage 1 and they comprised 24% of the stage-1 population. Therefore, it is reasonable to presume that the problem will become much more serious in the coming few decades when these patients progress to late stages of the disease. Even though the older age group (>70 years) comprises 37% of the CKD population in the United States [13], only about 10% of the patients in our cohort were >70 years old.

A similar age-structure pyramid in the CKD population has been reported from another Asian country, Iran, where 23.6% are aged <30 years, and only 6.0% are >70 years old [4]. In the present cohort, 37% of stage 3–5 CKDu patients were older than 60 years.

In our cohort, the probability of death in a 2-year period was 23% for a patient older than 65 years and that was 8% for a patient younger than 66 years. These values are almost double those reported by Manns et al. [14], who showed that the probability of death in a CKD cohort in Alberta, Canada, was 4.81% for patients aged <66 years and 13.29% for those aged >65 years. It is particularly interesting to note that 81% of the Canadian cohort had CKD stage 1 or 2, 18% had stage 3, and 1% had stage 4 or 5, whereas in our cohort, 50% had stage 1 or 2, 22% had stage 3, and 28% had stage 4 or 5. This demonstrated that the disease progression in our CKDu cohort was much faster than that in western countries, and we suspect that this could be a reason for the higher death rate in our cohort. The overall death rate in our cohort was 5.3% per 100 patient-years, which is higher than the rate reported in Taiwan [15]. Lack of adequate dialysis facilities to manage the CKDu patients in this region might have contributed to this higher mortality.

BMI is associated with CKD [16, 17]. Many researchers have reported that higher BMI is a risk factor for CKD in apparently healthy persons [18, 19]. By contrast, in our cohort, only 8% of the patients were overweight (25–29.9 kg/m2) or obese (>30 kg/m2), whereas 42% of the patients were underweight (<18.5 kg/m2). According to some studies, malnutrition has been shown to increase mortality risk more than high BMI [20, 21], and malnutrition could have also contributed to the higher death rate in our cohort.

In our study, men (58%) were at higher risk of CKDu than women were (42%). A similar association has been reported in a study in Taiwan [15], where 62% of the affected population was male. However, the majority of published studies have shown female-dominant CKD cohorts [4, 22, 23]. Further research is needed to determine the reasons for male dominance in our study area. However, factors such as genetic susceptibility and some behavioral risk factors, such as alcohol consumption and betel chewing, which could play a role in increasing disease occurrence, should not be neglected.

Diabetes and hypertension are known risk factors for CKD [24, 25]. According to the definition of CKDu, patients with diabetes and uncontrollable hypertension were not included in the present cohort. However, it should be noted that five patients developed diabetes and several developed hypertension during the follow-up period. At the same time, several patients had mild hypertension at the time of diagnosis of CKDu. Among the factors that were evaluated as suspected risk factors for disease progression in our cohort, hypertension was identified as the leading determinant of disease progression. Self-reported hypertension was a significant predictor (p < 0.05) for disease progression, with an HR of 3.38 by multivariate Cox proportional model analysis for stage 1–4 patients. Our results suggest a strong correlation between hypertension in the early stage of CKDu and disease progression. Sixty-six percent of stage 1 patients in our cohort were from the population aged <30 years, and hypertension showed an HR of 7.26 in a sex-stratified, age-adjusted Cox proportional model for stage 1 patients. A similar correlation was observed for stage 2–3 patients, with an HR of 26.16. Many researchers have shown that the prevalence of hypertension in South Asian populations is higher than that in western populations [26]. These results strongly suggest that tight control of blood pressure, particularly in the early stage of CKDu, has a beneficial effect to slow down disease progression.

CKDu stage was the only factor identified to have a correlation with death, with HRs of 3.0 and 2.94 by univariate (sex-stratified and age-adjusted) and multivariate Cox proportional model analysis, respectively. This result is supported by the fact that almost all reported deaths were in patients with stage 4 or 5 disease. In contrast, many studies have reported that CKD patients in western countries frequently die before developing end-stage CKD [27–29]. The reason for this could be the difference in age at disease initiation, because in many countries, CKD is common in the older age groups [30]. In the elderly population, eGFR is reduced in the normal aging process, but elderly patients have a greater mortality risk when compared with younger patients with the same eGFR range [30]. Therefore, it is likely that elderly patients will die in the middle of the disease process rather than reaching the end stage of CKD. However, our CKDu cohort was relatively younger than the cohorts that have been studied in other countries [15, 24, 25]. Thus, we can imagine that these CKDu patients would survive until they reach end-stage CKD, in contrast to the population who are already in the latter part of their life when they begin the disease process.

Our study had a number of limitations. The study was started at the initiation of community screening in this endemic area; therefore, we could recruit only a limited number of patients. We used the MDRD formula for estimation for GFR, which has not been validated for the Sri Lankan population and across the age range. Therefore, the use of eGFR for staging the patients might not have reflected the actual GFR for this population. For the exclusion of diabetes mellitus, we used a cut-off HbA1C level of 6.5%. However, previous studies have shown that such a cut-off level misses a substantial number of people with type 2 diabetes, including some with fasting hyperglycemia, as well as people with impaired glucose tolerance [31]. Therefore, there is a possibility that we might have recruited some diabetic patients in this cohort because of misdiagnosis.

Conclusions

In this study, we tried to determine factors possibly associated with progression and mortality of CKDu in a single-center cohort registered in the Medawachchiya regional clinic in Sri Lanka. We repeatedly found an association between disease progression and hypertension. Men were at higher risk of CKDu than women. A significant proportion of the patients in this cohort were underweight, indicating a need for future studies on the nutritional status of these patients. Compared with findings in western countries and other regions of Asia, younger age at disease onset, equal patient distribution in different age groups, faster disease progression in those with hypertension, deaths occurring mostly in end-stage CKDu, and relatively higher mortality rate are some of the characteristics that were identified in this cohort.