CKD of Uncertain Etiology: A Systematic Review

Abstract

Background and objectives

Epidemics of CKD of uncertain etiology (CKDu) are emerging around the world. Highlighting common risk factors for CKD of uncertain etiology across various regions and populations may be important for health policy and public health responses.

Design, setting, participants, & measurements

We searched PubMed, Embase, Scopus and Web of Science databases to identify published studies on CKDu. The search was generated in January of 2015; no language or date limits were used. We used a vote-counting method to evaluate exposures across all studies.

Results

We identified 1607 articles, of which 26 met inclusion criteria. Eighteen (69%) were conducted in known CKDu–endemic countries: Sri Lanka (38%), Nicaragua (19%), and El Salvador (12%). The other studies were from India, Japan, Australia, Mexico, Sweden, Tunisia, Tanzania, and the United States. Heavy metals, heat stress, and dietary exposures were reported across all geographic regions. In south Asia, family history, agrochemical use, and heavy metal exposures were reported most frequently, whereas altitude and temperature were reported only in studies from Central America. Across all regions, CKDu was most frequently associated with a family history of CKDu, agricultural occupation, men, middle age, snake bite, and heavy metal exposure.

Conclusions

Studies examining etiologies of CKDu have reported many exposures that are heterogeneous and vary by region. To identify etiologies of CKDu, designing consistent and comparative multisite studies across high-risk populations may help elucidate the importance of region–specific versus global risk factors.

Introduction

CKD is a growing health burden globally with many known and unknown etiologies (1). In high-income countries, it is most commonly associated with noncommunicable diseases, such as diabetes and hypertension. However, in low- and middle-income countries, it has several additional potential etiologies, such as infectious diseases and environmental toxins, but many remain unknown (1–3). CKD of uncertain etiology (CKDu) is a term that has been used to describe CKD that is not attributable to any traditional risk factor, such as diabetes, hypertension, or HIV. CKDu is being reported with increasing frequency across the globe, and in many parts of Central America, eastern Europe, and south Asia, it is being reported in epidemic proportions (4–6).

The term CKDu was first used in El Salvador in the early 2000s to describe a disease predominantly affecting agricultural communities from large-scale plantations (7,8). Subsequently, this disease pattern, which was similar to that of Balkan Endemic Nephropathy (9), was reported in other countries, including Nicaragua, Costa Rica, and Sri Lanka (6,10–12), and its prevalence is now as high as 16% in some of these areas (13). In El Salvador, ESRD that is largely attributable to CKDu has become the leading cause of hospital deaths (8), and in Sri Lanka, nearly 4% of the national health care budget is now spent on treating CKDu (14,15). As such, it has become a public health priority and the focus of many World Health Organization–sponsored epidemiologic studies (16–18).

One proposed common feature of CKDu in these regions is its occurrence mainly in agricultural communities and most conspicuously, among young male farmers (19,20). Environmental factors, such as heavy metal exposures (21), high seasonal temperatures (22), agrochemical use, mycotoxins (23), contaminated water supplies, and snake bite (24), have all been studied as potential causes of CKDu. However, the underlying etiologies have not been clearly identified in most regions (18), and highlighting common risk factors across these global settings and populations could be important for understanding and preventing CKDu. Therefore, we performed a systematic review of the existing literature to identify common exposures and potential risk factors for CKDu across different global populations.

Materials and Methods

Search Strategy

We searched the PubMed, Embase, Scopus, and Web of Science databases to identify published studies. Search terms included a combination of relevant subject headings and text words for kidney disease (e.g., CKD and CKDu) and the potential exposures (e.g., agriculture, heavy metals, and ground water) (Supplemental Appendix 1). We narrowed our search to randomized trials and observational studies using the Cochrane Highly Sensitive Search Strategy filter for identifying randomized, controlled trials (25), which we expanded to include observational trials. Without limits on language or publication dates, the search was generated in January of 2015. We supplemented the search by screening both the cited and citing references of all included studies.

Selection Criteria

All abstracts were retrieved by two authors (J.L. and M.V.I.). Two authors (J.L. and D.M.) then independently reviewed each title and abstract for inclusion. Any disagreement resulted in a joint review of the full article with reconciliation that led to inclusion or exclusion.

After review of the abstracts, each article was independently reviewed in full by two authors (J.L. and D.M.). Studies were included if they assessed the association between CKDu and potential exposures and reported a statistical association between CKDu and exposures. Randomized, controlled trials and nonrandomized studies, including cross-sectional studies, cohort studies (retrospective or prospective), and case-control studies, were eligible for inclusion. Editorials, reviews, and case reports without the required data were excluded. We excluded studies if CKDu was not differentiated from CKD by failing to make the distinction in the case definition and/or failing to account for traditional risk factors for CKD, such as diabetes and hypertension.

Quality Assessment and Data Extraction

Each selected study was independently assessed by two authors (J.L. and D.M.) using standard quality assessment scales for observational studies (26,27). Cross-sectional, case-control, and cohort studies were assessed for quality on the basis of selection, comparability, and outcome measures (Supplemental Appendix 2). On the basis of these criteria, each article received a numeric quality score on a scale ranging from zero (lowest) to nine (highest) for cohort studies and from zero to 10 for case-control and cross-sectional studies. Studies with quality scores eight or higher were considered high quality. Studies with scores six or seven were considered medium quality, and studies with scores five or lower were considered to be low quality. Any disagreement resulted in joint review of the article with reconciliation. In cases where reconciliation was not achieved, a third author (J.W.S.) independently reviewed the articles to provide majority consensus.

Two authors (J.L. and D.M.) independently extracted the data from all included studies, and they entered the data independently and in duplicate into a preformulated table. Errors in data extraction were resolved by joint review of the original articles. We extracted the study characteristics, including authors, study dates, study location, sample size, and study design, and we extracted the measured exposures for CKDu, including heavy metals, water hardness, ambient temperature, agrochemical use, heat stress, demographics (age, sex, and occupation), family history of CKDu, smoking, herbal use, snake bites, and diet. In cases of uncertainty or missing data, we attempted to contact the corresponding authors for additional information.

Statistical Analyses

We assessed inter-rater agreement for inclusion and quality assessment using Cohen κ-coefficient, and using an I² statistic, we investigated heterogeneity across the medium– and high–quality case-control studies for the most commonly reported risk factors from these studies. Because we wanted to describe the range of potential exposures that may contribute to CKDu, we used a vote-counting method to evaluate the exposures across the different studies. For application to studies on CKDu, this method had the advantage of providing a descriptive summary of a range of variables across heterogeneous studies. It did not provide estimates pertaining to the magnitude or influence of a particular variable, which would be expected in a meta-analysis (28).

We assigned a positive (+), negative (−), or null (0) value for each association between exposure and outcome (CKDu) that was measured in the individual studies. To receive a positive or negative value, the association had to be reported as significant in the positive or negative direction, respectively, at a P value <0.05. In cases where the association was reported as not statistically significant (P>0.05), a null value was assigned. For associations in which P values were not reported, no entry was assigned (Supplemental Appendix 3).

Results

We identified 1607 articles for review. Review of the title and abstracts yielded 89 articles for full-text review, and one additional article, which was in press at the time of the search but known to the authors, was included. The full-text review yielded 26 articles for final inclusion (Figure 1). Inter-rater agreement for the included articles was excellent (κ=0.90). Of 26 included studies, 18 (69%) were conducted in known CKDu-endemic countries, including Sri Lanka (10; 38%), Nicaragua (five; 19%), and El Salvador (three; 12%) (Tables 1 and 2). The other studies were from India, Japan, Australia, Mexico, Sweden, Tunisia, Tanzania, and the United States. All included studies were conducted between 2003 and 2015. Most were cross-sectional (12; 46%) or case-control (12; 46%), and only two were cohort designs (Table 1).

Figure 1.

Flow diagram illustrating included and excluded studies in the systematic review.

Table 1.

Characteristics of studies included in the systematic review

Authors	Country	Year	Sample Size, n	Design	Quality
South Asia
Siriwardhana et al. (47)	Sri Lanka	2014	200	Case-control	Medium
Nanayakkara et al. (39)	Sri Lanka	2014	815	Case-control	High
Bandara et al. (48)	Sri Lanka	2008	64	Case-control	Low
Chandrajith et al. (12)	Sri Lanka	2011	44	Case-control	Low
Athuraliya et al. (29)	Sri Lanka	2011	6153	Cross-sectional	High
Jayatilake et al. (17)	Sri Lanka	2013	627	Cross-sectional	High
Wanigasuriya et al. (21)	Sri Lanka	2011	886	Cross-sectional	High
Nanayakkara et al. (49)	Sri Lanka	2012	237	Case-control	Medium
Wanigasuriya et al. (24)	Sri Lanka	2007	383	Case-control	Medium
Rango et al. (45)	Sri Lanka	2015	134	Case-control	Medium
Siddarth et al. (38)	India	2013	668	Case-control	Medium
Central America
VanDervort et al. (50)	El Salvador	2014	16,384	Cross-sectional	High
Orantes et al. (4)	El Salvador	2011	775	Cross-sectional	Medium
Peraza et al. (32)	El Salvador	2012	664	Cross-sectional	High
Torres et al. (20)	Nicaragua	2010	1096	Cross-sectional	Medium
Laux et al. (22)	Nicaragua	2012	267	Cross-sectional	Medium
O'Donnell et al. (31)	Nicaragua	2011	771	Case-control	High
Sanoff et al. (51)	Nicaragua	2010	997	Case-control	Medium
Raines et al. (30)	Nicaragua	2014	424	Cross-sectional	High
Nonendemic
Maruzeni et al. (52)	Japan	2014	9446	Cohort	Medium
Ivey et al. (33)	Australia	2013	948	Cohort	Medium
Robles-Osorio et al. (53)	Mexico	2012	90	Cross-sectional	Medium
Sommar et al. (54)	Sweden	2013	496	Case-control	High
Abid et al. (55)	Tunisia	2003	1159	Case-control	Low
Stanifer et al. (37)	Tanzania	2015	481	Cross-sectional	High
Sontrop et al. (56)	United States	2013	3427	Cross-sectional	Medium

Table 2.

Study sites by region

Countries	Studies, n (%)
South Asia	11 (42)
Sri Lanka	10 (38)
India	1 (4)
Central America	8 (31)
Nicaragua	5 (19)
El Salvador	3 (12)
Other regions	7 (27)
Japan	1 (4)
Australia	1 (4)
United States	1 (4)
Tunisia	1 (4)
Tanzania	1 (4)
Sweden	1 (4)
Mexico	1 (4)

Ten (38%) studies met the criteria for high quality, 13 (50%) studies met the criteria for medium quality, and three (12%) were considered low quality (Table 1). The inter-rater agreement for study quality was excellent (κ=0.90). The common reasons for studies not meeting high-quality criteria were failure to report the nonresponse rate (24 [92%] case-control and cross-sectional studies), failure to justify sample size (21 [81%] studies), and inadequacy of follow-up (one [4%] cohort study). Several studies were also limited by their lack of validated measurements or clear descriptions of the methods used to measure risk factors (Supplemental Appendix 2). For medium– and high–quality case-control studies conducted in endemic regions, we noted significant heterogeneity across studies in the reported estimates for the two most commonly reported risk factors for CKDu (farming occupation: n=5; I²=84.3%; P<0.01 and men: n=5; I²=73%; P<0.01) (Supplemental Appendix 4).

Risk factors most frequently measured were age (n=13; 50%), occupation (n=12; 46%), sex (n=12; 46%), agrochemical use (n=10; 38%), heavy metals (cadmium, arsenic, lead, and aluminum; n=9; 35%), smoking (n=8; 31%), family history of CKDu (n=7; 27%), body mass index (BMI; n=7; 27%), alcohol intake (n=6; 23%), and heat stress (n=5; 19%). Other less frequently measured risk factors included water hardness (n=3; 12%), altitude (n=3; 12%), dietary exposure (e.g., rice and rice products, fish and seafood, legumes, and meats; n=3; 12%), tobacco chewing (n=2; 8%), snake bite (n=2; 8%), ochratoxin A (a mycotoxin; n=1; 4%), ambient temperature (n=1; 4%), and herbal medicines (n=1; 4%).

All studies (n=11) that reported temperature, altitude, dietary exposure, ochratoxin A, herbal use, and snake bite found a significant association between CKDu and these exposures (Figure 2A). These studies were all conducted in Central America and South Asia. Significant associations were also reported for family history of CKDu (n=6), age (n=10), men (n=9), and heat stress (n=4). Water hardness was not reported in any study as significantly associated with CKDu, and smoking was only reported as significant in one study (21).

Figure 2.

Frequency of measured exposures from studies across different regions. (A) Exposures measured in studies from all regions; (B) exposures measured in studies from South Asia; (C) exposures measured in studies from Central America. BMI, body mass index.

Heavy metals, heat stress, and dietary exposures were reported in studies across all geographic regions (Figure 2A); however, we did observe regional heterogeneity in the other exposures reported (Table 3). Family history (n=5), agrochemical use (n=4), and heavy metal exposures (n=6) were reported most frequently in studies from South Asia, whereas altitude (n=3) and temperature (n=1) were reported only in studies from Central America.

Table 3.

Five most frequently studied CKD of uncertain etiology risk factors by region

CKDu Risk Factors by Region	Frequency
South Asia (n=11; 42%)
Heavy metals	6
Occupation (farmer)	6
Family history	5
Agrochemical use	4
Smoking	4
Central America (n=8; 31%)
Age	7
Men	7
Agrochemical use	6
Occupation (farmer)	5
Heat stress	3
Other regions (n=7; 27%)
Body mass index	3
Heavy metals	2
Age	2
Dietary exposure	1
Heat stress	1

CKDu, CKD of uncertain etiology.

In South Asia, heavy metals (n=6), occupation (n=6), and family history (n=5) were studied most frequently. Family history (n=5), snake bites (n=2), dietary history (n=1), tobacco chewing (n=1), herbal use (n=1), and heat stress (n=1) were found to be positively associated with CKDu in all studies that reported these exposures. Exposures associated with a lower risk of CKDu were higher BMI (n=1), men (n=1), agrochemical use (n=1), and heavy metals (n=1) (Figure 2B). In the high-quality studies from this region, which were all conducted in Sri Lanka, the prevalence of CKDu among high-risk populations was reported to be 14.9% (17), 9.5% (29), and 7.3% (21).

In Central America, age (n=6), sex (n=6), and occupation (n=6) were studied most frequently. Age (n=6), men (n=6), low altitude (n=3), high ambient temperatures (n=1), and dietary history (n=1) were found to be positively associated with CKDu in all of the studies that reported them. Higher BMI and heat stress were negatively associated with CKDu in one third of the studies that reported them (Figure 2C). Among high-risk populations from this region, the prevalence of CKDu in high-quality studies was reported to be 25.9% and 13.0% in Nicaragua (30,31) and 18.0% in El Salvador (32).

In nonendemic countries (four of which were the high-income countries of the United States, Sweden, Japan, and Australia), BMI (n=2) and heavy metals (n=2) were studied most frequently as potential exposures related to CKDu. Heavy metals (n=2), heat stress (n=1), and ochratoxin A (n=1) were found to be positively associated with CKDu in all of the studies that reported these exposures. The single study that suggested that dietary intake was associated with a lower risk of CKDu examined the protective role of proanthocyanidins on kidney function among elderly Australian women (33).

Discussion

Despite an increasing global awareness of CKD and CKDu, many of the risk factors remain unknown (1). We identified many potential risk factors for CKD that may vary by region, but the heterogeneity in their reported associations with CKD limits conclusions about the etiologies of CKDu. In South Asia, family history, agrochemical use, and heavy metal exposures were studied most frequently, whereas altitude and temperature were studied only in Central America. However, many similarities also exist. Heavy metals, heat stress, and dietary exposures were reported in studies across all geographic regions, and family history, temperature, altitude, dietary exposure, ochratoxin A, herbal use, and snake bite were frequently reported in both South Asia and Central America.

Given the similarities and the differences observed in studies across the regions, the growing CKD burden may, in part, be driven by factors that are common across regions as well as unique within regions. Pathologic exposures can affect disease outcomes by interacting with a wide range of factors, including source emissions, transport and transformation, human contact, bioavailability, early expression of disease, and/or health effects (34). In low-income countries, for example, rapid urbanization has led to poor sanitation, unplanned infrastructure, overcrowding, and environmental pollution (35,36). For CKD, these exposures may interact with other urban risk factors, such as high rates of noncommunicable and communicable diseases, to increase CKD prevalence (37).

Likewise, in rural, low–income areas, extreme poverty and agricultural-based economies expose people to other CKD risk factors, such as dehydration, snake bite, water contamination, heavy metals, and agrochemicals, which can also interact with noncommunicable and communicable diseases as well as genetic factors to increase CKD risk (1,3). As such, the regional variation in reporting of risk factors for CKDu may reflect a complex interplay between different global and regional exposures and local factors, such as environment and lifestyle.

One example of regional variation in the reporting of risk factors for CKDu was observed with dehydration. Although dehydration is increasingly being posited as a potential etiologic factor for CKDu in endemic communities (22), assessment of the associations between dehydration and CKDu across studies was limited by inconsistent reporting of the measurements for assessing dehydration. For example, whereas all studies that measured variation in altitude and seasonal temperature reported a significant association between dehydration and CKDu, studies reporting heat stress infrequently reported an association between dehydration and CKDu. This highlights the importance of designing standardized measurements to consistently and comparatively assess the role of dehydration in the etiology of CKDu across the world.

Additionally, genetics may also play a role in the observed regional heterogeneity in the reported epidemiology of CKDu. Genetic differences in ethnicities are known to have a strong effect on the prevalence and risk of progression of CKD. In the United States, for example, Americans of African descent who carry the APOL1 genotype have higher rates and faster progression of CKD, especially when exposed to other augmenting factors, such as diabetes and hypertension. In India and Sri Lanka, there are specific polymorphisms (38) and single-nucleotide mutations (39) that have been associated with CKDu in some endemic communities, and in the Balkan states, particular chromosomal aberrations have been associated with Balkan Endemic Nephropathy (40,41). Therefore, more studies should be conducted to explore the role of genetic predisposition in the CKDu disease mechanism.

Agrochemical use has been widely considered to be a risk factor for CKDu, and it may be important to distinguish CKDu in agricultural communities from CKDu in other regions across the world (11–13). Agrochemicals, such as glyphosate, organochlorine compounds, and paraquat, are used across the globe in farming communities as pesticides and herbicides. However, despite their use, there are safety concerns about their health effects, including oxidative damage to the kidneys (42). In our review, only two studies from Sri Lanka and two studies from Central America reported a significant association between agrochemical use and CKDu. Agrochemical use was not reported in studies from other regions, suggesting the possibility of a distinct etiology for CKD in agricultural communities in CKDu-endemic regions (6,11–14). These findings may reflect variation in the regional exposures being studied as well as difficulties in defining and quantifying agrochemical exposure, but they also highlight the need for more conclusive epidemiologic studies examining causation, particularly in agricultural communities in CKDu-endemic regions.

Additionally, the causal association between heavy metals and CKDu is not well established, and we observed significant variability in the reported associations. For example, cadmium was inconsistently reported to be associated with CKDu, whereas arsenic was consistently associated with CKDu. We also observed regional heterogeneity in the reporting of heavy metals as a risk factor for CKDu. Only one study from Central America reported heavy metals, whereas six studies from South Asia reported it. It is likely that heavy metals do play a role in the development of CKDu in endemic areas, where levels are often present in the microenvironment and multiple other risk factors coexist (43–45). However, because the microenvironment and human levels of heavy metals from CKDu-endemic regions have been reported to be below the upper limits of detection, exploring mechanisms of interaction between heavy metals and other risk factors may be important in advancing CKDu research (12,39,44–46).

Our study has many strengths. Because of the limited epidemiologic data regarding CKDu, we chose not to limit our systematic review to any particular population, region, or time period, and our comprehensive search is one of the broadest assessments of CKDu globally to date. Also, our systematic approach provides a reproducible method and descriptive summary for other investigators examining CKDu.

We noted limitations as well. Because we observed significant heterogeneity among the studies, even within regions, we were not able to synthesize the data using meta-analytic techniques. Instead, we used a vote-counting method, which itself has several limitations. Typically, it assigns equal weights to the findings of different included studies. This can be problematic when quality is variable and effect sizes are not incorporated in any way. In the studies that we reviewed, sample sizes varied, and some may have been underpowered or limited in precision in their reporting of CKDu risk factors, which may have introduced a reporting bias. Finally, our findings are descriptive in nature. Studies were not weighted by quality or sample size, and the magnitudes of association were often not reported; therefore, conclusions about the relative importance of potential risk factors are limited. However, vote counting can provide a useful starting point for a systematic assessment of studies within a given research area, especially when there is a lack of agreement or consensus in the literature and when the nature of relationships across disparate sites and methodologies is not well described. Given that little is known about the causes of CKDu, this approach may be useful for generating hypotheses that can inform future epidemiologic studies as well as determine uniform causes to evaluate in future studies.

In conclusion, we describe potential risk factors for CKD that have been studied globally, many of which vary by region and are heterogeneous in their reported associations with CKDu. It remains unknown whether the regional variation in CKDu risk factors reflects inconsistencies in measurement across studies, poor quality of studies, including sample size limitations, or complexity in the interactions between global exposures and local factors, such as environment, genetics, and lifestyle. Thus, to identify the causative factors of CKDu, designing consistent and comparative multisite studies in high-risk populations may help elucidate the importance of region–specific versus global risk factors. Finally, to better characterize the global epidemiology of CKD more broadly, investigators should strive to identify the interactions between global, regional, and local factors.

Disclosures

J.W.S is supported by the American Kidney Fund Clinical Scientist in Nephrology Fellowship. Other authors have no relevant disclosures. Dr. Patel receives support from the National Institute of Diabetes, Digestive and Kidney Diseases of the National Institutes of Health under award numbers R01DK93938, R34DK102166, and P30DK096493.