Introduction
Residents of Sri Lanka’s dry zone face a high risk for kidney disease characterized predominantly by tubulointersititial kidney disease of as yet unknown etiology (1,2). Termed CKD of unknown etiology (CKDu), this condition has emerged as a leading cause of hospitalization and death, with significant political and economic upheaval in the region (3). Care for this single disease now consumes 5% of the total health care budget in Sri Lanka (4). The demographic features (2,5) and the timeline of CKDu’s emergence in Sri Lanka align with descriptions of Mesoamerican nephropathy (6–8) and Uddanam nephropathy in Andhra Pradesh, India (9).
Despite considerable efforts to improve the understanding of CKDu epidemiology and investigate its etiology, several questions remain unanswered, including whether a set of clinical features can inform reliable, noninvasive diagnosis for persons living in recognized “hot spots.” At the current time, the “gold standard” for a diagnosis of CKDu is biopsy-proven primary tubulointerstitial kidney disease with no evident predisposing condition and residence in an area with high prevalence of similar disease. However, kidney biopsies are impractical for large-scale surveillance, geographic mapping, and entry into a patient-control study. Thus, varying approaches have been applied for categorization and study inclusion. For example, the CKDu National Project Team surveyed nearly 5000 persons in a population-based study in three endemic regions and one nonendemic region of Sri Lanka (10). Persons who had albuminuria without a reduction in eGFR were classified as “CKDu cases,” despite numerous studies of CKDu pointing to tubulointerstitial kidney disease (2,6,7,11,12) as unlikely to present as albuminuria without reduced eGFR. Any assessment of exposures was biased (13). Other studies have accepted clinician- or self-reported diagnosis and restricted to variable thresholds of serum creatinine (5,14–16).
Working in a major tertiary care hospital servicing central Sri Lanka, we investigated whether a set of routinely measured clinical features was associated with biopsy-proven tubulointerstitial kidney disease and could inform standardized approaches to select patients for patient-control studies investigating CKDu etiology in endemic regions. To identify potential risk factors for further study, we also compare the demographic, occupational, environmental, and selected lifestyle exposures of patients with tubulointerstitial versus other kidney disease.
Materials and Methods
We systematically approached persons referred to Kandy Teaching Hospital nephrologists as new patients over a 1-year time period (October 1, 2016 to September 30, 2017). Persons serviced by Kandy Teaching Hospital reside primarily in the Central Province of Sri Lanka, which includes both CKDu endemic and non endemic areas. Participants provided informed consent for the study, which was approved by the ethics committees at Stanford University and Kandy Teaching Hospital.
Questionnaire Administration
A trained research coordinator administered a study questionnaire, which captured detailed data on residential, family, occupational, drinking water, and tobacco and alcohol use history. Patients were asked to report history of diabetes, hypertension, cardiovascular disease, or autoimmune disease and whether they were using any nonsteroidal anti-inflammatory agents (NSAIDs) or ayurvedic supplements routinely. At the completion of clinical visit, nephrologists reported preliminary/presumed cause(s) of kidney disease. The research coordinator abstracted BP, body weight, and laboratory measures (serum creatinine, basic metabolic panel, urine studies [including urine protein-to-creatinine ratio], and kidney ultrasound). In patients returning for kidney biopsy, the research coordinator abstracted repeat measurements. We entered data into Redcap, an open source, privacy-protected platform.
Biopsy Indications
The participants’ attending nephrologists determined the need for kidney biopsy. Standard clinical practice in this region is to refer patients with unexplained abnormal urine sediment or elevated serum creatinine for kidney biopsy, with the most common contraindication for biopsy being small kidneys (<9 cm in length on ultrasound measurement).
Biopsy Review
A single nephropathologist working with Kandy Teaching Hospital reviewed all biopsies. We triaged biopsy samples for light microscopy and immunofluorescence microscopy. The light microscopy sample was fixed, processed, and paraffin embedded. We obtained multiple hematoxylin and eosin– and silver methanamine–stained sections for examination, with additional sections stained with periodic acid–Schiff, Congo red, and Perl stain as needed. The immunofluorescence microscopy tissue was fresh frozen and stained with antisera to IgG, IgA, IgM, and C3. Electron microscopy examination was unavailable.
The pathologist made a diagnosis of primary tubulointerstitial kidney disease if the immunofluorescence study was predominantly negative and if active inflammation and/or chronic tubulointerstitial damage was the only or predominant pathology rather than glomerular and vascular abnormalities. Periglomerular fibrosis was considered a sequela of adjacent interstitial inflammation or disease. For the initial 6 months of the study period, a second nephropathologist coreviewed electronically transmitted images (n=45; 52% of biopsies) and traveled to Kandy Teaching Hospital for review of select patients. The two pathologists concurred on initial diagnoses. In patients with a diagnosis of tubulointerstitial kidney disease (n=43), the onsite pathologist reassessed histologic compartments and scored for severity (grades 1–3) of interstitial fibrosis, tubular atrophy, and tubulointerstitial inflammation using a standardized form (Supplemental Appendix) (n=41 [95%] with slides available for scoring).
Statistical Analyses
We present comparisons between patients recommended and not recommended for kidney biopsy and between patients with a diagnosis of tubulointerstitial versus other kidney disease on biopsy using t tests for continuous variables and chi-squared or Fisher exact tests for categorical variables. To evaluate the feasibility of a noninvasive approach to identification of CKDu, we preselected nine routinely available clinical and laboratory features hypothesized to be associated with presence or absence of tubulointerstitial kidney disease on kidney biopsy: age, sex, self-reported or physician-diagnosed diabetes mellitus, physician-diagnosed or measured hypertension, serum potassium and albumin, and urine microscopy characteristics (dipstick proteinuria, hematuria, and pyuria). We ran a logistic regression model; we computed the c statistic as a measure of discrimination and the calibration slope to evaluate model fit. In addition to a full model including nine preselected features, we tested five parsimonious versions to determine if a simpler model would achieve comparable predictive ability (17). In evaluating the models, we defined a patient as having probable CKDu if the predicted probability exceeded 0.5 (Supplemental Appendix 2).
To assess correlates of tubulointerstitial kidney disease, we used modified Poisson models to compute the risk ratio (and 95% confidence interval) for tubulointerstitial kidney disease after accounting for patients’ age (18). We used StataMP, version 13.1 (Stata Corporation, College Station, TX) and R (19) (version 3.4.3) package rms (20) to conduct the analyses.
Results
A total of 606 newly referred patients presented to Kandy Hospital nephrologists over the course of 1 year; 600 (99%) consented to participate in the study, and 87 (15%) underwent successful kidney biopsies (Figure 1). In contrast to patients presenting to nephrologists who were not recommended for biopsy, patients recommended for biopsy were younger, were more likely to have hematuria or pyuria, were more likely to have lower serum albumin concentrations, and were more likely to have kidney size <9 cm (Table 1). Patients recommended for biopsy were also less likely to carry diagnosis of diabetes mellitus and/or hypertension.
;)
Figure 1.: Flowchart of study participants. Of the 606 patients approached during the 1-year study period, 600 agreed to further participate; nephrologists recommended biopsy in 120, and 87 underwent successful biopsy. CKDu, CKD of unknown etiology.
Table 1. -
Characteristics of patients recommended or not recommended for
kidney biopsy at Kandy Teaching Hospital in 2016–2017
| Selected Characteristics |
Biopsy Recommended, n=120 |
Biopsy Not Recommended, n=480 |
| Age |
45 (12)
a
|
57 (12) |
| Women |
31 (26) |
122 (25) |
| Systolic BP, mm Hg |
128 (21) |
131 (20) |
| Diastolic BP, mm Hg |
80 (9) |
81 (10.4) |
|
Hypertension
b
|
54 (45)
c
|
300 (63) |
| Physician diagnosed |
27 (50) |
149 (50) |
|
Diabetes
b
|
20 (17)
c
|
157 (33) |
| Physician diagnosed |
17 (85) |
114 (73) |
|
Serum creatinine, mg/dl
|
1.9 (1.4) |
2.1 (1.5) |
| eGFR, ml/min per 1.73 m2
|
54 (29)
c
|
44 (25) |
| Missing |
|
15 (3) |
|
Serum albumin, g/dl
|
3.9 (0.9)
a
|
4.4 (0.6) |
| Low serum albumin
d
|
27 (23) |
15 (3) |
| Missing |
19 (16) |
209 (44) |
|
Hematuria
e
|
21 (18)
c
|
41 (9) |
| Missing |
2 (2) |
54 (11) |
|
Pyuria
e
|
50 (42)
c
|
116 (24) |
| Missing |
2 (2) |
54 (11) |
|
Urine dipstick protein negative
|
66 (55) |
281 (59) |
| Missing |
2 (2) |
53 (11) |
|
Small kidney
f
|
22 (18)
a
|
247 (52) |
| Missing |
20 (17) |
153 (32) |
Numbers are mean (SD) or N (%).
aP<0.001.
bPatients without physician-diagnosed hypertension had either a clinic measurement of systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg or self-reported hypertension. Participants without physician-diagnosed diabetes self-reported diabetes.
cP<0.05.
dSerum albumin <3.5 g/dl.
eHematuria was defined as more than three urine red blood cells per high power field. Pyuria was defined as more than three urine white blood cells per high power field. Urine dipstick protein trace or nil was categorized as negative.
fA participant was categorized as having a small kidney if one kidney was <9 cm.
Kidney Biopsy Findings
Of the 87 patients undergoing biopsy, 43 (49%) had a diagnosis of tubulointerstitial kidney disease. The tubulointerstitial inflammation was predominantly lymphocytic, with less frequent plasma cells and occasional neutrophils (Figure 2). Thirteen (30%) patients had moderate to severe (grades 2–3) tubulitis or interstitial inflammation in the nonatrophic cortex, which qualified as active tubulointerstitial disease. Six (15%) patients had moderate to severe chronic tubulointerstitial damage, tubular atrophy, or interstitial fibrosis. The remaining patients had mild (grade 1 interstitial fibrosis or tubular atrophy) chronic tubulointerstitial kidney disease. A high degree of global glomerulosclerosis (grades 2–3) was identified in six (15%) patients; arteriosclerosis was insignificant, with only mild changes observed. Tubular neutrophilic casts were observed in five (12%) patients; coexistent glomerular disease was evident in four (10%) patients, but on the basis of prominent tubulointerstitial inflammation with a mild degree of glomerular involvement, these patients were given the diagnosis of tubulointerstitial kidney disease. The most common diagnoses in patients without tubulointerstitial kidney disease were IgA nephropathy (n=12; 27%) and membranous nephropathy (n=7; 16%) (Supplemental Figure 1).
Figure 2.: Biopsy features of tubulointerstitial kidney disease in Sri Lanka demonstrating marked fibrosis and active interstitial inflammation. (A) Broad bands of tubular atrophy and interstitial fibrosis with relative sparing of glomeruli. In addition to global glomerulosclerosis, this atrophic cortex shows glomeruli with periglomerular fibrosis (arrows). The preserved glomeruli are unremarkable (inset), and the immunofluorescence microscopy is negative (methanamine silver stain, ×100). (B) Active interstitial inflammation in nonatrophic cortex. The infiltrate is composed of lymphocytes admixed with a few plasma cells and rare eosinophils. The glomerulus is normal, and the interlobular artery (arrow) shows mild arteriosclerosis (hematoxylin and eosin, ×200).
Characteristics of Patients with Tubulointerstitial Kidney Disease on Kidney Biopsy
Patients with tubulointerstitial kidney disease were geographically clustered in the eastern regions of Central Province (Figure 3). They were more likely to have spent most of their lifespan in the endemic region (Table 2). Family history of any kidney disease (but not necessarily kidney disease associated with death or dialysis) was more common in patients with tubulointerstitial kidney disease. More than 80% of patients with tubulointerstitial kidney disease were farmers at one point in their lives. Rice was the primary crop, and most reported doing all tasks associated with farming. More than 90% drank well water as a primary water source at one point. Self-reported tobacco use was high in patients with tubulointerstitial kidney disease.
Figure 3.: Patients with tubulointerstitial kidney disease on biopsy (red dots) are primarily from the Eastern part of Central Province in Sri Lanka. Kandy Teaching Hospital serves as a major tertiary referral center for the Central Province of Sri Lanka. Although patients with other kidney diseases on biopsy resided throughout the province, a vast majority of patients with CKD of unknown etiology (CKDu) on biopsy are residing in the eastern portion of the Central Province, the Uva Province, or the Eastern Province.
Table 2. -
Demographic, behavioral, and occupational characteristics of persons with
biopsy-proven tubulointerstitial
kidney disease
| Selected Characteristic |
Tubulointerstitial Kidney Disease, n=43 |
Other Kidney Disease, n=44 |
Age-Adjusted Risk Ratio
a
(95% CI) |
|
Demographics
|
|
|
|
| Education ≥10 yr |
17 (40) |
31 (71) |
0.7 (0.4 to 1.1) |
| Endemic providence
b
|
|
|
|
|
Birth
|
43 (100) |
33 (75) |
1.4 (1.1 to 1.7) |
|
Current residence
|
31 (72) |
31 (73) |
0.8 (0.6 to 1.1) |
|
≥80% of lifespan
|
39 (91) |
29 (66) |
|
| Family history |
|
|
|
|
Any kidney disease
|
16 (37) |
7 (16) |
2.6 (0.4 to 4.7) |
|
Death or dialysis requirement
|
5 (31) |
5 (71) |
— |
|
Occupation
|
|
|
|
| Current occupation |
|
|
|
|
Professional
|
3 (7) |
6 (14) |
— |
|
Service
|
2 (5) |
3 (7) |
— |
|
Factory worker or day laborer
|
1 (2) |
8 (18) |
— |
|
Farmer
|
28 (65) |
8 (18) |
— |
|
Never worked
|
2 (5) |
6 (14) |
— |
|
Not currently working
|
7 (16) |
12 (270) |
— |
| Ever farmer |
36 (84) |
17 (39) |
2.0 (1.2 to 2.9) |
|
Primary crop, rice
|
29 (81) |
12 (71) |
1.2 (0.8 to 1.6) |
|
Primary crop, fruits and vegetables
|
6 (17) |
2 (12) |
— |
|
Primary job, everything on farm
|
33 (92) |
12 (71) |
1.3 (0.9 to 1.7) |
|
Primary job, manual harvesting
|
3 (8) |
4 (24) |
— |
|
Water
|
|
|
|
| Current water source |
|
|
|
|
Public tap
|
11 (26) |
16 (36) |
0.5 (0.2 to 0.9) |
|
Wells
|
21 (49) |
9 (21) |
— |
|
Spring or surface
|
0 (0.0) |
9 (21) |
— |
|
Reverse osmosis, tanker, or bottle
|
11 (26) |
10 (23) |
1.1 (0.2 to 2.0) |
| Ever used well water |
39 (91) |
25 (57) |
1.5 (1.1 to 2.0) |
|
Lifestyle
|
|
|
|
| Tobacco use, ever |
35 (81) |
19 (43) |
1.7 (1.0 to 2.3) |
|
Chewable tobacco use
|
25 (71) |
10 (53) |
1.3 (0.7 to 2.0) |
| Alcohol use, ever |
27 (63) |
20 (46) |
1.3 (0.8 to 1.9) |
|
Local alcohol use
|
23 (85) |
13 (65) |
1.3 (0.8 to 1.8) |
| Ever used traditional medications |
2 (5) |
3 (7) |
— |
|
Clinical symptoms or history
|
|
|
|
| Acute symptoms of fever, back pain, or dysuria over the past 6 mo |
35 (81) |
29 (66) |
1.2 (0.9 to 1.5) |
| Dysuria, ever |
18 (42) |
13 (30) |
1.5 (0.5 to 2.6) |
| Kidney stones |
3 (7) |
3 (7) |
— |
| Snake bites |
7 (16) |
8 (18) |
— |
Numbers are N (%). 95% CI, 95% confidence interval.
aRisk ratios are performed for variables with sufficient sample (denominator and/or numerator with at least ten observations).
bEndemic province defined as Central Province, Northern Province, Uva Province, or Eastern Province.
Symptoms of recent fever, back pain or joint pain, and/or dysuria in the past 6 months were more common in patients with tubulointerstitial kidney disease, but there was a high background rate in patients with other kidney diseases as well (81% versus 66%). Serum urate concentrations were not higher in the patients with tubulointerstitial kidney disease (median of 6.4; interquartile range, 5.4–7.7 versus median of 7.1; interquartile range, 5.8–7.7).
Despite questioning in detail using local drug names, no patients reported using NSAIDs routinely (i.e., more than ten times a month). No patients with tubulointerstitial kidney disease self-reported a diagnosis of an autoimmune disease.
Prediction of Tubulointerstitial Kidney Disease in CKDu Endemic Regions
Forty-two (98%) patients with and 29 (66%) patients without CKDu on biopsy had eGFR<90 ml/min per 1.73 m2; 41 (95%) patients with and 26 (59%) patients without CKDu on biopsy had serum creatinine ≥1.2 mg/dl. Of the 43 patients diagnosed with tubulointerstitial kidney disease on biopsy, nephrologists had a priori diagnosed 25 (58%) as having CKDu (Figure 1).
Table 3 lists selected characteristics of patients with tubulointerstitial versus other diagnoses on biopsy as entered into the prediction model. The full model validation c statistic was 0.82; validation calibration slope was 0.59 (Supplemental Table 1). Of the parsimonious models, model E (including age, urine dipstick for protein, and serum albumin as key predictors, with an interaction term for the laboratory variables) had a validation c statistic similar to the one for the full model (0.84) and a validation calibration slope of 0.84, with sensitivity of 79%, specificity of 84%, positive predictive value of 83%, and negative predictive value of 80% (Supplemental Figure 2 has the relative ranking of strength of predictors, Supplemental Figure 3 shows calibration plots, and Supplemental Appendix 2 has model equations. Patients referred for kidney disease evaluation ages 35–60 years old with urine dipstick negative and normal serum albumin had high likelihood of tubulointerstitial kidney disease (predicted probability ≥0.6). For example, a 50-year-old patient with serum albumin of 4.0 g/dl and urine dipstick trace or negative for protein had a predicted probability of primary tubulointerstitial kidney disease of 0.85 (95% confidence interval, 0.74 to 0.99).
Table 3. -
Selected characteristics of persons with tubulointerstitial
kidney disease (versus other
kidney disease) on
kidney biopsy
| Selected Characteristic |
Tubulointerstitial Kidney Disease, n=43 |
Other Kidney Disease, n=44 |
| Age |
48 (11)
a
|
38 (12) |
| Women |
6 (14)
b
|
13 (30) |
| Hypertension
c
|
14 (33) |
19 (43) |
| Diabetes
c
|
7 (16) |
6 (14) |
|
Serum albumin, g/dl
|
4.2 (0.9)
b
|
3.5 (0.9) |
| Low serum albumin
d
|
6 (14) |
19 (43) |
| Missing |
4 (9) |
— |
|
Serum potassium, mEq/L
|
4.0 (0.7)
b
|
4.4 (0.8) |
| Low serum potassium
e
|
6 (14) |
3 (7) |
| Missing |
3 (7) |
— |
| Hematuria
f
|
5 (12) |
14 (32) |
| Pyuria
f
|
20 (47) |
21 (48) |
| Urine dipstick protein negative
f
|
35 (81)
a
|
13 (30) |
Numbers are mean (SD) or N (%). Diabetes was defined as self-reported or physician-diagnosed diabetes.
aP<0.001.
bP<0.05.
cHypertension was defined as self-reported or physician-diagnosed hypertension, systolic BP ≥140 mm Hg, or diastolic BP ≥90 mm Hg.
dSerum albumin <3.5 g/dl.
eSerum potassium <3.5 mEq/L.
fHematuria was defined as more than three urine red blood cells per high power field. Pyuria was defined as more than three urine white blood cells per high power field. Urine dipstick protein trace or nil was categorized as negative.
Presence or absence of comorbid diabetes mellitus or hypertension was not associated with the likelihood of tubulointerstitial kidney disease. However, there was clear physician selection in referral to biopsy: patients with diabetes or hypertension who were recommended for biopsy were younger and more likely to have low serum albumin, hematuria, or pyuria (Supplemental Table 2) than patients with these conditions not referred for kidney biopsy.
Discussion
In our study to characterize CKDu as seen in Sri Lanka’s endemic areas, we found that active tubulitis and interstitial nephritis and moderate to severe chronic tubulointerstitial damage were common findings. Compared with patients with other identified kidney diseases (e.g., autoimmune or primary glomerular disease), patients with tubulointerstitial kidney disease on biopsy had several characteristic exposures, including residence, farm work, well water use, and tobacco use. Three clinical parameters—age, urine dipstick for protein, and serum albumin—predicted presence of tubulointerstitial kidney disease on biopsy, with positive and negative predictive values exceeding 80%. If nephrologists referred patients with comorbid diabetes mellitus and/or hypertension, they did not experience significantly lower odds of tubulointerstitial kidney disease on biopsy.
Currently available biopsy studies from endemic regions have generally been small in size (n=8–64) (2,11,21) and/or have focused solely on patients with possible CKDu or Mesoamerican nephropathy (2,22). Their goal has been to describe pathology features in detail rather than to validate approaches for a clinical diagnosis. Clinical patient definitions developed by expert committees in Sri Lanka (23) and Mesoamerica (24) were not informed by data on sensitivity and specificity relative to a criterion standard, and they have large differences between them. Compared with clinical descriptions of CKDu, the current patient definitions cast a wider net for some parameters (e.g., including patients with renal tubular acidosis alone or albuminuria alone without reduction in eGFR) while leaving others out (e.g., excluding patients with diabetes or hypertension).
Accurately identifying persons with abnormal kidney function as potentially having CKDu using simple clinical tools can strengthen surveillance, geographic mapping, and entry into patient-control or cohort studies. Notably, knowledge of overall CKD (i.e., data on kidney disease parameters eGFR and albuminuria as defined by Kidney Disease Improving Global Outcomes criteria) is lacking in many of the regions presumed to have an epidemic of CKDu (25). Any population-based surveys of chronic diseases should integrate markers of kidney disease to gauge its overall and likely growing burden. An expert committee report on Mesoamerican nephropathy organized by the Pan American Health Organization echoes this recommendation (26). At the same time, to inform investigations into potential causes of the tubulointerstitial kidney disease afflicting select populations in select regions of Sri Lanka and Mesoamerica, it is also important to attempt to distinguish this specific disease entity from other forms of kidney disease. A standard approach to do so could allow for comparison of data across studies, over time, and across regions (27). Given the resource and research capacity constraints in the afflicted regions, a uniform patient definition could also facilitate pooling of smaller studies and yield better capacity to assess multiple risk factors.
Our work indicates that a clinical, noninvasive approach to identifying patients as having probable CKDu is feasible. The parameters seen commonly in patients with tubulointerstitial kidney disease in our study—that is, young to middle age, urine dipstick negative for proteinuria, and normal serum albumin—align with clinical descriptions of CKDu (2,22,28) and a published protocol aimed at studying CKD in disadvantaged populations (29). A negative result on urine dipstick for protein (rather than more sensitive measures of proteinuria) provided acceptable sensitivity and specificity in Sri Lanka’s endemic region. We also find that, in regions with known high prevalence of CKDu, persons with a diagnosis of diabetes or hypertension may still require consideration for possible CKDu. If a nephrologist recommended a patient with diabetes for biopsy (because for example, he or she was young and did not have accompanying hypertension), this patient was nearly equally as likely to have tubulointerstitial kidney disease as other diagnoses on kidney biopsy. In areas where the background prevalence of diabetic nephropathy and renovascular disease is low but that of CKDu is high, patients with diabetes or hypertension, CKD, and no proteinuria may require additional investigation with either ancillary testing to determine the extent of other end organ damage or kidney biopsy to determine a definitive diagnosis. Additional biopsy studies in endemic regions could better define the prevalence of tubulointerstitial kidney disease in patients with diabetes and/or hypertension.
Evaluation of the demographic, occupational, environmental, and selected lifestyle correlates of patients with tubulointerstitial kidney disease on biopsy shows that they have experienced certain exposures at a high frequency. These include birth and residence in endemic regions, farm work, well water use, and tobacco use. All four of these exposures were nearly ubiquitous (present in >80% of patients with tubulointerstitial kidney disease), and they point us in the direction of environment and lifestyle rather than genetics alone (because although a positive family history was more prevalent in patients with tubulointerstitial kidney disease, it was reported in only about one third of patients). The same correlates, with the exception of well water use, were observed in a recent study evaluating risk factors for postharvest eGFR decline in Guatemala (30). In contrast to patients with Mesoamerican nephropathy who seem to have clearly defined tasks on the sugarcane farm and may experience differential task-dependent risks for kidney disease (31,32), patients with tubulointerstitial kidney disease in Sri Lanka were essentially running the whole farm. Thus, task-based studies are not likely to be as relevant in this area. We also did not find high prevalence of traditional medication or NSAID use, and we did not observe higher serum urate as was described in one study from Mesoamerica (33).
Our study has several limitations. First, our results require additional external validation and apply only to persons living in regions with a high prevalence of CKDu in Sri Lanka. Furthermore, persons presenting to a nephrology clinic for additional evaluation may differ from persons who may not have consented to follow through on the large-scale screening efforts put forth by the Sri Lankan Ministry of Health. Second, we did not dictate kidney biopsy using a study protocol, but rather, we allowed physicians to evaluate for need for biopsy, thus leaving room for inconsistencies in selection for biopsy. Younger patients with fewer comorbidities underwent biopsies. Lack of data on the feasibility, risks, and benefits of kidney biopsy in this region meant that following standard of care rather than prescribing an external protocol adhered to best practices principles of research in low-resources settings (34). Because we observed that a substantial burden of tubulointerstitial kidney disease coexists with a high burden of diabetes and hypertension among those with kidney disease in this region, our work could provide rationale for studies with a priori biopsy protocols. Third, we assume that persons with tubulointerstitial kidney disease on biopsy have CKDu rather than other possible causes; none reported prior diagnosis of an autoimmune disease or regular NSAID use. Until specific molecular signatures of CKDu on biopsy are discovered, we were limited to this assumption. Furthermore, although urine protein by dipstick was among the variables identified as distinguishing CKDu from other causes of kidney disease, quantitative assessment of proteinuria could provide additional discrimination, but it was not routinely available. The high positive and negative predictive values obtained by our model result from the high prevalence of CKDu in the population studied, and they may not translate to other regions with lower prevalence of CKDu.
Within the constraints of a medical system servicing patients with a variety of clinical presentations, most traveling from significant distance, we performed standardized questionnaire administration and data abstraction with pathology review at two institutions in patients with and without tubulointerstitial kidney disease on biopsy. In addition to finding that specific environmental and lifestyle exposures are ubiquitous, we identified common clinical parameters that could serve as a first step toward a standardized noninvasive approach to identifying CKDu, which in turn, is essential for rigorous investigation into its cause.
Disclosures
None.
Acknowledgments
We thank the study participants, medical officers, and nurses contributing to our work.
S.A. is supported by National Institutes of Health grant K-23 5K23DK101826. We thank the Center for Innovation in Global Health, Stanford University for its seed grant. Redcap use is supported by National Institutes of Health grant UL1 TR001085.
Supplemental Material
This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.07430618/-/DCSupplemental.
Supplemental Table 1. Discrimination and calibration of all preselected characteristics: (full) model and parsimonious models
Supplemental Table 2. Comparisons of patients with diabetes and/or hypertension by physician recommendation for biopsy
Supplemental Figure 1. Distribution of diagnoses in patients without CKDu on biopsy.+
Supplemental Figure 2. Ranking of the relative strength of the nine preselected characteristics used in the full model.
Supplemental Figure 3. Calibration plots of the full model and five parsimonious models.
Supplemental Appendix 1. Definition of CKDu histologic scoring.
Supplemental Appendix 2. Methods for prediction modeling.
References
1. Athuraliya NT, Abeysekera TD, Amerasinghe PH, Kumarasiri R, Bandara P, Karunaratne U, Milton AH, Jones AL: Uncertain etiologies of proteinuric-chronic
kidney disease in rural
Sri Lanka.
Kidney Int 80: 1212–1221, 201121832982
2. Nanayakkara S, Komiya T, Ratnatunga N, Senevirathna ST, Harada KH, Hitomi T, Gobe G, Muso E, Abeysekera T, Koizumi A: Tubulointerstitial damage as the major pathological lesion in endemic chronic
kidney disease among
farmers in North Central Province of
Sri Lanka. Environ Health Prev Med 17: 213–221, 201221993948
3. Ismail A: The insurmountable socio-economic impact of CKDU.
Daily Mirror, 2017. Available at:
http://www.dailymirror.lk/article/The-Insurmountable-Socio-Economic-Impact-of-CKDU-129427.html. Accessed September 6, 2018
4. Ellepola Y: World
Kidney Day 2016: What More Can
Sri Lanka Do to Save Lives?
Sri Lanka, Institute of Policy Studies of
Sri Lanka, 2016. Available at:
http://www.ips.lk/talkingeconomics/2016/03/10/world-kidney-day-2016-what-more-can-sri-lanka-do-to-save-lives/. Accessed September 6, 2018
5. Nanayakkara S, Senevirathna ST, Abeysekera T, Chandrajith R, Ratnatunga N, Gunarathne ED, Yan J, Hitomi T, Muso E, Komiya T, Harada KH, Liu W, Kobayashi H, Okuda H, Sawatari H, Matsuda F, Yamada R, Watanabe T, Miyataka H, Himeno S, Koizumi A: An integrative study of the genetic, social and environmental determinants of chronic
kidney disease characterized by tubulointerstitial damages in the North Central Region of
Sri Lanka. J Occup Health 56: 28–38, 201424351856
6. Peraza S, Wesseling C, Aragon A, Leiva R, García-Trabanino RA, Torres C, Jakobsson K, Elinder CG, Hogstedt C: Decreased
kidney function among agricultural workers in El Salvador. Am J
Kidney Dis 59: 531–540, 201222300650
7. Torres C, Aragón A, González M, López I, Jakobsson K, Elinder CG, Lundberg I, Wesseling C: Decreased
kidney function of unknown cause in Nicaragua: A community-based survey. Am J
Kidney Dis 55: 485–496, 201020116154
8. Trabanino RG, Aguilar R, Silva CR, Mercado MO, Merino RL: [End-stage renal disease among patients in a referral hospital in El Salvador]. Rev Panam Salud Publica 12: 202–206, 200212396639
9. Ganguli A: Uddanam nephropathy/regional nephropathy in India: Preliminary findings and a plea for further research. Am J
Kidney Dis 68: 344–348, 201627209444
10. Jayatilake N, Mendis S, Maheepala P, Mehta FR; CKDu National Research Project Team: Chronic
kidney disease of uncertain aetiology: Prevalence and causative factors in a developing country. BMC Nephrol 14: 180, 201323981540
11. Wijkström J, Leiva R, Elinder CG, Leiva S, Trujillo Z, Trujillo L, Söderberg M, Hultenby K, Wernerson A: Clinical and pathological characterization of Mesoamerican nephropathy: A new
kidney disease in Central America. Am J
Kidney Dis 62: 908–918, 201323850447
12. Athuraliya TN, Abeysekera DT, Amerasinghe PH, Kumarasiri PV, Dissanayake V: Prevalence of chronic
kidney disease in two tertiary care hospitals: High proportion of cases with uncertain aetiology. Ceylon Med J 54: 23–25, 200919391454
13. Redmon JH, Elledge MF, Womack DS, Wickremashinghe R, Wanigasuriya KP, Peiris-John RJ, Lunyera J, Smith K, Raymer JH, Levine KE: Additional perspectives on chronic
kidney disease of unknown aetiology (CKDu) in
Sri Lanka--lessons learned from the WHO CKDu population prevalence study. BMC Nephrol 15: 125, 201425069485
14. Jayasumana C, Paranagama P, Agampodi S, Wijewardane C, Gunatilake S, Siribaddana S: Drinking well water and occupational exposure to Herbicides is associated with chronic
kidney disease, in Padavi-Sripura,
Sri Lanka. Environ Health 14: 6, 201525596925
15. Wanigasuriya KP, Peiris-John RJ, Wickremasinghe R, Hittarage A: Chronic renal failure in North Central Province of
Sri Lanka: An environmentally induced disease. Trans R Soc Trop Med Hyg 101: 1013–1017, 200717643458
16. Chandrajith R, Nanayakkara S, Itai K, Aturaliya TN, Dissanayake CB, Abeysekera T, Harada K, Watanabe T, Koizumi A: Chronic
kidney diseases of uncertain etiology (CKDue) in
Sri Lanka: Geographic distribution and environmental implications. Environ Geochem Health 33: 267–278, 201120853020
17. Harrell F: Regression Modeling Strategies: With Application to Linear Models, Logistic and Ordinal Regression, and Survival Analysis, 2nd Ed., New York, Springer, 2001
18. Zou G: A modified poisson regression approach to
prospective studies with binary data. Am J Epidemiol 159: 702–706, 200415033648
19. R Foundation for Statistical Computing: A Language and Environment for Statistical Computing, Vienna, Austria, 2017
20. Harrell FE Jr.: rms: Regression Modeling Strategies. R package version 5.1-2, Nashville, Tennessee, 2018
21. López-Marín L, Chávez Y, García XA, Flores WM, García YM, Herrera R, Almaguer M, Orantes CM, Calero D, Bayarre HD, Amaya JC, Magaña S, Espinoza PA, Serpas L: Histopathology of chronic
kidney disease of unknown etiology in Salvadoran agricultural communities. MEDICC Rev 16: 49–54, 201424878649
22. Badurdeen Z, Nanayakkara N, Ratnatunga NV, Wazil AW, Abeysekera TD, Rajakrishna PN, Thinnarachchi JP, Kumarasiri R, Welagedera DD, Rajapaksha N, Alwis AP: Chronic
kidney disease of uncertain etiology in
Sri Lanka is a possible sequel of interstitial nephritis! Clin Nephrol 86: 106–109, 2016
23. Rajapakse S, Shivanthan MC, Selvarajah M: Chronic
kidney disease of unknown etiology in
Sri Lanka. Int J Occup Environ Health 22: 259–264, 201627399161
24. Ferreiro A, Álvarez-Estévez G, Cerdas-Calderón M, Cruz-Trujillo Z, Mena E, Reyes M, Sandoval-Diaz M, Sánchez-Polo V, Valdés R, Ordúnez P: Confirmed clinical case of chronic
kidney disease of nontraditional causes in agricultural communities in Central America: A case definition for surveillance. Rev Panam Salud Publica 40: 301–308, 201628076578
25. Levey AS, de Jong PE, Coresh J, El Nahas M, Astor BC, Matsushita K, Gansevoort RT, Kasiske BL, Eckardt KU: The definition, classification, and prognosis of chronic
kidney disease: A KDIGO Controversies Conference report.
Kidney Int 80: 17–28, 2011
26. Pan American Health Organization: Epidemic of Chronic
Kidney Disease in Agricultural Communities in Central America: Case Definitions, Methodological Basis and Approaches for Public Health Surveillance, Washington, DC, PAHO, 2018
27. Krause G, Brodhun B, Altmann D, Claus H, Benzler J: Reliability of case definitions for public health surveillance assessed by Round-Robin test methodology. BMC Public Health 6: 129, 200616686946
28. Wijkström J, Jayasumana C, Dassanayake R, Priyawardane N, Godakanda N, Siribaddana S, Ring A, Hultenby K, Söderberg M, Elinder CG, Wernerson A: Morphological and clinical findings in Sri Lankan patients with chronic
kidney disease of unknown cause (CKDu): Similarities and differences with Mesoamerican Nephropathy. PLoS One 13: e0193056, 201829513702
29. Caplin B, Jakobsson K, Glaser J, Nitsch D, Jha V, Singh A, Correa-Rotter R, Pearce N: International collaboration for the epidemiology of eGFR in low and middle income populations - rationale and core protocol for the disadvantaged populations eGFR epidemiology study (DEGREE). BMC Nephrol 18: 1, 201728049448
30. Butler-Dawson J, Krisher L, Asensio C, Cruz A, Tenney L, Weitzenkamp D, Dally M, Asturias EJ, Newman LS:
Risk factors for declines in
kidney function in sugarcane workers in Guatemala. J Occup Environ Med 60: 548–558, 201829370016
31. García-Trabanino R, Jarquín E, Wesseling C, Johnson RJ, González-Quiroz M, Weiss I, Glaser J, José Vindell J, Stockfelt L, Roncal C, Harra T, Barregard L: Heat stress, dehydration, and
kidney function in sugarcane cutters in El Salvador–A cross-shift study of workers at risk of Mesoamerican nephropathy. Environ Res 142: 746–755, 201526209462
32. Laws RL, Brooks DR, Amador JJ, Weiner DE, Kaufman JS, Ramírez-Rubio O, Riefkohl A, Scammell MK, López-Pilarte D, Sánchez JM, Parikh CR, McClean MD: Changes in
kidney function among Nicaraguan sugarcane workers. Int J Occup Environ Health 21: 241–250, 2015
33. Kupferman J, Amador JJ, Lynch KE, Laws RL, López-Pilarte D, Ramírez-Rubio O, Kaufman JS, Lau JL, Weiner DE, Robles NV, Verma KP, Scammell MK, McClean MD, Brooks DR, Friedman DJ: Characterization of Mesoamerican nephropathy in a
kidney failure hotspot in Nicaragua. Am J
Kidney Dis 68: 716–725, 201627575010
34. Stanford University Center for Innovation in Global Health and the Johns Hopkins University Berman Institute of Bioethics:Bioethics SUCfIiGHJHUBIo: Ethical Challenges in Short-Term Global Health Training. Available at: http://ethicsandglobalhealth.org/. Accessed September 6, 2018
