Chronic kidney disease of undetermined cause (CKDu) describes impaired kidney function not explained by traditional risk factors such as hypertension, diabetes or heavy proteinuria. In a 2017 population-representative survey from Sri Lanka's North Central Province, low estimated glomerular filtration rate (eGFR) was observed in 11.2% of men and 3.7% of women in the absence of these conventional risks. Those cross-sectional findings prompted a prospective study to identify early kidney-function loss, track trajectories over time, and examine exposures associated with deterioration in a working-age population.
This investigation was a 6-year longitudinal cohort conducted in Puhudivula, Anuradhapura district, Sri Lanka, among adults aged 20–60 years. Individuals with evidence of diabetes, hypertension or pre-existing chronic kidney disease were excluded to concentrate on non-traditional causes of reduced kidney function. The analytic sample comprised 425 participants followed repeatedly to measure kidney function via eGFR and to collect exposure information.
To characterise underlying kidney-health states and transitions between them over time, the authors fitted hidden Markov models (HMMs) to serial eGFR data. HMMs estimate latent states that are not directly observed (here, healthy vs unhealthy kidney states) and the probabilities of moving between states at successive measurement points. The models also enabled assessment of associations between measured exposures and both state membership and transition events, including the use of lagged (prior-period) exposure variables when evaluating risk of transition to an unhealthy state.
The HMM analysis identified four distinct kidney-health trajectories within the cohort (n = 425):
Overall, almost one quarter of working-age adults in this population exhibited eGFR changes consistent with poor kidney health at some point during the study period.
Specific exposures were associated with either being in an unhealthy state or with transition to an unhealthy state. Use of smokeless tobacco, including betel quid, was associated with belonging to an unhealthy category (odds ratio 2.29; 95% CI 1.17–4.49). When examining lagged exposures and the risk of transition from a healthy to an unhealthy state, the study reported associations for prior smoking (OR 2.26; 95% CI 1.25–4.10), prior smokeless tobacco use (OR 1.98; 95% CI 1.13–3.48) and prior exposure to weedkiller (OR 1.72; 95% CI 1.15–2.59). These findings link both tobacco products and a herbicide exposure with subsequent deterioration in kidney-health state as estimated by the HMMs.
In this working-age Sri Lankan cohort, serial eGFR measurements modelled with hidden Markov methods revealed that a substantial minority experienced early loss of kidney function or persistent poor kidney health over six years, in the absence of diabetes or hypertension. The observed associations implicate smokeless tobacco, smoking and weedkiller exposure as potential contributors to transitions toward an unhealthy kidney state in this population. These results support targeted investigation of non-traditional environmental and behavioural risk factors in regions affected by CKDu and underscore the value of longitudinal monitoring with methods that can detect latent state transitions.
Given that this report is a preprint, the authors caution that findings have not been peer reviewed and should not yet be used to change clinical practice. The study approach, however, demonstrates a method for early detection of decline in kidney function at the population level and for identifying candidate exposures for further causal and interventional research.
Ethical approval for the study was granted (EC-25-170) by the Ethics Review Committee, Faculty of Medicine, University of Colombo, Sri Lanka. The authors confirm participant consent and adherence to relevant reporting guidelines. Information on data availability can be obtained by contacting Dr Thilanga Ruwanpathirana, Ministry of Health, Sri Lanka. The authors declared no competing interests in the preprint.
Declared funders included the National Science Foundation of Sri Lanka (RPHS/2016/CKDu 07), the Ministry of Health, Nutrition and Indigenous Medicine and the WHO Country Office Sri Lanka, the Medical Research Council (MR/P02386X/1) and the Colt Foundation. The preprint is made available under a CC-BY 4.0 International license. As noted, the manuscript is a preprint and has not undergone peer review.