Groundwater arsenic is a major public health concern globally  including India [1]. However, estimates of both the population in India exposed to high groundwater arsenic and the detrimental health outcomes arising from that exposure are highly varied [2,3,4]. Recently, [4] went some way to addressing this by quantitatively estimating the fraction of cardiovascular disease mortality in India attributable to groundwater arsenic. However, equivalent estimates for groundwater-arsenic attributable cancer mortality in India are lacking.

We combined our AI/machine learning modelling of groundwater arsenic distribution in India [2], with state-level drinking water source data to estimate the number of people in each state exposed to drinking water arsenic at a series of four different classes of arsenic concentration (viz. 0-10 , 10-20, 20-50, 50-80, > 80 µg/L) as reported by [4] and then combined those data with published dose-response data for arsenic-attributable cancers to estimate the population attributable fractions (PAFs) of those cancers due to groundwater arsenic. 

Plausible model PAFs so calculated for the whole of India are largely in the range 1 % to 5 % for selected cancers, albeit that dose-response relationships at low arsenic concentrations remain a key uncertainty to the modelling. Whilst still representing a substantial public health burden, these PAFs are much lower than those arising from other key factors (e.g. smoking) and confirms that the reported attribution of increased crude cancer mortality rates in India to groundwater arsenic exposure is largely unfounded (and better ascribed to (i) an aging population, (ii) massive improvements in healthcare decreasing mortality rates from other causes, and (iii) improved detection and reporting).

Acknowledgements: This research was funded in part by the Newton Fund, NERC(UK) (NE/R003386/1), and DST (India) (DST/TM/INDO-UK/2K17/55(C) & 55(G)) to DP and others (https://www/farganga.org  ) . With thanks, I acknowledge the use of unpublished data generated by Ruohan Wu for our co-authored paper [4]). I thank Michael Berg, DRS Middleton, Abhijit Mukherjee, Joel Podgorski,  Dipankar Saha and Joachim Schuz for discussions.

References:   [1] Mukherjee et al (2024) https://doi.org/10.1038/s43017-024-00519-z ; [2] Podgorski et al (2020) https://doi.org/10.1007/s10653-020-00655-7 ; [3] Mukherjee et al (2021) https://doi.org/10.1016/j.scitotenv.2020.143511 ; [4] Wu et al (2021) https://doi.org/10.3390/w13162232