In their short report, Ohgami et al 1 reported lithium levels in drinking water and linked them to the risk of suicide. Despite the report highlighting the pitfalls of drawing simple conclusions from large-scale ecological studies, a Google search shows that these findings have been widely disseminated in scientific and lay media.
A major concern, addressed only obliquely by the authors, is the likelihood of confounding in this scenario. As noted by Chandra & Babu, 2 sociological factors play an important role in suicide.
The lack of accounting for such potential confounders for the different districts in the study is a serious methodological omission, rendering the results of the study untenable from an epidemiological perspective. The demographics of the different areas (beyond age structure) are not addressed, thus ignoring important economic and social factors (like deprivation and unemployment) which contribute to suicide risk.
Adjusting for differences in age structures between centres using standardised mortality ratios (SMRs) is unlikely to account for all important sources of confounding, so that the possibility of residual confounding must be considered a major qualifier when considering these results, rather than details to be addressed in future studies. 3
The potential reasons behind the difference in lithium levels in the drinking water samples in the different municipalities are also not explained. Lithium levels in water sampled across a number of districts in New Zealand differ within municipal areas, depending where the sample is sourced. In this context, how valid is it then to use the mean value to represent the lithium exposure in that area? This would require the matching of lithium levels with suicide data from each discrete area of water supply and a loss of statistical power for such a relatively uncommon event as suicide.
The duration of exposure to a specific level of lithium in the drinking water was also not addressed. Apart from the issue of dietary intake of lithium noted in the letter by Desai & Chaturvedi, 4 there is the question of where people source most of their drinking water, and the use of bottled water.
In the context of the short report, it is also difficult to fully assess the suitability of the analysis methods used. It would have been useful to have more detail on the weighting structure used in the regression, alongside frequency data on the number of events observed in each locality. Also, the reported beta coefficient from the regression is not interpretable in the context of the presented figure or reported analysis methods.
Although the reported results were indeed intriguing, in the absence of more a developed approach to the research question it seems too early, and indeed misleading for a non-scientist audience, to even start speculating on the relationship between suicide rates and lithium in drinking water sources on the basis of these data. In this era of rapid information dissemination, the publishing of reports without rigorous scrutiny of the statistical method and due consideration of the confounding variables is a concern.
Edited by Kiriakos Xenitidis and Colin Campbell