Response to "Comment on 'Environmental Cadmium and Mortality from Influenza and Pneumonia in U.S. Adults'".

We thank Dr. Kawada for addressing our paper. Kawada raised three questions (Kawada 2021) given our findings of a statistically significant association with urinary cadmium (Cd) but no statistically significant association with blood Cd. We can best address these questions by providing additional explanation of our results based on significance testing.

In our study (Park et al. 2020), we observed that the adjusted hazard ratio (HR) comparing individuals in the 80th vs. 20th percentiles of urinary Cd (creatinine-corrected) was 1.15 {[95% confidence interval (CI): 1.05, 1.26]; }, and the corresponding HR for blood Cd was 1.14 [(95% CI: 0.96, 1.36); )]. These findings suggest that Cd body burden as assessed by either urine or blood concentrations had a similar magnitude of association but that the urine result was more precise because of the larger number of cases (141 influenza/pneumonia deaths of 7,173 participants for the urine analysis vs. 56 of 8,678 for the blood analysis). Given the smaller number of cases, the blood Cd analysis was relatively underpowered, resulting in a failure to reject the null hypothesis. However, this finding does not necessarily indicate no effect of Cd measured in blood. The results of effect modification by age should be interpreted with even more caution because cases were divided into three age groups. Therefore, the statistically significant associations in participants aged 65–84 y but not in other age groups should not be interpreted as if older individuals are necessarily more susceptible to Cd toxicity or age-dependent confounding was inadequately controlled for in that age group. Instead, those findings imply that our study was not sufficiently powered to detect age-dependent associations in a robust way.

Lack of statistical significance (i.e., ) is often incorrectly interpreted as lack of effect (Greenland et al. 2016; Rothman 2014; Stang et al. 2010). On the other hand, a weak effect is often misinterpreted as important if it is statistically significant. Although a weak effect may be of limited importance for individuals seen in clinical practice, it can have a large population impact if exposure is very wide. Low-level widespread population exposure to environmental toxicants, such as Cd, is such a case.

We agree with Kawada in several points he raised. First, given the observed low biomonitoring concentrations of Cd in the general U.S. population, a potential underlying biological mechanism is less likely related to Cd’s nephrotoxicity. Second, urinary Cd is probably superior to blood Cd as a biomarker of long-term Cd exposure. However, blood Cd still has value as a biomarker of Cd body burden, particularly in populations with very low environmental exposure (Järup et al. 1998; Järup and Åkesson 2009). In our view, it is preferable to use both biomarkers in epidemiologic studies of Cd exposure and pulmonary outcomes.