The cohort study examined the effects of arsenic exposure due to consumption of contaminated tube well water on infant death and fetal loss. The exposure was determined as the arsenic concentration in tube well water and the outcomes were reported as early fetal loss, late fetal loss, neonatal death, postnatal death and infant death.
The association between outcome and exposure levels were measured by using statistical analysis tools. More specifically a Cox Proportional Hazards Model was used. A number of covariates (such as seasons, socioeconomic status, etc.) were collected and their association with the outcome and exposure was determined using Spearman's correlation coefficient. The factors causing significant change in the association of exposure and outcomes were identified and included in the final model.
With the large data set that was dealt with in this study, statistical model and methods are the most appropriate for determining association between outcome and exposure level. The authors have been very thorough in the study design and have collected a very large data set and tried to minimize the bias (and other sources of error, such as multiple pregnancy, etc.). Inclusion of covariates with significant effect on the association also confirms the robust study design and data evaluation.
Hope this helps.