Table 6.

The Nine Bradford-Hill (Hill, 1965) Criteria and the Evidence that Supports or Meets Each Criterion.

Criterion	Supporting Evidence
1) Strength (e.g., effect size)	The effect size (d) of the score differences between exposed and controls in the T. M. Farahat et al. (2003) study was 0.55 for Trail Making A and 0.61 for B; both are moderate effect sizes per Cohen (1992). In the present study of a workforce from the same population, applicators completed the Trail Making A task an average of 19.7 (95% CI: 3.8 – 35.6; p = 0.015; d ~ 0.38) s slower than engineers and 12.7 (95% CI: 1.1 – 24.4; p = 0.032; d ~ 0.24) s slower than technicians. For Trail Making B, applicators were 30 (95% CI: 6.5 – 53; p = 0.012; d ~ 0.46) s slower than engineers and 3.2 s slower than technicians (p = 0.747; d ~ 0.05); technicians were 26 (95% CI: 9 – 44; p = 0.003; d ~ 0.41) s slower than engineers at each of the testing sessions. The effect sizes are model-based rather than derived from the mean and standard deviation as developed by (Cohen, 1992).
2) Consistency (e.g., reproducibility from multiple studies)	Demonstrated by the Munoz-Quezada et al. (2016) review identifying 8 studies that documented poorer exposed group performance on the Trail Making Test compared to control groups. The present study adds a ninth that replicates and extends one prior finding in the same population of pesticide applicators (T. M. Farahat et al., 2003).
3) Specificity (e.g., there is a specific population at a specific site with no other likely explanation)	Demonstrated by two studies (the present study and T. M. Farahat et al., 2003) with the same result in different samples from the same population primarily exposed occupationally to only one neurotoxic pesticide, CPF. There is limited evidence that exposure to pyrethroids, to which this population is also exposed, reduces performance on some cognitive tests (Hansen et al., 2017) though not those with a motor component as seen in the present studies.
4) Temporality (e.g., does the exposure precede the effect)	Demonstrated for chronic exposures by the differences between the controls and exposed and for recent exposures by Rohlman et al. (2016) in adolescent pesticide applicators in this same population that had better performance on the Trail Making test before and after applications by these young applicators (e.g., as a well-paying summer job).
5) Biological gradient (e.g., greater exposure should generally lead to greater incidence of the effect, or dose response)	Dose-related effect based job title demonstrated for the first time in multiple levels by the present study, as applicators, who had the greatest CPF exposure, had the poorest performance, while engineers/technicians who had CPF exposures lower than applicators but higher than controls, had intermediate performance and controls who had the lowest CPF exposures had the best performance. This is supported by the Meyer-Baron et al. (2015) meta-analysis of 22 studies relating duration of OP pesticide exposure in years (history/range of OP pesticides not documented) to performance deficits on a range of behavioral tests (including Trail Making).
6) Plausibility (e.g., a plausible mechanism of cause and effect)	While ACHE inhibition is known to mediate the acute neurotoxic effects of OPs, considerable evidence in the human and animal literature point to non-cholinergic mechanisms of neurotoxicity following chronic exposures (Voorhees et al., 2016). How chronic OP exposures cause behavioral deficits remains an area of active research, but there is evidence implicating neuroinflammation and oxidative stress: chronic OP exposures has been shown to trigger neuroinflammation and oxidative stress (Guignet and Lein, 2019; Naughton and Terry, 2018) and both of these neuropathological responses have been linked to behavioral deficits in humans and animals. Thus there is a plausible mechanism to link chronic OP exposures to behavioral deficits.
7) Coherence (viz., does not conflict with known facts of biological effects of OP chemicals)	As indicated under criterion 6 above, there is significant evidence in the experimental literature that under conditions of well-controlled exposures, OPs, including CPF, cause behavioral deficits via non-cholinergic mechanisms (Voorhees et al., 2016). Thus, the human data are coherent with the known facts of biological effects of OP chemicals.
8) Experiment (viz., agrees with experimental studies)	Studies in a rat model with exposures based on those reported for Egyptian pesticide workers in this study demonstrated behavioral deficits in a cued learning test in two contexts or settings associated with CPF exposures (as related to attention also measured by the Trail Making Test) that appeared only after repeated exposures (viz., 15 exposures but not after 5 or 10 exposures). The series of studies also presents evidence that AChE is a poor biomarker of the effects of repeated exposures (Lattal et al., 2010). Experimental studies suggest that CPF-induced behavioral deficits appear to be causally related to oxidative stress (Lein, unpublished data)
9) Analogy (e.g., similar evidence from related compounds or conditions)	The Munoz-Quezada et al. (2016) systematic review of studies of chronic OP exposure in adults revealed that 24 of 35 studies found adverse neurobehavioral effects, including 8 that used the Trail Making Test. There is independent evidence in children who apply the same pesticide in Egypt of adverse behavioral effects of CPF (Rohlman et al., 2016).