Skip to main content

Table 3 Bradford Hill Elements of Data Contributing to a Causal Association and EDCs

From: A path forward in the debate over health impacts of endocrine disrupting chemicals

Hill Elements1

Application to EDCs

Strength of Association

 

The examples used were testis cancer in chimney sweeps and lung cancer in smokers. In both examples, the strength of the associations were made by comparing death rates in a control group (men who were not chimney sweeps and non-smokers, respectively).

There are no groups of people unexposed to EDCs. Moreover, no one is exposed to a single chemical. Finally, endocrine diseases and disorders are clearly multicausal. Thus, the concept of strength of the association must be adjusted as it is applied to EDCs.

Consistency

 

The concept is that multiple studies should observe the same relationships between exposure and outcome.

In principle, there should also be consistent observations between relationships of interest. However, there are at times modifying factors that can change this. For example, perchlorate exposure is inversely related to serum thyroid hormone in populations with low iodine intake or in those who smoke cigarettes. However, this is not the case in populations with high iodine intake and/or who do not smoke.

Specificity

 

The example was that of nickel refiners of South Wales with a high incidence of cancer of the lung or nose. The specificity of this relationship could be used as evidence of causation. However, Hill cautioned about making too much of the specificity of the relationship and concluded that, “In short, if specificity exists we may be able to draw conclusions without hesitation; if it is not apparent, we are not thereby necessarily left sitting irresolutely on the fence.”

The specificity of relationships of interest with EDCs must be evaluated carefully because hormone systems are involved in a great many processes and this is life-stage specific. For example, androgens play an important role in development of the male reproductive system in the fetus, but in the adult, androgens are related to different processes in men and women. Likewise, transient hypothyroidism during fetal development can lead to lower IQ and attention deficit, but transient hypothyroidism in the adult can lead to weight gain that is reversible.

Temporality

 

Hill’s concept was to be cautious about the temporal relationship of associations with particular attention to the question of which element of the dyad came first? For example, do particular dietary habits lead to disease, or does the disease predispose those affected to prefer a specific diet?

The temporal relationship between exposure to an EDC and a specific endocrine-mediated adverse outcome may be quite complex. The classic example is that of DES exposure during fetal life and the production of reproductive tract cancer 20 years later (long after DES was gone). This relationship was observed because women were prescribed DES and there were specific records of exposure. This will not likely be the case for non-accidental exposures to EDCs. Thus, “temporality” may be important, but it may not be a concurrent relationship.

Biological Gradient

 

Hill noted that the linear increase in the death rate from lung cancer with number of cigarettes smoked daily added greatly to the simple evidence that the cancer rate was higher in smokers than non-smokers. But he didn’t discount a relationship in which the death rate is higher in people who smoke fewer cigarettes per day.

The shape of the dose–response is important for EDCs, but there may be more variability depending on the mechanism of disruption. For example, perchlorate should produce a typical S-shaped dose–response curve on thyroid hormone concentrations in the human population because it is a competitive inhibitor of iodine uptake into the thyroid gland. In contrast, BPA is likely to produce more of a “square wave” dose–response curve because it is an indirect antagonist on the thyroid hormone receptor.

Plausibility

 

Hill insisted that “it will be helpful” if the causation we suspect is biologically plausible. However, we cannot demand this. In short, the association we observe may be one new to science or medicine and we must not dismiss it too light-heartedly as just too odd.

Likewise for EDCs, biological plausibility will likely strengthen our confidence in the causal nature of relationships of interest. Moreover, our knowledge of hormone actions will likely drive us to evaluate specific relationships. However, there is a great deal we have to learn about the endocrine system, and requiring complete knowledge of the endocrine mechanism mediating a relationship of interest is unrealistic.

Coherence

 

Hill reasoned that the interpretation of a causal relationship between exposure and outcome should not conflict with generally known facts of the natural history and biology of the disease.

Coherence is also important for EDCs. Thus, the interpretation of causation should not conflict with generally known facts of the biology of the endocrine system under study.

Experiment

 

Hill reasoned that occasionally, confidence in a conclusion of causality could be strengthened by changing elements of the environments and observing a predicted change. For example, dust in the workplace could be reduced, oil changed, work conditions altered. He did not include animal or biochemical experiments.

For EDCs, animal and biochemical experimental evidence must be integrated with (or without) epidemiological data to consider that a chemical may produce an adverse outcome through an endocrine mechanism. This is a novel component of assessing the evidence and the logic guiding this has not been formally validated. Because of the complexity of hormone action, such experiments need to be properly designed with positive and negative controls, and must be properly interpreted based on principles of endocrinology.

Analogy

 

Hill reasoned that known causal relationships can reasonably be extended to other relationships that have similar characteristics. His example was that with the effects of thalidomide and rubella being known, it would be more likely to be reasonable to accept slighter but similar evidence with another drug or viral disease in pregnancy.

Likewise, it is reasonable in the EDC field to extend this to include analogous endpoints. For example, if we observe a relationship between phthalate exposure and anogenital distance in newborn boys, we can reasonably extend this relationship to other androgen-dependent endpoints. Moreover, if we know that a chemical has antiandrogenic properties in vitro, it is reasonable to tailor the endpoints that are evaluated in vivo to androgen-sensitive endpoints. Likewise, if we observe a relationship between PCB exposure and the expression of thyroid hormone-responsive genes in the placenta, we can reasonably extend this to thyroid hormone action in tissues we cannot obtain, such as the fetal/neonatal brain. And if we know that PCBs have anti-thyroid properties, we should evaluate thyroid-sensitive endpoints.

  1. 1These elements are taken from: Hill AB. The Environment and Disease: Association or Causation? Proc R Soc Med. 1965 May;58:295–300 [51].