Issue 33 – Grades of Evidence

The significance of different standards of evidence in determining the safety of pharmaceuticals against the safety of chemicals

What are the differences between how pharmaceuticals are tested for health as compared to chemicals? Does this limit the evidence which we can draw on to decide if a chemical is safe and set policies to protect health by restricting its use and marketing?

Clinical trials for pharmaceuticals are, depending on how they are classified, divided into five phases. The first phase is pre-clinical and involves in vitro and animal testing to determine whether there is any merit to developing a compound into a drug. Here initial information about efficacy, activity in a biological system and toxicity of a candidate pharmaceutical are gathered.

The next two phases, I and II, of drug development involve humans. Phase I is the first stage of clinical trials. Small groups of usually healthy volunteers are assembled to test for side-effects in humans, determine the maximum tolerable dose, and understand how a drug is processed in the body.

Phase II trials involve larger groups of up to 300 people and are designed to assess how well a drug works, as well as continue the Phase I safety and dosing assessments in a larger group of volunteers and patients. It is often at this stage where the drug development process fails, because the test compound either does not work as planned or has toxic side-effects. [Continues]

Don’t miss! H&E’s November update on developments around controversial food contact material BPA; or a new presentation about how the environment may be making people obese.

Phase III studies are randomised, controlled multicentre trials (RCTs) on large groups of patients. They are intended as the definitive assessment of how effective the drug is, normally in comparison with the current best treatment. Phase III trials are difficult and expensive to conduct. Normally a drug has to succeed in two Phase III trials for it to be approved in the US or EU.

Phase IV trials, also known as post-marketing surveillance, take place after a drug has been approved for marketing. They involve safety surveillance (pharmacovigilance) and reveal information not necessarily tested for in trials, such as interactions with other drugs and effects on patient subgroups such as pregnant women or patients who manifest diseases besides the one which the drug is licensed to treat.

In comparison to pharmaceuticals, allowing chemicals on the market resembles more of a direct jump from preclinical testing to occasional post-marketing surveillance: deliberate toxicity testing is carried out on animals, while follow-up research after marketing is sporadic, continuing with animals and epidemiological studies of its potential harm to humans.

The reason there are no equivalents to Phase I, II or III trials in chemical safety assessment is that, except in rare circumstances, it is not possible to conduct clinical trials of harm to health from chemicals. Pharmaceutical testing is ethically permissible because potential harm is off-set by the supposed benefits of a drug while the risk of harmful side effects is theoretically minimised by the way the trials are conducted.

Because with chemicals we only have access data equivalent to preclinical and Phase IV trials, greater weight must therefore be placed on collection, analysis and use of either the preclinical data or data from epidemiological research.

Limitations of epidemiology

One of the central challenges to epidemiology, and which RCTs are designed to eliminate, is the issue of control: because humans are exposed to a myriad of chemicals in a multitude of ways over a range of times, it becomes very difficult to isolate a single chemical exposure, control for other variables, and connect that exposure with a health effect.

Isolating exposures amidst the bustle of daily life and associating them with health effects is the art of epidemiology. It works best for obvious exposures with striking health outcomes, perhaps most notably for smoking and death by lung cancer and heart disease, as exemplified in Doll and Peto’s 1976 study of smoking habits and mortality among doctors (Doll & Peto, 1976).

However, for ubiquitous exposures where an unexposed population is difficult to find, where exposure includes not only the target chemical but a range of other chemicals, where the health end-points may be varied, subtle, and more than one chemical to which the population is exposed may have similar effects, it is very difficult through epidemiology alone to obtain associations strong enough to count as causal. (See Austin Bradford Hill’s criteria of causation in epidemiology for interest. Hill 1965.)

Huge association studies are possible with, for example, the US National Health and Nutrition Examination Survey (NHANES). The data set has been used to find associations between the plastic and epoxy resin hardener bisphenol-A (BPA) and heart disease (Melzer et al. 2010a), non-stick chemicals PFOAs and thyroid disease (Melzer et al. 2010b), and most recently BPA, triclosan and immune function (Clayton et al. 2010).

However, association does not prove causation, so epidemiologists cannot say for sure that a particular chemical caused particular effects. The most realistic way to achieve that is through what researchers refer to as “control”: isolating two groups of organisms, exposing one to a chemical and the other not, making sure no other factors in their biology or environment differ, and then watching for health effects as they develop. Humans cannot be subjected to such treatment, which is why experimentation is done on animals.

Limitations of animal research

Primates are the closest model for humans. However, there are still substantial biological differences between primates and humans, they are expensive, and their longevity also presents problems for identifying health effects within feasible lengths of time for a study. The higher intelligence and emotive faculties of primates also discourages their use as research subjects.

Rodents are much more popular as research subjects: the health effects of chemical exposure can be observed relatively quickly and individual differences between rodents can also be controlled relatively easily, with particular strains of mice and rats developed specially for laboratories. However, although experiments with rodents can often be applied to humans, there is debate about their usefulness as well.

“There has been huge opposition for a long, long time to using mice or rodents to predict effects in humans,” says Dr Julia Taylor, a chemicals and health researcher at the University of Missouri, USA. This stands in stark contrast to the standards applied in pharmaceutical testing, where relevance to humans is assumed and if in animal models a drug shows toxicity or fails to show efficacy then it is more than likely to be dropped.

Where next?

There is a temptation that in order to make a claim about something, then the evidence must be of the standard we see from RCTs. This is not unreasonable and in science it is essential. However, when it comes to health effects of chemicals on humans, this standard of evidence is unattainable: regulators have no choice but to make decisions on a less determinate evidence base than is available for pharmaceuticals.

Since as a society we are more-or-less cut off from proving causation, the way in which evidence is weighed and the importance placed on animal studies, along with how this is interpreted in evaluating a chemical’s safety via risk assessment, is therefore an issue of utmost importance in chemicals regulation.

There may be ways of calibrating how evidence is reviewed in order to tighten the standard of evidence we get from animal studies. More can be done with biomonitoring and NHANES-type data sets to catch problems which slip through the net. Studies could be carried out with set protocols which make them easier to reproduce and the findings more certain, with effort spent on ensuring that studies are replicated in, for example, multiple species.

Examining how evidence can be combined into a strong case for harm could also be productive. Research suggests that BPA causes decreased sperm count in mice (Al-Hiyasat et al. 2002), while there is a very recent study showing that people who work in the production of BPA are finding effects with human sperm count as well (Li et al. 2010). Diabetes is another example: in 2008 epidemiological data showed an association between BPA exposure and type-II diabetes in humans (Lang et al. 2008); recently these findings were corroborated by a study suggesting that BPA exposure causes insulin resistance in mice (Alonso-Magdalena et al. 2010).

As a final thought, it may be useful to bear in mind two questions during the process of assessing the evidence: firstly, how much can one rely on epidemiological evidence, given that it requires harm to humans to be done in order for associations between chemical exposure and harm to be drawn out? Secondly, is it better to err on the side of caution and mistakenly restrict exposure to some chemicals which are not in fact harmful, or lean the other way and risk allowing harmful chemicals onto the market for other benefits they may bring?

5&5: News and Science Highlights from November


WMA Statement on Environmental Degradation and Sound Management of Chemicals: The World Medical Association has published a series of recommendations for its members regarding sound management of chemicals, voicing concerns about low-dose and chronic exposure.

Lab animals and pets face obesity epidemics: Nature reports on unsettling evidence that it’s not just people that are getting fatter: a statistical analysis of more than 20,000 animals suggests that the obesity epidemic is spreading to family pets, wild animals living in close proximity to humans, and animals housed in research centres.

Scientists issue health warning on BFRs: A group of 145 scientists from 22 countries has signed a statement warning that brominated and chlorinated flame retardants (BFRs and CFRs) pose health and environment hazards while at the same time providing only limited fire safety benefits.

Chemists propose to design safer chemicals – right from the start: Chemists at Yale University are calling on others in their discipline to understand how to design and build safer, less toxic chemicals – right from the start.

Environmental toxin may play important role in multiple sclerosis: Researchers have found evidence that an environmental pollutant may play an important role in causing multiple sclerosis and that a hypertension drug might be used to treat the disease.


High concentrations of polybrominated diphenylethers (PBDEs) in breast adipose tissue of California women. PBDE data in this study are among the highest reported, exceeding data from the National Health and Nutrition Examination Survey and consistent with the high use of PBDEs in California.

Exposure patterns of UV filters, fragrances, parabens, phthalates, organochlor pesticides, PBDEs, and PCBs in human milk. The first time cosmetic UV filters, synthetic musks, parabens and phthalate metabolites have been analysed in the same sample, along with persistent organochlor pollutants (POPs). The study found widespread exposure to UV filters.

Postmenopausal Breast Cancer Is Associated with Exposure to Traffic-Related Air Pollution in Montreal, Canada. Case-control study finding “evidence of an association between the incidence of postmenopausal breast cancer and exposure to ambient concentrations of NO2”.

Endocrine Disrupting Chemicals and other Substances of Concern in Food Contact Materials: An Updated Review of Exposure; Effect and Risk Assessment. A review of hazards posed by food contact materials which does what it says on the tin.

Developmental exposure to TCDD reduces fertility and negatively affects pregnancy outcomes across multiple generations. Study “demonstrating reduced fertility and an increased incidence of premature birth in mice exposed in utero to TCDD as well as in three subsequent generations.”

Leave a Comment »

RSS feed for comments on this post.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at
Entries and comments feeds.

%d bloggers like this: