Tags: Cancer, chemicals, food, health, organic, organic food, pesticides
Organic food: does eating
it cut cancer risk?
A recent study published in the British Journal of Cancer (BJC) found that, in a group of over 600,000 British middle-aged women followed for just over 9 years, no decrease in cancer risk could be observed between women who described themselves as “always” or “usually” eating organic food, and women who described themselves as “never” eating organic food (Bradbury et al. 2014).
The finding was picked up by almost all major media outlets in the UK, usually under headlines along the lines that eating organic does nothing to reduce cancer risk among women (the Daily Mail) or that women who eat organic food are no less likely to develop cancer than women who do not (the Guardian and the Telegraph).
Three people were quoted in the media coverage: Professor Tim Key of Oxford University, one of the study authors; Dr Claire Knight, Cancer Research UK’s health information manager (CRUK funded the research and owns the British Journal of Cancer); and Peter Melchett, Policy Director of the Soil Association, the UK’s largest body promoting organic agriculture. Key said: “In this large study of middle-aged women in the UK we found no evidence that a woman’s overall cancer risk was decreased if she generally ate organic food.” Knight said: “This study adds to the evidence that eating organically grown food doesn’t lower your overall cancer risk.”
Melchett was critical of the study, suggesting that the researchers were too quick to dismiss an observed 21% decrease in non-Hodgkin’s lymphoma among women who ate organic food, pointed out that reasons for eating organic extend beyond cancer risk to bee protection and prohibition of genetically modified organisms, and also that CRUK’s advice that people wash vegetables if they were concerned about pesticide exposure would not work for systemic pesticides or pesticides in processed foods such as bread.
It would be reasonable to assume that Melchett’s response comes from concern that people reading about the study might be discouraged from eating organic food. The Cancer Prevention and Education Society (publisher of Health & Environment) also recommends that people eat an organic diet, or at least limit as far as possible their exposure to pesticides. We are therefore going to look at exactly what can be concluded from the study and consider whether or not this supports any sort of a case against the purchase and consumption of organic food.
What the study did
Historically, researchers investigating the prospective health benefits of an organic diet have mainly examined either the consequences of an organic diet on levels of pesticides in people’s bodies or differences in nutritional value between conventional and organic produce.
A systematic review of this data conducted in 2012 concluded that while there was good evidence that an organic diet reduces pesticide exposure, there was a lack of compelling evidence of increased nutritional value in organic food (Smith-Spangler et al. 2012): only three studies had examined clinical outcomes in relation to organic diet, finding no significant differences between incidence of allergic reactions or food poisoning in relation to conventional or organic diets.
This leaves a substantial research gap in terms of whether or not the reduced exposure to pesticides resulting from an organic diet translates into a measurable health benefit. It is to this which the BJC study was designed to contribute.
The BJC study is an analysis of data generated by the Million Women Study, a national investigation of the health of UK women over the age of 50. The Million Women Study periodically collects data on the health and lifestyles of over a million women volunteers in order to uncover connections between reproductive and lifestyle factors and women’s health (Million Women Study, 2013). The variety of questions being asked (see for example the third questionnaire, sent to participants 8 years after recruitment) means the study is large set of health and lifestyle data for analysis in prospective epidemiological research.
One of the questions in the surveys concerns the frequency with which the respondent eats organic food (possible responses being “never”, “sometimes”, “usually” or “always”). This allows a rough distinction to be made between people who eat more organic food and those who never eat it. When this is set against the health records of the respondents, in theory it enables researchers to observe differences in health outcomes between the group which eats only a conventional diet (which we shall call the “never-eats”) against the group which eats a larger proportion of organic food (the “usually/always-eats”).
What the study found
The study found that 30% of the women surveyed say they “never” eat organic, 63% “sometimes” do, and 7% either “usually” or “always” eat organic (the “usually” and “always” groups were combined throughout the analysis, so we do not know from the study what percentage stated they “always” eat organic).
The study found that the usually/always-eats were almost exactly as likely to get cancer as the never-eats: the relative risk (RR) of cancer for the usually/always-eats was 1.03 with a 95% confidence interval (CI) spanning 0.99 to 1.07. Additional analyses showed that usually/always-eats were a bit more likely to get breast cancer (RR of 10.9, CI: 1.02-1.15) than the never-eats and somewhat less likely to get non-Hodgkin’s lymphoma, although by exactly how much it was difficult to say (RR of 0.79, CI: 0.65-0.96).
The authors state that the results for non-Hodgkin’s lymphoma and breast cancer both need to be interpreted cautiously. Since the study looked at a large number of cancers, it is much more likely that any individually significant results are the result of chance rather than reflective of a real difference in cancer risk between the two groups. (This is because the researchers were looking for relationships which had a better than 95% chance of being true; because they did this may times over, they greatly increased the probability that any of the individual relationships they observed would be spurious. Rather like a 20-sided dice, if you rolled it once and got a 20 you might think something significant had happened, but if you rolled it 20 times and got a 20, you should be less likely to think something significant had happened – though you could not rule out that it had.)
The two issues worth discussing in detail are whether or not the difference in organic food consumption between the never-eats and the usually/always-eats is great enough to allow the study to detect a subsequent difference in cancer risk between the two groups, and the extent to which a study of middle-aged women allows inferences to be drawn about the putative health benefits of eating organic.
How often is “usually”? It is not clear from the study what it is the respondents mean when they say they “usually” eat organic. Does that mean 20% of their diet is organic? That they eat exclusively organic vegetables but not organic meat or dairy products? That 80% of their diet is organic? That they do not know exactly how often they eat organic but they usually try to do so? And so forth.
It is also not clear how long before the start-point of the study that participants had been eating organic diets. Given that cancers resulting from chronic exposure to low levels of environmental pollutants can take a long time to emerge, it could be the case that 10 years of eating organic since the age of 45 (which would mean participants starting a usually-organic diet in around 1993, when this was much more difficult than 10 years later) may be of small significance compared to eating only conventional food up to the age of 45.
The problem is, the less often the usually/always group actually eats organic, the more similar they are in pesticide exposure to the never group. And the closer they get, the more difficult it becomes for the study to observe subsequent differences in health outcomes – particularly if the effect being studied is small, as the contribution of organic diet to reduced cancer risk may be.
The data would be easier to interpret if an always-eats group was compared to the never-eats group. The likelihood is that the always group is too small to be usable (this would explain why the usually and always groups were merged). Hard data from blood samples taken from the respondents would allow the assumed difference between the never-eats and the usually/always-eats to be validated but this was not available to the researchers.
Without being able to get a fix on the real difference in diet between the two groups which are being compared, it is very difficult to get a sense of the likelihood that the study design was either unable to detect a real effect or correctly observed that there is no real difference in cancer incidence between the two groups. Since the difference in lifetime diet between the two groups could in fact be negligible, the possibility that this study in fact yields a false negative (that it finds nothing when in reality there is an effect) needs to be considered.
Relevance of the study group. The study only looked at middle-aged women. This group may not be particularly susceptible to the cancers most likely to be prevented by an organic-only diet – instead, men of a different age may be at greater risk, or children of both sexes, and so forth. Sub-groups in the study, such as vegetarians, may see a particular benefit in an organic diet (the study did not look at this).
Cancer is also only one health end-point among many which could be influenced by pesticide exposure; of these, the vulnerability of children to neurotoxic effects of pesticides is of particular interest (Young 2014). These considerations do not invalidate the study but caution should be exercised in drawing general about health lack of health benefits of an organic diet from the findings described here.
What do we conclude?
From a research perspective, this is a well-conducted study which is unique for looking at cancer risk in relation to consumption of organic food in the diet. It did not really find anything. It is difficult to judge if this is either because there was no significant effect to be found, or because the effect of eating organic food is too small to be detected, or instead because there was a lack of a real difference between the two groups being compared.
Therefore, it is difficult to draw a conclusion which goes further than saying that women who eat an unclear amount more of organic food for an unclear period of time (potentially starting quite late in life) see no benefit in terms of reduced cancer risk. As such, this study adds a little but not much to the evidence that there is no benefit to eating organic.
If there really is no actual benefit to eating organic in this group of women, it has to be remembered that a beneficial effect may still exist for other population groups or sub-groups in the study. Perhaps more importantly, dietary pesticides may affect other health end-points to a greater degree than they do cancer, with middle-aged women not necessarily being the part of the population most at risk of this harm.
The major limitation of the study, at least in this author’s opinion, is the uncertainty around difference in diet between the never-eats and usually/always-eats groups. Even if this was obscuring only a small beneficial effect, for a disease as common as cancer this could have important consequences for cancer prevention strategy. If, as may be the case, there is no significant exposure difference between the groups at all, even a fairly large beneficial effect could be obscured in this study design.
Future epidemiological studies should include biological sampling in order to give a clearer picture of chemical exposures in relation to diet and disease risks. If we are interested specifically in quantifying the health benefits of an organic diet, then epidemiology may well be ruled out as a useful tool: the authors state that confirming whether or not an organic diet reduces risk of non-Hodgkin’s lymphoma could require a 10-year study with 500,000 participants. It might be more realistic to study this in an animal model instead, if it was felt that the resulting data would provide sufficient ethical justification for conducting such research.
Even if that data were still to be inconclusive or show an absence of benefit, there are still other reasons for eating organic. The possibility that pesticides may play a role in increased incidence of neurodegenerative disorders in farm workers (van Maele-Fabry et al. 2012, Allen & Levy 2013) along with the acute hazards their use presents, means that pesticides pose some risk even if that is not transmitted down to the consumer. There is also the argument that organic farming practices are more sustainable, yielding more robust and varied agricultural ecosystems. It would therefore be a mistake to put too much weight on this particular study in building a case against the consumption of organic foods.
A Health & Environment retrospective: concerns that talking about causes of cancer can cause cancer; the difficulties of defining “endocrine disruptor”; shedding light on the obesogen hypothesis; and more.August 14, 2013 at 4:09 pm | Posted in H&E Features | Leave a comment
Tags: BPA, Cancer, chemicals, endocrine disruptor, environment, health, PFCs, science, thresholds of toxicological concern, triclosan
After 5 years and 60 editions of Health & Environment, we look back at our most popular articles.
Does discussion of environmental causes of cancer cause cancer? There is no obvious reason why the 2010 US President’s Cancer Panel examination of chemicals as potential, under-discussed causes of cancer should have caused so much controversy. The report was hardly unheralded while the recommendations were in line with much existing cancer prevention strategy. The conclusion that one is forced to, is critics of the report must believe that discussing environmental causes of cancer undermines cancer prevention strategy: in other words, that talk about environmental causes of cancer causes cancer. (July 2012)
False alarms or missed hazards: how should regulators define “endocrine disruptor”? The trade-offs entailed in how we define, for regulatory purposes, a chemical as an endocrine disruptor should make us very cautious about conflating the purpose of a regulatory definition of EDC with the purpose of a scientifically correct definition of EDC, and may even show us that the process of defining EDC is a democratic matter which cannot be decided by expert committees alone. (November 2012)
Even though BPA is a weak oestrogen, there is a mechanism by which low levels of BPA could have a powerful health effect. It is often argued that BPA is too weak a hormone to have an effect on cell function. Here, we sketch out a mechanism by which BPA could potentially have a strong effect via an indirect pathway rather than a direct effect on the part of the cell nucleus which responds to oestrogen. (July 2009)
Thresholds of Toxicological Concern: Evaluating an Initiative to Reduce Animal Testing. One rationale for reducing the burden of chemical toxicity testing is the application of thresholds of toxicological concern (TTCs), a pragmatic, probabilistic approach to risk assessment of substances for which toxicity data are unavailable. It holds that if a substance is unlikely enough to pose a risk to health, then toxicological testing of the substance is not required. (January 2012)
The Obesogen Hypothesis. Energy imbalance is the immediate cause of obesity, a combination of excess dietary calories and a lack of physical activity. However, the full set of reasons as to control over energy balance can be lost is complex – and it is now being hypothesised that chemical pollutants have a role to play. (March 2011)
Shifting the curve: how small changes in individuals have large effects in populations. “Shifting the Curve,” a video narrated by Dr. Bruce Lanphear of Simon Fraser University, shows graphically how small effects – like an increase in the number of individual children with ADHD-related behaviors – result in large increases in the prevalence of ADHD in the overall population. (October 2012)
Assessing Risk Posed by Chemicals in Mixtures. Current practice in risk assessment, although changing, is more-or-less grounded in the 1970s, when pollution from industrial smokestacks and waste outlets was seen as the primary source of risk. It is now recognised that people are exposed to a wide variety of chemicals from many sources, not just industrial pollution. Rather than anticipating the risk to health and the environment posed by a specific waste outflow, the problem has become one of understanding and managing the risk that multiple, everyday exposures may pose to health. (July 2011)
PFCs: A case study in favour of the precautionary principle. PFCs are an example of how production and marketing of a substance can outpace scientific research into its safety and placing regulatory restrictions on its use. In the case of PFCs, this has resulted in 3 generations of people being exposed to an unknown hazard while a complex consensus, based on weak data and economic interests, develops around restricting their use. (December 2011)
Triclosan or soap and water? Despite its widespread use, only a relatively small amount of research into the potential health effects of triclosan has been published. Much of the evidence for potential harm from triclosan comes from in vivo amphibian and fish studies and in vitro mammalian studies. (March 2013)
Tags: Cancer, cancer prevention, environment, female cancers, health, medicine, public health
Cancer prevention strategies are based on what is known about attributable causes of cancer. But does focusing on existing knowledge advance or hamper efforts to reduce cancer incidence? We evaluate two opposing perspectives to conclude that demanding highly robust data may in fact limit our ability to prevent cancers beyond the proportion caused by lifestyle choices.
In December, the British Journal of Cancer published an analysis of the fraction of cancers in the UK in 2010 which were attributable to lifestyle and environmental factors (Parkin 2011). Among the main findings were that 23% of male and 15% of female cancers were caused by tobacco use, that alcohol consumption was responsible for 3-5% of cancers in both males and females, and that obesity was responsible for 6% of female cancers [chart here].
Casual readers may be puzzled by two features of the report: firstly, although the report states it is about environmental causes of cancer, it deals almost exclusively with lifestyle choices; and secondly, although the report only identifies causes for 40% of cancers, it makes a clear statement that these factors ought to be prioritised as part of a cancer prevention programme.
Why so little discussion of the environment?
Lay readers (not epidemiologists) are likely to understand the term “environmental” as referring to anything with which the body involuntarily interacts with in the world, in particular pollutants, while lifestyle factors are likely to be interpreted as encompassing people’s choices, such as what they eat and how much they exercise.
There is a simple explanation as to why the report equates environmental causes of cancer with lifestyle causes, as Professor Max Parkin, lead author of the BJC report, explains: “Epidemiologists just split things up into environmental or genetic. Genetic is something built in from the word go, that you are born with. Everything else is an external influence, which is referred to as environmental.”
Of course, what many environmental organisations are interested in is the specific contribution which inadvertent exposure to environmental pollutants makes to the burden of cancer and other disease. On this, the report has very little to say, devoting only one chapter to occupational causes of cancer, and no space at all to environmental pollutants and chemicals as a potential cause of cancer.
Instead, the BJC report satisfies itself with identifying lifestyle causes of cancer and leaving the causes of the other 60% of cancers unaccounted for. Since most lifestyle factors are accounted for in the study, this leaves a significant proportion which must be environmental in the conventional sense. So why are they not in the report?
There is a simple explanation for this as well: there are no specific environmental causes of cancer for which there are sufficient data to calculate attributable risk. For calculating attributable risk, the BJC report required sufficient evidence on the presence and magnitude of likely causal associations with cancer risk from high-quality epidemiological studies, and data on risk factor exposure from nationally representative surveys.
“We could have written a little bit about air pollution,” says Parkin. “However, it is likely to be a minor component compared to those other things [assessed in the report]. It could be doubling the risk of cancer, but we don’t really have a very good hold on that.”
“It’s possible there are things we don’t know about which are substantial contributors to cancer,” Parkin adds. “But never say never: 35 years ago we didn’t know about HPV [human papilloma virus, a major risk factor in cervical cancer] so there might be something out there. It’s a bit odd that non-Hodgkin’s lymphoma is going up, and testicular cancer likewise. However, it is unlikely we will find [another major factor] like smoking.”
If only 40% of cancers are accounted for, why are these the most important causes to address?
The BJC figures are clearly intended to inform the UK’s cancer prevention strategy and place the emphasis on people’s choices about diet, exercise, alcohol consumption and so forth. Cancer Research UK’s Chief Executive, Dr Harpal Kumar, describes these “healthy habits” as the priority for cancer prevention (CRUK 2011). Tobacco cessation therefore continues to be the number-one priority, with reducing alcohol intake and encouraging more exercise and better diet next in the queue.
Some argue this approach does more harm than good. Professors Andrew Watterson and Rory O’Neill of the University of Stirling, Scotland, argue forcefully against the “victim-blaming” they see as implicit in a healthy habits strategy, diverts attention away from the broader, societal roots of many cases of cancer, in which “income and social class connect directly to […] poor diet […] poor housing near busy, polluting roads and […] dusty, dirty, chemical-laden jobs and long hours” play into cancer incidence.
Not everybody sees things this way, however. In a Personal View in March’s Lancet Oncology, Professor Bernard Stewart of the University of New South Wales, Australia, acknowledges that although a focus on lifestyle does not recognise the burden caused by the chemical industry and associated pollution, no intervention based around reducing involuntary exposure to pollutants (except for air pollution) should be prioritised over lifestyle choices in a cancer prevention strategy (Stewart 2012).
Stewart argues that public health strategies aimed at cancer prevention have to be based in solid data. To be solid enough, the data has to show (1) the circumstances of exposure; (2) a calculation the consequential risk of cancer; and (3) the effectiveness of measures in reducing that risk of cancer. Only if all three boxes are ticked can an intervention be properly described as preventive.
For Stewart, only air pollution has adequate data for describing exposure and establishing the burden of cancer. He describes industrial pollution and pesticide findings as “not unequivocal”. For EDCs, their role in human breast cancer is “mainly inferential” and as yet unsupported by epidemiological evidence. Nor, he says, is there any evidence that there is a case of cancer which would have been avoided had a consumer decided not to buy a particular product, or had regulators been more diligent.
This is the interpretation of the data on which the established view that personal choices provide the greatest opportunities for reducing cancer incidence, while controls on specific pollutants are understood to be individually so causally insignificant they cannot be interpreted as part of a cancer prevention strategy.
Does this lead to an effective prevention strategy?
Professor Richard Clapp of Boston (US) University’s School of Public Health is unimpressed with the findings of the BJC report and is opposed to the view that lifestyle choices should be a priority in a cancer prevention strategy. Although Professor Richard Peto says the BJC report will focus attention on the high priority areas, such as refocusing on tobacco and the continuing importance of tobacco control and efforts to change the UK diet (Peto 2011), to Clapp, this is “just more of the same”.
“The 60% unknown is the elephant in the room,” says Clapp, reiterating Parkin’s and Stewart’s view that, beyond obvious causes such as smoking, it is very difficult to attribute percentages to causes of cancer. However, he draws a very different conclusion, saying it is “counterproductive and pointless” to assign certain exposures as causing a specific fraction of cancer when it is clear that preventable occupational and environmental exposure fuel excess cancer cases and deaths. (See e.g. Clapp et al. 2007)
Clapp argues this is because the fundamental mechanism of cancer is both environmental and genetic. Exposures from outside the body combine with inherited genes and genetic mutations, all of which converge to produce cancer. Overall, there are six essential alterations which need to happen in order for the body’s defences against cancer to be overwhelmed (Hanahan & Weinberg 2011).
Clapp describes this as an integrated circuit, in which a combination of exposures is required to produce a tumour, then prevention of any one of these will prevent the tumour. While preventing any of the single major factors, such as a carcinogenic exposure, is therefore a preventive measure, it becomes impossible to calculate what proportion of cancers any particular such measure prevents.
That the effectiveness of an intervention cannot be measured is unimportant for Clapp, who says we do not need a hierarchy of interventions or to play one cause off against another; exposures from all sources should be systematically reduced: “It doesn’t matter if tobacco is responsible for 20% or 30% of cancers; if it’s a carcinogen, we should minimise the exposure.”
Is Stewart’s position of holding out on action until the data is certain help set the right priorities for a cancer strategy, or does it limit our ability to prevent cancer?
For one thing, human exposure to chemicals is uncontrolled. In the instance we find a control group, exposure is usually so thoroughly confounded with other chemicals, lifestyle choices, occupation, economic status etc. that it becomes extremely difficult to prove an association which could pass as causal.
It is therefore hard to see how we can attain Stewart’s required level of proof for preventive action, beyond the 40% we already know about (and have known about for over 30 years). Waiting for proof for a set of factors which Stewart and Parkin both acknowledge may remain unproven for many years to come, seems a missed opportunity for preventing many cancers – even if we do not know how many cancers such action will prevent.
So although Stewart argues against use of the precautionary principle, it seems if we are to progress beyond the 40% of cancers attributed to lifestyle choices, we will ultimately have no choice but to work with limited data: being measurable is important, but only if your measurements help you make the optimal decision, rather than simply the optimal measurable decision.
“It is probable that limiting exposure to tobacco smoke has reduced incidence of lung cancer; there are lots of opportunities to expand that approach,” says Clapp. Watterson and O’Neill say that “social, political and physical environmental factors all play into cancer incidence and prevalence, and should form part of a coherent cancer prevention strategy.”
Although individual exposures are unlikely to make much difference to cancer risk, if there are hundreds of exposures which can be eliminated through general pollution control programmes, then significant further progress could be made. As Parkin acknowledges, “healthy habits” will not help the majority of people who don’t smoke, aren’t overweight and will get cancer anyway.