select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A study published in Occupational & Environmental Medicine looks at occupational environment and ovarian cancer risk.
Prof Paul Pharoah, Professor of Cancer Epidemiology, Cedars-Sinai Medical Center, said:
“This paper reports the findings from an ovarian cancer case-control study. The authors report statistically significant associations with ovarian cancer for two classes of occupation – i) hairdressers and beautician and ii) accountants.
“The major problem with this study is the sample size. In total fewer than 500 women with ovarian cancer took part in the study. Consequently, the findings are not statistically robust. Given the number of occupations investigated, one or two statistically significant associations would have been expected by chance (which is what they found).
“I do not think it is possible to draw any firm conclusions about occupational exposures from the findings of this study. In particular, there is not strong evidence to suggest that either hairdressers or accountants are at increased risk of ovarian cancer and women in these occupations should not be concerned about their risk.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, Open University, said:
“This is a careful study that uses appropriate statistical methods, and is appropriately cautious on what it reports and concludes. But it’s particularly important to understand what the researchers were aiming to find, and what limitations arise from the data they had available and the methods they used.
“The researchers clearly state that their study was ‘exploratory’ and that it is ‘aimed at generating new hypotheses’. So, it is certainly not claiming that they have definitely found occupational groups, or exposures to chemicals and other agents, that are associated with ovarian cancer. Even less are they claiming that being in certain occupations, or being exposed to certain chemicals at work, causes an increased risk of ovarian cancer. Instead, they aimed at suggesting occupational groups, and agents to which women might be exposed at work, that possibly might be associated with ovarian cancer risk, and they say clearly that further research is needed to ‘give a more solid grounding’ to any conclusions that might be drawn about associations between what women do at work and their risk of getting ovarian cancer. The research paper is, rightly, full of mentions of ‘suggestive’ evidence, and it draws attention to unavoidable limitations on what they could conclude. The linked comment by Friesen and Freeman, also points out that ‘this study is best interpreted as hypothesis-generating’.
“All this uncertainty about the implications of their findings does not, by any means, indicate that the research was bad, or was not worth doing. In my view, it very much was worth doing. The pattern of occupations that women might work in over the course of their lives can be complicated. But it is still important to understand possible occupational cancer risks. However, when women may work in different occupations over their lives, and be exposed to different agents that could be linked to cancers, there are large numbers of occupational patterns to consider. A study like this, that points at associations that it might be important to research further, is a good way of pointing at future research questions that can be investigated in more depth, using more detailed methods. One simply cannot research everything, and this study can be seen as a kind of filter to try to ensure that future research is as productive as possible.
“So why can’t this exploratory study produce more definite results? There are several reasons.
“One reason is that the study is observational. Obviously, the researchers could not make the respondents work in specific jobs. They just observed what jobs they had done, along with some other characteristics, and the study included some women (the ‘cases’) who had had ovarian cancer and some others (the ‘controls’) who had not. So, there are bound to be differences between the women who had and had not had ovarian cancer, apart from differences in what jobs they had done and what possibly cancer-causing agents they had been exposed to at work. These other differences might be the true cause of whether they had cancer or not, rather than the occupational differences.
“As the paper and the press release point out, the researchers made statistical adjustments to allow for several other factors, and several associations between occupation and ovarian cancer risk were still present after these adjustments. But such adjustments can only go so far, and you can’t make statistical adjustments for factors on which you don’t have data. The researchers themselves point out that, when they were looking at possible reasons why working in accountancy was associated with higher ovarian cancer risk, they considered that sedentary behaviour in that kind of job might contribute to cancer risk. But the only data they had on physical activity was for recreational physical activity, which says little or nothing about how physically active accountants (and non-accountants) might be in the workplace. These issues mean that, even where there’s some evidence of association between working in an occupation, or being exposed to an agent at work, we just can’t be sure of whether the occupation or agent is the cause of any increased risk. It might be, or it might not be.
“There are other issues too, that limit what can be concluded from this work. One issue is that the number of women in the study was not immense – just under 500 women who had had ovarian cancer, and just under 900 who had not. Because these numbers were not large, the numbers in specific occupational groups was often small. For instance, there were only 20 women in this study who had worked in jobs classified as ‘Hairdressers, Barbers, Beauticians and Related Workers’ for ten or more years. So, although there is evidence that they were more likely to have had ovarian cancer than women who had never worked in that industry, the study just couldn’t be very precise about how big the difference in risk was. The best estimate was that the risk was about three times as big in those who had worked in those occupations for at least ten years, but the margin of statistical uncertainty around that estimate is large. The data is consistent with the risk only being not much higher (a quarter higher) than in women who hadn’t worked in those jobs, or over eight times as high. For accountants, where 44 women had worked in that job for at least ten years, the best estimate was that their risk of ovarian cancer was roughly double the risk for women who had never been accountants, but the data is consistent with a risk only 10% higher than for the non-accountants, or a risk nearly four times as high.
“For some of the other occupations or industries mentioned in the research paper (and often in the press release too), the position is even less clear. For women who had worked for at least ten years in the construction industry, the best estimate was that their ovarian cancer risk was almost three times the risk for women who had never worked in that industry. But the data is consistent with the risk for those working at least ten years in construction having only about half the risk of those who had never worked in construction, or having nearly five times the risk of those who hadn’t worked in that industry. There’s just not enough information to be at all precise about this. So yes, it could make sense to look further, in future research, at whether working in construction is associated with higher ovarian cancer risk, but this specific new study just can’t say on its own whether construction work is a risk factor. This study has generated a hypothesis about working in construction, but it can’t give a conclusion about that hypothesis.
“Another reason why this study can only be exploratory is that, despite looking separately at occupational groups and at possible agents that might be associated with ovarian cancer, it couldn’t really separate the two. That’s because workplace exposure to certain agents is concentrated in certain types of occupation. For example, substances such as some solvents, ammonia, and hydrogen peroxide is relatively common in women working in jobs classified as ‘Hairdressers, Barbers, Beauticians and Related Workers’, so the apparent increase in risk from ammonia that was found in this study could be due to the ammonia, or to some entirely different aspect of working in those jobs, or indeed to a risk that only arises when a person is exposed to ammonia alongside some of the other agents as well. Also, the fact that all these things are associated with one another makes it more difficult, statistically, to estimate the size of any increased risks. So again, this study can suggest hypotheses about agents, but it can’t go further than a suggestion.
“The researchers (and the press release) rightly point out another statistical issue. Pretty well all conclusions based on statistical methods are subject to some uncertainty. A finding that being in one occupation is associated with a cancer risk could be genuine (even if it isn’t an association of cause and effect), or, possibly, it could just be a statistical fluke. Statistical methods aim to distinguish the random flukes from the real associations, but they can’t do that perfectly. Other things being equal, the more associations that are examined in a piece of research, the more likely it is that some of them are just random flukes. It’s possible to make statistical adjustments to allow for this, but these researchers did not do so. That’s because they were aiming to look for hypotheses that could be worth exploring in further research, rather than trying to make reasonably watertight discoveries in this piece of research on its own. In my view, that’s not a problem if the researchers make it clear that that’s what they were doing, as they indeed did. I think it’s pretty likely that some of the associations found in this study are random flukes rather than being real – but we don’t know which they are, not until further research is done.
“Finally, I’ll just mention one further thing. When I can, in commenting on a study, I do try to give an estimate of the absolute level of risk involved. So, here, it would have been good to be able to say, for instance, that the risk of having ovarian cancer in women who never worked as an accountant is such and such a percentage, while the risk in women who worked as an accountant for at least ten years is such and such a percentage (which would be higher than for the non-accountants, indeed about double the risk for non-accountants according to the relative risk found in this study). But I can’t do that, because I don’t know the risk for non-accountants, and because this is what’s called a case-control study, it did not involve estimating that absolute risk. In the UK, Cancer Research UK say that the lifetime risk of ovarian cancer in women born after 1960 is about 1 in 50 (that is, 2%). But whether that applies to women in Montreal, Canada, where this new research was done, I have no real idea.”
‘Occupational environment and ovarian cancer risk’ by Lisa Leung et al. was published in Occupational & Environmental Medicine at 23:30 UK Time Monday 10 July 2023.
DOI: 10.1136/oemed-2022-108557
Declared interests
Prof Kevin McConway: ““I am a Trustee of the SMC and a member of its Advisory Committee. My quote above is in my capacity as an independent professional statistician.”
Prof Paul Pharaoh: I have no conflicts of interest to report
For all other experts, no reply to our request for DOIs was received.