January 7, 2020

expert reaction to study on talcum powder and ovarian cancer

Research, published in JAMA, reports that there is no statistically significant association between the use of talc, and other powders, in the genital area and ovarian cancer.

Prof Iain McNeish, Director of the Ovarian Cancer Action Research Centre at Imperial College London, said:

“This is a very well-conducted study by a highly respected group of researchers. Proving causation links of this type is incredibly difficult and the authors are very careful to highlight the potential limitations of their study. However, this research is robust, analysing data from 250,000 women followed for an average of over 11 years, and has concluded there is no statistically significant relationship between talc use and the development of ovarian cancer.”

Prof Justin Stebbing, NIHR Research Professor of Cancer Medicine and Medical Oncology, Imperial College London, said:

“A very well conducted rigorous investigation pooled results from 4 studies in over 250,000 women, to show that talcum powder didn’t cause ovarian cancer. There weren’t many cases of ovarian cancer in the group so it’s possible a small effect has been missed, but it doesn’t look like talc is a carcinogen which is an important and reassuring finding, especially as they also looked at duration and frequency of use, again finding no causative effects.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“One thing this research clearly demonstrates is how difficult it is to tie down whether something like this is indeed a risk factor for cancer. Despite this being a good, competent, careful study involving over quarter of a million women, it still leaves room for doubt about the association, if there is one, between using powder in the genital area and ovarian cancer. There is still uncertainty about whether any such association exists. If it does exist, there is uncertainty about whether the powder itself is what causes any increase in cancer risk. And there’s also uncertainty about what the size of the risk increase is, if it there is one. But what the research does establish, I’d say, is that if using talc or other powder on that part of a woman’s body does really increase the risk of ovarian cancer, the increase in risk is likely to be small. I’m not a woman, so can’t have concerns about my own health in these respects – but if I were a woman, this wouldn’t be high on my list of worries.

“Why is there uncertainty about whether any association exists? The study involved large numbers of women, followed up for many years, so surely it should have tied things down clearly? But what counts here is not really the total number of women involved, but the number who actually had ovarian cancer, and that number is not so large – just over 2,000. Estimating the difference in cancer risk between women who did and didn’t use powder, on the basis of only 2,000 cases, is not a precise business. Most of the risk differences that the researchers examined were not statistically significant. That doesn’t mean that these differences were definitely zero, only that they were small enough so that they could plausibly be due only to random variation – the women who used talc just happened, by chance, to be slightly more likely to get ovarian cancer that those who didn’t use talc. But just because these results could be explained by random variation, that doesn’t mean that there is definitely no risk difference – so there’s still uncertainty. The explanation in terms of random variation is plausible, but it isn’t the only possibility.

“A few of the findings, however, were statistically significant. For instance, the researchers looked separately at data from women who has a patent reproductive tract, that is, they had not had their wombs removed in a hysterectomy, or had their fallopian tubes tied for the purposes of sterilisation, so at the end of their time in the study there was still the possibility of talc applied to the genitals making its way to their ovaries. In those women, the risk difference between those who used powder and those who didn’t was statistically significant, though only by a small margin. But even if a finding is statistically significant, that doesn’t mean that it’s definitely real – there is still some uncertainty involved. Importantly, the researchers compared the risk difference in women who had a patent reproductive tract, with the risk difference in women who didn’t, and found that it remains statistically plausible that those two differences are actually equal, despite the fact that one is statistically significant and the other is not. Putting it crudely, there’s too much statistical variability, because of the relatively small number of cancer cases, to come to any very clear conclusions on which effects might be real and which not. A point that the researchers themselves make is that they tested rather a lot of differences for statistical significance, and they made no statistical adjustments for the fact that they had done all these tests. Each statistical test has a chance of coming to an incorrect conclusion – statistically significant when there’s really no difference, or not significant when there’s really a difference. The more tests you do, the greater the chance that a few tests will produce a statistically significant result when there’s really no difference. Because the researchers did not formally allow for this, they rightly say that “findings from subgroup and sensitivity analyses should be regarded as exploratory” (that is, they could well be overturned by future research) – and I’d say that applies to the analysis of data from women with a patent reproductive tract.

“But let’s suppose for now that there is a real risk difference between women who used and did not use powder, at least in some subgroups of women. Does that mean that the difference is actually caused by the use of powder? No, we can’t be sure of that either. There are many differences between the women who used powder and those who didn’t, apart from their powder use, and perhaps the difference in cancer risk was caused by some of these other differences, and not the powder at all. The researchers did make statistical adjustments to allow for some of these other differences, such as age, race, education, smoking use of oral contraception, and several more. But adjustments can be made only for factors for which the researchers have data, and they are careful to point out that it’s possible that these adjustments did not go far enough.

“If there is a real risk difference, how big is it likely to be? Again, the fact that the number of cancer cases was not very large makes it difficult to be precise. Only about 9 of every thousand women involved actually got ovarian cancer during the study. By allowing for their ages, and allowing for the other statistical adjustments that were made, the researchers calculated that about 11 or 12 out of every thousand would have had an ovarian cancer diagnosis by the age of 70. That’s not a high risk. In the UK, Cancer Research UK estimate that 2% of women (20 in every thousand) will get ovarian cancer at some time in their lives – not too different from the figure from the new American research, allowing for the fact that the UK figure covers all ages, not just up to 70. Several other cancers, particularly cancer of the breast, lung, and bowel, are considerably more common in women. The new research findings are consistent with the ovarian cancer risk (to age 70) being about 11 for every thousand women who never used powder in their genital region, and about 12 for every thousand women who did use powder in their genital region at some point. But there’s some statistical uncertainty in those figures – the risks for the women who did not use powder could plausibly even go the other way (that is, be a small amount higher than for those who did use it that those who didn’t), or the risk difference could be in the direction of the main estimate but a bit larger (11 per thousand for women who didn’t use powder, about 13 per thousand for women who did). Those are all pretty small differences in risk. These figures really only apply to women like those in this study, and (as the researchers point out), those women aren’t very typical of women in general in the US, let alone anywhere else. (They are more likely to be white, and well educated, and less likely to be overweight, than the general US female population.)  However I, for one, would be surprised if the scale of the risk differences was a lot larger in other populations than was found here – but that’s just my feeling rather than being based firmly on data from good studies, because such data mostly isn’t available.

“This might not be the first you’ve heard of possible associations between talcum powder use and ovarian cancer. There have been well-reported legal cases in the USA, and a World Health Organization agency has pronounced on the matter. That agency, the International Agency for Research on Cancer (IARC), decided in 2010 to classify the use of talc-based body powder in the genital area is ‘possibly carcinogenic to humans’. But that’s a conclusion on a very different basis from this new research. First, the fact that IARC made the classification of ‘possibly carcinogenic’, its weakest classification that actually comes down on the side of something possibly being able to cause cancer, means that they felt the evidence is not very clear. Then, the IARC classifies things according to the possibility that they could cause cancer at some level of use, possibly a very different level from what actually occurs in real life, so the classification means something different from what was examined in the new research, which looked at real women’s reported use of powder. And, importantly, the IARC findings were largely based on results of case-control research. In studies like that, a group of women who have already been diagnosed with ovarian cancer are compared with a group that is generally similar but who have not had an ovarian cancer diagnosis. Both groups are asked about their previous use of powder. A problem with this kind of research is that people’s recall of whether they had used talc-based powder in the past might be affected by whether or not they have cancer. People who actually have cancer are likely to be interested in what may have caused it, and may be more liable to remember previous use of something that might plausibly have caused the cancer than would be the case for people who do not have cancer. This so-called ‘recall bias’ is likely to be more marked if there has been publicity of a possible link between a substance, such as powder, and the type of cancer that a person has. The report on the new research points out an example of this found by some previous researchers (Schildkraut and colleagues), who found a very large difference in the percentage of women with ovarian cancer who reported previous powder use, depending on whether they were interviewed before or after 2014. That’s roughly the time when major US lawsuits on talc and cancer began (and began to be reported). The percentage went up from 37% before 2014, to 52% after 2014. In women who had not had ovarian cancer, the percentage who reported talc use was almost identical before and after 2014. This points to a possible very substantial recall bias after 2014. It says nothing direct about whether there was any such bias before 2014 – but there could well have been. The IARC report concluded that there may well have been recall bias in the studies it considered, but that there was no way to estimate how large the bias might be, and that’s a major reason why they felt they could not say more than it was possible (rather than probable or certain) that talc-based powder in the genital area was capable of causing cancer.

“A major strength of the new research is that it is not a case-control study. Instead, it has what’s called a cohort design. For the new research, all the women were asked about their use of powder when they entered the study, before any of them had had an ovarian cancer diagnosis. Therefore the type of recall bias that I have described could simply not occur. They were then followed up for years to see whether they developed ovarian cancer later.

“The researchers on this new study have, rightly, been assiduous in pointing out its limitations as well as its strengths. They mention all the issues I have mentioned, as well as some others. The excellent accompanying editorial by Dana Gossett and Marcela del Carmen also draws attention to many important points.”

‘Association of Powder Use in the Genital Area With Risk of Ovarian Cancer’ by Katie M. O’Brien et al. was published in JAMA at 16:00 UK time on Tuesday 7 January. 

DOI:10.1001/jama.2019.20079

Declared interests

Prof Iain McNeish: No declarations of interest

Prof Justin Stebbing: No declarations of interest

Prof Kevin McConway: Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.