Title, Date of Publication & Journal

Type and timing of menopausal hormone therapy and breast cancer risk: individual participant meta-analysis of the worldwide epidemiological evidence. 29 August 2019, The Lancet

Study’s main claims – and are they supported by the data

The paper does provide evidence to support the claim that menopausal hormone therapy (MHT) is associated with an increased risk of developing breast cancer in postmenopausal women. The claim that MHT might cause breast cancer could be supported by some aspects of the analysis, but because these findings come from observational studies, they cannot provide proof of causality.

Participants in observational studies are not randomised to receive a particular treatment (MHT) so the reason they are taking it could also be related to their risk for a particular outcome (breast cancer) that may have nothing to do with the treatment itself, and this ‘bias’ could be difficult to account for by statistical analysis. For example, some people may have a genetic trait that causes severe enough menopausal symptoms to require MHT and this trait may also raise their risk of breast cancer whether or not they actually take MHT; or if MHT has been previously promoted as preventing cancer, then the population of people taking it may represent a higher-risk population and this may explain the higher incidence compared with people not taking it. The authors did not study these factors, but they remain plausible explanations for why they may have found that MHT use associates with a higher risk of developing breast cancer if, in fact, the association was not causal.

The authors do not technically claim a causal association, but there is speculation around the public health impact ‘if the association is causal’, such as that one in fifty users of oestrogen + progestogen MHT who take it for five years from age 50 will develop breast cancer within 20 years. It is important to note that, because MHT is a relatively common treatment and breast cancer is a relatively common disease, these numbers are particularly sensitive to the relative risk (RR) estimates from which they were derived. The estimates themselves contain some uncertainty (reported as 95% confidence intervals) but the projections from them do not report a similar uncertainty interval. This interval may be wide and support numbers that would indicate a much smaller or much larger public health impact.

Causality is supported by biologically plausible explanations for some of the observed effects, such as smaller excess risks of developing breast cancer in previous users of MHT compared with current users; and the finding that obese women, whose high levels of adipose (fat) tissue produce enough oestrogen to raise their risk independent of MHT use, had a smaller excess risk on oestrogen-only HRT than lean or average weight women. Causality is also supported by dose-response relationships indicating higher relative risks for longer duration of MHT use.

Strengths/Limitations

Strengths:

This paper has many strengths including a pre-planned statistical analysis plan consistently applied to all studies and a large sample size covering many geographic regions. The findings were consistent across regions containing different health service delivery systems. The authors anticipated that their results may be sensitive to certain factors and have undertaken analyses to explore them, with little sensitivity found.

Limitations:

The major limitation here is the potential for bias in the risk estimates. The authors made good attempts to eliminate it, but even with all their efforts it could still be present because: unaccounted for factors (genetic?) may explain both propensity to take MHT and propensity to develop breast cancer; missing covariate data was handled using a method (‘missing indicator’) that has been shown to produce biased risk estimates in observational studies https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3414599/; and adherence to MHT treatment was not reported but was likely imperfect, which would likely have biased the results. Also, the authors do not specify whether they accounted for between-study variability in the pooled risk estimates they presented, which, if omitted, may produce overly confident and possibly biased risk estimates.

Glossary

MHT: menopausal hormone therapy, either oestrogen-only or oestrogen plus progestogen.

Bias: the phenomenon whereby a statistical estimate, such as a relative risk, is higher or lower than the unknown but true value it is trying to find because some factors relating to both the treatment and the disease are not accounted for in the analysis.

Any specific expertise relevant to studied paper (beyond statistical)?

I am a medical statistician with no particular expertise on the subject of MHT.