Tuesday September 30 2008
Data on birth weight, length and head circumference were used
”Women who were bigger than average at birth are at greater risk of breast cancer”, reports the Daily Mirror. The newspaper, along with several others, says that research summarising 32 studies, and 22,058 cases of breast cancer among a total of more than 600,000 women from developed countries has confirmed the link. The authors suggest that exposure to oestrogen in the womb both affects growth and, in some way, increases future cancer risk.
This study found that the increase in breast cancer risk due to birth size was moderate or small. For baby girls who weighed 2.5kg (5.5lbs) to 3kg (6.6lbs) at birth, there was a 9.4% risk of cancer by the age of 80 years, compared with 11.6% for those who weighed 3.5kg (7.7lbs) to 4kg (8.8lbs). Discovering links such as these in observational studies and researching the underlying mechanisms is often the first step to an understanding of the causes of disease. The limitation is that these study designs cannot prove causation, but confirming another risk factor for such an important and common cancer will point to other avenues for research.
Where did the story come from?
Professor Isabel dos Santos Silva from the Department of Epidemiology and Population Health, and colleagues from the London School of Hygiene & Tropical Medicine in London conducted this research, which was funded by a Cancer Research UK programme grant and Training Fellowship. It was published in the Public Library of Science peer-reviewed and open-access journal, PLoS Medicine.
What kind of scientific study was this?
This was a systematic review with meta-analysis of individual-level data from 32 studies. This type of study involves the researchers re-analysing the raw data from published and unpublished studies to obtain more precise estimates of the ‘birth size–breast cancer’ association. In some cases, this meant contacting the authors of the primary research to find out details on specific women, rather than relying on the published literature alone. Any data sent to the researchers remained anonymous.
The researchers included studies that collected information on at least one measure of birth size and also recorded new-onset breast cancers. They identified cohort studies and case-control studies (that were themselves part of larger cohort studies), by a search of the usual databases, including PubMed and Embase, up to the end of June 2007. They identified further studies by searching through reference lists and by personal communication with cancer researchers. In this way a total of 27 published and seven unpublished cohort and case-control studies were identified. Some studies were excluded from the analysis if, for example, they had contributed data to other included studies, or if the individual-level data could not be retrieved. At the end of this selection process the researchers had individual participant data from 32 studies, comprising 22,058 breast cancer cases.
As the babies tended to be smaller in studies of twins and of premature/low birthweight babies, the researchers analysed these separately from the studies reporting data on single babies. Individual participants were excluded from all analyses if they had a known history of cancer other than non-melanoma skin cancer at the start of the study. They were also excluded if all birth size data were missing.
The researchers used a statistical technique known as a random effect model to combine the estimates of effect for the studies. This model assumes that the studies are not so similar that a similar effect would be expected. Birth size was measured by weight (kg), length (cm) and head circumference (cm) at birth. The researchers looked at the effect on the rates of breast cancer of increases in these measurements in steps of about one standard deviation, that is 0.5kg (1.1lbs) for weight, 2cm (0.8inches) for length and 1.5cm (0.6inches) for head circumference.
What were the results of the study?
Birth weight was positively associated with breast cancer risk in studies based on birth records. For each step increase in birth weight (0.5kg) there was an increase risk of 6% (RR 1.06, 95% confidence interval 1.02 to 1.09). There was a steady increase in the risk of breast cancer with increasing weight at birth. Compared with women who weighed 3 to 3.499kg, the risk was lower in those who weighed less than 2.5kg, and greater in those who weighed 4kg or more. Birth length and head circumference from birth records were also positively associated with breast cancer risk.
When the researchers adjusted for all three birth size variables, they showed that length at birth was the strongest independent predictor of risk. The established breast cancer risk factors, number of children and socioeconomic factors, did not appear to interfere statistically with the estimates. These were not modified by including age or menopausal status into the equation either.
What interpretations did the researchers draw from these results?
The researchers say that the “pooled analysis provided evidence of moderate positive trends in the risk of breast cancer among studies based on birth records, with risk increasing with increasing birth weight, length and head circumference”.
They comment that the source of birth size data was the main source of differences between the studies (heterogeneity). They say that the positive association of birth size with breast cancer risk was found only in data from birth records but not in data from self-reports or maternal recalls when the women were adults, suggesting that their approach to analysing the recorded data only was less prone to bias.
Adjustment for weight, length and head circumference in their analysis showed that length at birth was the strongest predictor of risk, despite the fact that it tends to be measured less accurately than weight or head circumference.
The birth size effect did not appear to be confounded or modified by known breast cancer risk factors. The association between birth size and breast cancer risk was observed consistently in women born over a period of several decades, and in different geographical areas.
What does the NHS Knowledge Service make of this study?
This was a large study including a large amount of birth data on women who go on to develop cancer. As the authors say, this means that the statistical power – the ability to detect an effect if one exists – is higher, therefore the study can be expected to give a more precise estimate of the strength of any link.
Heterogeneity, that is, the underlying difference between studies which can sometimes prevent valid pooling of the results, was partly addressed by the researchers by obtaining data on individual women, and defining and coding the measurements of interest (weight, length and head circumference) in a standard way, and by choosing some factors to control for across all individuals. These measurements and adjustments may have been treated differently in the original primary publications, and the ability to use raw data to maintain a standard approach is a strength of an individual-level meta-analysis such as this.
The researchers also acknowledge some limitations and biases that need consideration:
- Publication bias can be a problem with pooled analysis because studies reporting negative findings may be published less often than those that report positive results. The authors argue that as inclusion in this pooled analysis was not dependent on publication, their re-analysis is less likely to have been affected by publication bias than meta-analyses of the published literature.
- The researchers relied on direct measurements of birth size, rather than those reported by the women. This means that any measurement error or reporting bias could be lower than if they had relied on recall in questionnaires, for example. Despite this, there is still a small possibility that birth size, or other measured factors, may have been incorrectly recorded, or that breast cancers may have been misclassified.
- The researchers adjusted for the potential confounding factors on which they had information, such as maternal age, number of children and socioeconomic status. By comparing the effect estimates in the unadjusted and adjusted analysis, they show that the results showed little variation. It is important that this was done, but it cannot completely exclude residual or unmeasured confounding by these or other factors.
Overall, this is a reliable summary of observational studies, which adds precision to the estimate of the strength of a risk factor link to breast cancer. The link shown is modest at best, and is comparable to other known risk factors, such as increasing age, not having children and having a late menopause. The biological mechanisms behind the association will need further evaluation. In particular, to determine whether oestrogen alone is the common factor determining birth size and breast cancer risk or, as the authors also mention, if there is a complex interplay of several hormonal and non-hormonal factors.