Jocalyn Clark takes you through the paper and explains what it means

To be published as a research pointer in the BMJ, studies must be fascinating and not scientifically ridiculous, and they dont have to have direct clinical relevance. This study is interesting and it considers an important public health issue.

Whats the paper about?

This paper is concerned with whether radiation received for breast cancer adversely affects the heart, which might raise the risk of cardiovascular problems later in life. Just like radiation exposure from nuclear power incidents is believed to cause serious health problems such as thyroid disease and cancer, the authors thought that radiotherapy for breast cancer might increase cardiovascular disease. They hypothesised that radiotherapy to the left breast—because the heart is more exposed to the radiation—would raise the risk of cardiovascular death more than radiation to the right breast.

What did the authors do?

This study used data kept for administrative purposes. The authors linked information from a Swedish cancer registry with information contained in the countrys death records. No patients were directly involved, but because the authors were using confidential (but anonymous) medical information they had to get approval from a research ethics committee. They focused on the years 1970 to 1996.

To ensure they were studying only the records of women who received radiation treatment for their breast cancer, the authors left out the records of women whose cancer was diagnosed at the time of the autopsy and those who had other cancers. Once these data were excluded, the study sample totalled 89 407 women aged 18 to 79 years of age. Even in the world of epidemiology, this is an extremely large sample size, which means that study had a lot of statistical “power.”

The paragraph that explains statistical methods is complex. Essentially they looked at womens deaths in the years following radiation and compared women who had died within 10 years of their breast cancer diagnosis with women who had died more than 10 years after diagnosis. For each time block, the authors compared the death rates of women with left breast versus right breast tumours to test their hypothesis. Regression analysis (a statistical method used to predict relations) was used to assess the strength of the relation between the location of the breast tumour (left versus right) and cardiovascular death. Thats why you see in the results reported as a ratio comparing left versus right.

What were the findings?

The table shows that deaths from breast cancer were the same for women with right and left sided breast cancers, and these did not differ by time period. But as the authors predicted, women with left rather than right breast tumours had higher death rates from heart disease. This was particularly true for deaths that occurred more than 10 years after diagnosis.

What are the limitations?

The authors remind us that in spite of the cleverness of their statistical analyses, their results should be read with caution. The confidence interval is the range of values in which we can be 95% sure the true value lies. When the authors state, then, that the confidence intervals are wide they mean that their full confidence in the accuracy of the numbers is rather low. Thats why they state the true hazard of breast radiotherapy for cardiovascular death is still uncertain. More research will have to be done. This is particularly important because advances in radiation treatments have been made.

Dont blindly trust everything you read

It is good practice to always read research studies with a critical eye. While this study is massive in size and employs sophisticated statistical tests to make its conclusions, it can be criticised for not being up to date. The study reports death rates that are pronounced for women dying more than 10 years after their diagnosis. This implies that the old style radiotherapy of the 1970s played a part, but the authors do not comment much about this aspect of their findings. New developments—such as positioning women with their arms above their head while lying down—alter the position of the soft tissues to be irradiated and removes the heart from the radiation field.¹ The study can also be criticised because it does not consider confounders—that is, factors that distort the true relation between the exposure (radiation) and the outcome (cardiovascular death) because they themselves are related to the outcome. For example, the authors did not investigate whether women had pre-existing cardiovascular disease or other risk factors for heart disease such as smoking and diabetes. In addition, the authors did not look at how much of the heart was exposed when the radiation was done. This would have varied among the women and may have been an important factor in womens risk of heart disease.

Jocalyn Clark editorial registrar
Email: jclark@bmj.com

1. Electronic responses. Mortality from cardiovascular disease more than 10 years after radiotherapy for breast cancer: nationwide cohort study of 90 000 Swedish women. BMJ 2003. http://bmj.com/cgi/eletters/326/7383/256 (accessed 3 Mar 2000).