October 2, 2015

height and cancer incidence in the Swedish population

An abstract presented at the European Society for Paediatric Endocrinology annual meeting has examined cancer incidence in the Swedish population and reports an association between height and risk of developing cancer. Roundup comments accompanied this analysis.

Title, Date of Publication & Journal

Unpublished work (conference presentation).

Abstract title = ‘Positive association between height and cancer in the Swedish population’ by Emelie Benyi et al.

To be presented at the European Society for Paediatric Endocrinology Annual Meeting (under embargo until 23:01 UK time on Thursday 1 October 2015).

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

It is impossible to assess how robust the study is and whether the main claims are supported by the data, because the data is not presented (only an abstract is presented, and the study had not been published or been through peer review).

The Cancer Research UK website already lists height as a known risk factor for breast cancer, citing a meta-analysis and a pooled analysis:  http://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/breast-cancer/risk-factors#heading-Twenty-Three

If this is indeed a true association then height likely acts as a marker for some other, unmeasured variable.

Strengths/Limitations

To be able to assess strengths and limitations of the study, questions that would need to be answered would include:

1. How did the authors treat people who had already died (and therefore would no longer be at risk of cancer)?

It would be interesting to understand the height distribution of this population.

2. How many cancer outcomes were studied?  The authors say they pre-specified breast cancer and melanoma as being of interest (this is good), but did they examine anything else?

When melanoma and breast cancer are excluded from analyses, are the Hazard Ratios (HRs) for male & female for All Other Cancers still significant?

3. Were non-cancer outcomes also examined, in order to assess robustness of data and methodology?

With such a large study size (5.5 million adults), statistical significance would not be difficult to attain.  One issue is whether inclusion of further important confounders (measured or unmeasured) would adjust the HRs down towards 1.

4. It was impossible to adjust for some known confounders, e.g. duration of exposure to the sun, for melanoma.  These may have altered estimates somewhat.

Glossary

HR = Hazard Ratio (in survival analysis, the hazard ratio is the ratio of the hazard rates corresponding to the conditions described by two levels of an explanatory variable. For example, in a drug study, the treated population may die at twice the rate per unit time as the control population.)

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

None

Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.