PHP147 - Astrological Signs and Health Problems; Demonstrating the Multiple Comparisons Problem

Abstract

Astrological signs have been linked to numerous health problems, not only in pseudoscientific but also in high impact, peer-reviewed scientific literature. The aim of this study was to use a large prescription database to assess associations between astrological star signs and health problems. This study may demonstrate important statistical issues that arise when doing multiple comparisons. For this cross-sectional study, patient and drug data were collected from a large prescription database in The Netherlands. Patients were followed for one year, starting from their birthday in 2010. Outcomes were drug prescriptions for one of four health problems: cardiovascular disease, depression, erectile dysfunction in men and fertility problems in women. The “determinant” was patients’ astrological sign (based on Western-culture horoscope). Differences in prevalence rates between star signs were tested using logistic regression, adjusted for age and gender. Significance was set at P=0.05. Almost 300,000 persons were included for analysis. In our analysis, persons born under the star sign Aquarius were less likely to suffer from cardiovascular disease (P=0.001), whereas patients born under Scorpio or Sagittarius were more likely (P=0.011 and P=0.017). There were no effects on depression or erection dysfunctions. Women born under Gemini were 37% less likely to use fertility drugs (P=0.016). This study found several associations between star signs and health problems, none of which had any biologic plausibility. Interestingly, datasets filled with randomly generated star sign data also resulted in significant findings. Multiple comparisons without prespecified hypotheses, biological plausibility and adjustments for familywise error rates, can lead to false-positive findings. Multiple comparisons are common in outcomes research, appropriate use of statistics therefore is important.