Sir,

We thank Mrs Stratton for highlighting that the figure of 0.7% for the self-reported history of measles infection in UK Biobank participants1 is far lower than the likely true prevalence in this cohort.

Most UK Biobank participants were born between 1937–1970, and routine measles vaccination2 in the UK began in 1968. Thus, most participants would not have been vaccinated against measles during childhood. Prior to vaccination, ~99% of children were seropositive for measles antibodies, suggesting that exposure to the virus was ubiquitous.2

Self-reported measles infection was calculated from the following two interview questions, firstly, ‘Has a doctor ever told you that you have had any other serious medical conditions or disabilities?’ (the ‘other’ referring to cancer, which was discussed separately during the interview), and secondly, ‘In the touch screen you selected that you have been told by a doctor that you have other serious illnesses or disabilities, could you now tell me what they are?’. We suspect the phrase ‘serious medical conditions’ contributed to the low self-reports of measles, since for most participants a measles infection may not have been perceived as serious.

We observed1 that high myopia was more common in participants who did vs did not report having measles before 17 years of age (OR=1.48, 95% CI=1.07–2.07). Since childhood measles infection was nearly ubiquitous, this association likely reflects, in reality, an association between high myopia and an unusually serious or debilitating measles infection. In support of this, we saw similar associations with reports of certain other febrile illnesses.

Self-report is a widespread tool in epidemiology with recognised strengths and limitations.3 Accuracy can range widely, for example, sensitivity: 83% for cataract and 31% for colon polyps in NHANES.4 We hypothesize that reports for a severe childhood febrile illness are likely to be highly specific, but relatively insensitive. Such misclassification bias is likely to have reduced the power of our analyses. Methods to detect antibodies to viruses, for example, VirScan,5 would provide greater accuracy.

We are grateful to Mrs Stratton for flagging this important point regarding the strengths and weaknesses of analyses using large population studies.