The earth may move, but let's keep our feet on the ground

Havelock Ellis, the Edwardian Dr Ruth, was concerned with what he called “the problem of sexual abstinence.” Abstinence was, he claimed, responsible for neurasthenia, spinal irritation, hysteria, hypochondriasis, myalgia, and anorexia. This led him to ask a surprising question: would it be right for a physician to “prescribe” sex, even out of wedlock? He answered this question with a firm “no,” the justification of which reads like modern evidence based medicine: “In giving such a prescription the physician has in fact not the slightest knowledge of what he may be prescribing. He may be giving his patient a venereal disease [or] an illegitimate child, the prescriber is quite in the dark.”

Meanwhile non-epidemiologists will dismiss or endorse Davey Smith and colleagues' study according to their own prejudice. “Epi-sceptics” will doubt the association with an argument akin to “I don't care what the evidence shows, my grandfather smoked until he was 98 and it never did him any harm.” In contrast, “epi-enthusiasts” will greet it as a justification for their habits along similar lines to the drinker's justification: “It has been scientifically proved that drinking half a bottle of wine every day makes you live longer.” As you read this, we confidently predict that “Sex makes you live longer” will occupy more newspaper inches over the holiday period than the Queen's speech. The public will hear what they want to hear, and they will be deaf to the problems of bias, confounding, reverse causality, or chance, the quartet of spoilsport alternatives epidemiologists use to judge each others' associations.

This paper is especially susceptible to confounding and reverse causality. Confounding occurs when a risk factor which independently causes the outcome is associated with the exposure under study. In this case age is a good example: the less sexually active subjects were also older. The authors ignore some of the traditional confounders (physical activity and alcohol) and also fail to address psychological confounders—for example, depressed mood and “vital exhaustion.” Both of these may be risk factors for early death, and they are certainly predictors of reduced sexual activity. Reverse causality occurs when the “exposure” is really an early sign of the outcome. Early heart disease is likely to lead to reduced sexual activity and death. Sexual activity is downstream in the direction of causality from disease. Although the authors claim to have accounted for coronary heart disease by using baseline reporting of chest pain, this is a blunt instrument. It would not take many cases of early undetected heart disease to give the results reported here.

Thus we salute an elegantly written analysis that for once deals with activities and outcomes close to all our hearts. We suspect that in the longer run Davey Smith and colleagues have not provided evidence to satisfy Havelock Ellis, but instead they may have provided an excellent worked example for students of evidence based medicine to grapple with the issue of confounding. We would also draw attention to an important failing. The authors begin with a literary guide to adolescent angst. We feel that their omission of Holden Caulfield and, most particularly, Alex Portnoy, would have given both these characters something to complain about.