Personality, Stress, Disease, and Bias in Epidemiologic Research

between the two factors, x2(5) = 0.76, p = .979. Unfortunately, Table 11 is incomplete, so a full reevaluation cannot be carried out, but the evidence for synergism is again less than conclusive. Having himself stated that it is certainly completely inadmissible to use simple univariate analyses for risk-factor studies . . . multivariate analyses are an absolute must, the analyses presented are an enigma. Not only are all the evaluations essentially univariate, they are all based on a false premise. Consequently, they have led to totally incorrect conclusions about the interactions between the risk factors investigated. No evidence of statistical synergism exists in the data reported. The implications deduced for prophylaxis and prevention are, therefore, indefensible, although intuitively attractive. This is particularly serious with respect to smoking. The data, when analyzed correctly, show that the risks due to smoking increase significantly with consumption level and that this phenomenon is true for all personality types (not just for stressed individuals, as Professor Eysenck erroneously concludes). Note