Figure 1

Variation in RR of ST-associated oropharyngeal cancer by study type and period of publication. For each of 19 studies, separated by study type and, for case-control studies, by period of publication, the individual study RR and 95% CI estimates, taken from the Lee and Hamling review [2] (see also Table 2), are shown numerically and also graphically on a logarithmic scale. In the graphical representation, the RR is indicated by a solid square, with the area of the square proportional to the weight (inverse-variance) of the estimate. Also shown are the combined estimates, derived by random-effects meta-analysis, for the three subgroups and overall. Here the sizes of the four squares corresponding to the RRs are also proportional to the weight of the estimate, though the constant of proportionality differs from that for the individual RRs.