We read with interest the recent meta-analysis of seven studies that reported that the age-adjusted incidence of Parkinson’s disease was 1.5 times greater in men than in women.¹ However, this meta-analysis excluded several informative studies (such as those with <50 patients and those restricted to older cohorts) and did not explore heterogeneity in the male to female (M:F) ratios. We were also unable to replicate some of the data used in the meta-analysis. For example, we believe it misquoted some crude incidence rates as being age-adjusted (eg, the studies from Ferrara and Olmstead County).

In the process of updating our previous systematic review of studies on the incidence of Parkinson’s disease,² we performed a new meta-analysis of age-adjusted M:F incidence ratios for Parkinson’s disease and attempted to identify the causes of heterogeneity. Additional studies published between January 2002 and April 2005 were identified using the same search strategy and inclusion criteria as those in the previous review.² Where possible, the age-standardised M:F incidence ratio for each study was calculated with the Confidence Interval Analysis software V.1 by applying the age-specific female incidence rates to the corresponding male population. In three studies, an age-adjusted M:F relative risk (RR) given in the original publication was used instead.

Meta-analysis of the natural logarithm of age-standardised M:F ratio (logSR) estimates was performed using a DerSimonian and Laird random effects model. Heterogeneity was formally assessed using the I² statistic and explored using retrospective meta-regression analyses, which assessed the influence of study location (dichotomised Western v Asian) and the mean age of onset in the incident parkinsonian cohort (mean age of onset/diagnosis <70 years of age v onset diagnosis ⩾70 years of age). Additionally, we assessed publication bias with a funnel plot of the logSR. The influence of individual studies was assessed by omitting each study one-by-one and recalculating the pooled estimates. All meta-analyses were carried out using STATA V.8 software.

In all, 17 relevant studies including more than 2500 people with Parkinson’s disease were identified (table 1). No evidence of publication bias was found. The pooled estimate for the age-standardised M:F ratio was 1.46 (95% confidence interval (CI) 1.24 to 1.72, p<0.001), but there was a high level of heterogeneity (I² = 85%). Removal of individual studies altered the pooled estimate only marginally (between 1.42 and 1.51), with the 95% CIs all comfortably excluding the null value. Meta-regression analysis showed that studies from the West gave significantly greater M:F ratios than those from Asia (table 1; RR from meta-regression 1.58, 95% CI 1.12 to 2.22, p = 0.009) and that studies in which the mean age of subjects with Parkinson’s disease was ⩾70 years (pooled M:F ratio 1.67) gave significantly greater M:F ratios than those with a mean age <70 years (pooled M:F ratio 1.23, RR from meta-regression 1.38, 95% CI 1.06 to 1.81, p = 0.018). However, heterogeneity remained high within the <70 and ⩾70 years age subgroups (I² = 80% for both) and for the Western studies subgroup (I² = 84%).

Our analysis on a larger dataset gave a very similar overall result as that of the previous meta-analysis,¹ with an M:F ratio of 1.46 versus an RR of 1.49. However, there was significant heterogeneity between studies, which was only partly explained by mean age of onset and study location. We found evidence that the M:F difference in the incidence of Parkinson’s disease increases with age of onset. Several individual studies have reported similar findings with little difference between the incidence in men and women <60 years of age.^5,6,8 The reason for this remains unclear. It may be due to postmenopausal hormonal changes in women or differential exposure of men and women to environmental risk factors in later life.

We also showed that the M:F difference is significantly greater in Western populations than in Asian populations (specifically Chinese or Japanese) where no significant M:F difference was found. Although this subgroup analysis should be interpreted cautiously (as it was based on only three studies), it is supported by findings from an incidence study from California, which also showed no difference in the M:F ratio in an Asian subpopulation.⁵ This finding argues against a fundamental protective effect of oestrogen, which would be expected to be present in all populations. The fact that the M:F ratio in Asian populations remains different from that in other ethnic groups with regard to those who move to a Western country suggests that there may be genetic influences on the M:F risk of developing Parkinson’s disease. However, further high-quality studies on incidence are required both from Asia and from the West to confirm that these ethnic differences are indeed real.

In summary, therefore, although there is good evidence that men are, in general, about 1.5 times more likely to develop Parkinson’s disease than women, this difference is not the same across different studies, and is more pronounced in (and possibly restricted to) people with an older age of onset and in Western populations.