Letter in response to Salari M,et al. Epidemiology and clinical features of Huntington's disease in MENASA region: A systematic review and meta-analysis

Dear Editor,

We read with interest the article by Salari et al. published ahead of print. ¹ The systematic review of studies from the Middle East, North Africa and South Asia is valuable; however, we have concerns about the inclusion of the pooled average prevalence of Huntington's disease (HD) from this geographical area. ¹ Publication of this article has coincided with our systematic analysis of the prevalence of HD and a guide to its interpretation which has also been published ahead of print. ²

In our article we make the case that there is so much heterogeneity in the prevalence studies of HD that whilst a pooled average with a 95% confidence interval can be presented, this is not the best way to summarize the data. Observational studies of disease prevalence typically have very large sample sizes (often the whole population of interest). When synthesised in a meta-analysis the estimate of the pooled average may consequently appear very precise with a narrow confidence interval. However, we argue that this can be misleading as the individual studies may have a very wide spread of results and the confidence intervals illustrated on a forest plot may not overlap. A 95% prediction interval may be a more conservative method of synthesising heterogeneous results, although this requires a sample of at least 10 studies. ³ In our study, the 95% prediction interval for all studies published after 1993 was 0.32 - 37.55 per 100,000 and for Europe it was 1.64–19.18 per 100,000; both of these are wide and limit their utility. We used a meta-regression analysis to identify some of the causes of the heterogeneity: both Asian and South American based studies had a significantly reduced point prevalence compared to the European reference group and each year post-1993 was associated with approximately a 7% increase in the point prevalence of HD. ²

The meta-analysis of Salari et al. was based on four studies. The study from Egypt came from a governate with a population of 42,000 and was undertaken prior to 1993. ⁴ There were 9 patients giving a prevalence of 21.43 per 100,000 (95% CI 10.58–41.42). A study with a small denominator may not be representative of a larger geographical area with the resulting prevalence figure being unusually high or unusually low. In our meta-analysis we chose to avoid the possibility of the reporting of a cluster of cases by excluding studies with a denominator of less than 100,000.

Salari et al. noted that the only study of patients from the South Asia subcontinent came from a report of 17 patients from Pakistan, Punjab and Gujarat living in the UK. ⁵ That study noted there were no patients from Bangladesh despite a significant number of people from that part of South Asia living in the UK. The prevalence in 1988 was 1.35 (95% CI 0.83–2.18) A note at the end of that paper states that since submission a further 6 patients were identified. The paper was published in 1990 so it may be reasonable to include these patients in which case the prevalence would be 1.83 per 100,000 (95% CI 1.2–2.75). It is reasonable to note that this is the only reported HD prevalence among South Asians, but we question whether this migrant population is representative of the population of the South Asia subcontinent.

The study from Israel was taken from an abstract which reported a period prevalence between 2016 and 2018 for adult patients of 4.45 per 100,000 ⁶ In our study, we used the published paper which reported a point prevalence for adults of 4.36 per 100,000 (95% CI 3.44–5.53). ⁷

The more recent study from Oman was used in our study and had a point prevalence of 7.36 per 100,000 (95% CI 5.4–10.02). ⁸

We have redrawn the forest plot to include the 95% confidence intervals (figure 1). Given the significant heterogeneity between these populations (I² = 94.6%) we caution against presenting potentially misleading pooled prevalence estimates, instead recommending presentation of the forest plot without quantitative synthesis.

OPEN IN VIEWER

Figure 1.

Forest plot presenting HD point prevalence estimates for populations from the MENASA region. Error bars are 95% confidence intervals.