Analysing increasing trends of Guillain-Barré Syndrome (GBS) and dengue cases in Hong Kong using meteorological data

Abstract

Background: Guillain-Barré Syndrome (GBS) is a severe paralytic neuropathy associated with virus infections such as Zika virus and Chikungunya virus. There were also case reports of dengue fever preceding GBS. With the aim to understand the mechanisms of GBS and dengue outbreaks, this ecological study investigates the relationships between GBS, dengue, meteorological factors in Hong Kong and global climatic factors from January 2000 to June 2016.

Methods: The correlations between GBS, dengue, Multivariate El Niño Southern Oscillation Index (MEI) and local meteorological data were explored by Spearman's Rank correlations and cross-correlations. Three Poisson regression models were fitted to identify non-linear associations among GBS, dengue and MEI. Cross wavelet analyses were applied to infer potential non-stationary oscillating associations among GBS, dengue and MEI.

Findings and conclusion: We report a substantial increasing of local GBS and dengue cases (mainly imported) in recent year in Hong Kong. The seasonalities of GBS and dengue are different, in particular, GBS is low while dengue is high in the summer. We found weak but significant correlations between GBS and local meteorological factors. MEI could explain over 17% of dengue's variations based on Poisson regression analyses. We report a possible non-stationary oscillating association between dengue fever and GBS cases in Hong Kong. This study has led to an improved understanding about the timing and ecological relationships between MEI, GBS and dengue.

Fig 1. Trends and seasonality of GBS and dengue cases (scaled by number of population in Hong Kong).

Panel (a), Annual cases of GBS and dengue cases show a sudden increase in recent years. Panel (b), Monthly cases of GBS and dengue cases. The grey shaded area of panel (a,b) is MEI. Panel (c), Boxplot of GBS cases per day. Panel (d), Boxplot of dengue cases per day.

Fig 2. Cross-correlation coefficient results among dengue, MEI and GBS.

Panel (a) shows cross-correlation coefficients between dengue and MEI. Panel (b) shows cross-correlation coefficients between GBS and MEI. Panel (c) shows the cross-correlation coefficients between GBS and dengue. In all three panels, we consider time lags from 0 to 11 months. In this plot, the lag (namely l) of, for example, X vs. Y represents that Y lags l month(s) behind X (i.e., X_t+l is corresponding to Y_t). The vertical black bars are 95% CI. The squares in the middle are the mean estimate of cross-correlation coefficients. The blue dotted line is p-value of each cross-correlation coefficient. The horizontal dashed light blue lines on all panels indicate the 0.05 significance level.

Fig 3. Poisson regression results among dengue, MEI and GBS.

Panel (a) shows regression coefficients between dengue and MEI, panel (b) shows regression coefficients between GBS and MEI and panel (c) shows regression coefficients between GBS and dengue. In all three panels, we consider time lags from 0 to 11 months. The vertical black bars are 95% confidence intervals and the squares in the middle are the mean estimate of regression coefficients. The blue dotted line is p-value of each correlation coefficient. The horizontal dashed light blue lines on all panels indicate the 0.05 significance level. The red dotted line is R² of each regression coefficient. The horizontal dashed pink lines represent the median level of all R².

Fig 4. Wavelet analyses of MEI, dengue and GBS from 2000-2016 in panels (a), (b), and (c).

(i) Left panels, mean spectrum plots at 5% (blue) and 10% (red) thresholds. (ii) the right panels are the wavelet power spectrum contour plots. The colour scheme is from blue to red, which represents increasing wavelet power level. The white line represents the 95% CI and the white shaded region is due to the edge effects.

Fig 5. Wavelet coherence and phase plots of dengue and GBS data from 2000-15 in Hong Kong.

Panel (a) is dengue time series with peaks shaded in grey. Panel (b) are phase plots of dengue and GBS. Data are shown in red and blue, and the black dashed line shows phase difference. Panel (c) shows cross wavelet average power level and wavelet coherence plots of dengue and GBS, which shares the same plot code as Fig 4. The horizontal axis labels of 5, 10 and 15 represent year 2005, 2010 and 2015.