作者
Jiao Pei,Mandi Li,Chen-Chou Wu,Mingqing Xu,Ting Shu,Chuandong Zhu
摘要
In a recent article, Lin et al1 analyzed correlations between cervical cancer incidence and mortality and the Human Development Index (HDI), depicted epidemic trends for cervical cancer over the 10 most recent years in 31 countries, and projected incidence and mortality rates for the next 15 years in 27 countries based on the latest data from the GLOBOCAN, Cancer Incidence in Five Continents Plus, and World Health Organization mortality databases. The article does not make clear exactly why the trending is different between the temporal and predicted incidence and mortality curves in the same time periods for Chile and Denmark, respectively. Specifically, comparing Figures 2 and 4 in their article, we see that during the period 2007-2012, the observed age-standardized incidence rate of cervical cancer for Chile in the temporal curve was inconsistent with that in the predicted curve, and so was the age-standardized mortality rate for Denmark. In addition, it may not be appropriate to use only the average annual percent changes in cervical cancer incidence and mortality for the most recent decade for some countries with significant fluctuating trends, such as Kuwait, Iceland, and Malta. A joinpoint regression program from the National Cancer Institute can be applied to describe such fluctuations, and the grid-search method can be used to find the number of significant joinpoints by the performance of several permutation tests.2 Therefore, for these countries, joinpoint regression models should be fitted, under the assumption that the change in the outcome variable is constant over each time-partition-defined joinpoint but varies among different time partitions, to identify the various trends of different time periods partitioned by joinpoints (in years). It is worth mentioning that although the average annual percent change for cervical cancer incidence in China was significantly negative, the temporal curve of incidence showed no declining trend. Within the context of the article by Lin et al,1 only Pearson correlation analyses were conducted to calculate the correlation coefficients (r) between cervical cancer incidence and mortality and HDI, which failed to reveal the relationships between cervical cancer epidemic trends and HDI. In longitudinal correlation analyses for each country, time-series data of annual cervical cancer incidence/mortality and HDI were used, so autocorrelation should be given fair consideration, which is not seen in the article. Multivariate time series models3 may be a more appropriate way to explore the dynamic relationship between incidence/mortality rates and HDI. Moreover, for cross-sectional correlation analyses, the scatter plots of morbidity/mortality rates and HDI did not show approximate linear trends. Therefore, we believe that using these linear correlation coefficients is not suitable for either cross-sectional correlation analyses or longitudinal correlation analyses. This study was supported by funding from the National Key R&D Program of China (Grant No. 2016YFC1302505-2). The authors made no disclosures.