The prediction of nasopharyngeal carcinoma mortality based on soil element levels in China.

The relationship between the mortality of nasopharyngeal carcinoma (NPC) and soil trace elements of 29 regions of China was investigated. A total of 29 elements (i.e., Mn, Na, K, Mg, Ca, Sr, Ba, Hg, Pb, Se, In, Yb, Lu, Th, U, Sn, Ti, Zr, Hf, Bi, Ta, Te, Br, I, As, Cr, Cu, Fe, and Zn) were considered. A hybrid strategy called genetic algorithm-partial least squares was used to screen out important elements. As a result, only six elements, i.e., Mn, Ti, Mg, K, Na, and I, were picked out, based on which, a PLS model containing two latent variables exhibited the best performance. According to whether the mortality is larger than 2/100,000 (2 × 10(-5)), all the 29 regions were divided into the low-mortality group with 23 regions and the high-mortality group with six regions. Based on the optimal PLS model, all high-mortality regions were successfully classified while only two low-mortality regions were misclassified, i.e., an accuracy of 93%, implying that the selected six elements are effective and successful for predicting the NPC mortality of a region.