I'd suggest such epidemiological measurements are generally pretty noisy - especially so in light of the different definitions of diabetes (and changes in those definitions): FBG, HbA1c, OGTT, ... .
That said, I do expect that there will be significant growth in diabetes in China etc. Growth from 10 to 20+ pct in 15 years seems entirely plausible. But trying to measure a 3 pct difference across two different epi studies not performed to the same protocol seems not highly likely to be accurate.