Background/Aims: Season has been associated with serum cholesterol, A1c, and affective disorders. The mechanisms acting behind these endpoints have been hypothesized to be latitude, temperature, and holiday overeating. We observed similar seasonal patterns and amplitude variation by latitude, and correlate these measures with mean daylight hours and temperatures to elucidate underlying relationships.
Methods: We used data from The Journey for Control of Diabetes: The IDEA Study (described elsewhere). The study included adults with type 2 diabetes 1 year or greater duration and suboptimal glycemic control (A1c =7%). Subjects were recruited from health care systems in Minneapolis (MIN), and Albuquerque (ABQ), after being identified using claims data; unenrolled subjects were included from ABQ. Data on hours of daylight and average temperatures were from the National Weather Service. We produced LOESS plots for A1c trends at both sites.
Results: We observed no statistically significant differences in the distribution of total test result counts by season. All 4 seasons and study site were significant predictors for A1C in full regression models; subsequent models were site-stratified. Overall mean A1c’s were 8.1 for MIN and 7.6 for ABQ. Using winter as the referent season, we observed differences in amplitude of seasonal change across the 2 sites with spring, summer, and fall values for MIN/ABQ parameter estimates of -0.32/-0.09, -0.31/-0.14, -0.18/-0.08 (p<.0001). In site-controlled models with terms for daylight and temperature, only temperature remained a significant predictor (p=.004); the stratified model for ABQ showed a retrograde estimate for light (0.02, p=0.31) which may be explained by ABQ’s later peak A1c date of March 1, compared to December 31 for MIN.
Conclusions: The trends observed in this study reflect those published previously and indicate a strong seasonal element to A1c scores with latitudinal amplification. LOESS plots reveal two distinct “humps” starting at the solstice and ending early October where the slope of the line flattens for ABQ, and reverses for MIN from the equinoctial peak. ABQ’s late A1c peak is also close to the spring equinox, suggesting that perhaps 12-hour days act as an independent signal above and beyond the greater trend of increasing or decreasing daylight.