Terminal trajectory of HbA1c for 10 years supports the HbA1c paradox: a longitudinal study using Health and Retirement Study data

Abstract

Objectives: We aimed to assess the potential time-varying associations between HbA_1c and mortality, as well as the terminal trajectory of HbA_1c in the elderly to reveal the underlying mechanisms.

Design: The design is a longitudinal study using data from the Health and Retirement Study.

Setting and participants: Data were from the Health and Retirement Study. A total of 10,408 participants aged ≥50 years with available HbA_1c measurements at baseline (2006/2008) were included.

Methods: Longitudinal HbA_1c measured at 2010/2012 and 2014/2016 were collected. HbA_1c values measured three times for their associations with all-cause mortality were assessed using Cox regression and restricted cubic splines. HbA_1c terminal trajectories over 10 years before death were analyzed using linear mixed-effect models with a backward time scale.

Results: Women constitute 59.6% of the participants with a mean age of 69 years, with 3,070 decedents during the follow-up (8.9 years). The mortality rate during follow-up was 29.5%. Increased mortality risk became insignificant for the highest quartile of HbA_1c compared to the third quartile (aHR 1.148, 1.302, and 1.069 for a follow-up of 8.9, 6.5, and 3.2 years, respectively) with a shorter follow-up, while it became higher for the lowest quartile of HbA_1c (aHR 0.986, 1.068, and 1.439 for a follow-up of 8.9, 6.5, and 3.2 years, respectively). Accordingly, for both decedents with and without diabetes, an initial increase in HbA_1c was followed by an accelerating terminal decline starting 5-6 years before death.

Conclusions and implications: The time-varying association between HbA_1c and mortality mapped to the terminal trajectory in HbA_1c. High and low HbA_1c may have different clinical relationships with mortality. The HbA_1c paradox may be partially explained by reverse causation, namely, early manifestation of death.

Keywords: glycated hemoglobin (HbA1c); mortality; terminal decline; time-varying; trajectory.

Figure 1

Nonlinear association of HbA_1c measured in 2006/2008 (A), 2010/2012 (B), and 2014/2016 (C) with mortality for total sample. Data were derived from the Cox models with restricted cubic splines. Analyses were adjusted for HbA_1c; sex; age; race; marital status; education; physical activity; smoking; drinking; body mass index; history of hypertension, diabetes, and heart diseases; and multimorbidity score. HbA_1c, glycated hemoglobin; HR, hazard ratio; CI, confidence interval.

Figure 2

Trajectories of HbA_1c over 10 years before death (decedents, n = 3,070) or end of follow-up (survivors, n = 7,338). Estimated mean values were from linear mixed-effect models for total sample (A), participants without diabetes (B), and those with diabetes (C). Analyses were adjusted for sex, age at time 0, race, marital status, education, smoking, drinking, body mass index, physical activity, hypertension, diabetes [not for figure (B) or (C)], heart diseases, multimorbidity score, survival status, time terms (time and time 2), and interactions of survival status and time terms. HbA_1c, glycated hemoglobin.