Published online Oct 15, 2026. doi: 10.4239/wjd.124090
Revised: July 21, 2026
Accepted: September 8, 2026
Published online: October 15, 2026
Processing time: 128 Days and 23.8 Hours
Major guidelines recommend active management of prediabetes, including drug treatment for many individuals, a strategy that could impose a significant burden on primary care systems. Previous studies showed that some individuals with prediabetes could regress to normoglycemia without treatment, but whether this transition was associated with reduced risks of diabetic complications, the ulti
To investigate the rates of regression to normoglycemia and progression to dia
Participants aged ≥ 45 years with prediabetes in 2011 and untreated during 2011-2015 were identified from The China Health and Retirement Longitudinal Study. The proportions of regression to normoglycemia, persistent prediabetes and pro
A total of 2461 participants with prediabetes in 2011 were included, of whom 21.5% regressed to normoglycemia and 15.2% progressed to diabetes by 2015. Over the subsequent five years, the progression group showed non-statistically significant higher risks of CVD [adjusted odds ratio (aOR) = 0.97, 95%CI: 0.62-1.52] and kidney disease (aOR = 2.01, 95%CI: 0.90-4.48). For all-cause mortality, the progression group showed a statistically significantly higher risk (aOR = 2.36, 95%CI: 1.18-4.72), corresponding to an attributable risk of 4.35% in that group and 0.66% of the total prediabetes population. Regression to normoglycemia was not statistically significantly associated with risks for the three outcomes during follow-up. Various sensitivity analyses yielded consistent results.
Given the modest five-year absolute risk of diabetic complications attributable to progression, prediabetes ma
Core Tip: In the absence of drug treatment over four years, the majority (approximately 85%) of individuals with prediabetes in this study remained stable or regressed to normoglycemia, while 15.2% progressed to diabetes. Progression to diabetes was associated with an increased risk of all-cause mortality in the subsequent five years, but the absolute risk attributable to progression was modest (4.35% of the progression group or 0.66% of the total prediabetes population). The associations between progression and the risks of cardiovascular and kidney disease within 5 years were not statistically significant.
- Citation: Zhang JY, Xu YH, Wang ST, Huang Y, Zheng WX, Xiao Y, Yang ZY. Regression to normoglycemia, progression to diabetes, and their association with health outcomes in Chinese adults with untreated prediabetes. World J Diabetes 2026; 17(10): 124090
- URL: https://www.wjgnet.com/1948-9358/full/v17/i10/124090.htm
- DOI: https://dx.doi.org/10.4239/wjd.124090
Prediabetes is a condition in which blood glucose levels are higher than normal but not high enough for a diagnosis of type 2 diabetes. The American Diabetes Association defines prediabetes as the presence of fasting plasma glucose (FPG) of 100-125 mg/dL (5.6-6.9 mmol/L), 2-hour post-load plasma glucose (2hPG) of 140-199 mg/dL (7.8-11.0 mmol/L), or glycated hemoglobin (HbA1c) of 5.7%-6.4% in individuals who have not been diagnosed with diabetes[1]. According to these criteria, 35.2% of the general adult population in China and 38.0% in the United States were affected by prediabetes[2,3]. In clinical practice, FPG and HbA1c are more commonly used than 2hPG due to their convenience. Previous studies showed that these two measures could identify over 90% of prediabetes[4].
As prediabetes is associated with modestly increased risks of diabetic complications such as cardiovascular disease (CVD), kidney disease, and mortality[5], some clinical guidelines recommend aggressive treatment of prediabetes. For instance, the American Diabetes Association suggests that metformin should be considered in those aged 25-59 years with body mass index (BMI) ≥ 35 kg/m2, FPG ≥ 6 mmol/L, HbA1c ≥ 6.0%, or prior gestational diabetes mellitus[6]. The Chinese Expert Consensus has gone even further to recommend those meeting any of the over 15 specified criteria, which represents a substantial proportion of those with prediabetes, should be considered for drug therapy[7].
However, previous studies showed that prediabetes was a heterogeneous condition, with the potential of remaining in the prediabetic state or even regression to normoglycemia in the absence of drug treatment[8]. Furthermore, randomized trials showed that drug therapy was not better than lifestyle intervention in reducing the risks of diabetic complications (the ultimate goal of glucose management)[9], which is explicitly recognized by current guidelines[6,7]. To develop optimal management strategies of prediabetes, it is highly important to understand its natural history, including the transitions (e.g., regression or progression) of glucose level per se and the risks of diabetic complications associated with these transitions.
Although many studies have investigated the transition rates of prediabetes[8], few have examined the transitions in different subgroups as well as their associations with diabetic complications. Because of this, the United States Preventive Services Task Force explicitly called for more research on the natural history of prediabetes in its recent guidelines[10]. Against this background, the present study aimed to (1) Investigate the rates of regression to normoglycemia and pro
The participants in this study were selected from The China Health and Retirement Longitudinal Study (CHARLS), which is an ongoing, nationally representative, open cohort of individuals aged ≥ 45 years and their spouses. The protocol of CHARLS has been published elsewhere in detail[11]. Briefly, approximately 17708 participants were recruited from 450 villages or urban communities across 28 provinces in mainland China using the multistage stratified probability proportional-to-size sampling method. Analysis showed that these participants were representative of the national population of the same age range[11].
The baseline survey of CHARLS was conducted in 2011, and follow-up surveys were conducted every two years (i.e., one wave) thereafter through face-to-face, computer-assisted personal interview. Questionnaire surveys on social, eco
The protocol of CHARLS was approved by the ethical review committee (institutional review board) of Peking Uni
As the first objective of this study, which was to quantify the rates of prediabetes transitions over time, wave 1 (2011) was treated as baseline, and data from wave 3 (2015) were used to ascertain prediabetes transitions, e.g., regression or pro
The 2hPG level was not used to define prediabetes in this study, as the CHARLS did not conduct the oral glucose tolerance test for participants. According to previous studies, isolated 2hPG elevation accounted for approximately 8% of all prediabetes[4]. There has been no consensus on the optimal interval between two glycemic measurements for ascertaining prediabetes transitions. A previous study of 19255 prediabetic individuals found that the mean duration of prediabetes before it progressed to diabetes (if ever) ranged from 3.1 years to 4.8 years, depending on the age group[13]. This suggests that a follow-up interval of 3-5 years should be able to capture clinically significant transitions for most participants.
As the second objective of this study, which was to preliminarily examine the association of prediabetes transitions with diabetic complications, the participants were further required to be free of CVD and chronic kidney disease by 2015 (when prediabetes transitions were ascertained) and have follow-up data on CVD, chronic kidney disease and/or mortality in either wave 4 (2018) or wave 5 (2020), in addition to the three criteria listed above. Other diabetic complications such as retinopathy were not examined in this study, as the CHARLS did not collect data on these parameters.
For blood sample collection, participants were asked to fast overnight and blood was collected in the morning. The fasting status of each participant was recorded. Three tubes of venous blood were collected from each participant using a standard protocol and ethylenediaminetetraacetic acid-K2 anticoagulant vacuum tubes. FPG was analyzed using an enzymatic colorimetric test. HbA1c was analyzed by boronate affinity high performance liquid chromatography. High-sensitivity C-reactive protein was also measured from frozen plasma samples and analyzed by an immunoturbidimetric assay[11].
Data on demographics, lifestyle factors and health conditions were collected by uniformly trained interviewers using a questionnaire, and included age, gender, marital status, education level, total household income (< 10000; ≥ 10000 Chinese yuan), health insurance, smoking status, drinking status, physical activity, self-rated general health status, history of glucose-lowering therapy, antihypertensive therapy and lipid-lowering therapy, and presence of hypertension, dyslipidemia, CVD, and other comorbidities. No data on dietary factors were collected in the CHARLS. Current smokers were defined as those who were currently smoking. Current alcohol drinkers were defined as those who ‘drank more than monthly in the last year’[14]. Participants were categorized as “active” if they engaged in vigorous or moderate physical activity for at least 30 minutes on three or more days per week[15]. Existing diseases were determined by asking the participants whether they had ever been told by a doctor that they had specific diseases. Height and weight were measured by trained staff. BMI was calculated as kg/m2 and categorized into underweight/normal weight (BMI < 24 kg/m2) or overweight/obesity (BMI ≥ 24 kg/m2) according to the current Chinese standard[16].
For the first research question, the outcomes of interest included regression to normoglycemia and progression to diabetes. Normoglycemia was defined as an FPG level < 5.6 mmol/L (100 mg/dL) and an HbA1c level < 5.7% at the follow-up measurement in 2015 (wave 3). Diabetes was defined as an FPG level ≥ 7.0 mmol/L (126 mg/dL), an HbA1c level ≥ 6.5%, or both at the follow-up measurement in 2015 (wave 3)[1].
For the second research question, the outcomes of interest included CVD, kidney disease and all-cause mortality. CVD and kidney disease were ascertained by the respondent’s answer to the questions regarding whether or not they were told by a doctor that they had a specific condition in wave 4 (2018) and wave 5 (2020) of the CHARLS. Mortality was det
Baseline characteristics of participants are described as mean (SD) or median (interquartile range) for continuous variables and frequency with percentage for binary or categorical variables. The included participants were divided into three groups according to their prediabetes transitions: (1) Regression to normoglycemia; (2) Persistent prediabetes; and (3) Progression to diabetes. The proportions of group 1and group 3 out of all the included participants were calculated. The 95%CI for proportions were calculated using the asymptotic normal approximation of the binomial distribution (Wald method). The proportions across different strata defined by major risk factors of diabetes (e.g., age, gender, BMI) were also calculated.
Baseline characteristics were compared across the three groups. Characteristics of the included participants (n = 2461) and those who had prediabetes in 2011 but were excluded (n = 1867) due to unclear glycemic status in 2015 for various reasons (e.g., death, lost to follow-up, and non-fasting blood samples) were compared to facilitate the assessment of potential selection bias. In all these comparisons, one-way analyses of variance were performed for continuous variables and χ² tests for categorical variables.
As the included participants differed from those excluded in some characteristics, efforts were made to predict the excluded participants’ glycemic status in 2015 according to their baseline characteristics in 2011. In the predictive equation, the dependent variable was the glycemic status in 2015 (regression to normoglycemia, persistent prediabetes, or progression to diabetes), while independent variables were baseline characteristics. The coefficient for each characteristic (covariate) was derived from data of the 2461 included participants through logistic regression. Classification of the predicted outcomes (i.e., which category a participant belonged to) followed the maximum probability rule[17,18]. Based on predicted outcomes of the excluded participants, we conducted a pooled analysis of the included and excluded participants and re-calculated the proportions of regression to normoglycemia and progression to diabetes, respectively. Comparison of these “new” proportions with those from the main analysis could help assess the magnitude and direction of potential selection bias.
Binary logistic regression model was adopted to identify the factors associated with regression to normoglycemia and progression to diabetes, respectively, using persistent prediabetes as the reference group. The following baseline factors collected in 2011 were adjusted for in the regression model: (1) Age; (2) Gender; (3) Marital status; (4) Educational level; (5) Total household income; (6) Health insurance; (7) Smoking status; (8) Drinking status; (9) Physical activity; (10) BMI; (11) Self-reported current general health status; (12) Existing diseases; and (13) High-sensitivity C-reactive protein level. Among the existing diseases, hypertension, dyslipidemia, and CVD were each included in the model as separate covariates, while the other five diseases were integrated into a covariate named “number of comorbidities” (categorized as “0” or “≥ 1”) for inclusion into the regression model. The association between each covariate and outcome was measured by odds ratio (OR) and 95%CI.
The binary logistic regression model was adopted to preliminarily examine associations between prediabetes transitions with CVD, kidney disease and all-cause mortality, respectively. The Cox proportional hazards regression model was used in sensitivity analyses only and not in the primary analyses, as in the CHARLS, onset of the three outcomes was only recorded at the year level, with no data available on the month or day, and the length of follow-up was modest (median 5 years), which significantly limited the statistical power and precision of the Cox regression analysis. In the logistic regression analysis, prediabetes transition was treated as a categorical variable, with persistent prediabetes as the reference group. Age, gender, marital status, educational level, total household income, health insurance, smoking status, drinking status, physical activity, self-rated general health status, hypertension, dyslipidemia, number of comorbidities, BMI, antihypertensive therapy, lipid-lowering therapy and glucose-lowering therapy in 2015 (when prediabetes transitions were ascertained) were adjusted for in the multivariable regression model. A crude OR and an adjusted OR (aOR) along with their 95%CIs were estimated for each variable. The covariates data collected in 2015 were adjusted for because, in this analysis, the exposure was the “type of prediabetes transition”, which could not be ascertained until 2015. The covariates data collected in 2011 were adjusted for in sensitivity analyses to examine the robustness of results from our primary analyses.
In all the regression models, the missing-indicator method was adopted to handle missing data of covariates[19]. A covariate with missing values was treated as a multiple categorical variable and all missing values of the covariate were assigned to a newly created category. By doing so, no imputation (which was based on various assumptions) was required, and no participants would be excluded due to missing data (thus maintaining the statistical power of analysis). To examine the robustness of results produced by this method, we additionally performed a sensitivity analysis using the multiple imputation method.
Several sensitivity analyses were conducted to assess the robustness of the primary findings. First, we conducted a multinomial logistic regression model to identify the factors associated with prediabetes transitions. This model simultaneously estimated the odds of all three outcomes (regression to normoglycemia, persistent prediabetes, and progression to diabetes) within a single framework, using persistent prediabetes as the reference group, which could theoretically reduce the chance of false-positive findings due to multiple testing in the main analyses. All covariates included in the primary analysis were adjusted for in this model. Second, we redefined the baseline prediabetes population using two alternative criteria: (1) FPG-defined prediabetes: FPG of 5.6-6.9 mmol/L (100-125 mg/dL) and HbA1c ≤ 6.4%; and (2) HbA1c-defined prediabetes: FPG of ≤ 6.9 mmol/L (125 mg/dL) and HbA1c of 5.7%-6.4%. All primary analyses were repeated separately within each of these two alternatively defined cohorts, using the same statistical methodologies and covariate adjustment strategies described above. Third, to address the concern regarding potential bias from the missing-indicator method for dealing with missing data, we performed sensitivity analyses using the multiple imputation method. Multivariate imputation by chained equations was implemented using the “mice” package in R, generating 5 imputed datasets with 20 iterations each. The imputation model included all covariates from the main analysis as well as the outcome variables to improve imputation accuracy. Variables with a missing proportion exceeding 30% (e.g., household income) were not imputed but retained with their original category structure. After imputation, all regression models were replicated exactly as in the primary analysis. Estimates from the 5 imputed datasets were pooled using Rubin’s rules. Fourth, in the analyses of the associations between prediabetes transitions and health outcomes, the covariates collected in 2011 (as opposed to 2015 in primary analyses) were adjusted for to address the concern that 2015 metabolic covariates might lie on the causal pathway (i.e., act as potential mediators) from prediabetes transitions to health outcome. Fifth, instead of using the binary logistic regression model (as performed in the primary analyses), we conducted Cox proportional hazards regression to preliminarily examine associations between prediabetes transitions with CVD, kidney disease and all-cause mortality, respectively. Given the aforementioned limitations in the data (i.e., lack of the exact month and/or date of outcome occurrence), we estimated the timing of outcome events using a standardized midpoint imputation approach based on the survey intervals. Specifically, for participants with a known year of diagnosis, the event time was assigned to July 1 of that year (the midpoint of the calendar year). For those with an unknown diagnosis year, event times were imputed as the midpoints of the respective inter-survey intervals (i.e., the midpoint between the 2015 and 2018 surveys, or between the 2018 and 2020 surveys). Participants who remained event-free until the end of the study were censored at the 2020 survey date (September 2020), while those lost to follow-up were censored at the midpoints of the corresponding survey waves. Crude and adjusted hazard ratios (HRs) with 95%CIs were estimated using the same covariate set as in the primary logistic regression models. Furthermore, to address the potential bias introduced by competing events, we performed an additional sensitivity analysis using the Fine-Gray subdistribution hazard model for non-fatal outcomes (CVD and kidney disease). In these models, all-cause mortality was treated as a competing risk event. Subdistribution HRs with 95%CIs were estimated. The adjustment set of covariates and the handling of missing data were kept consistent with the primary logistic regression models. All analyses were conducted using R version 4.3.1 and a P value < 0.05 was considered statistically significant.
A total of 11417 participants aged 45 years or above with blood test results were recruited in the baseline survey of the CHARLS in 2011, and 4328 of them had prediabetes defined by FPG and HbA1c. By 2015, 58 (1.3%) participants had died, 493 (11.4%) were lost to follow-up, 910 (21.0%) had no follow-up measurement of FPG or HbA1c, and 406 (9.4%) were not fasting, leaving 2461 with valid data on glycemic status for inclusion in this study (Figure 1). The included participants’ characteristics are shown in Table 1. Their ages ranged from 45 to 90 years, with a mean of 59 years. Males accounted for 46.0% of the participants. The proportions of participants with overweight/obesity, hypertension, dyslipidemia and CVD were 44.0%, 29.0%, 11.2% and 14.9%, respectively. The baseline FPG and HbA1c levels were 6.0 mmol/L and 5.2%, respectively.
| Characteristics | Total | Regression to normoglycemia | Progression to diabetes | Persistent prediabetes | P value |
| Number of participants | 2461 | 530 (21.5) | 373 (15.2) | 1558 (63.3) | |
| Baseline FPG (mmol/L), mean (SD) | 6.0 (0.4) | 5.9 (0.3) | 6.1 (0.4) | 6.0 (0.4) | < 0.001 |
| Baseline HbA1c (%), mean (SD) | 5.2 (0.4) | 5.0 (0.4) | 5.4 (0.4) | 5.2 (0.4) | < 0.001 |
| Age (years), mean (SD) | 59.0 (8.6) | 57.7 (8.6) | 60.1 (8.8) | 59.2 (8.4) | < 0.001 |
| Age category | |||||
| 45-59 years | 1375 (55.9) | 331 (62.5) | 191 (51.2) | 853 (54.7) | 0.001 |
| ≥ 60 years | 1086 (44.1) | 199 (37.5) | 182 (48.8) | 705 (45.3) | |
| Males | 1133 (46.0) | 283 (53.4) | 145 (38.9) | 705 (45.3) | < 0.001 |
| Married | 2080 (84.5) | 457 (86.2) | 306 (82.0) | 1317 (84.5) | 0.230 |
| Education level: High school or above | 217 (8.8) | 60 (11.3) | 28 (7.5) | 129 (8.3) | 0.064 |
| Total household income (CNY) | |||||
| < 10000 | 777 (50.3) | 171 (51.5) | 128 (51.6) | 478 (49.5) | 0.744 |
| ≥ 10000 | 768 (49.7) | 161 (48.5) | 120 (48.4) | 487 (50.5) | |
| Health insurance | |||||
| No | 114 (4.6) | 32 (6.0) | 13 (3.5) | 69 (4.4) | 0.165 |
| Yes | 2339 (95.4) | 497 (94.0) | 359 (96.5) | 1483 (95.6) | |
| Current smokers | 678 (27.9) | 152 (28.8) | 94 (25.7) | 432 (28.1) | 0.563 |
| Current alcohol drinkers | 454 (18.5) | 106 (20.0) | 57 (15.3) | 291 (18.7) | 0.19 |
| Physical activity | |||||
| Active | 958 (38.9) | 198 (37.4) | 136 (36.5) | 624 (40.1) | 0.312 |
| Other | 1503 (61.1) | 332 (62.6) | 237 (63.5) | 934 (59.9) | |
| BMI (kg/m2) | |||||
| < 24 | 1206 (56.0) | 297 (62.9) | 135 (40.5) | 774 (57.4) | < 0.001 |
| ≥ 24 (overweight/obese) | 947 (44.0) | 175 (37.1) | 198 (59.5) | 574 (42.6) | |
| Self-rated current health status | |||||
| Poor/very poor | 647 (26.3) | 138 (26.0) | 118 (31.7) | 391 (25.1) | 0.034 |
| Very good/good/fair | 1811 (73.7) | 392 (74.0) | 254 (68.3) | 1165 (74.9) | |
| Disease status | |||||
| Hypertension | 712 (29.0) | 147 (27.9) | 156 (41.8) | 409 (26.3) | < 0.001 |
| Dyslipidemia | 270 (11.2) | 48 (9.2) | 72 (19.6) | 150 (9.8) | < 0.001 |
| CVD | 365 (14.9) | 72 (13.6) | 83 (22.3) | 210 (13.6) | < 0.001 |
| Number of comorbidities | |||||
| 0 | 1588 (64.5) | 330 (62.3) | 225 (60.3) | 1033 (66.3) | 0.045 |
| ≥ 1 | 873 (35.5) | 200 (37.7) | 148 (39.7) | 525 (33.7) | |
| High-sensitivity C-reactive protein (mg/L), mean (SD) | 2.5 (6.8) | 1.9 (4.3) | 3.4 (7.6) | 2.6 (7.3) | 0.008 |
| Baseline FPG category | |||||
| 5.6-6.0 mmol/L | 1310 (55.1) | 338 (65.0) | 139 (38.7) | 833 (55.6) | < 0.001 |
| 6.1-6.9 mmol/L | 1067 (44.9) | 182 (35.0) | 220 (61.3) | 665 (44.4) | |
| Baseline HbA1c category | |||||
| 5.7%-5.9% | 92 (35.0) | 9 (52.9) | 25 (31.6) | 58 (34.7) | 0.246 |
| 6.0%-6.4% | 171 (65.0) | 8 (47.1) | 54 (68.4) | 109 (65.3) |
At baseline, 2377 (96.6%) had FPG-defined prediabetes (5.6-6.9 mmol/L), 263 (10.7%) had HbA1c-defined prediabetes (5.7%-6.4%), and 179 (7.3%) met both criteria. During a median follow-up of 4 years, 530 participants (21.5%, 95%CI: 19.9%-23.2%) regressed to normoglycemia, 1558 (63.3%, 95%CI: 61.4%-65.2%) had persistent prediabetes, and 373 (15.2%, 95%CI: 13.7%-16.6%) progressed to diabetes. Proportions of the three transition patterns in various subgroups defined by baseline characteristics are shown in Supplementary Table 1. Among the 2377 participants with FPG-defined prediabetes, 520 (21.9%, 95%CI: 20.2%-23.6%) regressed to normoglycemia, 1498 (63.0%, 95%CI: 61.1%-65.0%) had persistent pre
As shown in Supplementary Table 2, the 1867 participants who had prediabetes at baseline but were excluded due to unclear glycemic status in 2015 differed from the included participants in terms of baseline FPG, age, gender, marriage status, educational level, health insurance, physical activity and dyslipidemia. As predicted using the methods described above, four of the 1809 who were still alive in 2015 might have regressed to normoglycemia, 1796 remained in the pre
Table 2 shows the results of multivariable logistic regression analyses of the associations between baseline characteristics and prediabetes transitions. Briefly, hypertension (aOR = 1.56), dyslipidemia (aOR = 1.72) and overweight/obesity (aOR = 1.78) were statistically significantly associated with a greater probability of progression to diabetes, while younger age [< 60 years vs ≥ 60 years (aOR = 1/0.67 = 1.49)], male (aOR = 1.58) and absence of overweight/obesity (aOR = 1/0.78 = 1.28) were statistically significantly associated with a greater probability of regression to normoglycemia. Multinomial logistic regression analysis yielded consistent results (Supplementary Table 3). When confined to the FPG-defined prediabetes cohort, the results were also similar (Supplementary Table 4). The results based on the HbA1c-defined prediabetes cohort (n = 263) were not presented due to the limited sample size and insufficient outcome events (17 for regression; 79 for progression) relative to the number of covariates, which precluded a robust analysis. The sensitivity analysis using multiple imputations to deal with missing covariates yielded consistent results (Supplementary Table 5).
| Factors | Regression to normoglycemia | Progression to diabetes | ||
| Adjusted OR (95%CI)a | P value | Adjusted OR (95%CI)a | P value | |
| Age (≥ 60 vs < 60) | 0.67 (0.54-0.83) | < 0.001 | 1.10 (0.86-1.40) | 0.468 |
| Male vs female | 1.58 (1.23-2.03) | < 0.001 | 0.76 (0.55-1.03) | 0.079 |
| High school or above vs middle school or below | 1.29 (0.92-1.81) | 0.137 | 0.95 (0.60-1.48) | 0.839 |
| Married vs other | 1.04 (0.78-1.41) | 0.780 | 0.84 (0.61-1.16) | 0.281 |
| Total household income (≥ 10000 vs < 10000, CNY) | 0.89 (0.69-1.16) | 0.390 | 0.87 (0.65-1.17) | 0.352 |
| Health insurance (yes vs no) | 0.71 (0.46-1.12) | 0.128 | 1.19 (0.66-2.30) | 0.585 |
| Current smoker vs others | 0.79 (0.60-1.04) | 0.089 | 1.29 (0.93-1.81) | 0.129 |
| Current alcohol drinker vs other | 0.94 (0.71-1.24) | 0.669 | 0.99 (0.69-1.41) | 0.971 |
| Active physical activity vs other | 0.89 (0.72-1.10) | 0.282 | 0.83 (0.65-1.06) | 0.135 |
| Disease status (yes vs no) | ||||
| Hypertension | 1.19 (0.94-1.50) | 0.156 | 1.56 (1.21-2.02) | 0.001 |
| CVD | 0.97 (0.71-1.32) | 0.865 | 1.27 (0.93-1.74) | 0.132 |
| Dyslipidemia | 0.96 (0.66-1.36) | 0.810 | 1.72 (1.23-2.40) | 0.001 |
| Number of comorbidities (≥ 1 vs 0) | 1.16 (0.94-1.44) | 0.166 | 1.23 (0.96-1.58) | 0.101 |
| BMI (≥ 24 kg/m2 vs < 24 kg/m2) | 0.78 (0.61-0.98) | 0.033 | 1.78 (1.36-2.32) | < 0.001 |
| Self-reported general health (very good/good/fair vs poor/very poor) | 0.95 (0.75-1.22) | 0.693 | 0.89 (0.68-1.18) | 0.414 |
| High-sensitivity C-reactive protein (every 1-mg/L increase) | 0.98 (0.95-1.00) | 0.075 | 1.01 (0.99-1.02) | 0.209 |
Among the 2461 participants whose prediabetes transitions could be ascertained in 2015, 800 had pre-existing CVD or kidney disease by 2015 (467 at baseline, 333 during 2011-2015), leaving 1661 participants in the analyses to investigate the associations of prediabetes transitions with CVD (n = 1581), kidney disease (n = 1576), and mortality (n = 1661) (Figure 1). During a median follow-up of 5 years from 2015 to 2020, the cumulative risks of CVD, kidney disease, and mortality were 14.6% (95%CI: 12.9%-16.4%), 3.7% (95%CI: 2.8%-4.6%), and 3.7% (95%CI: 2.8%-4.6%), respectively. In the progression group, 16.1% (34/210) received antidiabetic treatment. As shown in Table 3, for both CVD and kidney disease, the progression group showed a significantly higher risk in unadjusted models (CVD: OR = 1.57, 95%CI: 1.06-2.31; kidney disease: OR = 2.24, 95%CI: 1.13-4.46) compared to the persistent group. However, these associations became statistically insignificant after adjustment for covariates [CVD (aOR = 0.97, 95%CI: 0.62-1.52); kidney disease (aOR = 2.01, 95%CI: 0.90-4.48)]. For all-cause mortality, the progression group showed a statistically significantly higher risk in both unadjusted and adjusted models (aOR = 2.36, 95%CI: 1.18-4.72). The regression group was not statistically significantly different from the persistent group in terms of the risks for the three outcomes in either unadjusted or adjusted models (Table 3). In sensitivity analyses based on the FPG-defined prediabetes cohort, the results were consistent with those from the primary analyses (Supplementary Table 6). Sensitivity analyses based on the HbA1c-defined prediabetes cohort was limited by the small sample size and should be regarded as exploratory only (the association with mortality could not be estimated due to the small number of events, Supplementary Table 7). Sensitivity analyses adjusting for covariates collected in 2011 or using multiple imputations to deal with missing covariates yielded consistent results with the primary analyses (Supplementary Table 8). The results of sensitivity analyses using Cox regression and competing-risk models (for non-fatal outcomes) were also consistent with those from the primary analyses using logistic regression (Supplementary Table 9). Survival curves for CVD, kidney disease, and all-cause mortality during the 2015-2020 follow-up, stratified by prediabetes transition groups (Supplementary Figure 1).
| Outcome/variables | n | N | Cumulative risk (95%CI) | Crude OR (95%CI) | P value | Adjusted OR (95%CI)a | P value |
| CVD | |||||||
| Regression to normoglycemia | 43 | 346 | 12.4 (9.0-15.9) | 0.86 (0.60-1.23) | 0.403 | 0.93 (0.64-1.36) | 0.716 |
| Persistent prediabetes | 148 | 1041 | 14.2 (12.1-16.3) | 1 (reference) | 1 (reference) | ||
| Progression to diabetes | 40 | 194 | 20.6 (14.9-26.3) | 1.57 (1.06-2.31) | 0.024 | 0.97 (0.62-1.52) | 0.899 |
| Kidney disease | |||||||
| Regression to normoglycemia | 16 | 346 | 4.6 (2.4-6.8) | 1.63 (0.88-3.03) | 0.123 | 1.47 (0.78-2.78) | 0.236 |
| Persistent prediabetes | 30 | 1038 | 2.9 (1.9-3.9) | 1 (reference) | 1 (reference) | ||
| Progression to diabetes | 12 | 192 | 6.2 (2.8-9.7) | 2.24 (1.13-4.46) | 0.022 | 2.01 (0.90-4.48) | 0.088 |
| Mortality | |||||||
| Regression to normoglycemia | 12 | 359 | 3.3 (1.5-5.2) | 1.04 (0.54-2.03) | 0.898 | 1.12 (0.56-2.23) | 0.747 |
| Persistent prediabetes | 35 | 1092 | 3.2 (2.2-4.2) | 1 (reference) | 1 (reference) | ||
| Progression to diabetes | 14 | 210 | 6.7 (3.3-10.0) | 2.16 (1.14-4.08) | 0.018 | 2.36 (1.18-4.72) | 0.015 |
In this cohort study of 2461 Chinese middle-aged and elderly participants with prediabetes, 21.5% regressed to normo
The prevalence of prediabetes among the CHARLS participants was 37.9% (4328/11417, Figure 1), similar to the 35.2% reported by larger-scale nationally representative surveys in China. The annualized rate of progression observed in this study (15.2%/43.8%) was in the range reported by previous studies (3.5%-10%)[8]. Notably, the majority of prediabetes patients did not progress to diabetes during four years, and the rate of regression to normoglycemia (21.5%) was 1.4 times that of progression (15.2%), consistent with the findings of previous studies[20]. In many other cohort studies, the rates of regression to normoglycemia were even higher, ranging from 33% to 59% within one to five years follow-up, and from 17% to 42% for 6 to 11 years of follow-up[8].
Our finding that regression to normoglycemia was not statistically significantly associated with lower risks of CVD, kidney disease or mortality aligned with previous studies[21-23]. One potential reason for these null findings is the relatively young population (mean age 35-59 years) and short follow-up (all ≤ 10 years). Recently, the long-term follow-up data of two trials (median follow-up 20 years and 30 years, respectively) showed that regression to normoglycemia was associated with a significantly reduced risk of cardiovascular death or hospitalization for heart failure (HRs = 0.41 and HRs = 0.49)[24]. The lack of statistical significance of the risks of CVD and kidney disease in the progression group might be due to the relatively short follow-up (5 years) and small sample size (< 200) of the progression group. Both issues could have led to the small number of outcome events and insufficient statistical power. Previous studies found that the progression group had higher risks of CVD and mortality as compared with the persistent prediabetes group alone or plus the regression to normoglycemia group[21,25]. However, the relative increase in risks was modest, mostly between 10%-20%, corresponding to even smaller absolute risks.
Possible explanations for the regression to normoglycemia observed in this study as well as many other previous studies included “regression to the mean” and effects of lifestyle modifications (e.g., weight reduction) of participants. To reduce the impact of regression to the mean, repeated measurements of blood glucose levels (as opposed to current practice that requires only one test) may be considered when defining prediabetes. Regardless of the driving factors, however, it was clear that the regression to normoglycemia in this study was achieved in the absence of drug treatment for prediabetes, suggesting that a less aggressive strategy of management may be suitable for many people.
Our findings on the association between prediabetes progression with health outcomes (i.e., diabetic complications) further support this argument. Based on the relative risk of mortality (aOR = 2.36) in the progression group and the risk in the persistent group, the absolute risk of mortality attributable to progression to diabetes was estimated to be 4.35%, respectively. As the progression group accounted for 15.2% of all prediabetes, it can be further estimated that the participants who died due to progression accounted for 0.66% of all individuals with prediabetes, respectively. These findings suggest that the majority of individuals with prediabetes would unlikely benefit from aggressive interventions (including drug treatment) in the first five years in terms of preventing diabetic complications. This is consistent with randomized trials which evaluated the effectiveness of screening (i.e., early detection and early treatment) for type 2 diabetes and found that screening did not reduce the risks of diabetes complications or mortality in a median follow-up of 9.6-12 years[26,27]. Taken together, current evidence suggests that the timing of intervention for hyperglycemia is worth further discussion.
The management strategies of prediabetes could also be improved in terms of the targeted population, e.g., by adopting a total risk approach to target those with a higher risk of diabetic complications rather than over-emphasizing glycemic levels per se. The total risk approach has been adopted for the management of hypertension and dyslipidemia for many years. It entails a holistic management of all major cardiometabolic factors such as overweight/obesity, hypertension, and dyslipidemia, which were found to be associated with a higher probability of progression in our analysis (Table 2). Although current guidelines also recommend taking other risk factors into account when making treatment decisions, it is mainly based on the presence of individual risk factors, rather than the risk of diabetic complications (as predicted by multiple risk factors). For example, the Chinese guidelines recommend that individuals with prediabetes and meeting any of the over 15 specified criteria should be to considered for drug treatment[7]. Given the huge population affected and the small absolute benefit, the cost-effectiveness of such an approach is debatable.
This study adds to the evidence base on the prognosis of prediabetes without drug treatment, which can inform the development of management strategies (especially those regarding drug treatment) for the condition. However, the following limitations should be borne in mind when interpreting our findings. Firstly, the sample size of the regression group was relatively small, which, at least partially, led to the small outcome events in that group and insufficient statistical power. Secondly, CVD and kidney disease were self-reported in the CHARLS, which was less accurate than those diagnosed by doctors. For the same reason, it was not possible to accurately differentiate the subtypes of CVD and kidney disease. As some subtypes (e.g., rheumatic heart disease, kidney stone) of the two conditions are unlikely to be caused by hyperglycemia, and inclusion of such outcome events in all groups might have biased the association between prediabetes transition and outcomes towards null. Due to the above two limitations, our analyses of the associations between prediabetes transitions and health outcomes should be regarded as exploratory in nature. Larger studies with more accurate data on outcomes (e.g., diagnosed by physicians) are warranted to obtain more robust results. Thirdly, the majority of prediabetes in this study met the FPG-based definition but not the HbA1c-based definition. According to previous studies[28,29], FPG elevations appeared not to be independently associated with diabetic complications after controlling for 2hPG and HbA1c, which indicated that it might not be the optimal measure for defining hyperglycemia, including prediabetes. As the number of participants with HbA1c-defined prediabetes was small in this study, future studies with a larger sample size meeting this criterion are warranted to compare the performance of different definitions. It would also be relevant to compare the association of prediabetes defined based on single glucose testing with that based on repeated measurements. Fourthly, due to lack of data on the exact timing of the occurrence of health outcomes, binary logistic regression (which could not account for the time-to-event) was used in our primary analyses. However, the sensitivity analyses using Cox regression models yielded consistent results. Thus, we believe the choice of model had little (if any) impact on the results of this study. Lastly, as with other observational studies, residual confounding (e.g., that caused by dietary factors which were not measured in this study) may exist in the comparison between different transition groups.
In this cohort of middle-aged and elderly Chinese participants with prediabetes, 85% remained in the prediabetic state or even regressed to normoglycemia in the absence of drug treatment over a 4-year period. The rate of regression to normoglycemia was 1.4 times that of progression to diabetes. Progression to diabetes was associated with a higher risk of mortality in the subsequent 5 years, but the absolute risk attributable to progression per se was modest and accounted for a small fraction of the total prediabetes population. These findings call for more precise management of prediabetes in terms of the timing and target population of intervention in order to improve its cost-effectiveness, e.g., by using pre
We thank the China Center for Economic Research, the National School of Development of Peking University for providing the data.
| 1. | American Diabetes Association Professional Practice Committee. 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2025. Diabetes Care. 2025;48:S27-S49. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 1062] [Cited by in RCA: 868] [Article Influence: 868.0] [Reference Citation Analysis (1)] |
| 2. | Li Y, Teng D, Shi X, Qin G, Qin Y, Quan H, Shi B, Sun H, Ba J, Chen B, Du J, He L, Lai X, Li Y, Chi H, Liao E, Liu C, Liu L, Tang X, Tong N, Wang G, Zhang JA, Wang Y, Xue Y, Yan L, Yang J, Yang L, Yao Y, Ye Z, Zhang Q, Zhang L, Zhu J, Zhu M, Ning G, Mu Y, Zhao J, Teng W, Shan Z. Prevalence of diabetes recorded in mainland China using 2018 diagnostic criteria from the American Diabetes Association: national cross sectional study. BMJ. 2020;369:m997. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 1223] [Cited by in RCA: 1176] [Article Influence: 196.0] [Reference Citation Analysis (8)] |
| 3. | Centers for Disease Control and Prevention. National Diabetes Statistics Report. 2024. Available from: https://www.cdc.gov/diabetes/php/data-research/index.html. |
| 4. | Xue X, Li J, Zheng W, Zhang B, Wang S, Zhang J, Yang Z. Contributions of 2-h post-load glucose, fasting blood glucose and glycosylated haemoglobin elevations to the prevalence of diabetes and pre-diabetes in adults: A systematic analysis of global data. Diabetes Obes Metab. 2025;27:7285-7298. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in RCA: 1] [Reference Citation Analysis (0)] |
| 5. | Cai X, Zhang Y, Li M, Wu JH, Mai L, Li J, Yang Y, Hu Y, Huang Y. Association between prediabetes and risk of all cause mortality and cardiovascular disease: updated meta-analysis. BMJ. 2020;370:m2297. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 550] [Cited by in RCA: 525] [Article Influence: 87.5] [Reference Citation Analysis (1)] |
| 6. | American Diabetes Association Professional Practice Committee. 3. Prevention or Delay of Diabetes and Associated Comorbidities: Standards of Care in Diabetes-2025. Diabetes Care. 2025;48:S50-S58. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 54] [Cited by in RCA: 77] [Article Influence: 77.0] [Reference Citation Analysis (0)] |
| 7. | Chinese Society of Endocrinology, Chinese Diabetes Society; Chinese Endocrinologist Association. [Intervention for adults with pre-diabetes: a Chinese expert consensus (2023 edition)]. Zhonghua Tangniaobing Zazhi. 2023;15:484-494. [DOI] [Full Text] |
| 8. | Richter B, Hemmingsen B, Metzendorf MI, Takwoingi Y. Development of type 2 diabetes mellitus in people with intermediate hyperglycaemia. Cochrane Database Syst Rev. 2018;10:CD012661. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 109] [Cited by in RCA: 148] [Article Influence: 18.5] [Reference Citation Analysis (1)] |
| 9. | Diabetes Prevention Program Research Group. Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: the Diabetes Prevention Program Outcomes Study. Lancet Diabetes Endocrinol. 2015;3:866-875. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 869] [Cited by in RCA: 782] [Article Influence: 71.1] [Reference Citation Analysis (4)] |
| 10. | US Preventive Services Task Force, Davidson KW, Barry MJ, Mangione CM, Cabana M, Caughey AB, Davis EM, Donahue KE, Doubeni CA, Krist AH, Kubik M, Li L, Ogedegbe G, Owens DK, Pbert L, Silverstein M, Stevermer J, Tseng CW, Wong JB. Screening for Prediabetes and Type 2 Diabetes: US Preventive Services Task Force Recommendation Statement. JAMA. 2021;326:736-743. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 112] [Cited by in RCA: 363] [Article Influence: 72.6] [Reference Citation Analysis (3)] |
| 11. | Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43:61-68. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 4266] [Cited by in RCA: 4110] [Article Influence: 342.5] [Reference Citation Analysis (2)] |
| 12. | Phillips D, Green H, Petrosyan S, Shao K, Wilkens J, Lee J. Harmonized CHARLS documentation, version D (2011-2018). The Gateway to Global Aging Data. 2021. Available from: https://charls.pku.edu.cn/en/team/Codebook.pdf. |
| 13. | Davoodian N, Lotfaliany M, Huxley RR, Lee CMY, Pasco JA, Adams RJ, Azizi F, Bertoni AG, Björkelund C, Colagiuri S, Fahimfar N, Gabriel R, Giedraitis V, Gill TK, González C, Gregg EW, Hadaegh F, Hange D, Harland JW, Hodge AM, Holloway-Kew KL, Hosseini SR, Jacobs DR Jr, Khalili D, Magliano DJ, Mongraw-Chaffin M, Mishra GD, Lissner L, Mehlig K, Najafipour H, Nieto-Martinez R, Ostovar A, Shadkam M, Shaw JE, Sundh V, Schreiner PJ, Sakurai M, Wittert GA, Yatsuya H, Zethelius B, Mohebbi M. Prediabetes transitions to normoglycaemia or type 2 diabetes and associated risk factors in the Obesity, Diabetes and Cardiovascular Disease Collaboration: an individual-level pooled analysis of 19 prospective cohort studies. Lancet Glob Health. 2025;13:e1533-e1542. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 13] [Cited by in RCA: 18] [Article Influence: 18.0] [Reference Citation Analysis (0)] |
| 14. | Cheng HG, Chen S, McBride O, Phillips MR. Prospective relationship of depressive symptoms, drinking, and tobacco smoking among middle-aged and elderly community-dwelling adults: Results from the China Health and Retirement Longitudinal Study (CHARLS). J Affect Disord. 2016;195:136-143. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 83] [Cited by in RCA: 141] [Article Influence: 14.1] [Reference Citation Analysis (0)] |
| 15. | Xu X, Xu Y, Shi R. Association between obesity, physical activity, and cognitive decline in Chinese middle and old-aged adults: a mediation analysis. BMC Geriatr. 2024;24:54. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 2] [Cited by in RCA: 46] [Article Influence: 23.0] [Reference Citation Analysis (0)] |
| 16. | Pan XF, Wang L, Pan A. Epidemiology and determinants of obesity in China. Lancet Diabetes Endocrinol. 2021;9:373-392. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1300] [Cited by in RCA: 1164] [Article Influence: 232.8] [Reference Citation Analysis (9)] |
| 17. | van der Heijden GJ, Donders AR, Stijnen T, Moons KG. Imputation of missing values is superior to complete case analysis and the missing-indicator method in multivariable diagnostic research: a clinical example. J Clin Epidemiol. 2006;59:1102-1109. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 390] [Cited by in RCA: 420] [Article Influence: 21.0] [Reference Citation Analysis (1)] |
| 18. | Little RJA. Regression with missing X's: A review. J Am Stat Assoc. 1992;87:1227-1237. [RCA] [DOI] [Full Text] [Cited by in Crossref: 106] [Cited by in RCA: 61] [Article Influence: 1.8] [Reference Citation Analysis (0)] |
| 19. | Groenwold RH, White IR, Donders AR, Carpenter JR, Altman DG, Moons KG. Missing covariate data in clinical research: when and when not to use the missing-indicator method for analysis. CMAJ. 2012;184:1265-1269. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 184] [Cited by in RCA: 398] [Article Influence: 28.4] [Reference Citation Analysis (0)] |
| 20. | Rooney MR, Rawlings AM, Pankow JS, Echouffo Tcheugui JB, Coresh J, Sharrett AR, Selvin E. Risk of Progression to Diabetes Among Older Adults With Prediabetes. JAMA Intern Med. 2021;181:511-519. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 155] [Cited by in RCA: 152] [Article Influence: 30.4] [Reference Citation Analysis (2)] |
| 21. | Kim SM, Lee G, Choi S, Kim K, Jeong SM, Son JS, Yun JM, Kim SG, Hwang SS, Park SY, Kim YY, Park SM. Association of early-onset diabetes, prediabetes and early glycaemic recovery with the risk of all-cause and cardiovascular mortality. Diabetologia. 2020;63:2305-2314. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 30] [Cited by in RCA: 30] [Article Influence: 5.0] [Reference Citation Analysis (0)] |
| 22. | Cao Z, Li W, Wen CP, Li S, Chen C, Jia Q, Li W, Zhang W, Tu H, Wu X. Risk of Death Associated With Reversion From Prediabetes to Normoglycemia and the Role of Modifiable Risk Factors. JAMA Netw Open. 2023;6:e234989. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 19] [Cited by in RCA: 27] [Article Influence: 9.0] [Reference Citation Analysis (1)] |
| 23. | Lee G, Kim SM, Choi S, Kim K, Jeong SM, Son JS, Yun JM, Park SM. The effect of change in fasting glucose on the risk of myocardial infarction, stroke, and all-cause mortality: a nationwide cohort study. Cardiovasc Diabetol. 2018;17:51. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 52] [Cited by in RCA: 45] [Article Influence: 5.6] [Reference Citation Analysis (0)] |
| 24. | Vazquez Arreola E, Gong Q, Hanson RL, Wang J, Sandforth L, He S, Sandforth A, Qian X, Giacca M, Bornstein SR, Fritsche A, Stefan N, Preissl H, Gregg EW, Marx N, Jumpertz-von Schwartzenberg R, Li G, Birkenfeld AL. Prediabetes remission and cardiovascular morbidity and mortality: post-hoc analyses from the Diabetes Prevention Program Outcome study and the DaQing Diabetes Prevention Outcome study. Lancet Diabetes Endocrinol. 2026;14:137-148. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 34] [Cited by in RCA: 24] [Article Influence: 24.0] [Reference Citation Analysis (0)] |
| 25. | Liu X, Wu S, Song Q, Wang X. Reversion From Pre-Diabetes Mellitus to Normoglycemia and Risk of Cardiovascular Disease and All-Cause Mortality in a Chinese Population: A Prospective Cohort Study. J Am Heart Assoc. 2021;10:e019045. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 11] [Cited by in RCA: 66] [Article Influence: 13.2] [Reference Citation Analysis (1)] |
| 26. | Simmons RK, Echouffo-Tcheugui JB, Sharp SJ, Sargeant LA, Williams KM, Prevost AT, Kinmonth AL, Wareham NJ, Griffin SJ. Screening for type 2 diabetes and population mortality over 10 years (ADDITION-Cambridge): a cluster-randomised controlled trial. Lancet. 2012;380:1741-1748. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 150] [Cited by in RCA: 165] [Article Influence: 11.8] [Reference Citation Analysis (0)] |
| 27. | Rahman M, Simmons RK, Hennings SH, Wareham NJ, Griffin SJ. How much does screening bring forward the diagnosis of type 2 diabetes and reduce complications? Twelve year follow-up of the Ely cohort. Diabetologia. 2012;55:1651-1659. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 34] [Cited by in RCA: 38] [Article Influence: 2.7] [Reference Citation Analysis (0)] |
| 28. | Zhang X, Yang Z, Zhang M, Zhu Y, Zhao Z, Huang Z, Li C, Zhou M, Farmer AJ, Tang J, Wang L. Independent effects of 2hPG, FPG and HbA1c on cardiovascular risk: Analysis of a nationally representative sample from China. Diabetes Res Clin Pract. 2021;173:108672. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1] [Cited by in RCA: 8] [Article Influence: 1.6] [Reference Citation Analysis (0)] |
| 29. | Zhang J, Zheng W, Wang S, Gao X, Xiao Y, Yang Z. Associations of three diagnostic glycemic measures with all-cause and cardiovascular mortality in people not on antidiabetic medications: A prospective cohort study. J Diabetes Complications. 2025;39:108970. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 1] [Reference Citation Analysis (0)] |