Incidence of diabetes remission and association with microvascular and macrovascular complications in patients with newly diagnosed type 2 diabetes mellitus: a retrospective cohort study using real-world data in Thailand

Abstract

Objective

To estimate the incidence of diabetes remission and relapse among patients with type 2 diabetes (T2DM) and to assess the association between remission status and cardiorenal outcomes using a multistate analysis of long-term real-world data in Thailand.

Design

Retrospective cohort study using real-world data.

Setting

Single tertiary care hospital located in Bangkok, Thailand.

Participants

A total of 29 706 patients newly diagnosed with T2DM who received at least one antidiabetic medication during the study period.

Primary and secondary outcome measures

Diabetic remission, relapse after remission and cardiorenal outcomes. Diabetes remission was defined as achieving a glycated haemoglobin level <6.5% for at least 90 days after discontinuation of glucose-lowering medications.

Results

Among 29 706 newly diagnosed patients with T2DM, the 1-year probability of achieving a first diabetes remission was 3.44% (95% CI 3.22 to 3.67), increasing to 8.32% (95% CI 7.94 to 8.71) at 5 years. However, remission was frequently transient, with 38.38% (95% CI 36.22 to 40.63) of patients who achieved first remission experiencing disease relapse within 1 year. The incidence of cardiorenal outcomes was significantly lower among patients who achieved remission than those without remission (HR 0.53; 95% CI 0.46 to 0.61).

Conclusions

Our study provides real-world evidence that diabetes remission, although often transient, was significantly associated with a decreased risk of cardiorenal complications. These findings highlight the potential benefit of targeting remission early in newly diagnosed T2DM to decrease the risk of cardiorenal complications.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Long-term follow-up (up to 14 years) in a large real-world Asian cohort.

• Use of a multistate model with a clock-reset approach as a key methodological strength.

• Outcomes verified using objective laboratory and clinical markers rather than International Classification of Diseases, 10^th Revision (ICD-10) codes alone.

• Routine clinical data collection limited the capture of detailed lifestyle factors.

• Limited generalisability beyond East Asian populations with distinct metabolic phenotypes.

Introduction

Diabetes mellitus (DM) is a significant public health problem and a leading cause of death worldwide.¹ Among its various forms, type 2 diabetes mellitus (T2DM) is the most prevalent² and is strongly linked to an increased risk of both microvascular and macrovascular complications, including chronic kidney disease (CKD) and cardiovascular disease (CVD).³

DM is a lifelong progressive disease because insulin resistance, the primary underlying mechanism of DM, is permanent and cannot be reversed. However, it has been suggested that blood glucose levels in patients with diabetes can return to a healthy range and remain stable even after discontinuing antidiabetic medications. This concept, known as ‘diabetes remission’, has gained prominence since 2009. Diabetes remission can be achieved through lifestyle modifications, as well as medical and surgical interventions.^4–6 Previous randomised controlled trials have shown that approximately 40%–80% of patients with diabetes who adhered to intensive lifestyle interventions achieved diabetes remission within 1 year. However, in real-world practice, the incidence of diabetes remission among patients with T2DM is much lower. Cohort studies from Western countries have reported remission rates ranging from 2.8 to 9.7 per 1000 person-years,^{7 8} while data from a nationwide patient registry in Japan indicate an incidence of 10.5 per 1000 person-years among Japanese individuals.⁹

Despite achieving remission, many patients experience a relapse of diabetes over time. A previous study in Japan found that two-thirds of patients who achieved diabetes remission relapsed within 1 year.⁹ However, evidence on the longitudinal patterns of diabetes remission and relapse in real-world settings remains limited.

Furthermore, findings from landmark studies such as the UK Prospective Diabetes Study (UKPDS) and the Veterans Affairs Diabetes Trial demonstrated that intensive glucose control significantly reduces the risk of both microvascular and macrovascular complications in patients with T2DM.^{10 11} Therefore, achieving diabetes remission is expected to lower the risk of these complications. However, long-term exposure to hyperglycaemia may have persistent harmful effects on various tissues, a phenomenon known as the ‘legacy effect’. As a result, the risk of diabetic complications may still be present even after achieving remission.

Given these uncertainties, extrapolating previous findings to characterise the long-term impact of diabetes remission remains challenging and there is a need for real-world evidence with long-term follow-up to confirm this hypothesis. Therefore, this study aims to estimate the incidence of diabetes remission and relapse in patients with T2DM and assess their effects on the risk of diabetic complications using a multistate analysis of long-term real-world data from Thailand.

Methods

Study participants and setting

This retrospective cohort study used data from the T2DM database developed by the Department of Clinical Epidemiology and Biostatistics, Faculty of Medicine Ramathibodi Hospital, Mahidol University. Details for the development of this database are described elsewhere.¹² Briefly, this database included patients with T2DM treated at Ramathibodi Hospital between 1 January 2010 and 30 June 2024. Patients with T2DM were identified by ICD-10 codes (E110-119), fasting plasma glucose (FPG) ≥126 mg/dL (≥7 mmol/L) or glycated haemoglobin (HbA1c) ≥ 6.5% on at least two consecutive observations, or use of any antidiabetic medications.

Patients were eligible for this study if they (1) were newly diagnosed with T2DM between 1 January 2010 and 30 June 2024, (2) were treated with at least one antidiabetic medication, (3) did not have CKD and CVD prior to or at the date of initiating antidiabetic medication and (4) had at least two follow-up visits after initiating antidiabetic medication. Patients were excluded if they used oral or systemic steroids for at least 180 days within 1 year preceding the initiation of antidiabetic drugs.

Data collection

Demographic data (ie, age, sex, marital status and reimbursement schemes), underlying diseases (ie, hypertension (HT), dyslipidaemia (DLP), atrial fibrillation (AF), obstructive sleep apnoea (OSA)) and physical examination (ie, systolic blood pressure (SBP), diastolic blood pressure (DBP), height and body weight) at the date of initiating antidiabetic medications (referred to as the index date) were collected. Demographic data and physical examination were retrieved from the hospital information systems, while underlying diseases were identified using ICD-10 codes.

Information about antidiabetic medications (ie, metformin, sulfonylureas (SUs), glinides, thiazolidinediones, dipeptidyl peptidase-4 (DPP-4) inhibitors, sodium glucose co-transporter (SGLT)-2 inhibitors, glucagon-like peptide-1 receptor agonist (GLP-1 RA) and insulin injections) and other relevant medications including antihypertensive drugs (eg, ACE inhibitors, angiotensin receptor blockers, beta-blockers, calcium-channel blockers, diuretics) and lipid-lowering agents (eg, statins) were retrieved from the computerised provider order entry (CPOE) system and billing databases based on generic/trade names, Thai Medical Terminology and in-house codes. The details of dose, frequency/day, prescription amount and special notes for each medication were retrieved from the CPOE system. Additionally, a history of bariatric surgery was identified from the hospital database.

Laboratory data, including FPG, HbA1c, serum triglyceride, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol and serum uric acid, were retrieved from the hospital laboratory information system using Logical Observation Identifiers, Names, and Codes.

Outcomes of interest and verification

The outcomes of interest included diabetes remission, relapse after diabetes remission and incidence of cardiorenal outcomes. Diabetes remission was defined according to the American Diabetes Association criteria as achieving an HbA1c level <6.5% for at least 90 days after discontinuation of glucose-lowering medications.¹³ Relapse after diabetes remission was defined as an HbA1c level ≥6.5% on two consecutive occasions over a period of at least 90 days. Cardiorenal outcomes were defined as a composite of CKD and CVD. CKD was identified by an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² on two consecutive measurements at least 90 days apart, or by the presence of ICD-10 codes for CKD (ie, N18, N19) or ICD-9 procedure codes indicating renal replacement therapy. CVD events included incident coronary artery disease (CAD) and stroke. CAD was identified using ICD-10 codes (ie, I20–I25), ICD-9 diagnosis codes, and procedure codes for percutaneous coronary intervention and coronary artery bypass graft surgery (’00.4, ’00.66, ’17.55, ’36.0, ’36.1, ’36.31, ’36.99). In addition, CAD was confirmed by elevated troponin-T levels (>14 ng/mL) in conjunction with positive ECG findings occurring within 2 weeks before or after the troponin test. Stroke was identified using ICD-10 codes for ischaemic stroke (ie, I63, I64) and haemorrhagic stroke (ie, I60–I62, I69).

Sample size estimation

The sample size was estimated based on the precision of the estimated proportion of diabetes remission. Assuming a two-sided CI of 0.05 (Z=1.96), a desired margin of error of 1% and an expected diabetes remission rate of 7%,¹⁴ the minimally required sample size was approximately 2500 patients with newly diagnosed T2DM. However, all patients in the cohort who met the predefined eligibility criteria were included in this study to maximise the utilisation of available data and improve the predefined precision of the estimates.

Statistical analysis

We applied a multistate modelling framework to estimate the probabilities of diabetes remission, relapse after diabetes remission and cardiorenal outcomes among patients with newly diagnosed T2DM. The multistate model comprised an initial state of newly diagnosed T2DM; transient states including first remission, first relapse, second remission and second relapse; and an absorbing state representing cardiorenal outcomes, as illustrated in figure 1. A total of nine possible transitions were specified: T1 (initial state to first remission), T2 (initial state to cardiorenal outcomes), T3 (first remission to first relapse), T4 (first remission to cardiorenal outcomes), T5 (first relapse to second remission), T6 (first relapse to cardiorenal outcomes), T7 (second remission to second relapse), T8 (second remission to cardiorenal outcomes) and T9 (second relapse to cardiorenal outcomes). Sensitivity analysis for the multistate modelling framework was performed using less stringent diabetic remission criteria, defined as achieving either an HbA1c level <6.5% and/or an FPG level <126 mg/dL.¹³ This sensitivity analysis was performed to enhance the applicability of our findings to situations where HbA1c was considered an unreliable biomarker of chronic glycaemic control or HbA1c testing was unavailable, as may be the case in resource-limited settings.

Figure 1. A multistate model of diabetes remission, relapse, and cardiorenal outcomes among patients with newly diagnosed diabetes mellitus. There are six states with a total of nine possible transitions. The numbers in the left box along each transition indicate the number of patients who moved from the state at the tail of the arrow to the state at the head of the arrow. For example, in transition T1, 2153 patients progressed from initial state (tail state) to first diabetes remission (head state). The numbers in the right box within each state denote the number of patients who remained in that state without transition. T2DM, type 2 diabetes mellitus.

Time from the index date to the first remission state or the cardiorenal outcome state was estimated, whichever occurred first. Among patients who achieved first remission, the time from one transient state to another transient or absorbing state was subsequently estimated. Kaplan-Meier estimators with a clock-reset approach were used to calculate transition probabilities for each transition. Patients who did not experience cardiorenal outcomes were censored at the end of the study period (30 June 2024) or at the date of their last follow-up, whichever occurred first.

The risks of cardiorenal outcomes between patients who achieved diabetes remission and those who did not were compared using multivariable Cox proportional hazards models. Covariates, including age, sex, reimbursement scheme, body mass index (BMI), history of HT, DLP, AF, OSA, FPG and eGFR with a P value<0.10 in univariable Cox regression analyses, were included in the multivariable models to assess the independent association between diabetes remission and risk of cardiorenal outcomes.

All statistical analyses were performed using STATA V.18 (StataCorp, College Station, Texas, USA). A two-sided p value<0.05 was considered statistically significant for all analyses.

Patient and public involvement statement

None.

Results

A total of 66 906 patients with newly diagnosed T2DM between 1 January 2010 and 30 June 2024 were initially identified. Of these, 37 200 patients were excluded due to the absence of antidiabetic medications at diagnosis (n=24 474), a follow-up duration of less than 1 year (n=7274), pre-existing cardiorenal complications before the index date (n=3795) and prolonged steroid use of more than 180 days before or after the index date (n=1657). Finally, 29 706 patients were included in this study. The median follow-up time (IQR) of this cohort was 4.06 years (1.80–7.36), and the median number of hospital visits during the follow-up period (IQR) was 9 (5–16).

Baseline characteristics

Baseline characteristics of the study population are summarised in table 1. The mean age of participants was 56.4±13.2 years, and 62.4% were women. The mean BMI was 28.4±5.7 kg/m². HT and DLP were present in 76.4% and 78.3% of participants, respectively. The median FPG level was 138.0 mg/dL (115.0, 183.0), the mean HbA1c was 8.0±2.1% and the median eGFR was 97.4 mL/min/1.73 m² (86.5, 107.7).

Table 1. Baseline characteristics of the study participants.

Characteristics	Participants (n=29 706)
Age, mean (SD), years	56.4 (13.2)
Sex, n (%)
Male	11 164 (37.6)
Female	18 542 (62.4)
BMI, mean (SD), kg/m²	28.4 (5.7)
Underlying diseases
Hypertension, n (%)
Yes	22 697 (76.4)
No	7009 (23.6)
Dyslipidaemia, n (%)
Yes	23 268 (78.3)
No	6438 (21.7)
Obstructive sleep apnoea, n (%)
Yes (not using CPAP)	242 (0.8)
Yes (using CPAP)	579 (2.0)
No	28 885 (97.2)
Atrial fibrillation, n (%)
Yes	1222 (4.1)
No	28 484 (95.9)
Laboratory at baseline
FBS, median (IQR), mg/dL	138.0 (115.0–183.0)
HbA1c, mean (SD), %	8.0 (2.1)
Total cholesterol, mean (SD), mg/dL	210.0 (60.8)
HDL-C, median (IQR), mg/dL	48.0 (41.0–56.0)
LDL-C, mean (SD), mg/dL	131.2 (42.9)
Triglycerides, median (IQR), mg/dL	143.0 (104.0–201.0)
eGFR, median (IQR), mL/min/1.73 m²	97.4 (86.5–107.7)
Medication, n (%)
Antidiabetic medications
Biguanides
Yes	26 119 (87.9)
No	3587 (12.1)
Thiazolidinediones
Yes	2955 (9.9)
No	26 751 (90.1)
Sulfonylureas
Yes	7383 (24.9)
No	22 323 (75.1)
DPP-4 inhibitors
Yes	4310 (14.5)
No	25 396 (85.5)
SGLT-2 inhibitors
Yes	286 (1.0)
No	29 420 (99.0)
GLP-1 RAs
Yes	958 (3.2)
No	28 748 (96.8)
Insulin
Yes	3922 (13.2)
No	25 784 (86.8)
Antihypertensive medications
ACE inhibitors
Yes	4275 (14.4)
No	25 431 (85.6)
ARBs
Yes	6996 (23.6)
No	22 710 (76.4)
Calcium channel blockers
Yes	7830 (26.4)
No	21 876 (73.6)
Beta blockers
Yes	6400 (21.5)
No	23 306 (78.5)
Spironolactone
Yes	556 (1.9)
No	29 150 (98.1)
Diuretics
Yes	3275 (11.0)
No	26 431 (89.0)
Lipid-lowering medications
Statins
Yes	17 623 (59.3)
No	12 083 (40.7)

ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker; BMI, body mass index; CPAP, continuous positive airway pressure; DPP-4, dipeptidyl peptidase-4; eGFR, estimated glomerular filtration rate; FBS, fasting blood glucose; HbA1c, glycated haemoglobin; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; GLP-1 RA, glucagon-like peptide-1 receptor agonist; SGLT-2, sodium-glucose co-transporter 2.

Regarding initial glucose-lowering therapy, biguanides were the most frequently prescribed agents (87.9%), followed by SUs (24.9%), DPP-4 inhibitors (14.5%), insulin (13.2%) and thiazolidinediones (9.9%). SGLT-2 inhibitors and GLP-1 RAs were prescribed as initial therapy in 1.0% and 3.2% of participants, respectively.

Transitional probability of having diabetes remission and relapse after remission

For the analysis of diabetes remission, defined as achieving an HbA1c level <6.5%, 27 097 patients had available HbA1c measurements and were included. The transitional probabilities of each state are presented in online supplemental table 1 and figure 1. Among newly diagnosed patients with T2DM, 2153 achieved a first diabetes remission during follow-up (figure 1). The 1-year probability of achieving first diabetes remission was 3.44% (95% CI 3.22 to 3.67), see online supplemental table 1. The probability of having remission increased to 8.32% (95% CI 7.94 to 8.71) at 5 years and 13.32% (95% CI 12.67 to 13.99) at 10 years. However, approximately 40% of patients who achieved first remission experienced disease relapse within 1 year (38.38%; 95% CI 36.22 to 40.63), and almost all patients relapsed within 6 years (99.67%; 95% CI 98.29 to 99.97), see online supplemental table 1.

Following the first relapse, 14.85% (95% CI 12.81 to 17.18) of patients achieved a second remission of diabetes within 1 year. The probability of having a second remission increased to 30.67% (95% CI 27.36 to 34.27) at 5 years and 33.67% (95% CI 29.69 to 38.01) at 10 years. Nonetheless, disease relapse after second remission remained high, with 47.39% (95% CI 41.18 to 54.03) and 75.13% (95% CI 68.50 to 81.30) of patients experiencing a second relapse within 1 and 2 years, respectively (see online supplemental table 1).

The transitional probability of experiencing cardiorenal events was higher among patients without diabetes remission compared with those who achieved remission. At 1 year, the probability of cardiorenal events was 7.75% (95% CI 7.44 to 8.08) in patients without remission, compared with 3.81% (95% CI 3.00 to 4.82) and 1.25% (95% CI 0.40 to 3.84) among patients with first and second diabetes remissions, respectively. In addition, the probabilities of having cardiorenal outcomes at 1 and 5 years among patients with relapses after having first diabetes remission were lower than those in patients without diabetes remission. The transitional probabilities from disease relapse state to cardiorenal state at 1 and 5 years were 4.14% (95% CI 3.08 to 5.55) and 14.56% (95% CI 11.76 to 17.96), compared with 7.75% (95% CI 7.44 to 8.08) and 16.01% (95% CI 15.52 to 16.51) from initial state to cardiorenal state (see online supplemental table 1).

For the less stringent criteria using either HbA1c or FPG levels, the probability of achieving first diabetes remission was 5.61% (95% CI 5.08 to 5.34) at 1 year, which increased to 11.5% (95% CI 10.67 to 11.08) and 16.91% (95% CI 15.58 to 16.24) at 5 and 10 years, respectively (see online supplemental table 2). Similar to the more stringent criteria based on HbA1c alone, approximately half of the patients achieving remission experienced disease relapse within 1 year (46.39%; 95% CI 44.55 to 48.28) and almost all relapsed within 3 years after disease remission (98.15%; 95% CI 95.55 to 99.40), see online supplemental table 2).

Association between diabetes remission and risk of cardiorenal events

During 14 years of follow-up, 4644 patients developed cardiorenal complications, corresponding to an incidence rate of 3.49 per 100 person-years (95% CI 3.39 to 3.59). The incidence of cardiorenal outcomes was lower among patients who achieved diabetes remission than among those without remission (figure 2). Factors associated with an increased risk of cardiorenal outcomes included older age, male sex, type of healthcare coverage, a history of HT, DLP, OSA and AF, higher FPG levels and lower eGFR (see online supplemental table 3).

Figure 2. Cumulative probability of having cardiorenal outcomes between patients having diabetes remission and not having diabetes remission.

After adjustment for these potential confounders in a multivariable analysis (table 2), diabetes remission remained significantly associated with a lower risk of cardiovascular and renal outcomes (HR 0.53; 95% CI 0.46 to 0.61). Increasing age (per 5-year increment: HR 1.25; 95% CI 1.23 to 1.27; p<0.001) and male sex (HR 1.70; 95% CI 1.60 to 1.81; p<0.001) were associated with a higher risk. Comorbidities, including HT (HR 3.78; 95% CI 3.26 to 4.40; p<0.001), DLP (HR 1.78; 95% CI 1.59 to 2.00; p<0.001), AF (HR 2.42; 95% CI 2.21 to 2.65; p<0.001) and OSA (HR 1.39; 95% CI 1.19 to 1.64; p<0.001), were also independently associated with increased risk. In addition, higher FPG levels (per 10 mg/dL increase: HR 1.02; 95% CI 1.01 to 1.02; p<0.001) were associated with an elevated risk of cardiorenal outcomes.

Table 2. Factors associated with cardiorenal outcomes: a multivariable analysis.

Factors	HR (95% CI)	p value
Having diabetes remission		<0.001
No	1
Yes	0.53 (0.46 to 0.61)
Age, every 5 years increased	1.25 (1.23 to 1.27)	<0.001
Sex
Females	1
Males	1.70 (1.60 to 1.81)	<0.001
Hypertension
No	1
Yes	3.78 (3.26 to 4.40)	<0.001
Dyslipidaemia
No	1
Yes	1.78 (1.59 to 2.00)	<0.001
Atrial fibrillation
No	1
Yes	2.42 (2.21 to 2.65)	<0.001
Obstructive sleep apnea
No	1
Yes	1.39 (1.19 to 1.64)	<0.001
FBS level, every increase by 10 mg/dL	1.02 (1.01 to 1.02)	<0.001

FBS, fasting blood glucose.

Discussion

Our findings suggest that the 1-year probability of achieving first diabetes remission was 3.44%. The probability increased to 8.32% and 13.32% at 5 years and 10 years, respectively. Moreover, our results indicate that diabetes remission was significantly associated with a lower risk of cardiovascular and renal outcomes in patients with T2DM. Patients with remission had 47% lower risk of cardiovascular and renal outcomes than patients without remission. Despite these benefits, the relapse rate was substantial: around 40% of the patients experienced a disease relapse within 1 year, and almost all patients relapsed within 6 years.

Our findings provide real-world evidence on the incidence of type 2 diabetes remission. We observed a cumulative incidence of diabetes remission of 3.44% within the first year, which steadily increased to 13.32% over a 10-year follow-up period. The 10-year cumulative incidence of remission in our study (13.32%) was notably higher than that reported in Western real-world cohorts.^6–8 This discrepancy is likely attributable to differences in study populations, as prior Western studies included individuals with varying durations of diabetes. In contrast, our cohort was comprised exclusively of newly diagnosed patients. Our findings are consistent with evidence from British primary care practice,⁶ which suggests that the earliest stages of diagnosis offer the greatest potential for remission. Furthermore, a previous remission prevalence of approximately 5% was frequently associated with intentional weight loss and surgical interventions.⁶ In contrast, our cohort achieved a 10-year cumulative remission rate of 13.32% through routine clinical management, underscoring the potential for remission even without widespread access to surgical interventions.

Within the East Asian context, data from the Japan Diabetes Clinical Data Management registry reported a diabetes remission rate of 10.5 per 1000 person-years, with key predictors including a shorter duration of diabetes and a reduction in BMI of greater than 10% within the first year.⁹ Similarly, recent large-scale data from Hong Kong reported an incidence rate of 7.8 (95% CI 7.5 to 8.1) per 1000 person-years. They highlighted intensive weight control during the first year as the most critical strategy for achieving diabetes remission. Participants who achieved a weight reduction of at least 10% had a 3.28-fold higher likelihood of remission compared with those who did not.¹⁵ While weight loss has been identified as a major driver of remission across multiple cohorts,^{4 16–20} our findings suggest that in routine clinical practice, early diagnosis and prompt initiation of standard therapy may also facilitate remission by reducing glucotoxicity and permitting β-cell recovery, even in the absence of substantial documented weight loss.

Several studies have reported that although remission is attainable, it is frequently accompanied by relapse.^{7 9 21} This instability was also evident in our cohort, with approximately 40% of patients relapsing within the first year and almost all experiencing diabetes recurrence within 6 years. The high relapse rate (99.67% within 6 years) observed in our study may be fundamentally explained by the distinct pathophysiology of T2DM in East Asian populations. According to the Twin Cycle Hypothesis, remission is achieved by clearing ectopic fat from the liver and pancreas.²² However, for Asian populations, this state of remission is extremely fragile given lower insulin secretory capacity and limited beta-cell reserve, in which the ‘Pancreas Cycle’ can be easily overwhelmed.²² Unlike Western populations, where significant insulin resistance with compensatory hyperinsulinaemia is more commonly observed, Asian populations have lower insulin secretory capacity from early disease stages, resulting in a limited beta-cell reserve. Consequently, even modest changes in lifestyle behaviours or small increases in body weight may overwhelm the already constrained beta-cell function,^{23 24} leading to rapid loss of glycaemic control and earlier relapse of diabetes compared with Western populations. This susceptibility is further exacerbated by the unique metabolic profile of nonobese Asians, where ectopic fat accumulation - particularly in the liver and pancreas - serves as a critical determinant of glucose dysregulation. Furthermore, low aerobic fitness further impairs this delicate balance by reducing peripheral insulin sensitivity. Together, these factors place an unsustainable metabolic demand on the limited beta-cell reserve. Therefore, even minimal weight regain or a decline in physical activity can trigger cycles of metabolic stress,²⁵ leading to a higher incidence of relapse observed in this cohort. Importantly, evidence from the randomised Diabetes Remission Clinical Trial²⁶ suggests that, following the achievement of diabetes remission, continued support for weight-loss maintenance and healthy lifestyle behaviours may help sustain remission and reduce the risk of diabetes relapse. In contrast, our study used real-world data reflecting routine clinical practice, in which structured interventions or dedicated programmes providing ongoing support for weight maintenance and healthy lifestyle behaviours were not available. The absence of such structured and continuous support may partly explain the relatively high rate of diabetes relapse observed in our study. Therefore, implementing structured interventions or dedicated programmes that provide ongoing support for weight maintenance and healthy lifestyle behaviours are likely to be important for achieving and maintaining diabetes remission and reducing the risk of disease relapse in routine clinical practice. A recent nationwide claims-based study demonstrated that patients who achieved diabetes remission had a significantly lower risk of CVD, with a 29% reduction in risk (HR 0.71; 95% CI 0.57 to 0.89) compared with those who remained in an uncontrolled diabetic state. Consistent with these findings, our study showed that achieving even a single episode of remission was associated with a 47% reduction in the risk of composite cardiovascular and renal events (HR 0.53; 95% CI 0.46 to 0.61). These results emphasise the concept that diabetes remission is an achievable and clinically meaningful therapeutic target in routine practice, offering substantial long-term protection against major complications, even in real-world settings where intensive lifestyle interventions are less tightly controlled than in clinical trials.

Moreover, our study extends prior observations by evaluating composite cardiorenal outcomes over a prolonged 14-year follow-up, thereby providing a more comprehensive assessment of the long-term ‘legacy effect’²⁷ than that afforded by shorter follow-up periods. It is noteworthy that the use of SGLT-2 inhibitors and GLP-1 RAs was minimal in our cohort, with only 1.0% and 3.2% of patients receiving these agents as initial therapy, respectively. Given the established cardiorenal benefits of these newer drug classes, the 47% risk reduction observed in our study—predominantly under metformin-based regimens—strongly suggests that achieving glycaemic remission itself serves as a potent independent protector. Evidence from the ADVANCE trial demonstrated that intensive glucose-lowering therapy targeting an HbA1c≤6.5% significantly preserved renal function, reducing the risk of end-stage renal disease by 65% and microalbuminuria by 9% compared with standard care.²⁸ Intensive glycaemic control (HbA1c<7.0%) remains the cornerstone for preventing microvascular complications and reducing major cardiovascular events.²⁹ This is further supported by the 44-year follow-up of the UKPDS (2024), which confirmed a sustained ‘legacy effect’ whereby early intensive glycaemic control continues to reduce the risk of microvascular disease, myocardial infarction and all-cause mortality long after HbA1c levels converge.³⁰ These findings suggest that achieving remission during the ‘golden window’ of newly diagnosed T2DM may prevent the early accumulation of glycaemic burden and interrupt the progression of vascular damage. Importantly, even transient remission has been associated with meaningful protection against microvascular and macrovascular complications,²¹ consistent with our observation that cardiorenal benefits persist despite high relapse rates. Collectively, these data indicate that the 47% risk reduction observed in our cohort is not contingent on permanent disease reversal but rather reflects the clinical benefit of interrupting chronic hyperglycaemia during a period of maximal physiological plasticity.

A major strength of this study is its long-term follow-up of up to 14 years in a large, real-world Asian cohort, providing robust evidence on the durability of the diabetes remission ‘legacy effect’. By focusing exclusively on newly diagnosed individuals, we captured the critical early ‘golden window’ of metabolic plasticity that is often obscured in studies including mixed disease durations. The application of a multistate model with a clock-reset approach represents a key methodological strength, enabling dynamic modelling of transitions between diabetes, remission, relapse and subsequent remission while effectively minimising immortal time bias and allowing precise estimation of cardiorenal risk from the onset of each metabolic state. Outcome validity was strengthened through rigorous verification using objective laboratory and clinical markers rather than relying solely on ICD-10 codes.

Nevertheless, several limitations warrant consideration. As an observational study, causal inference cannot be definitively established despite multivariable adjustment for key confounders. In addition, reliance on routine clinical data may have limited the capture of detailed lifestyle factors, and variable testing intervals influenced the timing of remission and relapse between glycaemic assessments. The number and frequency of hospital visits varied among study participants, which may have resulted in misclassification or under-ascertainment of diabetes remission and relapse. In particular, participants with less frequent follow-up visits may have had fewer opportunities for glycaemic assessment, and transient episodes of remission and relapse occurring between assessments may therefore have been missed. In our cohort, the median number of hospital visits during the follow-up was 9 (IQR: 5–16), with a median follow-up duration of 4.06 years (IQR: 1.80–7.36), corresponding to approximately two per patient-year on average. Although this indicates that patients generally had repeated opportunities for glycaemic assessment, it does not rule out the possibility of missed episodes of remission or relapse. Consequently, the observed rates of remission and relapse may underestimate the true frequency of these events. In addition, all-cause mortality was not modelled as a competing risk, and the findings may not be fully generalisable beyond East Asian populations with distinct metabolic phenotypes.

Conclusions

Our study provides real-world evidence that diabetes remission, although often transient, was significantly associated with a decreased risk of cardiorenal complications. These findings highlight the potential benefit of targeting remission early in newly diagnosed T2DM to decrease the risk of cardiorenal complications.

Data availability statement

Data are available upon reasonable request.