Prolonging Partial Remission in Type 1 Diabetes: The Role of Low-Carbohydrate Diet: A Literature Review

Abstract

Background

Type 1 diabetes mellitus (T1DM) is an autoimmune disease characterized by progressive destruction of insulin-producing β-cells, resulting in lifelong insulin dependence and impaired quality of life. Although a transient phase of partial remission may occur after diagnosis, sustained preservation of β-cell function remains exceptional. Recently, low-carbohydrate diets (LCD) have gained attention among patients, partly driven by anecdotal reports describing prolonged remission.

Objective

The aim of this review was to critically examine the existing clinical evidence for prolonged partial remission under LCD in patients with T1DM and to summarize the biological mechanisms proposed in the literature to explain such observations.

Methods

A literature review was conducted focusing on reports of prolonged remission in individuals with T1DM adopting an LCD, as well as experimental studies exploring mechanistic pathways potentially modulated by carbohydrate restriction.

Results

Only six individual cases describing prolonged partial remission under LCD were identified in the literature. The diagnostic certainty of T1DM in these reports was variable, and misclassification (eg, latent autoimmune diabetes in adults or monogenic diabetes) cannot be excluded. The mechanistic hypotheses proposed—including reduction of glucotoxicity, modulation of inflammation, attenuation of oxidative stress, and possible β-cell regeneration—are derived predominantly from animal models, in vitro studies, and non-T1DM populations. Currently, no controlled clinical trial has demonstrated that LCDs prolong remission or preserve β-cell function in patients with T1DM.

Conclusion

Evidence supporting prolonged partial remission under LCD in patients with T1DM remains based on rare case reports of uncertain diagnostic validity. Mechanistic data are largely indirect and extrapolated from heterogeneous experimental contexts. At present, LCD cannot be considered as a therapeutic strategy for inducing or maintaining remission in T1DM. This review highlights the need for well-designed longitudinal studies with stringent diagnostic criteria and mechanistic validation in authentic T1DM models.

Core Tip

• Prolonged partial remission in type 1 diabetes mellitus (T1DM) under low-carbohydrate diets (LCDs) has been reported only in a very small number of cases worldwide.

• The diagnostic validity of these cases is uncertain, and misclassification with latent autoimmune diabetes in adults (LADA) or monogenic diabetes (MODY) cannot be excluded.

• Proposed mechanisms—including reduced glucotoxicity, modulation of inflammation, decreased oxidative stress, and β-cell regeneration—are largely based on experimental and animal studies rather than human T1DM data.

• No controlled clinical trials currently demonstrate that LCDs prolong remission or preserve β-cell function in T1DM.

• LCDs should not be considered a therapeutic strategy for inducing remission in T1DM based on current evidence.

• Future studies are required using strict diagnostic criteria, longitudinal C-peptide monitoring, and mechanistic validation in authentic T1DM populations.

Background

Type 1 Diabetes Mellitus

The incidence of type 1 diabetes mellitus (T1DM) is significantly increasing worldwide. In 2021, researchers from the NGO Juvenile Diabetes Research Foundation have estimated that almost 510.000 new cases of T1DM will be diagnosed worldwide on a yearly basis. According to recent estimates, this disease will affect 17.4 million people by 2040.¹

As an autoimmune disorder, T1DM is characterized by chronic hyper-glycemia resulting from T-cell-mediated destruction of insulin-producing pancreatic β-cells. T1DM can be complicated by acute events such as diabetic ketoacidosis or severe hypo-glycemia, as well as vascular disease because of chronic hyper-glycemia. A model developed using data from the Swedish National Diabetes Registry has demonstrated that patients with T1DM can expect to live with one or more complications for approximately 40% of their remaining lifetime, with a simulated life expectancy approximately 13 years lower than that of the sex- and age-matched general population.² These complications also incur considerable health care costs. A recent study has reported that T1DM is estimated to cost the United States of America around $14.4 billion a year in terms of medical costs and lost income.³ The impact on patients’ longevity is highly significant.⁴ It is therefore essential to minimize the complications of T1DM through early interventions that not only stabilize glycemia but also slow down β-cell destruction.

Dietary Recommendations in T1DM

Nutritional recommendations for individuals with T1DM, as issued by leading national and international diabetes organizations, are broadly consistent with those provided to the general population and are centered on maintaining nutritional balance, dietary variety, and long-term metabolic health.⁵

Guidance from the American Diabetes Association (ADA) and the International Society for Pediatric and Adolescent Diabetes (ISPAD) emphasizes the consumption of carbohydrates from whole and minimally processed foods, including fruits, vegetables, whole grains, legumes, and dairy products. In parallel, both organizations recommend adequate fiber intake and limiting exposure to sugar-sweetened beverages and industrial trans fats.^6,7

According to recent ADA statements, there is no single macronutrient distribution that can be considered optimal for all individuals with T1DM, highlighting the need for personalized dietary approaches.⁶ In contrast, ISPAD provides general quantitative targets for macronutrient intake, suggesting that carbohydrates should represent approximately half of total energy intake, fats less than one-third (with saturated fat below 10%), and proteins of the remaining proportion.⁷

Traditionally, it is recommended that 50% of the total caloric intake comes from carbohydrates, 30% from lipids, and 20% from proteins.⁸

Partial Remission or “Honeymoon Phase” in T1DM

Definition

For 60% of patients with T1DM, a few weeks after diagnosis and initiation of insulin treatment, exogenous insulin requirements are reduced by an average of 50% owing to a rebound in the activity of β-cells, resuming insulin production. In some cases, temporary insulin independence may be achieved.⁹ This phenomenon, lasting from a few weeks to a year (average of 9 months), is commonly referred to as partial remission.¹⁰

Residual β-cell activity can be assessed by measuring serum C-peptide levels. During proinsulin conversion, equimolar production of C-peptide and insulin occurs. C-peptide is considered a reliable marker (yet with several limitations)¹¹ for assessing endogenous insulin secretion, even after the initiation of insulin therapy. C-peptide concentrations can be measured under basal conditions or after stimulation, for example, following an intravenous injection of glucagon or ingestion of a standardized mixed meal. The latter method is commonly used in clinical research and intervention studies to monitor changes in β-cell function in people with T1DM.^10,12 In this context, partial remission is defined as a stimulated plasma C-peptide concentration of ≥0.3 nmol/L at 120 minutes in some studies, while others use a threshold of >0.3 nmol/L measured at 90 minutes.¹²

However, the implementation of this dynamic testing in routine clinical practice remains challenging due to its complexity, time requirements, and cost. As a result, researchers have attempted to develop alternative clinical formulas that can estimate partial remission without the need for direct C-peptide measurement.^12–14 Currently, there is no established consensus on the definition of partial remission.

Indeed, numerous authors have attempted to propose standardized definitions and to compare the various approaches for quantifying β-cell function, particularly those incorporating HbA1c levels and exogenous insulin requirements.¹² For example, Schölin et al have proposed defining partial remission as the maintenance of an HbA1c level <6.5% combined with a daily insulin requirement <0.4 IU/kg/day for at least one month.¹³

To further refine this definition, Mortensen et al have introduced in 2009 a more precise clinical model—the insulin dose-adjusted HbA1c (IDAA1C)—calculated as HbA1c (%) + 4 × insulin dose (IU/kg/day), with a value ≤9 indicating partial remission.¹⁴ In a cohort of 275 children and adolescents under 16 years of age newly diagnosed with T1DM, the authors have demonstrated a correlation between an IDAA1C ≤9 and C-peptide levels ≥0.3 nmol/L. Despite its potential limitations,¹⁵ this method remains the most widely used approach in routine clinical practice to assess partial remission in patients.¹²

Intervention to Prolong Partial Remission in T1DM

Recent advancements have led to the development of innovative strategies aimed at preserving pancreatic β-cell function and promoting their regeneration in individuals at risk for or newly diagnosed with T1DM. Notably, early immunomodulatory interventions—such as anti-CD3 monoclonal antibodies (eg., teplizumab) and β-cell-depleting agents like anti-CD20—have shown potential when combined with immunoregulatory fusion proteins, including etanercept (a TNF-α inhibitor) and abatacept (CTLA4-Ig). These therapies aim to modulate the autoimmune response while preserving residual β-cell mass. Additionally, emerging approaches focus on β-cell protection and functional support through agents targeting inflammatory pathways and cellular stress responses.^16,17 While these multifaceted strategies show promise in delaying disease progression, there is currently no universally accepted or standardized therapeutic protocol. Another emerging therapeutic approach involves the early reduction of hyperglycemia to preserve residual β-cell function within the islets of Langerhans.

Dietary Intervention to Prolong Partial Remission in T1DM

Partial remission phase represents a window of opportunity to preserve β-cell activity and reduce insulin dependence. Early interventions based on specific dietary measures have shown to be beneficial in specific circumstances.

For example, there is strong evidence that nutritional interventions and significant weight loss can lead to remission of type 2 diabetes mellitus (T2DM) in a substantial number of patients. This has been convincingly shown in the Diabetes Remission Clinical Trial (DiRECT) study by Lean M et al conducted by 49 general practitioners in Scotland and England between 2014 and 2017. This study included 298 people with T2DM: 149 in the intervention group and 149 in the control group. Participants followed a strict hypocaloric diet, starting with a 3-to-5-month liquid diet phase with a daily calorie intake of 825–853 kcal, followed by a gradual reintroduction of solid foods over a period of 2–8 weeks. Results after one year have revealed that 46% of the participants were in remission, defined as an HbA1c level below 6.5% without the use of anti-diabetic drugs. At two years, the remission rate was 36%. Notably, among patients who lost ≥15 kg, the remission rate reached 86%. However, it is crucial to note that maintaining remission required long-term persistence of lifestyle changes.¹⁸ Several similar studies corroborate these findings.^19–22

Although T1DM and T2DM are fundamentally distinct diseases with different pathophysiological mechanisms, findings from dietary intervention studies in T2DM have nonetheless attracted attention and have encouraged interest in applying similar nutritional strategies to T1DM.

Various strategies have been explored, including diets with a low n-6/n-3 essential fatty acid ratio (ie., reduced omega-6 relative to omega-3 intake), ergocalciferol supplementation and low-carbohydrate diet (LCD), aiming to modulate inflammation and preserve β-cell function.^23,24

Misclassification Bias and Diagnostic Uncertainty in Reported T1DM Remission

The reliability of diagnosing T1DM based primarily on autoantibody testing is limited and likely contributes to a significant overestimation of partial remission rates reported in the literature.²⁵ Indeed, up to 10–15% of adults initially classified as having T1DM are negative for diabetes-related autoantibodies and may in fact have undiagnosed monogenic diabetes (MODY).²⁵ In addition, latent autoimmune diabetes in adults (LADA), which accounts for approximately 5–10% of diabetes cases diagnosed in adulthood, is characterized by the presence of autoantibodies and a slow rate of β-cell destruction over several years, potentially mimicking “prolonged remission”.²⁶ Likewise, individuals with MODY typically retain β-cell function for decades, creating the appearance of stable disease or indefinite remission.²⁵

The absence of systematic testing using a complete autoantibody panel (GAD, IA-2, ZnT8, and IAA), combined with insufficient longitudinal monitoring of C-peptide levels (eg., at 6, 12, 24, and 36 months after diagnosis) and limited access to genetic testing when indicated, likely results in the inclusion of approximately 15–25% of misclassified cases in some cohorts. In clinical practice, persistent detectable C-peptide beyond two years after diagnosis should prompt diagnostic reassessment.²⁷ These considerations underscore the need for strict inclusion criteria in studies of partial remission in T1DM, including the presence of at least two diabetes-associated autoantibodies, clear evidence of insulin deficiency at disease onset (diabetic ketoacidosis or high insulin requirements), and exclusion of cases labelled as “remission” beyond 18–24 months without diagnostic re-evaluation.^27,28 Collectively, these limitations must be carefully considered when interpreting remission data in T1DM.

Dietary Interventions in T1DM: Low-Carbohydrate Diet

Despite the absence of precise guidelines concerning the optimal proportion of macronutrients to be adopted, many people with T1DM voluntarily opt for the LCD.²⁹

LCD is a highly restrictive dietary pattern, predominantly composed of fats, with a moderate amount of proteins and a very low intake of carbohydrates. Although no universally accepted definition of this type of diets exists, A. Neyman et al have proposed definitions that are generally accepted.³⁰ According to their criteria, moderate carbohydrate restriction is defined as an intake of 26–44% of total daily calories from carbohydrates, whereas LCD provide <26%, very low-carbohydrate diets (VLCDs) consist of 20–50 g of carbohydrates per day, and ketogenic diets restrict carbohydrate intake to <20 g per day. This carbohydrate restriction is referred to as “ketogenic” because it induces nutritional ketosis. This metabolic state mimics the physiological adaptations observed during fasting but without any caloric restriction.

Foods such as cereals, potatoes, rice, fruits, starchy vegetables, pulses, and bread are replaced by other food sources rich in fats (particularly monounsaturated fatty acids and omega 3), combined with an increase in protein consumption (typically of animal origin).³¹

During ketosis, the liver produces ketone bodies (β-hydroxybutyric acid, acetoacetate, and acetone) from fatty acids through the process of ketogenesis. These ketone bodies serve as a source of ATP to compensate for the marked reduction in glucose. Fatty acids are derived from the lipolysis of triglycerides, which are present in large quantities because of high fat consumption.^32–34

Ketogenesis, as well as other compensatory mechanisms for glucose deficiency, are stimulated primarily by the lower blood insulin concentration, a direct consequence of low glycemia. This metabolic process enables the body to maintain its energy balance in the absence of available glucose, by mobilizing lipid reserves to produce ketone bodies that can be used by various tissues, notably the brain and muscles, as an alternative source of energy³¹ (Figure 1).

Figure 1.

Ketone body metabolism and energy production during the LCD. Created with BioRender.com. During the LCD, dietary fat intake increases while carbohydrate intake decreases. This shift leads to a reduction in insulin secretion. Lower insulin levels stimulate hormone-sensitive lipase (HSL) in white adipose tissue, promoting the breakdown of triglycerides into free fatty acids (FFa), which are released into the bloodstream and transported to the liver. In hepatocytes, FFA are converted into fatty acyl-CoA, then into acetyl-CoA, and subsequently into acetoacetyl-CoA. The decline in insulin levels upregulates HMG-CoA synthase, which catalyzes the conversion of acetoacetyl-CoA into HMG-CoA. This intermediate is then cleaved by HMG-CoA lyase to produce acetoacetate, a primary ketone body. Acetoacetate can be further metabolized into β-hydroxybutyrate (β-OHB) and acetone. β-OHB and acetoacetate circulate in the bloodstream to reach peripheral tissues, where they are reconverted into acetyl-CoA. This acetyl-CoA then enters the Krebs cycle (citric acid cycle) to generate ATP, providing an alternative energy source during periods of carbohydrate restriction.

Effects of LCD on HbA1c and Glycemic Parameters in T1DM

LCD has been investigated to evaluate its impact on glycemic control in patients with T1DM, with particular emphasis on HbA1c and continuous glucose monitoring (CGM)–derived parameters.

In this context, the systematic review conducted by Turton et al in 2018 examined eight studies assessing the effects of LCD on HbA1c.³⁵ One study reported follow-up HbA1c values without baseline data and was therefore excluded from the main analysis. Among the seven eligible studies, four reported no significant change in HbA1c under an LCD, while three demonstrated a statistically significant reduction (p < 0.05). Furthermore, in the two studies comparing an LCD with a high-carbohydrate diet, no significant difference between groups was observed at follow-up. With regard to insulin requirements, five studies reported total daily insulin dose. Two studies evaluating LCD observed a statistically significant reduction in insulin dose under LCD, one of which also demonstrated a significant difference compared with the high-carbohydrate control group. In the remaining studies, statistical significance could not be determined due to insufficient sample size or lack of individual participant data. Overall, the authors concluded that the current level of evidence is limited and emphasized the need for high-quality prospective trials to determine the true impact of LCD in T1DM management.

A second systematic review conducted by Paul et al in 2024 including 15 quantitative studies, assessed the effects of LCD and VLCD on HbA1c. VLCDs (≤50 g/day) were associated with a mean reduction in HbA1c of 2.9%, with statistically significant results in eight of the nine included studies evaluating this dietary pattern.³⁶ In contrast, more moderate LCDs (<130 g/day) were associated with a smaller mean HbA1c reduction of 0.4%, which reached statistical significance in only three of the six relevant studies. According to the ADA and the National Institute for Health and Care Excellence (NICE), a reduction in HbA1c of at least 0.5% is considered clinically meaningful, suggesting a clinically significant effect of VLCD in most of the included studies. However, this same review concluded that current evidence remains insufficient to definitively establish the effectiveness of LCD and VLCD in improving HbA1c among adults with T1DM. Substantial methodological heterogeneity across studies, particularly regarding dietary definitions, insulin adjustment protocols, and outcome measures, limits the generalizability of the findings.

A third systematic review by Sousa et al in 2025 included nine clinical trials evaluating the effects of LCD on CGM-derived glycemic parameters.³⁷ The findings showed a significant increase in time in range (+5.68%, p = 0.01), accompanied by a reduction in time below range, defined by thresholds of <54 mg/dL (–0.53%, p = 0.01) and <70 mg/dL (–2.28%, p = 0.01). In addition, a significant reduction in glycemic variability, as measured by the coefficient of variation (–5.51%, p = 0.01), and in total daily insulin dose (–8.39%, p = 0.01) was observed. These results suggest that LCD may not only improve mean glycemic outcomes but also enhance glycemic stability and reduce hypoglycemia risk.

Available evidence provides only limited support for a potential association between LCD and improvements in selected glycemic parameters in individuals with T1DM. Given the methodological heterogeneity and the overall low certainty of evidence, no firm conclusions regarding efficacy can currently be drawn. Further high-quality randomized controlled studies are required before any clinical implications can be considered.

In addition, significant concerns regarding the safety and monitoring of such dietary approaches have been raised by healthcare professionals.³¹ These include the risk of diabetic ketoacidosis, including euglycemic forms, potential effects on nutritional status and growth, as well as possible psychological consequences such as disordered eating behaviors or increased anxiety related to food intake.^5,38 These aspects are discussed in greater detail in the “Limitations” section of this work.

Materials and Methods

Protocol

This work was conducted as a systematic review. The objective of the study was to identify and analyze reported cases suggesting a prolongation of partial remission associated with LCD in patients with T1DM, and to explore the potential biological and metabolic mechanisms underlying the maintenance of residual β-cell function during this period.

This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, including the use of a flow diagram to document study selection, and methodological quality and risk of bias were assessed using the Joanna Briggs Institute (JBI) Critical Appraisal Tools.^39,40

Eligibility Criteria

Inclusion Criteria

Articles were selected according to the PICOS method. Inclusion criteria were: concerning the target population: patients diagnosed with T1DM, regardless of age or T1DM duration; concerning the intervention: adoption of a LCD, defined as a carbohydrate intake <26% of total daily energy intake and other more restrictive carbohydrate diets; concerning the comparison group: when available, comprised patients following standard nutritional recommendations for T1DM according to ADA/EASD guidelines, generally based on a balanced diet with insulin therapy adjusted accordingly. However, the presence of a control group was not mandatory for study inclusion. Concerning the outcomes, two outcomes were considered. The primary outcome was evidence of partial remission, assessed using C-peptide levels and/or the insulin-dose–adjusted HbA1c (IDAA1c), calculated as: HbA1c (%) + 4 × insulin dose (IU/kg/day). The secondary outcome was biological and metabolic mechanisms potentially involved in the prolongation of partial remission, including reduction of glucotoxicity, attenuation of inflammation and oxidative stress, and immunomodulatory processes.

Initially, eligibility criteria were restricted to studies conducted in individuals with T1DM. However, due to the limited number of publications specifically addressing biological mechanisms of LCD within this population, the inclusion criteria for the secondary outcome were expanded. Additional studies conducted in other populations, experimental animal models of diabetes, and in vitro models mimicking diabetic conditions were therefore included, if they investigated biological or metabolic mechanisms potentially relevant to T1DM. Importantly, the same search strategy and search equation were applied throughout the selection process to preserve consistency and limit deviation from the initial research question.

For the secondary outcome, in addition to LCD interventions, studies investigating exogenous β-hydroxybutyrate administration or fasting-based dietary regimens (including intermittent fasting and fasting-mimicking diets) were also considered, given their potential to induce nutritional ketosis and modify metabolic pathways comparable to those observed with LCD.

Eligible publications included original research articles reporting clinical observations or mechanistic data related to partial remission in the context of LCDs. Other eligibility criteria were defined as follows: randomized or non-randomized clinical trials, observational studies (cohorts, case series, cross-sectional), experimental animal or in vitro studies, systematic or narrative reviews, with full text in English, published between 2010 and 2025.

Exclusion Criteria

For the primary outcome, studies were excluded if the population did not consist exclusively of patients with T1DM. Studies involving pregnancy or other forms of diabetes (T2DM, LADA, MODY, gestational, or secondary diabetes) were excluded.

Studies were excluded if the intervention did not meet the definition of LCD (<26% of total energy intake from carbohydrates), if carbohydrate intake could not be reliably quantified, or if diet was not the primary intervention. For the secondary outcome, studies based on β-hydroxybutyrate administration or fasting-based regimens were excluded from clinical outcome analysis, as they represent indirect models of LCD.

Studies combining LCD with pharmacological treatments, supplements, vitamins, or immunomodulatory therapies were excluded.

Studies lacking relevant clinical or mechanistic outcomes, as well as editorials, conference abstracts, and publications without original data, were excluded. Only full-text articles in English within the defined time period were considered.

Methodological Consideration for Secondary Outcome

Because mechanistic studies in individuals with T1DM following LCD were scarce, the inclusion criteria for the secondary outcome were broadened to include animal models of diabetes, in vitro studies performed under diabetic conditions and human studies outside the T1DM population when mechanistic relevance was demonstrated. In addition to LCD interventions, studies investigating exogenous β-hydroxybutyrate administration or fasting-based regimens were also considered, given their ability to induce nutritional ketosis and influence metabolic pathways relevant to carbohydrate restriction.

Although this approach expanded the biological coverage, it introduced indirectness, since mechanistic data derived from experimental models or non-dietary ketogenic interventions may not directly translate to human disease processes in T1DM. Such findings should be viewed as hypothesis-generating rather than as evidence of clinical efficacy.

Search

The literature search was conducted using computerized databases, with search strategies adapted for each database. Keywords derived from the PICOS framework were used to construct the search equation (Table S1).

The search strategy was built as follows: « (« Type 1 diabetes » OR « T1DM » OR « Type 1 diabetes mellitus » OR « Diabetes mellitus, type 1 ») AND (« Low-carbohydrate diet » OR « Very-low-carbohydrate diet » OR « Ketogenic diet » OR « Low-carb diet » OR « Carbohydrate restriction » OR « Carbohydrate-restricted diet » OR « High-fat, low-carb diet » OR « fasting » OR « fasting-mimicking diet » OR « Diet, Carbohydrate-Restricted ») AND (« partial remission » OR « clinical remission » OR « clinical partial remission » OR « honeymoon phase » OR « remission » OR « Remission, Spontaneous » OR « beta-cell function » OR « Beta-cell preservation » OR « Immune modulation » OR « Insulin-Secreting Cells» OR « Mitochondria » OR « Oxidative balance » OR « Antioxidant defense » OR « Reactive Oxygen Species » OR « Oxidative Stress » OR « glucotoxicity » OR « Glucose toxicity » OR « NLRP3 inflammasome » OR « Microbial Communities » OR « Microbiota » OR « effect »).

This search strategy was adapted for the three databases used: PubMed, Scopus, and Cochrane. Results were limited to publications from the past 15 years. The searches were conducted in November 2024 (Table S2).

Study Selection

The literature search and study selection were conducted by a single reviewer. Duplicate records were first removed. Titles and abstracts were then screened according to the predefined PICOS criteria. Full-text articles were subsequently assessed for eligibility when required. When full-text publications were not accessible, attempts were made to contact the corresponding authors (Figure 2).

Figure 2.

PRISMA flow diagram of study selection. Flow chart outlining the identification, screening, eligibility assessment, and inclusion of studies in the systematic review. Records were identified through database searches (PubMed, Scopus, and Cochrane), followed by duplicate removal, title and abstract screening, and full-text eligibility assessment according to the predefined PICOS criteria. The diagram details the number of records excluded at each stage and the main reasons for exclusion, including the absence of low-carbohydrate diet exposure, lack of relevant outcomes related to partial remission or mechanistic pathways, pharmacological or immunotherapeutic interventions, and inappropriate study design.

Data Collection Process

A table was created to summarize the characteristics of the patients and the main findings of each included study, as well as to describe the reported effects of LCD on partial remission and residual β-cell function. This data extraction step was performed by a single reviewer (Table S3).

Risk of Bias in Individual Studies

Methodological quality and risk of bias were assessed using the JBI Critical Appraisal tools. Given the heterogeneity of study designs included in this review, different JBI checklists were applied according to study type (eg, case reports, case series, observational studies, and experimental studies). Each study was evaluated using the JBI checklist corresponding to its design. Narrative reviews were not included in the methodological quality assessment, as they do not provide primary data and were used only for contextual and interpretative support. The complete appraisal criteria for each study design are provided in Table S4–S8.

Literature Review

Case Presentations

A targeted literature search identified six individual case reports describing a prolonged partial remission in patients with T1DM following implementation of an LCD. These reports, summarized as Patient 1 through Patient 6 in Table 1, represent the entirety of the published clinical evidence directly addressing this question to date.^41–44

Table 1.

Clinical Characteristics and Metabolic Outcomes of Patients with T1DM Following LCD

	Patient 1 Thewjitcharoen et al	Patient 2 Bouillet et al (P1)	Patient 3 Bouillet et al (P2)	Patient 4 Bouillet et al (P3)	Patient 5 Ozoran et al	Patient 6 Benido Silva et al
Age	24 years	36 years	40 years	38 years	37 years	33 years
Sex	Male	Male	Male	Male	Male	Male
Diet initiation after diagnosis	At diagnosis	1 month post-Dx	2 month post-Dx	9 month post-Dx	6 month post-Dx	At diagnosis
Autoantibodies at diagnosis:
ANTI-GAD	+	+	+	+	+	+
ANTI-IA2	+	/	+	+	/	-
ANTI-ZNT8	+	/	/	/	/	-
Type of diet	LCD	VLCD	VLCD	VLCD	LCD (6 months); Ketogenic diet	LCD
Insulin dose (IU/kg):
- at diagnosis	0	0.25	0.19	0.58	0.3–0.5	0.5
- After 1 year	0	0	0	0	0.025–0.037	0
- After 2 years	0	0	0	0	0	0
- After 4 years	0	0	/	0.17	0.025	0
HbA1c (%):
- at diagnosis	9.4	8.9	13.8	11.5	11.6	10.6
- After 1 year	5.5	5.1	5.5	5.5	6.5	5.3
- After 2 years	5	5.3	/	5.8	5.8	5.4
- After 4 years	6	5.5	/	4.5	6.2	5.3
C-peptide (nmol/L):
- At diagnosis	/	0.33	/	0.17	0.50	0.10
- After 1 year	/	0.33	0.3	/	/	/
- After 2 years	2.15	0.3	/	0.5	/	/
- After 4 years	/	0.21–0.46	/	0.23–0.19	0.24–0.294	0.10–0.31

Notes: +: Positive. -: Negative. /: Not assessed.

All patients were diagnosed with T1DM and initiated the LCD, which enabled them to maintain optimal glycemic control (HbA1c <6.5%) for several years, often with reduced or often absent needs for insulin therapy, while preserving β-cell function, as assessed by C-peptide levels (≥0.3 nmol/L). These patients thus met the criteria for partial remission as defined by Schölin et al and Mortensen et al^13,14 (Table 1).

However, it should be emphasized that the diagnosis of T1DM in these patients was not documented uniformly, and in several cases it did not meet current consensus criteria for defining T1DM. The potential contribution of diagnostic misclassification (eg., LADA or monogenic diabetes) and its implications for the interpretation of “prolonged remission” are addressed in detail in the Discussion section.

Prolonging Partial Remission with LCD: Exploring Mechanisms

The LCD has attracted increasing interest among patients with T1DM, partly driven by a small number of anecdotal clinical reports describing prolonged partial remission in selected individuals.^41–44 Considering this growing attention, it appears relevant to summarize the biological mechanisms that have been proposed in the literature to explain why LCDs might hypothetically be associated, in specific contexts, with improved glycemic control or a transient preservation of β-cell activity.

These mechanistic hypotheses mainly involve modulation of inflammation, oxidative stress, and glucotoxicity—pathways known to contribute to β-cell dysfunction and diabetes-related complications.⁴⁵

However, it must be emphasized that most of the mechanistic evidence relies on in vitro experiments, animal models, or data from populations with metabolic conditions distinct from T1DM, such as obesity, T2DM or epilepsy. As such, extrapolation of these findings to T1DM remains speculative. Robust clinical evidence supporting a causal role of LCDs in prolonging partial remission in T1DM is currently lacking.

This section therefore does not present established clinical effects, but rather outlines mechanistic hypotheses derived from heterogeneous experimental settings. Its objective is to examine the biological plausibility of these mechanisms and to highlight the methodological limitations and uncertainties surrounding their relevance to T1DM.

Reduction of Glucotoxicity

Glucotoxicity refers to the non-physiological and potentially irreversible damage to β-cells caused by prolonged exposure to elevated glucose concentrations. Brereton et al, in 2014, demonstrated in a transgenic mouse model that chronic hyperglycemia induces significant β-cell loss and functional impairment, effects that were reversible upon normalization of blood glucose, highlighting a direct and modifiable relationship between excessive glycemia and islet viability.⁴⁶

More recent studies by Tang et al in 2018 exposed female Wistar rats and male C57BL/6J mice, either wild-type or JNK1 knockout, to prolonged hyperglycemia via continuous intravenous glucose infusion. In rats, 48 hours of severe hyperglycemia (~20–22 mM) led to impaired glucose-stimulated insulin secretion, reduced islet insulin content, and activation of the c-Jun N-terminal kinase (JNK) signaling pathway, without significant apoptosis. Similarly, in mice, moderate but sustained hyperglycemia (~13 mM for 96 hours) reduced both insulin secretion and content, as well as the disposition index, a well-established marker of β-cell compensation to insulin resistance. Pharmacological inhibition of JNK with SP600125 or genetic deletion of JNK1 preserved the expression of key β-cell functional genes, such as Ins2 and Pdx1, demonstrating that JNK activation by hyperglycemia induces β-cell dysfunction independently of other metabolic factors.⁴⁷

These findings align with earlier work by Robertson et al in 2003, who investigated the mechanisms of glucotoxicity in HIT-T15 cells in vitro and in the Zucker Diabetic Fatty (ZDF) rat in vivo. They showed that prolonged exposure to supraphysiological glucose concentrations reduced PDX-1 mRNA maturation and insulin promoter activity, leading to decreased insulin production. Additionally, β-cells were found to be particularly susceptible to oxidative stress due to their low expression of antioxidant enzymes, including superoxide dismutase (SOD), catalase, and glutathione peroxidase (GPx). In animal models, glycemic normalization or treatment with antioxidants, such as N-acetylcysteine and aminoguanidine, preserved PDX-1 expression and insulin secretion, confirming the central role of oxidative stress in glucotoxicity.⁴⁸

Experimental evidence further indicates that prolonged hyperglycemia impairs β-cell function prior to significant apoptosis. In both rat and mouse models, reductions in glucose-stimulated insulin secretion and islet insulin content precede cellular loss, suggesting that glucotoxicity initially induces functional deficits, which may progress to cell death if hyperglycemia persists.⁴⁸

In this context, Daems et al, in 2020, demonstrated that early treatment with empagliflozin, an SGLT2 inhibitor, or GABA, a neurotransmitter with β-cell regenerative properties, preserved β-cell mass and improved glucose tolerance in streptozotocin-induced diabetic mice. This protection was associated with decreased islet inflammation and endoplasmic reticulum stress, further supporting the role of glucotoxicity in β-cell decline.⁴⁹

Collectively, these data from both in vitro and in vivo models indicate that chronic hyperglycemia directly impairs β-cells through mechanisms involving oxidative stress, JNK activation, and altered expression of key insulin genes, and that these effects may be at least partially reversible through glycemic normalization or antioxidant interventions.^46–49 In T1DM, these mechanisms contribute to the partial remission phase and the progressive deterioration of β-cells, underscoring the importance of early and stringent glycemic control to mitigate the deleterious impact of glucotoxicity on pancreatic islets. Although there are currently no direct studies investigating the impact of LCD specifically, in T1DM, it is logical to hypothesize that dietary reduction of glucose intake could decrease chronic hyperglycemia and, consequently, reduce glucotoxic stress on β-cells. By limiting postprandial glucose excursions, an LCD may help preserve β-cell function and prolong the partial remission phase, but this hypothesis remains untested and requires rigorous clinical and experimental validation.

Anti-Inflammatory Effect

The NLRP3 inflammasome is a key multiprotein complex of the innate immune system, involved in the production of the pro-inflammatory cytokines IL-1β and IL-18. Its chronic activation is central to numerous inflammatory, metabolic, and neurodegenerative disorders, such as obesity, T2DM, atherosclerosis, Alzheimer’s disease, Parkinson’s disease, and T1DM.⁵⁰ Studies have shown that in obesity and diabetes, NLRP3 becomes activated in macrophages infiltrating tissues—particularly adipose tissue and the pancreas—in response to signals such as saturated fatty acids, oxidative stress, hyperglycemia, and pancreatic amyloid deposits.⁵⁰ This activation contributes to chronic low-grade inflammation, insulin resistance, β-cell destruction, and vascular complications.⁵⁰

In vivo studies in a streptozotocin-treated murine model and in vitro studies on human monocytes have shown that β-hydroxybutyrate (BHB), which is produced in significant amounts during ketogenic diets and fasting, leads in culture to reduced NLRP3 inflammasome activation, resulting in decreased IL-1β and IL-18 secretion and reduced NF-κB phosphorylation.^51,52

However, in healthy humans, the effects of BHB on the NLRP3 inflammasome remain controversial. In contrast to results obtained in rodents or cell culture—where BHB clearly inhibits NLRP3 activation—several human studies have demonstrated the opposite effect in acute settings.^50,53 For example, combined administration of BHB and lipopolysaccharide (LPS) from Gram-negative bacteria, used experimentally to stimulate the immune system, in healthy volunteers led to increased IL-1β secretion and caspase-1 activation, without changes in NLRP3 or IL1B mRNA levels. This discrepancy is partly explained by differences in NLRP3 activation mechanisms: in humans, leukocytes can be directly activated by LPS, whereas animal models require a two-step process (LPS followed by ATP). These findings suggest that the anti-inflammatory effects of BHB may primarily manifest in conditions of chronic low-grade inflammation, rather than in acute settings or in healthy individuals, and that their direct applicability to T1DM remains uncertain.⁵⁰ Consistent with this hypothesis, in vivo studies conducted in obese individuals—who typically exhibit a chronic inflammatory state—have reported more consistent and significant anti-inflammatory effects associated with BHB exposure.⁵³ However, to date, no clinical studies have specifically evaluated these effects in individuals with T1DM, highlighting an important gap in the current literature.

Rebalancing the Intestinal Microbiota

Diet is a well-established factor influencing the composition of the gut microbiota.³²

From a clinical perspective, patients with T1DM have been widely reported to exhibit increased intestinal permeability, commonly referred to as a “leaky gut” phenotype. Structural dysbiosis of the gut microbiota has been observed in T1DM, characterized by reduced microbial diversity, a decreased Firmicutes-to-Bacteroidetes ratio, depletion of short-chain fatty acid (SCFA)-producing bacteria, reduced butyrate production and bile acid metabolism, and increased lipopolysaccharide (LPS) biosynthesis. In a streptozotocin-induced murine model of T1DM, butyrate and LPS were shown to exert opposite effects on pancreatic islet structure and function, with LPS aggravating pancreatic inflammatory responses, whereas butyrate promoted β-cell–related gene expression, including Insulin1 and Insulin2.⁵⁴ Numerous recent studies have therefore investigated how the gut microbiome is altered by different lifestyle habits, including LCD.^55–60

The main changes reported included, on the one hand, an increase in butyrate-producing Firmicutes such as Allobaculum, Roseburia, and Eubacterium. On the other hand, a decrease was observed in certain Firmicutes, such as Ruminococcus gnavus, which are associated with intestinal inflammation in Bacteroidetes specialized in the degradation of complex polysaccharides and in other species with negative connotations, such as Mucispirillum.^56–58

For example, Shen X. et al conducted an experimental in vivo study in 2024 in diabetic ApoE^−/− mice induced by streptozotocin. A total of 24 diabetic ApoE^−/− mice were randomly allocated into four groups: ketogenic diet (no carbohydrate diet, KD group, n = 6), low-carbohydrate diet (LCD group, n = 6), medium-carbohydrate diet (MCD group, n = 6), and high-carbohydrate diet (HCD group, n = 6). During the feeding period, caloric intake was adjusted to ensure equal energy intake across all groups. After 6 months on their respective diets, jejunal and ileal samples were collected for analysis. These analyses demonstrated that the ketogenic diet enriched the microbiota with Allobaculum, a beneficial butyrogenic bacterium, while reducing potentially pathogenic species such as Mucispirillum and Ruminococcus gnavus.⁵⁷

However, some studies report divergent findings. In pediatrics, studies conducted in epileptic children under LCD demonstrated an increase in Bacteroides (+24% after one week of KD in infants) as well as a decrease in Firmicutes after six months of dietary intervention.^59,60 Such variations suggest that outcomes may depend on the study population, the duration of the intervention, and on weight loss itself, which independently improves the gut microbiome.^56,58 Outcomes may also depend on the type of diet implemented and the specific consumption of certain foods, such as almonds, trace elements, or the proportion of dietary fiber, which may positively influence the microbiome.^56,58

The observed modifications therefore particularly concern genera that produce SCFAs, such as Roseburia, Eubacterium rectale, and Allobaculum, whose protective role is well established. These bacteria, belonging to the Firmicutes phylum, produce butyrate. The study by Shen X. et al not only demonstrated the increase of these bacteria with ketogenic diets but also highlighted their functional effects on the intestinal barrier. Indeed, murine models fed a ketogenic diet exhibited, through increased butyrate production among other mechanisms, enhanced intestinal barrier integrity by upregulating tight junction proteins such as ZO-1, occludin, and claudin-1, thereby limiting bacterial and endotoxin translocation. This contributes to anti-inflammatory effects by reducing immune activation and oxidative stress.⁵⁷

These mechanisms are particularly relevant in the context of T1DM, where a dysbiosis characterized by reduced lactate- and butyrate-producing bacteria has been reported, with butyrate considered protective against β-cell autoimmunity.^61–63 Experimental studies in animal models have shown that modulation of the microbiota by antibiotics or probiotics can protect against T1DM onset by downregulating oxidative stress response proteins (Gpx1, GR, Cat), decreasing pro-inflammatory cytokines (IFN-γ, iNOS), and upregulating IL-10 and tight junction proteins.⁶¹

These findings indicate that targeting the gut microbiota could contribute to protection against T1DM by influencing oxidative stress at the mucosal level, modulating the pro- and anti-inflammatory balance, and possibly re-establishing the integrity of the intestinal mucosal barrier.⁶¹

Thus, although long-term human data remain limited and experimental evidence suggests that LCD may enrich butyrogenic bacteria and strengthen the intestinal barrier, it is difficult to determine whether these effects translate in vivo to individuals with T1DM, as no studies have yet been conducted under such conditions.

Furthermore, the mechanisms underlying these beneficial effects remain poorly understood and appear to depend largely on the fiber intake quality, as an excessive reduction in dietary fiber in some LCD may reduce butyrate-producing bacteria and protective metabolites.⁵⁶

Antioxidant Effect

The antioxidant properties of ketone bodies, particularly BHB, represent a central mechanism of cellular protection against oxidative stress. Numerous studies have demonstrated that excessive production of reactive oxygen species (ROS) plays a key role in the destruction of pancreatic β-cells, contributing to the progression of T1DM.^64–68 ROS-mediated damage leads to lipid peroxidation, protein oxidation, DNA damage, and the exacerbation of inflammation and apoptosis.

In this context, the LCD induces the production of ketone bodies, shifting the body’s primary energy substrate from glucose to ketones. This metabolic switch reduces mitochondrial ROS production, as ketone metabolism generates fewer electron leaks in the mitochondrial electron transport chain compared to glucose metabolism. This mechanism has been demonstrated in vivo by reduced staining of H2O2 in peripheral nerves of murine models fed a ketogenic diet. Since H2O2 is a by-product of superoxide radicals, it serves as a reliable marker of oxidative stress.^69,70

Beyond reducing ROS production, BHB also improves the cellular redox balance by decreasing the mitochondrial [NAD+]/[NADH] ratio. This shift prevents the accumulation of reductive intermediates that could promote further ROS generation. Additionally, BHB enhances the oxidation of the coenzyme Q/QH2 couple (ubiquinone/ubiquinol), thereby reducing the formation of the semiquinone radical, a major contributor to mitochondrial ROS production.⁷¹

Through these effects, BHB not only limits oxidative damage but also helps preserve the functionality of the mitochondrial NADH dehydrogenase (Complex I), preventing its inhibition by excessive reduced intermediates from the respiratory chain. This protection is critical for maintaining efficient mitochondrial metabolism.⁷²

These antioxidant effects have been confirmed in experimental in vitro models, including cultures of primary mesencephalic dopaminergic neurons and primary hippocampal neurons from rat embryos, in the context of neurodegenerative diseases.⁷¹

In addition to reducing ROS generation, BHB reinforces the cells’ endogenous antioxidant defense systems. It increases the expression of key antioxidant enzymes such as superoxide dismutase (SOD) and glutathione peroxidase, while simultaneously boosting intracellular levels of reduced glutathione (GSH), the primary scavenger of hydrogen peroxide. Furthermore, BHB activates key metabolic regulators, including AMP-activated protein kinase (AMPK) and sirtuins, which play pivotal roles in enhancing cellular resilience to oxidative stress and modulating the cellular response to ROS.^69,73

Although these protective effects have been observed in mouse models, there is currently no direct evidence from human studies demonstrating that adopting a ketogenic diet reduces ROS production and protects β-cells in people with T1DM, or that the mechanisms observed in nerve tissue also occur in pancreatic β-cells.

Changes in Amino Acid Metabolism

LCDs modify amino acid intake and metabolism. Branched-chain amino acids (BCAAs), such as valine, leucine, and isoleucine, tend to increase in the circulation due to the higher protein intake associated with LCDs. These circulating BCAAs have been suggested to play a role in the formation and expansion of regulatory T lymphocytes, which are crucial for maintaining immune tolerance and limiting autoimmune responses.⁷⁴

In line with this hypothesis, a study conducted by Elizabeth J. Mayer-Davis et al reported that higher intake and supplementation of branched-chain amino acids were associated with better preservation of β-cell function, as reflected by higher C-peptide levels, in a cohort of 1,316 youth with T1DM.⁷⁵

On the other hand, the levels of glucoforming amino acids such as proline, alanine and glutamine decrease in the bloodstream, as they are more consumed as an energy source. Glutamine is linked to the differentiation of pro-inflammatory Th1 and Th17 lineages, while reducing Treg.⁷⁴ Thus, variations in amino acids have a significant influence on the immune system.⁷⁴

Changes in Lipid Metabolism

Concerns have been raised regarding the ketogenic diet and its potential adverse effects on lipid profiles. However, available data remain inconsistent, and only a limited number of studies have specifically addressed this issue in the context of T1DM, despite heterogeneity among reported findings.^33,76–81 Some studies have reported increases in total cholesterol and LDL cholesterol, a major cardiovascular risk factor, sometimes exceeding recommended thresholds.^33,78–80 Other studies have found no significant differences in lipid profiles between individuals following LCD and reference populations.⁸² In contrast, two investigations have described more favorable changes, particularly an increase in HDL cholesterol without a concomitant rise in triglycerides.^76,77

One study proposed that the observed elevation in LDL cholesterol may result from a shift in LDL particle distribution toward larger, less atherogenic particles, potentially mitigating overall cardiovascular risk.³³ Furthermore, VLCD may preferentially increase large, buoyant LDL subfractions, which are less strongly associated with cardiovascular disease than small, dense LDL particles.^83,84 In individuals experiencing an increase in LDL cholesterol, concurrent improvements in triglycerides and HDL cholesterol may partially offset the potential adverse impact of elevated LDL levels, although this remains speculative.⁸⁵

Data from adolescents and young adults enrolled in the Diabetes Control and Complications Trial (DCCT) and its long-term observational follow-up, the Epidemiology of Diabetes Interventions and Complications (EDIC) study, provide important insights.^86,87 These analyses identified HbA1c as the strongest modifiable predictor of cardiovascular events, followed by triglycerides, whereas LDL cholesterol showed a weaker association.⁸⁴ This is consistent with the 2014 consensus statement jointly issued by the ADA and the American Heart Association (AHA), which emphasized a weaker relationship between elevated LDL cholesterol and cardiovascular events compared with poor glycemic control or other risk factors.⁸⁵ Moreover, LDL cholesterol can generally be more readily managed with pharmacological therapies, often with fewer adverse effects, whereas components of the metabolic syndrome may respond more favorably to carbohydrate restriction.⁸⁸

The quality of dietary fat plays a decisive role in shaping metabolic outcomes. For example, LDL cholesterol levels did not increase when most saturated fats in the diet (around 75%) were substituted with polyunsaturated fats.⁸⁹ In addition, a randomized controlled trial indicated that replacing carbohydrates with saturated fats did not alter hepatic fat content. However, when saturated fats were replaced with unsaturated fats, a significant reduction in liver fat was observed.⁸⁹

Nevertheless, the overall impact of these lipid alterations on cardiovascular health remains debated. A pragmatic nutritional strategy would therefore be to prioritize unsaturated fats over saturated fats, an approach likely to yield meaningful cardiometabolic benefits.

Finally, another important aspect of LCD relates to omega-3 fatty acids. End-products of their metabolism, such as eicosatetraenoic acid (ETA) and docosahexaenoic acid (DHA), exert anti-inflammatory effects, thereby contributing to the potential protective benefits of such dietary patterns.⁹⁰

Despite these findings, the impact of LCD on lipid profiles remains a major concern among healthcare professionals, who emphasize both the importance of the quality of dietary fats—favoring unsaturated over saturated fats—and the need for close clinical and biological monitoring in patients who choose to adopt such dietary approaches.^91,92

β-Cell Regeneration Due to Fasting-Mimicking Diet

An experimental study by Cheng H. et al in 2017 analyzed the effect of fasting-mimicking diets (FMD) on pancreatic islet β-cells. Their research was based on the principle that animals can survive food deprivation through an adaptive response that results in the atrophy of many tissues and organs to minimize energy expenditure. This atrophy, and its reversion after a return to a normal diet, involves stem cell-based regeneration in the hematopoietic and nervous systems.

Experimental studies were first carried out in mouse models, where a 4-day low-calorie (40% of the recommended daily caloric intake), low-protein, low-carbohydrate but high-fat fasting-mimicking diet was implemented in cycles. In C57BL/6J mice treated with streptozotocin (a chemical compound commonly used in research to induce a T1DM model), FMD cycles reduced the activity of PKA and mTOR proteins, promoting the expression of transcription factors that increase insulin production and the reprogramming of pancreatic islet cells, with gene expression comparable to that observed during fetal development.⁹³

Ex vivo experiments with human pancreatic islets have shown that treatment with serum from FMD subjects simultaneously inhibits PKA and mTOR protein kinase in T1DM pancreatic islets. This induces Sox2 and Ngn3 expression, suggesting that FMD may promote lineage reprogramming and β-cell regeneration.⁹³

These overall results indicate that fasting-mimicking diets have significant therapeutic potential for pancreatic β-cell regeneration in mice by modulating key signaling pathways and promoting cellular reprogramming.^45,93 However, caution is warranted when extrapolating these findings to humans, as rodent β-cells display a markedly higher proliferative capacity and a more dynamic cell cycle turnover compared to human β-cells, making in vivo regeneration much more challenging in clinical settings. Despite promising findings, additional studies are required to assess the applicability of these ex vivo results in humans (Figure 3).

Figure 3.

Beneficial effects of a LCD in T1DM. Created with BioRender.com. This figure illustrates the potential mechanisms by which the LCD could act in T1DM, categorized by macronutrient intake. (1) Reduced carbohydrate intake leads to increased β-hydroxybutyrate levels, which inhibit the NLRP3 inflammasome and promote regulatory and memory T cell expansion (anti-autoimmune effect). (2) Lower carbohydrate intake also reduces hyperglycemia, thereby decreasing glucotoxicity and ROS production (toxicity reduction). (3) Fasting-mimicking diets associated with LCD may induce reprogramming of pancreatic β-cells via modulation of signaling pathways such as PKA, mTOR, Sox2, and Ngn3 transcription factors. (4) Increased LDL-cholesterol, a cardiovascular risk factor. (5) Increased lipid intake, particularly omega-3 fatty acids, enhances levels of ETA and DHA, contributing to anti-inflammatory effects. (6) Higher protein intake shifts energy metabolism away from gluconeogenesis from amino acids and suppresses pro-inflammatory Th1 and Th1 cells (anti-autoimmune effect). (7) Increased intake of branched-chain amino acids (BCAAs) supports the biosynthesis and expansion of regulatory T cells (anti-autoimmune effect). Color coding indicates the level of evidence supporting each mechanism: data derived from human T1DM, human studies with inconsistent findings, STZ-induced diabetic mice, or healthy human populations. Most mechanisms remain supported by indirect or experimental evidence and should be interpreted as hypothetical.

Discussion

Comments on Cases

Comparison with General T1DM Population

The clinical trajectories observed in these cases are highly atypical compared to those generally reported in individuals with T1DM in whom the partial remission phase lasts only a few weeks to one year.¹⁰ In contrast, the patients described in this review maintained near-normal glycemic control (HbA1c < 6.5%) over several years, with reduced or even absent insulin requirements, while preserving pancreatic β-cell function, as indicated by sustained C-peptide levels ≥0.3 nmol/L.

By comparison, a large international study including over 320,000 individuals with T1DM showed that only a minority achieved an HbA1c below 58 mmol/mol (<7.5%), with wide variability across age groups: 15.7% to 46.4% in children under 15 years, 8.9% to 49.5% in adolescents and young adults aged 15–24 years, and 20.5% to 53.6% in adults aged 25 years and older.⁹⁴ These findings highlight how difficult it is—despite appropriate insulin therapy—to achieve long-term glycemic stability in most patients with T1DM.

Methodological Limitations and Sources of Bias

Diagnostic Uncertainty and Misclassification Bias

The cases of prolonged partial remission of T1DM reported in the literature should be interpreted with considerable caution, as they may reflect diagnostic misclassification rather than genuine remission of T1DM. According to the diagnostic framework proposed by the UK T1D Immunotherapy Consortium in 2022, a robust diagnosis of T1DM requires: (i) the presence of at least two pancreatic autoantibodies (GAD, IA-2, ZnT8, IAA); (ii) a C-peptide concentration <0.2 nmol/L approximately three years after diagnosis; (iii) either diabetic ketoacidosis at onset or an insulin requirement exceeding 0.7 IU/kg/day; (iv) progressive insulin dependence.²⁸ In classical T1DM, β-cell destruction is rapid, and the partial remission phase typically lasts no longer than 3–12 months; persistence beyond 18–24 months is considered discordant with the natural course of autoimmune T1DM and should prompt re-evaluation of the initial diagnosis.^95,96

LADA represents a major diagnostic confounder in this context. Defined by adult-onset autoimmune diabetes with slower β-cell decline, LADA is characterized by single-autoantibody positivity (most commonly GAD), delayed insulin dependence, and preservation of endogenous insulin secretion for several years.^97,98 In contrast to T1DM, in which C-peptide levels usually become undetectable within two to three years, individuals with LADA often retain measurable C-peptide concentrations for five to ten years, closely mimicking a “prolonged remission” phenotype.⁹⁹ Likewise, MODY patients preserve β-cell function over decades and may be erroneously classified as T1DM in the absence of genetic testing.²⁵

A critical analysis of the six published “prolonged remission” cases identified in this review reveals several major inconsistencies with contemporary diagnostic standards. Two of six patients underwent incomplete autoantibody testing, limited to single GAD measurement, thereby precluding confirmation of autoimmune T1DM.^42,43 None presented with diabetic ketoacidosis at diagnosis despite marked hyperglycemia, and all maintained detectable C-peptide levels several years after diagnosis.^41–44 Moreover, the reported durations of “remission” (15 months to over six years) closely resemble the natural course of LADA far more than that of classical T1DM.^28,97–99

These findings strongly suggest that several, if not most, published cases labeled as “prolonged remission of T1DM” likely represent misclassified LADA or MODY rather than disease modification induced by LCD. The inclusion of such cases introduces substantial ascertainment bias and leads to overestimation of the true frequency and duration of remission in T1DM. Furthermore, because all patients concurrently adopted dietary restriction, it is impossible to dissociate a putative dietary effect from the intrinsic metabolic trajectory of slowly progressive diabetes.

In the absence of systematic autoantibody profiling, longitudinal C-peptide monitoring, and genetic testing when indicated, remission studies in T1DM risk include up to 15–25% false diagnoses. Future investigations should therefore apply stringent inclusion criteria, including multiple autoantibody positivity, objective evidence of insulin deficiency, and standardized C-peptide follow-up, in order to avoid conflating prolonged insulin independence with true autoimmune remission.

Geographic Bias and Limited Generalizability

Another limitation of this review relates to geographic bias. Most of the available studies were conducted in high-income Western countries or in Asia, which may limit the generalizability of these findings to other regions with different healthcare systems, dietary patterns, and access to medical support.^41–44

Uncontrolled Confounding Factors

Finally, multiple confounding factors—including treatment adherence, intensified medical follow-up, concomitant lifestyle changes (eg., physical activity, weight loss), and individual genetic characteristics—may have independently influenced the outcomes.^100,101 At this stage, these findings should therefore be regarded as anecdotal observations rather than clinical evidence and do not justify any clinical recommendation in favor of LCD for prolonging partial remission in T1DM.

Comments on the Proposed Mechanisms

Scope and Purpose of the Mechanistic Analysis

Among the mechanisms involved in the prolonged partial remission observed with the LCD, four primary and complementary pathways have been identified: the reduction of glucotoxicity, anti-inflammatory effects, decreased oxidative stress, and potentially regenerative mechanisms.

It is important to clarify that the purpose of this section is not to promote LCD as a therapeutic strategy in T1DM, but rather to describe and critically examine the biological mechanisms that have been reported in the literature. The mechanisms discussed herein predominantly reflect pathways that have been hypothesized to explain potential beneficial effects observed in selected experimental or clinical contexts. Their inclusion does not imply that these mechanisms are proven, clinically relevant, or universally applicable to individuals with T1DM.

Furthermore, the present work does neither aim to provide dietary recommendations, nor to assess the overall clinical efficacy or safety of LCD. Instead, it focuses on analyzing how LCD-related metabolic changes might influence inflammation, oxidative stress, immune regulation, and β-cell function, based on currently available evidence. Many of these mechanisms are derived from in vitro studies, animal models, or populations distinct from T1DM, and therefore remain speculative when extrapolated to clinical practice.

Consequently, the description of predominantly “positive” mechanisms should not be interpreted as an endorsement of LCD, but as a reflection of how the existing literature has framed these hypotheses. A critical perspective is maintained throughout, and the limitations, uncertainties, and contradictions within the available evidence are explicitly acknowledged.

Methodological Constraints and Population Selection

Initially, the inclusion criteria of this review were designed to be highly restrictive, focusing exclusively on studies involving individuals with T1DM, diabetic animal models, or in vitro experiments conducted in diabetic cellular environments. However, during the literature screening process, it became evident that the number of available studies meeting these criteria was extremely limited and insufficient to support a comprehensive mechanistic analysis.

Consequently, the inclusion criteria were broadened to incorporate studies conducted in other populations or experimental models if they addressed biological mechanisms relevant to LCD. This methodological adaptation allowed for a more extensive exploration of potential mechanistic pathways but inevitably introduced greater heterogeneity in the study population and experimental conditions.

This broadened scope reinforces the need for caution when interpreting these findings, as mechanisms described in non-T1DM contexts may not be directly translatable to individuals with T1DM. Therefore, the mechanistic insights presented in this work should be interpreted as hypothesis-generating rather than as evidence of clinical efficacy.

Limitations and Safety Directions

Limitations

The purpose of this review was to describe the biological mechanisms underlying LCD; nevertheless, potential risks associated with such dietary patterns should be clearly communicated.

While current findings remain uncertain and often mixed, LCDs are sometimes accused of increasing the risk of short-term complications such as hypoglycemia and diabetic ketoacidosis (DKA). However, available data have not shown a significantly higher incidence rate of these complications compared to normal levels.^{31,78,102–104} Despite low rates of acute T1DM-related complications in some studies, the study by Buehler et al underscores the potentially serious risks of unmanaged ketosis, particularly in patients with T1DM.³⁸ It suggests that continuous glucose monitoring could be employed to prevent episodes of hypoglycemia or DKA when following a LCD.³¹

In the long-term, LCDs are also suspected of causing adverse effects, including dyslipidemia, nutritional deficiencies, growth disturbances in children, bone metabolism disorders in adults, electrolyte imbalances, and psychological disorders.^{5,31,105–109} Poor adherence to this type of restrictive diet is another significant concern associated with LCD over the long-term.³¹

As explained above, one of the main concerns of healthcare providers is the increased fat intake used to compensate for carbohydrate reduction in LCD.

However, the exact impact of these lipid modifications on cardiovascular health remains subject to debate.^33,76–81 An alternative approach is to prioritize unsaturated fats as substitutes for saturated fats, a strategy that could provide significant cardiometabolic health benefits.⁸⁹

These observations underscore the need for rigorous nutritional monitoring, with particular attention paid to the composition of fat and dietary fibres, to minimize risks and maximize health benefits for patients following the LCD.⁷⁸

Another frequently reported risk was nutritional deficiencies, particularly in vitamins, minerals, and fibres, due to restrictions on carbohydrate-containing foods such as fruits, vegetables, and whole grains. These foods are key sources of nutrients like iron, calcium, thiamine, vitamin B6, iodine, and fibres, which may become insufficient in individuals following the LCD.^107,110,111 To mitigate these risks, close dietary supervision is essential. Dietitians should assess nutritional intake, identify potential deficiencies, and recommend appropriate dietary substitutions or supplements when necessary.^31,33,78

This type of restrictive diet also poses significant challenges related to psychological and social implications, as well as long-term adherence. Such diets come with social constraints, including isolation in social settings involving carbohydrate-rich foods, particularly for children and adolescents. These restrictions may also lead to family conflicts or an increased risk of eating disorders.^{5,31,76,109,112}

Long-term adherence to LCDs remains a major issue. In a four-year study conducted by Nielsen JV et al less than 50% of participants maintained adherence after two years, leading to increased HbA1c levels. The lack of adequate psychological and nutritional support exacerbates these challenges, as participants often report misunderstanding or a lack of support from healthcare professionals.⁷⁶

Safety Directions

Healthcare professionals often express reluctance to recommend LCDs due to the lack of large-scale, well-designed, and long-term randomized studies. Additionally, existing studies frequently suffer from biases, notably the absence of reliable data on patient adherence to their diets. Furthermore, patient monitoring and follow-up protocols are rarely well-described and lack standardized guidelines.

Drawing inspiration from recommendations for epileptic patients on ketogenic diets, specific guidelines have been adapted by Seckold, R. et al for monitoring children with T1DM on LCD, although these have yet to achieve expert consensus.⁷⁸ These recommendations emphasize regular and multidisciplinary monitoring.

From a nutritional perspective, they include an initial assessment by a dietitian, followed by quarterly evaluations of weight, height, BMI, adequacy of the dietary plan (calories, protein, fibres, and fluids), and the need for vitamin and mineral supplements such as iron, calcium, and B vitamins.

Adherence to the diet, including management of social situations (eg, school, parties), is also reviewed every three months.

From a medical standpoint, quarterly consultations with an endocrinologist assess glycemia, insulin dosage, and overall effectiveness of the diet relative to parental expectations.

Comprehensive laboratory tests (hematology, electrolytes, liver, kidney, and lipid profiles, as well as urinalysis) are conducted initially, at three months, and then annually. Additional evaluations, such as renal ultrasound and bone density scans (DEXA), are performed at baseline and renewed annually to monitor potential long-term side effects.

These recommendations were primarily developed for pediatric populations. Although certain principles may be adapted to adult patients, there are currently no formal, evidence-based guidelines specifically established for adults. However, their implementation requires rigorous monitoring and individualized support to provide optimal care for patients living with diabetes.

Conclusion

In summary, this work does not demonstrate that LCDs prolong partial remission in T1DM, nor does it provide evidence supporting their clinical effectiveness in preserving β-cell function. Instead, it highlights a set of biological mechanisms that have been proposed in the literature to explain how metabolic changes induced by carbohydrate restriction might influence inflammation, oxidative stress, immune pathways, and β-cell homeostasis.

The mechanistic framework presented here is largely based on experimental models, indirect inferences, and heterogeneous clinical populations. In the absence of controlled trials, standardized diagnostic criteria, and long-term outcome data in T1DM, these mechanisms must be regarded as theoretical rather than causal. Moreover, the rarity of reported cases, the high likelihood of publication bias, diagnostic uncertainty, and the presence of multiple confounding factors preclude any reliable conclusion regarding a beneficial clinical effect of LCD in this population.

Rather than establishing LCD as a therapeutic option, this review underscores major gaps in knowledge and highlights how poorly current mechanistic hypotheses translate into evidence-based clinical practice. The value of this work therefore lies in organizing existing mechanistic concepts, identifying inconsistencies in the literature, and clarifying research priorities, rather than in supporting a dietary intervention.

Future research should focus on rigorously designed clinical trials, mechanistic studies conducted in authentic T1DM models, and standardized definitions of remission, to determine whether any of the proposed pathways translate into clinically meaningful outcomes.

Funding Statement

This research received no external funding.

Abbreviations

8-OHdG, 8-hydroxy-2′-deoxyguanosine; 8-epi-PGF2α, 8-epi-prostaglandin F2α; ROOH, Lipid hydroperoxide; NFκB, Nuclear Factor kappa-light-chain-enhancer of Activated b cells; AP-1, Activator Protein 1; IFN-γ, Interferon-gamma; TNF-α, Tumor Necrosis Factor-alpha; HLA, Human Leukocyte Antigen; CTLA4-Ig, Cytotoxic T-Lymphocyte Antigen 4 Immunoglobulin; SGLT2, Sodium-Glucose Cotransporter 2; GABA, Gamma-Aminobutyric Acid; AMPK, AMP-Activated Protein Kinase; PKA, Protein Kinase A; mTOR, Mechanistic Target of Rapamycin; Sox2, SRY (Sex-Determining Region Y)-Box 2; Ngn3, Neurogenin 3.

Data Sharing Statement

No new data were generated or analyzed in this study. Data sharing is not applicable to this article.

Acknowledgements

Mariem Gdoura has moved to a new institute and wishes to acknowledge the Pediatric Clinical Investigation Center, Cliniques universitaires Saint Luc in Brussels, Belgium.

Author Contributions

Margaux Rittweger: Conceptualization, methodology, investigation, data curation, writing – original draft.

Prof. Lysy: Conceptualization, supervision, resources, validation, writing – review and editing.

Dr. Gdoura: Methodology, validation, writing – review and editing.

All authors gave final approval of the version to be published; have agreed on the journal to which the article has been submitted and agreed to be accountable for all aspects of the work.

Disclosure

The authors declare no conflicts of interest.