Defining a ketone threshold for weight loss: evidence from 14 day daily β-hydroxybutyrate monitoring in 217 subjects on a ketogenic diet

Abstract

Background

The ketogenic diet (KD) is a widely used nutritional intervention for weight loss. The beneficial effects of the KD are intrinsically linked to the state of physiological ketosis, where ketone bodies (KBs) raise, even though minimal effective threshold of blood ketone concentration that correlates with significant weight loss remains unclear. Therefore, the main purpose of this study was to identify the optimal β-hydroxybutyrate (βHB) threshold associated with weight loss in individuals with overweight or obesity undergoing a KD.

Methods

This secondary analysis included 217 participants (111 males and 106 females) with overweight or obesity, who followed a KD for 14 days. Time to Ketosis (TtK)—defined as the number of days needed to reach and maintain a given ketone concentration—was calculated for each threshold.

Results

Regression analysis showed that a βHB concentration of ≥0.5 mmol/L was the most associated with significant weight loss. Moreover, body weight and gender significantly influenced TtK, suggesting interindividual variability in achieving effective ketosis.

Conclusions

Achieving and maintaining a ketonemia of at least 0.5 mmol/L may represent a clinically meaningful threshold to optimize weight loss in individuals undergoing a KD. Monitoring βHB levels and reducing TtK may improve individual responsiveness to KD-based interventions.

Introduction

Globally, nearly 40% of adults are classified as overweight or obese, representing a critical public health concern. Well over five million deaths annually are attributable to non-communicable diseases and are linked to overweight or obese conditions, including cardiovascular diseases, various forms of cancer, Alzheimer’s disease, and type 2 diabetes [1]. Currently, Body Mass Index (BMI) remains the most popular and fastest criterion to assess the condition of overweight (defined as a BMI between 25.0 and 29.9 kg/m²) and obesity (classified as a BMI of 30.0 kg/m² or higher) [2].

However, obesity is considered a complex multifactorial disease characterized by excess adiposity with causes that are still incompletely understood. Clinical obesity has been recently defined as a chronic and systemic illness, with alterations in the function of tissues, organs or the organism as a whole, where its development and progression are driven by an interplay of environmental, behavioral, and genetic factors [2]. Among modifiable factors, dietary patterns play a pivotal role in influencing the progression and management of obesity [3].

Among these, the ketogenic diet (KD) is considered a safe nutritional intervention to lose weight and improve global metabolic health [4]. KD is characterized by a significant reduction in carbohydrate intake (below 30 to 50 grams per day, or less than 5% of total daily energy intake) while maintaining an adequate protein and variable fat intake [5].

Interestingly, although KDs are frequently described in the scientific literature under a single umbrella term, they actually encompass a wide range of protocols with substantial variability in both caloric content and macronutrient composition. While the general principle of high fat and very low carbohydrate intake remains a common denominator, dietary regimens are often labeled as “ketogenic” even though they are simply a low carbohydrate diet (more than 50 grams of carbohydrate per day) or very low-calorie diets. Moreover, macronutrients are often reported as percentages of total energy intake rather than absolute grams, which introduces additional variability: in high-calorie diets, even a 5% carbohydrate allocation can correspond to a relatively high absolute intake of carbohydrates, potentially compromising the degree of ketosis. Similarly, fat intake can range widely depending on whether the diet is isocaloric, hypocaloric, or hypercaloric. This heterogeneity complicates comparisons between studies and underscores the importance of specifying the precise dietary composition when interpreting the effects of KDs on health and performance outcomes [6, 7].

In an effort to clarify the existing terminological confusion, Trimboli and colleagues proposed a clear and linear classification of KDs, categorizing them as follows: (I) very-low-calorie diet (VLCD), with energy intake below 700–800 kcal/day and fat intake less than 30–40 g/day; (II) low-calorie diet (LCD), with 700–800 kcal/day and fat intake above 30–40 g/day; and (III) isocaloric KD, aligned with total energy expenditure (TEE), where 70–80% of energy intake is derived from fat [8]. It has been suggested that KD may promote weight loss through several interrelated processes, including appetite suppression, increased fat oxidation, and increased energy expenditure due to the thermic effect of protein and gluconeogenesis [9].

However, due to methodological limitations such as not following a randomized controlled trial design [10], poor description of the diet composition [11] (e.g. diet low in carbs and inadequate in fat content), absence of control of blood or urinary ketone levels [12], and inconsistent adherence [12], the real health benefits of KDs compared to traditional diets remain the subject of ongoing debate.

The beneficial effects of the KD are intrinsically linked to the state of “physiological ketosis,” a concept introduced by Hans Krebs in 1966 [13]. Physiological ketosis, induced by carbohydrate restriction or fasting [14], is characterized by normal systemic pH, low but physiologically regular insulin and blood glucose levels, and ketone body (KB) concentrations between 0.3 and 4 mmol/L [6]. During ketosis, metabolism shifts from glucose to fatty acids and ketones as primary fuel sources. In contrast, pathological ketosis, typically seen in decompensated diabetes, is characterized by a low pH, insulin deficiency, hyperglycemia, and KB levels above 7–8 mmol/L [15]. While cutoffs ranging from 0.3 to 0.5 mmol/L are typically used [16–19], higher levels may be required to elicit therapeutic effects in several pathological conditions. For instance, in patients on KD for drug-resistant epilepsy [20] and GLUT1 deficiency syndrome [21], a serum ketone above 2 mmol/L is recommended. Conversely, a higher ketone level ( > 4 mmol/L) has been proposed for metabolic therapy in glioblastoma patients [22], recognized as the most aggressive primary brain tumor in adults.

Till now, the precise threshold for defining when a person is “in ketosis” remains unsettled.

Despite previous studies have reported associations between ketone levels and metabolic effects, these works did not address a clinically relevant question: what is the minimal βHB concentration required to effectively predict weight loss during a KD?

To our knowledge, no study has systematically evaluated this issue in a large human cohort with daily direct blood ketone monitoring. Currently, no clear data are available about the minimal level of KBs needed to exert a significant effect on weight loss. Moreover, an underexplored yet potentially critical variable in ketogenic therapies interventions is the Time to Ketosis (TtK), defined as the number of days required to reach the minimal level of blood ketones required for different therapeutic aims. Individual variability in the rate of ketosis onset could significantly impact the overall effectiveness of KD for weight loss.

Therefore, this study aimed to:

1. explore the influence of individual characteristics, such as gender, age, BMI, and different ketogenic protocols, on both TtK and weight loss;

2. identify the most appropriate threshold of blood ketones (among 0.3, 0.4, and 0.5 mmol/L) that better correlate with weight loss;

3. assess whether TtK influences the magnitude of weight reduction.

In this sense, the present analysis does not aim to describe mechanistic pathways of ketone metabolism, but to define, for the first time in a large real-life sample, a practical βHB associated with weight loss success under KD conditions. This threshold may serve as a pragmatic biomarker of adherence and effectiveness, allowing clinicians and dietitians to monitor patient progress and anticipate weight loss outcomes during KD interventions.

Materials and methods

Description of the protocol

In this study, we conducted a secondary analysis of previously published data from our group [23–27]. Additionally, we included further unpublished dataset collected in our laboratory between 2020 and 2022 (NCT04163120). Inclusion criteria included a BMI greater than 25 kg/m² and an age between 20 and 68 years. Exclusion criteria were as follows: (i) those receiving treatment for conditions such as diabetes, cardiovascular diseases, or depression; pregnancy and lactation; hormonal therapy; and (ii) individuals engaging in more than two hours of exercise per week. All subjects analyzed adhered to one of two dietary protocols: a “classical” KD (CKD) or a KD enriched in phytoextract (KEMEPHY). Ketonemia was reported daily for two weeks. Finally, 217 subjects—111 males and 106 females were included in the analysis. Body mass was measured daily, after an overnight fast, to the nearest 0.1 kg using an electronic scale (Tanita BC-545N; Amsterdam, Netherlands). Height was measured to the nearest 1 cm using a wall-mounted stadiometer (GIMA S.p.a., Milan, Italy). Blood ketone levels were assessed in all studies, after the overnight fast and before breakfast, using the same instrument and standardized procedure (Precision Xtra Blood β-Ketone Test Strips and Precision Xtra - Abbott Laboratories, Chicago, Illinois) [28]. Briefly, blood samples were collected by puncturing the skin with single-use lancing devices. Before each puncture, the skin was meticulously cleaned and dried. The first droplet was wiped away with a cotton swab, and the subsequent droplet was collected by the electrochemical strip for analysis. Then, the electrochemical strip was inserted into the meter, and the βHB concentration was displayed in mmol/L after 30 s.

Dietary protocol

All subjects followed a KD, one subgroup followed a “classical” KD (CKD) (n = 71), while the other one followed a Mediterranean KD with phytoextracts (KEMEPHY) (n = 146). In both diets, summarized in Table 1, daily carbohydrate intake was restricted to below 30 g/day, protein intake was around 1.2–1.5 g/kg/day, and fats provided between 50% and 70% of total daily energy intake. Energy intake ranged from approximately 1000 to 1600 Kcal in both protocols. The KEMEPHY dietary protocol, previously described in detail in earlier studies [24, 25, 29], included several commercially available ready-to-eat ketogenic products and specific phytoextracts (Tisanoreica, Gianluca Mech SpA, Asigliano Veneto, VI, Italy) [30]. The composition of these extracts has been reported in detail elsewhere [29]. Dietary protocols, together with related information including dietary instructions and the methods used to assess dietary intake and adherence during the intervention, are described in depth in each study [23–27].

Table 1.

Macronutrient composition/distribution of KEMEPHY and CKD (data are expressed as mean and SD).

	KEMEPHY group	CKD group
Kcal/day	1252 ± 109	1433.8 ± 145
Protein (% total daily Kcal)	36.6 ± 3.6	30.9 ± 2.4
Fat (% total daily Kcal)	55.9 ± 6.6	63.1 ± 2.7
Carbohydrate (% total daily Kcal)	8.7 ± 1.7	6.1 ± 1.0
Protein (g/day)	109.8 ± 10.2	109.3 ± 4.6
Protein (g/Kg body weight/day)	1.2 ± 0.8	1.3 ± 0.2
Fat (g/day)	79.7 ± 9.7	100.7 ± 10.8
Carbohydrate (g/day)	26.3 ± 3.4	22.3 ± 3.0

Statistical analysis

Our dataset consists of 217 observations and 14 variables:

Type of diet (CKD, KEMEPHY); Gender (male, female); Age (years); Weight (kilos); Height (cm); BMI; Percentage change of weight loss: [(weight 21^th day – weight 1^st day)/weight 1^st day]*100; Number of days in ketosis with thresholds of 0.3 mmol/L; Number of days in ketosis with thresholds of 0.4 mmol/L; Number of days in ketosis with thresholds of 0.5 mmol/L; First day in ketosis with thresholds of 0.3 mmol/L; First day in ketosis with thresholds of 0.4 mmol/L; First day in ketosis with thresholds of 0.5 mmol/L; Daily amount of weight loss (kg).

A descriptive analysis of the sample was first conducted, and the normal distribution of the variable “percentage change of weight loss” was verified. To address the research questions presented in the introduction section, the following analyses were performed:

1. The appropriate level of ketosis was determined through linear regression analysis, with weight loss as the response variable and time to ketosis at the considered level (0.3, 0.4, or 0.5) as explanatory variables, along with covariates such as gender, age, type of diet, BMI, weight, and height. After conducting the three regression models, the level of ketosis associated with the highest regression parameter estimate was chosen as the appropriate level, indicating the level of ketosis resulting in the greatest weight loss. Moreover, we have considered the R-squared goodness of fit metrics for all three models to better substantiate the selection of the model.

2. To define which variables (and to what extent) influence entry into ketosis, a Poisson generalized linear model was employed, with the day of entry into ketosis at the chosen level of ketosis from point 1 as the response variable, and gender, age, initial BMI, initial weight, height, and type of diet as explanatory variables. Possible interactions among these variables were also analysed.

3. Finally, to determine which variables (and to what extent) influence weight loss, a linear regression analysis was conducted with weight loss as the dependent variable and gender, age, initial BMI, initial weight, height, type of diet, first day in ketosis and days in ketosis as explanatory variables. Possible interactions among these variables were also analysed. In this analysis we are particularly interested in whether the type of diet leads to different outcomes in terms of weight loss.

Findings from an additional sensitivity analysis indicated that including both BMI and body weight in the regression models does not influence the estimation of the association between TUK and weight loss. Consequently, despite the high correlation between some variables, such as BMI and body weight, multicollinearity does not distort the parameter of interest in our analysis.

The analyses were performed using R and SAS software (version 9.4). All data are presented as mean and standard deviation (SD). Statistical significance was set at p < 0.05.

Results

Sample description

A total of 217 subjects, 111 males and 106 females.

Characteristics of the participants are shown in Table 2.

Table 2.

Descriptive statistics for age, as well as the initial values of weight, height, and BMI for all subjects and categorized by gender.

	n	Age (yrs)	Body mass (Kg)	Height (cm)	BMI (Kg/m2)	% Weight loss
All subjects	217	46.53 ± 10.35	88.60 ± 15.54	167.36 ± 8.44	30.34 ± 4.01	6.96 ± 2.70
Females	106	43.95 ± 10.33	83.73 ± 12.03	164.10 ± 6.98	30.34 ± 4.01	6.29 ± 2.99
Males	111	48.99 ± 10.10	94.78 ± 17.30	174.17 ± 6.76	31.10 ± 4.39	7.60 ± 2.22

Data are presented as mean and standard deviation where applicable, along with the percentage of weight loss.

Level of ketosis

According to the performed regression analyses, the appropriate threshold level where the greatest weight loss occurs, seems to be at 0.5 mmol/L. Specifically, the regression parameters of the three models are 0.330 (p-value = 0.0003, R² = 0.2067) at the 0.3 mmol/L level, 0.456 (p-value < 0.0001; R² = 0.3241) at the 0.4 mmol/L level, and 0.525 (p-value < 0.0001, R² = 0.4002) at the 0.5 mmol/L level.

Of interest 59.91% of subjects reached 0.5 mmol/L βHB between 5 and 8 days. The mean was 7.7 days with SD of 2.62.

Factors influencing entry in ketosis

The results of the Poisson model referred to the day of entry into ketosis at 0.5 mmol/L are presented in Table 3.

Table 3.

Poisson model of the day of entry into ketosis at 0.5 mmol/L.

Variable	Estimate	Std error	p-value
Intercept	-0.4768	3.3290	0.8861
Gender (Female)	-0.0679	0.0689	0.3242
Diet (Classic)	0.2552	0.0717	0.0004
Age (years)	0.0041	0.0025	0.1066
Initial BMI (kg/m2)	0.0330	0.0523	0.5282
Initial body mass (kg)	-0.0108	0.0187	0.5623
Height (cm)	0.0129	0.0200	0.5199
Gender*Diet	0.2914	0.1066	0.0062

The type of diet, which strongly interacts with gender, seems to strongly impact on TtK. Specifically, adoption of the CKD postpones entry into ketosis compared with the KEMEPHY diet in both genders, but for females the postponement is far more (the logarithm of the day of entry into ketosis is greater than 0.2552 for the classical diet compared with the KEMEPHY diet in males, and is greater than 0.5466 in females).

Factors influencing weight loss

Our data suggests that gender and the number of days in ketosis (βHB > 0.5 mmol/L) can influence weight loss (Fig. 1), but the type of diet interacts with gender (Table 4). Indeed, we note that women are at a disadvantage compared to men, and that the differences depend on the type of diet. In fact, with the KEMEPHY diet, women have more than 1 percentage points less weight reduction than men. In contrast, with the CKD, the disadvantage of women is almost 3 percentage points (2.9795 = 1.0057 + 1.9738).

Fig. 1. The influence of gender and days in ketosis on weight loss.

A Boxplots for the % of weight loss based on the number of days in ketosis above 0.5. B Boxplots for the % of weight loss based on the first day at 0.5. C Boxplots for the % of weight loss based on gender. D Boxplots for the % of weight loss based on type of diet and gender.

Table 4.

Linear regression model for the percentage weight loss.

Variable	Estimate	Std error	p-value
Intercept	-12.1108	20.1898	0.5493
Gender (Female)	-1.0057	0.3814	0.0090
Diet (Classic)	0.5269	0.4436	0.2363
Age (yrs)	-0.0196	0.0146	0.1795
Initial BMI (kg/m2)	0.1979	0.3077	0.5209
Initial weight (kg)	-0.0707	0.1099	0.5203
Height (cm)	0.1005	0.1192	0.4003
Gender*Type	-1.9738	0.6806	0.0041
Number of days in ketosis	0.5063	0.1457	0.0006
Day of entry into ketosis at 0.5 mmol/L	0.0156	0.1469	0.9155

As the number of days in ketosis increases, weight loss increases (Fig. 2). It should be noted that the performance of the model is decidedly good (R² = 0.424), given the small number of explanatory variables.

Fig. 2. Heat map depiction of the dataset.

The covariate most correlated with weight loss is TUK 0.5 (Time Under Ketosis), defined as the number of days in ketosis with a threshold above 0.5 mmol/L. Heat map depiction of the dataset, showing that the covariate most correlated with weight loss is TUK 0.5 (Time Under Ketosis), defined as the number of days in ketosis with a threshold above 0.5 mmol/L. TtK (Time to Ketosis), defined as the number of days needed to reach and maintain a given ketone concentration, was included for each threshold. The term ketonemia refers to the levels of ketone bodies in the blood, measured at the capillary level.

Discussion

Our results suggest that, during a KD, 0.5 mmol/L of blood KB may be considered a reliable threshold (i.e., minimum value) when the weight loss is the primary outcome. Furthermore, it seems that gender and type of KD could influence the time, in terms of days, to achieve the state of nutritional ketosis. Even though in the last 20 years there has been a surge of scientific research about the KD, there is a relative dearth of data about the precise blood KB concentration to define the reaching of ketosis. In addition, considering the different mechanisms involved in KB’s effects on seizures [31], anorexigenic effects [6], anti-inflammatory effects [32], etc., different thresholds for “ketosis” could be needed. Based on that, the minimum KB level to be considered under ketosis during a weight loss program is still not well defined.

From a biochemical standpoint, plasma ketone body concentrations reflect the dynamic balance between hepatic ketogenesis and extrahepatic ketolysis [33]. According to Volek and Phinney, the hallmark of KD lies in the induction of ketosis and stabilization of insulin levels. Lower insulin availability promotes lipolysis by reducing the inhibitory effect of insulin on hormone-sensitive lipase [13]. As such, ketonemia may be considered a proxy for lipid mobilization and oxidation [34]. Nonetheless, β-hydroxybutyrate (βHB), the main KB, may exert direct metabolic and endocrine effects. Several studies have reported that βHB can influence the secretion of appetite-related hormones, particularly ghrelin, whose circulating levels appear inversely correlated with βHB concentrations [35]. Sustained levels of βHB may modulate resting energy expenditure (REE). As reported by Ebbeling et al., under isoenergetic conditions, carbohydrate-restricted diets were more effective than low-fat diets in maintaining REE during the weight loss phase [36]. Collectively, these observations support the hypothesis that higher and sustained levels of βHB contribute to greater weight loss via a synergistic effect of increased lipolysis, suppressed appetite, and preserved energy expenditure [34]. In line with other studies, a minimum value of 0.5 mmol/L is indicative of stable ketosis, and it is also linked to the resolution of the typical initial symptoms, including nausea, malaise, and constipation, also known as “keto flu” [37]. Nevertheless, inter-individual variability in response to KD are reported. Indeed, factors such as age, gender, body composition, metabolic rate, physical activity level, caloric intake, and genetic factors significantly influence the magnitude and timing of βHB elevation [38]. For example, in the studies considered, differences in protein intake between the diets were observed (Table 1). Protein intake may influence the magnitude of ketosis [39]. In our study, the type of diet, which strongly interacts with gender, seems to strongly impact on TtK. The variability in TtK observed in our participants may be attributed to both intrinsic and extrinsic factors, including differences in diet and body composition. The quantity, quality, and timing of food that is consumed likely impact the time frame in which metabolic fuel switches from primarily carbohydrate to greater fat utilization [40]. In this case, our study showed that the KEMEPHY protocol was associated with a faster induction of ketosis compared to the CKD. These differences may be partially explained by differences in protein sources, micronutrient intake, and overall dietary adherence. Indeed, the efficacy of KD for adults with overweight/obesity can be limited by the poor adherence to the diet and facilitated by ready-to-eat ketogenic products as shown by previous research [37]. Greater adherence to the diet, and, in turn, a lower availability of carbohydrates, may also have led to a greater reduction in liver glycogen. Indeed, the depletion of hepatic glycogen is recognized as the key trigger for the metabolic switching. Starting a fast with less liver glycogen can impact on TtK by accelerating the onset of ketosis [37]. However, in our studies, we didn’t measure hepatic glycogen content at the baseline. Moreover, the type of diet strongly interacted with gender. In this regard, other authors have reported that genotype variation could explain different responses in terms of body mass loss induced by a KD [41, 42]. For example, the minor G allele of the angiotensin II receptor type 2 (AGTR2) gene was associated with a greater reduction in body fat percentage. Considering that the AGTR2 gene is X- linked, the impact of this gene variant on body mass loss with KD may be higher in men than in women [43]. Furthermore, factors such as reduced muscle mass and higher adiposity, particularly in individuals with overweight or obesity, may impair mitochondrial function and delay this transition. This is particularly relevant, first because our participants were with overweight or obesity, and second, we found that gender strongly impacts weight loss (Fig. 1C, D). In this sense, women have less muscle mass compared to the men. However, gender-specific effects of KDs on body composition are deeply described by Ashtary-Larky and colleagues [44].

The novelty of our analysis lies in the identification in a large cohort of 217 individuals of a pragmatic and reproducible cut-off ( ≥ 0.5 mmol/L βHB) that is strongly associated with weight reduction.

By relying on capillary βHB measurements, we were able to demonstrate that maintaining this level of ketonemia represents not only a marker of metabolic adaptation but also an actionable threshold for clinicians and dietitians to monitor adherence and anticipate outcomes. While mechanistic studies are crucial to understanding ketone metabolism, the translation of such findings into clinical practice requires actionable parameters. By demonstrating that achieving and maintaining βHB levels of at least 0.5 mmol/L is associated with weight loss, our study bridges the gap between mechanistic research and clinical applicability, offering a straightforward tool to optimize the management of overweight and obesity with KDs.

TtK was significantly correlated with weight loss magnitude (Fig. 1A, B). Each additional day spent in ketosis ( ≥ 0.5 mmol/L) was associated with an average weight reduction of ~0.5 percentage points, emphasizing the relevance of achieving and maintaining ketosis over time. Thus, TtK may be considered both a proxy of adherence and a predictor of treatment responsiveness, emphasizing the relevance of monitoring not only the achievement of ketosis but also the speed and consistency with which it is maintained.

Limitations

Apart from its strengths such as the number of subjects involved and the daily blood ketones measurements, this study has also several limitations that warrant consideration: first and foremost, we were not able to measure changes in body composition but only of body mass; second, the amount of spontaneous physical activity was not monitored and finally, the duration of the intervention (14 days) limits the extrapolation of these findings to longer-term outcomes. Since no biomarkers (e.g., enzyme expression, ketone fluxes) were collected, future studies should be conducted to elucidate physiological mechanisms.

Conclusion

Our results support the use of 0.5 mmol/L of βHB as a threshold to optimize ketosis in the context of weight loss. These findings provide a threshold and underline the importance of ketone monitoring to enhance the effectiveness of ketogenic interventions.

Author contributions

AP conceived the idea for the article. GB and SP performed the statistical analyses. AP, GC, DC, JG, AB, GB, PB, LC, and TM contributed to writing the draft and approved the final version of the manuscript.

Funding

This research was supported by intramural grants from the Department of Biomedical Sciences, University of Padova.

Data availability

Data will be made available on reasonable request through the corresponding author, Giuseppe Cerullo, at giuseppe.cerullo@unipd.it

Competing interests

Gianluca Mech S.p.A. (Asigliano Veneto, VI, Italy) provided the proprietary ketogenic meal products. The company had no role in participant recruitment, intervention delivery, data collection, data analysis, data interpretation, or in the preparation or review of the manuscript.

Ethics approval and consent to participate

The present study is a secondary analysis of previously published data from our group [23–27]. Additionally, we included further unpublished dataset collected in our laboratory between 2020 and 2022 (NCT04163120), and informed consent to participate was obtained from all participants.