Ketosis Suppression and Ageing (KetoSAge): The Effect of Suppressing Ketosis on GKI and Liver Biomarkers in Healthy Females

Abstract

Background: As the growing global population continues to age, the risk of chronic metabolic diseases, including cardiovascular disease, neurodegenerative disorders, type 2 diabetes mellitus, and fatty liver disease, increases considerably. Driven largely by lifestyle factors and metabolic dysfunction, this escalating health crisis is placing mounting pressure on healthcare systems and contributing to significant economic costs. Insulin resistance and hyperinsulinaemia are major drivers of these disorders, emphasising the need for early detection and intervention. Changes in liver enzymes, such as alanine aminotransferase (ALT) and gamma-glutamyl transferase (GGT), commonly assessed in routine laboratory testing, can serve as biomarkers of early-stage insulin resistance, offering a potentially underutilised window for intervention and disease prevention. Correspondingly, low-carbohydrate ketogenic diets have shown to be effective in reversing insulin resistance, metabolic disease, and liver disease. Objectives: We chose to explore the relationship between suppressing ketosis and changes in liver enzymes in the Ketosis Suppression and Ageing cohort. Methods: Ten lean (BMI 20.5 kg/m² ± 1.4), healthy young women (age 32.3 ± 8.9 years) who habitually followed a ketogenic diet maintaining nutritional ketosis (NK) for an average of 3.9 years (±2.3) were exposed to a higher carbohydrate diet, in line with standard healthy eating guidelines for a 21-day phase and then transitioned back to a ketogenic diet. Results: Carbohydrate challenge and suppression of ketosis increased insulin resistance score HOMA-IR by 2.13-fold (p = 0.0008), GKI by 22.28-fold (p = 0.0024), and liver markers ALT by 1.85-fold (p = 0.0010), GGT, 1.29-fold (p = 0.0087) and the ALT/AST, 1.30-fold (p = 0.0266), reflecting an adverse pattern suggestive of hepatic insulin resistance. Conclusions: These results support the clinical utility of liver markers as early and directional signs of hyperinsulinaemia.

1. Introduction

Recent years have seen significant advancements in the understanding of non-alcoholic fatty liver disease (NAFLD), now redefined as metabolic-dysfunction-associated steatosis liver disease (MASLD) to better reflect its metabolic origins. Many chronic non-communicable diseases, such as Alzheimer’s disease (AD), cardiovascular disease (CVD), cancer, type 2 diabetes mellitus (T2DM), metabolic syndrome (MetS), and MASLD, result from lifestyle factors that induce chronic hyperinsulinaemia [1,2,3]. MASLD is closely linked with other clinical features of metabolic syndrome and studies have reported MASLD prevalence in up to 75% of individuals with obesity, metabolic syndrome, and T2DM [4,5].

The liver plays a central role in both the regulation of glucose and lipid metabolism, as well as facilitating the absorption of fat-soluble vitamins A, D, E, and K, and essential fatty acids omega 3 and 6. It also plays a vital role in sulphonic acid taurine synthesis and uptake through enterohepatic bile reabsorption, whilst removing blood waste transferred from the spleen into bile [6,7,8]. The liver is also a major source of ketone body synthesis, both acetoacetate and beta-hydroxybutyrate (BHB; Figure 1). The energy and nutrient status signalling metabolite BHB is synthesised by hepatocytes when insulin is low enough to no longer inhibit ketogenesis. BHB is released into the bloodstream to be used in extra-hepatic tissues, to generate the necessary ATP a cell needs in order to survive and function when glucose is not provided exogenously [9]. When plasma BHB concentration is elevated, it acts as a signalling molecule from hepatocytes to the extra-hepatic tissues, altering cellular gene expression and behaviour to adapt to the nutrient scarcity signal, and utilises internal energy stores from fatty acids converted to BHB via ketogenesis. BHB signalling coordinates a systemic shift in cellular phenotypes towards conservation (slowing down the cell cycle/decreasing mitotic rate) and increasing recycling of cellular materials (upregulation of autophagy, mitophagy, and increasing mitochondrial biogenesis), culminating in greater cellular health and survival in the face of starvation [10,11]. Beyond metabolic flexibility, BHB exerts anti-inflammatory and antioxidative effects through inhibition of the NLRP3 inflammasome and enhancement of mitochondrial function [12], contributing to reduced hepatic inflammation and improved insulin sensitivity.

Ketones are synthesised from fatty acids derived either from adipocytes or from fat intake, primarily by hepatocytes (enterocytes, renal cells, retinal epithelial cells, and astrocytes are also known to perform ketogenesis) (Figure 1) [13,14,15,16,17]. Fasting-mimicking diets, such as eating in a narrow time frame within a 24 h window, or very low carbohydrate-healthy high-fat with moderate protein diets, known as ketogenic diets, are also able to induce ketosis in individuals, without the conscious effort of calorie restriction [10]. The result is decreased insulin secretion from the pancreas due to decreased glucose nutrient sensing, the primary driver of insulin secretion, which is predominantly driven by dietary sugars and starchy carbohydrates found in bread, pasta, rice, flour, corn, and fruit. Ketosis results in lower plasma concentrations of insulin, lower/well-regulated blood glucose (euglycaemia), and increased plasma ketone BHB (euketonaemia). If sustained over time, this induces a series of adaptive changes within cells, to shift their intracellular machinery to upregulate their ability to utilise fat and ketones instead of glucose for fuel; however, additional metabolic adaptations are induced by the presence of ketones and decreases in chronic excess insulin signalling in cells [18,19].

The liver is a major target of insulin’s action. In the development of subclinical hyperinsulinaemia (SCHI, insulin levels within the common population reference range that chronically inhibit ketosis) [1,20], early hepatic expression may be detected via changes in biomarkers, such as gamma-glutaryl transferase (GGT), and the transaminases alanine aminotransferase (ALT) and aspartate transferase (AST) [21]. Hyperglycaemia with a consequent diagnosis of insulin resistance may not be detected for years due to SCHI keeping blood glucose and haemoglobin A1C (HbA1c) within acceptable normative ranges [1,22]. In pathological conditions, including chronic non-communicable diseases, insulin fails to properly regulate liver metabolism, leading to excessive glucose production despite accelerated lipid synthesis—referred to as hepatic insulin resistance [23]. Therefore, both increased glucose production and de novo lipogenesis have been well-studied in insulin-resistant individuals [23,24,25]. In addition, elevated hepatic inflammatory markers are associated with an increased risk of extra-hepatic cancers [26]. The glucose ketone index (GKI) was initially developed as a tool to monitor nutritional ketosis in cancer patients and has emerged as a valuable indicator of metabolic status [27].

We have previously published the effects of suppressing ketosis in the KetoSAge cohort, on ageing- and chronic-disease-associated biomarkers, including insulin, insulin-like growth factor 1 (IGF-1), glucose, BHB, GGT, homeostasis model assessment for insulin resistance (HOMA-IR), osteocalcin (OCN), and leptin [1,2,28]. Given the protective role of ketosis in metabolic health, the KetoSAge cohort was designed to examine the consequences of suppression of ketosis on hepatic biomarkers and metabolic health. This study explores whether temporary reversion to a high-carbohydrate diet induces measurable metabolic stress, as reflected in liver enzyme markers and insulin resistance. In this report, we examined additional biomarkers associated with morbidity, such as albumin, AST, ALT, alkaline phosphatase (ALP), GGT, iron, total protein, total bilirubin, direct bilirubin, and creatine kinase NAC (CK-NAC). This study investigated the impact of long-term sustained nutritional ketosis (NK), or euketonaemia, and the suppression of ketosis, which is driven by lifestyle factors that increase insulin demand, and insulin’s subsequent inhibition of ketogenesis and its effects on biomarkers associated with the liver.

2. Materials and Methods

2.1. Participants and Study Design

Ten lean (BMI 20.5 kg/m² ± 1.4), healthy young female KetoSAge participants (age 32.3 ± 8.9 years) took part in the KetoSAge study, an open-labelled, non-randomised cross-over trial with three phases: baseline nutritional ketosis (NK) (Phase 1; P1), suppression of ketosis (SuK) (Phase 2; P2), and removal of intervention, returning to NK (Phase 3; P3), as previously published in Part 1 and Part 2 [1,2]. The participants self-reported their adherence to a lifestyle that sustained NK for ≥6 months (mean 3.9 ± 2.3 years), ensuring sufficient time for metabolic adaptations. Ketosis adaptation was proven during a 6-month lead-in period, where the participants were required to take a once-daily capillary BHB reading between 4 and 6 p.m., before their evening meal, prior to the commencement of the study. This standardised evening measurement was chosen due to it being a more rigorous threshold to pass in order to be judged as sustaining NK over the majority of the 24 h day, in comparison to morning fasted measurements, thus increasing confidence in the participants maintaining NK most of the time. Readings were taken with a Keto-Mojo™ GKI multi-function meter (Keto-Mojo, Napa, CA, USA; [29]). The baseline characteristics of these participants were described in our earlier publication [1,2,28]. Following each study phase, participants attended the laboratory for one day at 8 a.m. after a 12 h overnight fast to undertake anthropometric measurements and blood sampling (Figure 2). Further details on macronutrient breakdown with statistical analysis between phases may be found in our previous publication on this cohort [28].

2.2. Anthropometric Measurements

Upon arrival at the laboratory, height (to the nearest 0.1 cm) was measured using a stadiometer (Marsden HM-250P Leicester Height Measure), and body weight (to the nearest 0.1 kg), and waist and hip circumference measures were obtained with a non-stretch anthropometric circumference measuring tape (Seca^® 201) while participants stood upright on both feet. The average value (cm) of three measurements was used for analysis. All measurements were taken following a 12 h fast, wearing standardised clothing [1,2].

2.3. Blood Collection and Measurement

As previously described, blood was drawn into tubes anti-coagulated with lithium heparin (LH; BD, Oxford, UK) before being centrifuged at 3857× g for 10 min at 4 °C (Hettich Zentrifugen, Universal 320 R, Tuttlingen, Germany). Blood was also drawn into serum SST™ II Advance tubes with thrombin rapid clot activator and separation gel (BD, Oxford, UK) and left for 30 min at room temperature. Serum tubes were then centrifuged at 3857× g for 10 min at room temperature. Samples were either sent to SYNLAB Belgium (Alexander Fleming, 3–6220 Heppignies–Company No: 0453.111.546) for analysis or aliquoted into cryovial tubes under sterile conditions and stored at −80 °C for later analysis by Randox, Clinilabs (London, UK) or in-house testing [1,2].

2.4. Blood Marker Analysis

Markers of insulin, glucose, BHB, albumin, ALT, AST, ALP, GGT, iron, total protein, total bilirubin, direct bilirubin, and CK-NAC were assessed in all KetoSAge participants. Serum insulin was measured via a Simple Plex Assay (Ella™, Bio-Techne, Minneapolis, MN, USA). Whole blood glucose concentration was measured using a Biosen C-Line Clinic Glucose and Lactate analyser (EKF-Diagnostic, GmbH, Barleben, Germany). BHB was measured using capillary blood on the Keto-Mojo^TM meter. GGT was measured in serum by SYNLAB, and iron was measured in serum by Randox and Clinilabs [1,2]. Serum ALP was measured in 5 participants by SYNLAB and in 5 participants by HORIBA Pentra C400 Clinical Chemistry Analyser (Horiba Medical, Northampton, UK). The GKI was calculated from whole blood readings taken using a Keto-Mojo^TM meter, GKI = glucose mmol/L × ketones mmol/L.

Validation on the HORIBA Pentra C400 was conducted in-house to assess variance and precision between samples. Intra-assay precision was performed in 5 replicates within a day, against the manufacturers’ targets in both normal control (C-N) and patient control (C-P) (albumin, C-N CV = 1.25% and C-P CV = 0.27%; ALT, C-N CV = 0.71% and C-P CV = 0.78%; AST, C-N CV = 1.66% and C-P CV = 1.04%; ALP, C-N CV = 1.53% and C-P CV = 0.85%; GGT, C-N CV = 7.53% and C-P CV = 0.77%; iron, C-N CV = 3.32% and C-P CV = 0.15%; total protein, C-N CV = 2.89% and C-P CV = 2.12%; total bilirubin, C-N CV = 4.10% and C-P CV = 1.75%; direct bilirubin, C-N CV = 1.84% and C-P CV = 0.92%; CK-NAC, C-N CV = 0.88% and C-P CV = 1.61%). Inter-assay precision was performed in 5 consecutive days, in triplicate per day, against the manufacturers’ targets in both normal control (C-N) and patient control (C-P) (albumin, C-N CV = 2.83% and C-P CV = 4.79%; ALT, C-N CV = 3.43% and C-P CV = 5.10%; AST, C-N CV = 4.13% and C-P CV = 4.43%; ALP, C-N CV = 4.56% and C-P CV = 4.33%; GGT, C-N CV = 4.42% and C-P CV = 1.54%; iron, C-N CV = 1.88% and C-P CV = 3.34%; total protein, C-N CV = 4.63% and C-P CV = 3.64%; total bilirubin, C-N CV = 6.72% and C-P CV = 3.22%; direct bilirubin, C-N CV = 2.48% and C-P CV = 1.16%; CK-NAC, C-N CV = 1.98% and C-P CV = 4.21%). After daily validation, serum or plasma albumin, ALT, AST, ALP, GGT, iron, CK-NAC, total protein, and total and direct bilirubin were assayed in triplicate on HORIBA Pentra C400.

2.5. Statistical Analysis

Data were checked for normality using the Shapiro–Wilk test. Depending on the results of the normality tests, either KetoSAge participants were compared using repeated measures (RM) one-way ANOVA with Tukey’s correction for multiple comparisons, or the Friedman test with Dunn’s correction for multiple comparisons. When the sphericity of data was not met, Geisser–Greenhouse corrections were also added. Data are presented as mean ± SD. Data were analysed and graphed using GraphPad Prism (Version 10.1.0).

Additionally, mixed-effects models and their statistical analyses were performed using R version 4.43. To account for our study’s RM design, we used linear mixed-effects models with a random effect for each participant to compare the influence of different physiological variables on liver biomarkers across the study phases. All models used the R function lmerTest::lmer. Hypothesis testing for these models used Satterthwaite’s method for estimating the degrees of freedom and hypothesis testing.

3. Results

Key findings from our analysis of liver markers in KetoSAge participants are presented, as previously reported, in Parts 1 and 2 [1,2]. Markers of BMI, fat mass, insulin, glucose, HOMA-IR, GKI, leptin, and GGT significantly increased from the P1 to P2 phase, and this trend reversed following P3 (

Table 1. BMI, fat mass, fasted insulin, glucose, BHB, HOMA-IR, GKI, leptin, albumin, ALT, AST, ALP, GGT, CK-NAC, total protein, iron, and total and direct bilirubin across all phases in KetoSAge participants. Measurements were taken following each of the study phases: baseline nutritional ketosis (NK), P1; intervention suppress ketosis (SuK), P2; and removal of SuK returning to NK, P3. Measurements were taken at 8 a.m. after a 12 h overnight fast (n = 10). Values are presented as mean ± SD.

	P1	P2	P3	p Value	P1 vs. P2	P2 vs. P3	P1 vs. P3
BMI (kg/m2)	20.52 (±1.39)	21.54 (±1.30)	20.82 (±1.46)	<0.0001	<0.0001	<0.0001	0.0734
Fat Mass (kg)	14.21 (±2.55)	15.88 (±2.23)	14.78 (±2.20)	<0.0001	<0.0001	0.0018	0.1102
Insulin (µIU/mL)	4.95 (±1.24)	9.06 (±2.14)	5.62 (±1.83)	<0.0001	<0.0001	0.0001	0.5686
Glucose (mmol/L)	4.36 (±0.53)	5.12 (±0.59)	4.41 (±0.30)	0.0063	0.0110	0.0417	>0.9999
BHB (mmol/L)	2.43 (±1.28)	0.18 (±0.13)	2.31 (±0.71)	0.0002	0.0012	<0.0001	0.9638
HOMA-IR	0.97 (±0.32)	2.07 (±0.61)	1.11 (±0.41)	<0.0001	<0.0001	<0.0001	0.4074
GKI (Lab Day)	2.23 (±1.20)	49.68 (±42.62)	1.99 (±0.60)	0.0001	0.0024	0.0024	>0.9999
GKI (21 Day Average)	2.82 (±1.34)	56.30 (±30.01)	2.76 (±1.15)	<0.0001	0.0010	0.0052	>0.9999
Leptin (ng/mL)	4.50 (±3.67)	15.08 (± 8.00)	4.57 (±3.48)	<0.0001	0.0010	0.0052	>0.9999
ALT (U/L)	13.71 (±3.64)	25.42 (±12.26)	13.16 (±2.69)	<0.0001	0.0010	0.0052	>0.9999
AST (U/L)	18.65 (±4.15)	26.18 (±8.77)	19.63 (±3.11)	0.0265	0.0570	0.0915	0.5481
GGT (U/L)	9.60 (±3.13)	12.40 (±2.55)	9.70 (±2.50)	0.0021	0.0044	0.0059	0.9904
ALP (U/L)	52.98 (±11.43)	66.98 (±15.60)	56.39 (±15.28)	0.0160	0.0160	0.0742	0.7328
ALT/AST	0.74 (±0.14)	0.96 (±0.30)	0.69 (±0.17)	0.0017	0.0266	0.0047	0.2387
Albumin (g/L)	41.82 (±3.66)	40.49 (±2.08)	42.37 (±2.10)	0.1774	0.3881	0.1681	0.8480
CK-NAC (U/L)	55.03 (±24.15)	77.30 (±43.75)	61.30 (±24.24)	0.2223	0.2209	>0.9999	>0.9999
Total Protein (g/L)	69.4 (±9.58)	66.73 (±6.36)	67.25 (±3.79)	0.8302	>0.9999	>0.9999	>0.9999
Iron (μmol/L)	16.62 (±7.27)	14.40 (±8.70)	11.76 (± 11.78)	0.1873	>0.9999	0.2209	0.3526
Total Bilirubin (μmol/L)	7.88 (±4.50)	6.72 (±1.67)	5.59 (±2.29)	0.4362	>0.9999	>0.9999	0.5391
Direct Bilirubin (μmol/L)	1.82 (±1.10)	1.69 (±0.59)	1.65 (±0.44)	0.7909	0.9290	0.9893	0.8352

). Here, we further examined biomarkers associated with morbidity: Albumin, ALT, AST, ALP, CK-NAC, GKI, total protein, iron, total bilirubin, and direct bilirubin (Table 1).

3.1. Suppression of Ketosis Is Associated with Increases in GKI

Following P2, both GKI measured on participant Lab Day and the average across the 21 days prior significantly increased. GKI (Lab Day) significantly increased from 2.23 (±1.20, P1) to 49.68 (±42.62, P2; p = 0.0024; Figure 3A). This trend reversed following P3, where GKI (Lab Day) significantly decreased and returned to the participants’ baseline levels of 1.99 (±0.60, P3; p = 0.0024) compared to P2. GKI (21-Day Average) also significantly increased from 2.82 (±1.34, P1) to 56.30 (±30.01, P2; p = 0.0010; Figure 3B). This trend reversed following P3, where GKI (21-Day Average) significantly decreased and returned to the participants’ baseline levels of 2.76 (±1.15, P3; p = 0.0052) compared to P2.

3.2. Suppression of Ketosis Is Associated with Increases in ALT and GGT

Following P2, ALT significantly increased from 13.71 U/L (±3.64, P1) to 25.42 U/L (±12.96, P2; p = 0.0010; Figure 4A). This trend reversed following P3, where ALT significantly decreased and returned to the participants’ baseline levels of 13.16 U/L (±2.69, P3; p = 0.0052) compared to P2. A trend towards increased AST levels was also observed following P2, with values rising from 18.65 U/L (±4.15, P1) to 26.18 U/L (±8.77, P2; p = 0.0570; Figure 4B), although this did not reach statistical significance. This trend reversed following P3, where AST decreased to 19.63 U/L (±3.11, P3; p = 0.0915) compared to P2. As previously reported, at the end of P2, GGT significantly increased from 9.60 U/L (±3.13, P1) to 12.40 U/L (±2.55, P2; p = 0.0044; Figure 4C). This trend reversed following P3, where GGT significantly changed and returned to participants’ baseline levels of 9.70 U/L (±2.50, P3; p = 0.0059) compared to P2. ALP changed significantly from 52.98 U/L (±11.43, P1) to 66.98 U/L (±15.60, P2; p = 0.0160; Figure 4D). Following P3, this trend almost returned to participants’ baseline levels of 56.39 U/L (±15.28, P3; p = 0.0742) compared to P2.

3.3. Suppression of Ketosis Is Associated with Changes in the Ratio of Aminotransferases

Following P2, the ALT/AST ratio significantly increased from 0.74 (±0.14, P1) to 0.96 (±0.30, P2; p = 0.0266). This trend reversed following P3, where the ALT/AST ratio significantly returned to participants’ baseline levels of 0.69 (±0.17, P3; p = 0.0047) compared to P2 (Figure 5).

3.4. Levels of Albumin, Total Protein, Bilirubin, CK, and Iron Do Not Significantly Change

Levels of albumin remained consistent throughout all phases of this study. Following P2, albumin did not significantly change after P1, changing from 41.82 g/L (±3.66, P1) to 40.49 g/L (±2.08, P2; p = 0.3881; Figure 6A). Similarly, following P3, no significant change was observed compared to P2, with levels measured at 42.37 g/L (±2.10, P3, p = 0.1681).

Following P2, CK-NAC did not significantly change from P1 to P2, increasing from 55.03 U/L (±24.15, P1) to 77.30 U/L (±43.75, P2; p = 0.2209; Figure 6B). With the removal of the intervention for SuK and a return to NK for 21 days at the end of P3, there was no significant change in CK-NAC compared to P2 at 61.30 U/L (±24.24, P3, p ≥ 0.9999).

Following P2, total protein did not significantly change after P1, from 69.64 g/L (±9.58, P1) to 66.73 g/L (±6.36, P2; p ≥ 0.9999; Figure 6C). Similarly, results in P3 remained consistent, with no significant change compared to P2; levels were recorded at 67.25 g/L (±3.79, P3, p ≥ 0.9999).

Following P2, iron did not significantly change after P1, from 16.61 μmol/L (±7.27, P1) to 14.40 μmol/L (±8.70, P2; p ≥ 0.9999; Figure 6D). Similarly, following P3, there was no significant change compared to P2, with levels observed at 11.76 μmol/L (±11.78, P3, p = 0.2209).

Changes in total bilirubin and direct bilirubin did not reach significance in this study. Following P2, total bilirubin levels changed from 7.88 μmol/L (±4.50, P1) to 6.72 μmol/L (±1.67, P2; p ≥ 0.9999; Figure 7A). Similarly, following P3, there was no significant change compared to P2, with levels decreasing to 5.59 μmol/L (±2.29, P3, p ≥ 0.9999). Lastly, following P2, direct bilirubin levels changed from 1.82 μmol/L (±1.10, P1) to 1.69 μmol/L (±0.59, P2; p = 0.9290; Figure 7B). Similarly, following P3, there was no significant change compared to P2, with levels further decreasing to 1.65 μmol/L (±0.44, P3, p = 0.9893).

3.5. Liver Markers ALT, AST, GGT Change as Basal Insulin, HOMA-IR, and GKI Change

Data presented in

Table 2. Changes in biomarkers ALT, AST, GGT, ALP, and ALT/AST with changes in insulin across all phases in KetoSAge participants. Measurements were taken following each of the study phases: baseline nutritional ketosis (NK), P1; intervention suppress ketosis (SuK), P2; and removal of SuK returning to NK, P3. Measurements were taken at 8 a.m. after a 12 h overnight fast (n = 10). All models were structured as follows: lmerTest::lmer (liver_marker~insulin+(1|id).

A
Model	Effect Estimate	p Value
ALT~Insulin	2.0185	0.0017
AST~Insulin	1.3511	0.0028
GGT~Insulin	0.6271	0.0001
ALP~Insulin	1.4090	0.1410
ALT/AST~Insulin	0.0429	0.0033
B Log-transformed data
Model	Effect Estimate	p Value
ALT~Insulin	0.7047	0.0001
AST~Insulin	0.4010	0.0009
GGT~Insulin	0.3703	0.0027
ALP~Insulin	0.1498	0.1510
ALT/AST~Insulin	0.3008	0.0064

show the relationship between liver markers (ALT, AST, GGT, and ALP) and changes in insulin levels across the three phases, accounting for individual variability among the KetoSAge participants. Significant positive associations were observed between insulin levels and ALT (effect estimate = 2.0185, p = 0.0017), AST (1.3511, p = 0.0028), and GGT (0.6271, p = 0.0001). In contrast, the relationship between insulin and ALP was not significant (1.4090, p = 0.1410), indicating no notable change in ALP in response to insulin levels. The ALT/AST ratio had a significant positive association (0.0429, p = 0.0033; Table 2A; visualised in Figure 8).

When data were log-transformed to account for the scale differences across variables, the observed relationships remained consistent, with significant positive associations between insulin and both ALT (effect estimate = 0.7047, p = 0.0001) and AST (0.4010, p = 0.0009). The association between insulin and GGT also remained significant in the log-transformed model (0.3703, p = 0.0027), while ALP continued to show no significant association (0.1498, p = 0.1510). A similar significant pattern was also observed in the ALT/AST ratio after log transformation (0.3008, p = 0.0064) (Table 2B; visualised in Figure 9).

Data presented in

Table 3. Changes in biomarkers ALT, AST, GGT, ALP, and ALT/AST with changes in HOMA-IR across all phases in KetoSAge participants. Measurements were taken following each of the study phases: baseline nutritional ketosis (NK), P1; intervention suppress ketosis (SuK), P2; and removal of SuK returning to NK, P3. Measurements were taken at 8 a.m. after a 12 h overnight fast (n = 10). All models were structured as follows: lmerTest::lmer (liver_marker~HOMA_IR+(1|id).

A
Model	Effect Estimate	p Value
ALT~HOMA-IR	8.4960	0.0003
AST~HOMA-IR	4.6950	0.0061
GGT~HOMA-IR	2.2300	0.0004
ALP~HOMA-IR	5.1260	0.1530
ALT/AST~HOMA-IR	0.1978	0.0001
B Log-transformed data
Model	Effect Estimate	p Value
ALT~HOMA-IR	0.5984	<0.0001
AST~HOMA-IR	0.3194	0.0012
GGT~HOMA-IR	0.3092	0.0019
ALP~HOMA-IR	0.1451	0.0822
ALT/AST~HOMA-IR	0.2757	0.0016

demonstrate the relationship between liver markers (ALT, AST, GGT, and ALP) and changes in HOMA-IR across the three phases, accounting for individual variability among the KetoSAge participants. Significant positive associations were observed between HOMA-IR and ALT (effect estimate = 8.4960, p = 0.0003), AST (4.6950, p = 0.0061), and GGT (2.2300, p = 0.0003). In contrast, the relationship between HOMA-IR and ALP was not significant (5.1260, p = 0.1530), suggesting no notable change in ALP in response to HOMA-IR. The ALT/AST ratio had a significant positive association (0.1978, p = 0.0001; Table 3A; visualised in Figure 10).

When data were log-transformed to account for the scale differences across variables, the relationships remained consistent, with significant positive associations between HOMA-IR and both ALT (effect estimate = 0.5984, p < 0.0001) and AST (0.3194, p = 0.0012). The relationship between HOMA-IR and GGT also remained significant in the log-transformed model (0.3092, p = 0.0019), while ALP continued to show no significant association (0.1451, p = 0.0822). A similar significant pattern was also observed in the ALT/AST ratio after log-transformation (0.2757, p = 0.0016) (Table 3B; visualised in Figure 11).

Data presented in

Table 4. Changes in biomarkers ALT, AST, GGT, ALP, and ALT/AST with changes in GKI (Lab Day) across all phases in KetoSAge participants. Measurements were taken following each of the study phases: baseline nutritional ketosis (NK), P1; intervention suppress ketosis (SuK), P2; and removal of SuK returning to NK, P3. Measurements were taken at 8 a.m. after a 12 h overnight fast (n = 10). All models were structured as follows: lmerTest::lmer (liver_marker~GKI+(1|id).

A
Model	Effect Estimate	p Value
ALT~GKI (Lab Day)	0.0743	0.1510
AST~GKI (Lab Day)	0.0188	0.6140
GGT~GKI (Lab Day)	0.0423	0.0003
ALP~GKI (Lab Day)	0.0818	0.2570
ALT/AST~GKI (Lab Day)	0.0032	0.0030
B Log-Transformed Data
Model	Effect Estimate	p Value
ALT~GKI (Lab Day)	0.1593	0.0001
AST~GKI (Lab Day)	0.0731	0.0130
GGT~GKI (Lab Day)	0.0939	0.0002
ALP~GKI (Lab Day)	0.0593	0.0081
ALT/AST~GKI (Lab Day)	0.0878	0.0001

show the relationship between liver markers (ALT, AST, GGT, and ALP) and changes in GKI (Lab Day) levels across the three phases and accounting for individual variability among the KetoSAge participants. Significant positive associations were observed between the GKI and GGT (effect estimate = 0.0423, p = 0.0003) and ALT/AST (0.0032, p = 0.0030); Table 4A, visualised in Figure 12. ALT, AST, and ALP were not statistically significant.

When data were log-transformed to account for the scale differences across variables, all liver markers were statistically significant (ALT, effect estimate = 0.1593, p = 0.0001; AST, 0.0731, p = 0.0130; GGT, 0.0939, p = 0.0002; ALP, 0.0593, p = 0.0081; and ALT/AST, 0.0878, p = 0.0001; Table 4B, visualised in Figure 13).

Data presented in

Table 5. Changes in biomarkers ALT, AST, GGT, ALP, and ALT/AST with changes in GKI (21-Day Average; 4 evenly spaced tests per day, totalling 84 tests) across all phases in KetoSAge participants. Measurements were taken following each of the study phases: baseline nutritional ketosis (NK), P1; intervention suppress ketosis (SuK), P2; and removal of SuK returning to NK, P3. Measurements were taken at 8 a.m. after a 12 h overnight fast (n = 10). All models were structured as follows: lmerTest::lmer (liver_marker~GKI+(1|id).

A
Model	Effect Estimate	p Value
ALT~GKI (21-Day Average)	0.1397	0.0077
AST~GKI (21-Day Average)	0.0807	0.0329
GGT~GKI (21-Day Average)	0.0400	0.0017
ALP~GKI (21-Day Average)	0.1482	0.0412
ALT/AST~GKI (21-Day Average)	0.0032	0.0055
B Log-Transformed Data
Model	Effect Estimate	p Value
ALT~GKI (21-Day Average)	0.1742	<0.0001
AST~GKI (21-Day Average)	0.0871	0.0034
GGT~GKI (21-Day Average)	0.0913	0.0006
ALP~GKI (21-Day Average)	0.0604	0.0087
ALT/AST~GKI (21-Day Average)	0.0888	0.0002

show the relationship between liver markers (ALT, AST, GGT, and ALP) and changes in GKI (21-Day Average) levels across the three phases, accounting for individual variability among the KetoSAge participants. Significant positive associations were observed in all liver markers measured; ALT (effect estimate = 0.1397, p = 0.0077), AST (0.0807, p = 0.0329), GGT (0.0400, p = 0.0017), ALP (0.1482, p = 0.0412), and ALT/AST (0.0032, p = 0.0055) (Table 5A; visualised in Figure 14).

When data were log-transformed to account for the scale differences across variables, all relationships remained consistent, with liver markers being statistically significant; ALT (effect estimate = 0.1742, p < 0.0001), AST (0.0871, p = 0.0034), GGT (0.0913, p = 0.0006), ALP (0.0604, p = 0.0087), and ALT/AST (0.0888, p = 0.0002) (Table 5B; visualised in Figure 15).

4. Discussion

Many liver changes have been documented in ageing and chronic disease, such as a decrease in liver mass and liver blood volume in older adults, potentially causing a distribution shift in liver marker values as age and chronic illnesses progress. Furthermore, many standard biochemical and liver function tests remain within normal ranges. Additionally, differences, such as increased liver fat, total body and visceral fat, and differences in hepatic glucose metabolism, are also evidenced. Many chronic non-communicable diseases show the importance of the liver’s role in modulating insulin demand and secretion.

In this open-labelled cross-over trial, we found significant changes in the effects of long-standing ketosis and the suppression of ketosis on liver markers. It is widely recognised that liver function tests measuring AST, ALT, ALP, and GGT are used as diagnostic markers for many diseases, including hepatic steatosis, MASLD, CVD, T2DM, and MetS, encompassing insulin resistance and hyperinsulinaemia (Figure 16) [30,31,32,33]. We documented statistically significant changes in ALT, GGT, and the AST/ALT ratio after the suppression of ketosis and a return to nutritional ketosis. These notable increases in ALT and GGT suggest acute hepatic stress and oxidative damage triggered by impaired fatty acid oxidation and elevated insulin signalling during ketosis suppression. Importantly, the return to nutritional ketosis reversed these elevations, indicating a restorative effect on hepatic health.

Aminotransferases are a group of intracellular enzymes, and thus, abnormal plasma levels are highly suggestive of cell damage, whilst GGT is believed to reflect liver dysfunction and is also associated with systemic oxidative stress and inflammation, which are central to the development of chronic disease. During a state of oxidative stress, oxidised glutathione (GSSG) increases, which in turn induces hepatic expression of GGT. GGT recycles GSSG, releasing cysteine, which is a precursor for GSH synthesis (Figure 17) [34]. GGT activity, even within physiological ranges, has been linked to insulin resistance, a key component of MetS and a precursor to T2DM [35]. The reduction in GGT upon the reintroduction of ketosis in P3 suggests that nutritional ketosis not only alleviates oxidative stress but may also enhance mitochondrial efficiency and restore cellular redox balance. These findings are consistent with ketogenic metabolism’s role in dampening pro-inflammatory signalling [36,37] and improving hepatic resilience [38]. Hyperinsulinaemia increases mitochondrial ROS production via the generation of ceramides [39] concomitantly depleting NAD+, leading to decreasing ROS management capacity, resulting in diminished manganese superoxide dismutase 2 (MnSOD2) and a decreased GSH:GSSG ratio [22]. Depletion of hepatocyte GSH generates GGT, which then results in further hepatocyte inflammation and subsequent cellular release of ALT and AST into the circulatory system.

Physiological reference ranges for GGT in women have been documented as <30 U/L [40]. However, abnormal liver function parameters may only occur in up to 4% of asymptomatic patients [41]. Elevated GGT levels are associated with an increased risk of certain cancers, particularly liver, prostate, and breast cancers; furthermore, higher levels are associated with cancer progression and prognosis. In a large Swedish cohort from the Apolipoprotein Mortality Risk Study (AMORIS), n = 545,460, with 37,809 primary cancers, a positive association between GGT and overall cancer risk was found [42]. Significant correlations between GGT levels and prognosis in many cancers, including breast, ovarian, glioma, gastric, colorectal, hepatic, and melanoma cancer, have been found [43,44,45,46,47], and elevated levels of GGT in patients with ovarian cancer are associated with worse prognosis and indicate advanced disease [48]. In a study of 219 patients with early-stage non-small-cell lung cancer, pre-operative GGT levels were found to be a significant prognostic indicator for median survival with an Hazard Ratio of 2.0 (CI 1.0–3.9, p = 0.03) [49]. Our results suggest that at least part of those associations could be related to the association between GGT and insulin, which is a well-known cell proliferation and growth signal for tumours.

In this open-labelled cross-over trial, participants’ GGT levels significantly increased during P2, after sustaining hypoketonaemia for 21 days using dietary carbohydrates to stimulate an increase in insulin demand and secretion, thereby downregulating ketogenesis enzymes and inhibiting ketogenesis. In consequence, markers of BMI, waist circumference, and HOMA-IR also increased (but not outside conservative ranges), corroborating previous research. Despite occurring within normal reference ranges, these significant increases do not necessarily reflect a healthy physiological state, given that these markers returned to their baseline P1 levels after returning to a sustained euketonaemia state. Given that the average GGT levels in the KetoSAge cohort in euketonaemia were 9.60 U/L (±3.13) in P1 and 9.70 U/L (±2.50) in P3, for females, we propose that an optimal health reference range should be <12 U/L.

The rapid normalisation of ALT, GGT, and AST upon resumption of ketosis underlines the potential of nutritional ketosis as a reversible and protective intervention against transient hepatic stress. This adaptability suggests that early metabolic intervention through dietary modulation could mitigate long-term damage. We hypothesise that such an increase in insulin demand and signalling and loss of ketones, if sustained chronically instead of for 21 days, may have a greater negative impact on liver health and other organ systems concurrently. A recent study demonstrated that in non-diabetic individuals, insulin resistance was a strong predictor for abnormal liver function measures, particularly in those who are obese or overweight. Levels of ALT, AST, and GGT were also positively correlated with BMI, waist circumference, fasting insulin, and HOMA-IR, as well as HOMA-IR being a determinant of these liver enzymes after adjusting for age, sex, BMI, triglyceride, and cholesterol levels [21]. Our trial data suggest that liver markers increase as insulin and HOMA-IR increase (Figure 8, Figure 9, Figure 10 and Figure 11), which may enable the earlier detection of negative health trends, enabling individuals and healthcare providers to take action that would be more preventive rather than reactive.

Elevated serum levels of liver aminotransferases are associated with liver damage. The transfer of the amino group from L-alanine to alpha-ketoglutarate to produce L-glutamate and pyruvate, a critical process in the tricarboxylic acid cycle, is catalysed by the enzyme ALT (with essential coenzyme pyridoxal phosphate, commonly referred to as vitamin B6). ALT is found in the cellular cytosol of the muscles, intestines, prostate, adipose tissues, colon, and brain, but especially in the cytosol of hepatocytes at a 3000 times higher concentration than blood serum [50]. Metabolic disorders, such as T2DM, obesity, and hyperlipidaemia, are independently associated with mild-to-moderate ALT elevation. An analysis of 15,676 participants from the Third National Health and Nutrition Examination Survey (1988–1994) found 69.0% with unexplained elevation of aminotransferase. Upon further investigation, elevated aminotransferases ALT and AST were significantly associated with an increased BMI, fasting insulin and triglycerides, higher waist circumference, and lower high-density lipoprotein (HDL) concentration, all p < 0.05. Furthermore, where elevated aminotransferase could not be explained by hemochromatosis, consumption of alcohol, or viral hepatitis, a strong association with metabolic syndrome features was found and could be considered metabolic dysfunction associated with fatty liver disease [51]. In a retrospective cohort study of 91 lean women with polycystic ovarian syndrome (PCOS) and 45 healthy controls, statistically significantly higher levels of ALT were positively correlated in the lean PCOS women, indicating liver inflammation independent of adiposity, and further linking hyperinsulinaemia, female reproductive health, and liver function [52,53].

Reference ranges for ALT in males and females vary, but most commonly, the normal levels referenced are <50 U/L for males and <30 U/L for females. However, research has shown that higher levels of ALT and AST, even within normal ranges, were associated with an increased risk of metabolic syndrome [54]. Therefore, it has been proposed that the upper limit for ALT should be set at 19 U/L for women and 30 U/L for men to better represent healthy levels in hypoketonaemic individuals [41]. In this open-labelled cross-over trial, we observed average values of ALT at 13.71 U/L (±3.63) in NK P1, 25.42 U/L (±12.26) in SuK P2, and 13.16 U/L (±2.69) in NK P3, showing significant increases in ALT after the suppression of ketosis (hypoketonaemia) for 21 days. The higher values in P2 reflect elevated ALT levels above healthy ranges (19 U/L), indicating that the suppression of ketosis has a negative effect on liver function. Thus, our proposed optimal health reference ranges for ALT should be <16 U/L. It is clear that if females are assessed using pooled male and female reference ranges, they will be poorly served.

Previous studies have shown that serum ALT levels decreased on a low-carbohydrate diet, as well as decreasing insulin resistance. This open-labelled cross-over trial also found that low-carbohydrate diets are more effective than low-fat diets at reducing ALT levels [55]. Additionally, de Luis et al. (2014) reported that a 3-month hypocaloric diet intervention, whether low-fat or low-carbohydrate, led to improvements in biochemical and anthropometric parameters, such as weight, BMI, fat mass, waist circumference, systolic blood pressure, total cholesterol, LDL cholesterol, leptin, insulin, and HOMA-IR levels in obese individuals with GLP-1R polymorphism [56,57]. The Korean National Health and Nutrition Examination Surveys (KNHANES), including 19,749 participants, found that aminotransferase activity increased in correlation with the higher proportion of carbohydrates in the diet [57]. Multivariable analysis also revealed that abnormal ALT levels were strongly associated with the percentage of carbohydrates, even after adjusting for BMI, alcohol intake, fasting glucose, and triglyceride levels [57]. Across various models, carbohydrate percentage, but not fat percentage, was positively correlated with abnormal ALT and AST levels in both men and women. This suggests that carbohydrate consumption has a more significant impact on aminotransferase levels in the liver and should inform nutritional education strategies, especially in vulnerable patient groups, such as those with cancer undergoing treatments that already increase liver toxicity. Restricting carbohydrates, leading to decreased insulin action on the liver, where therapeutic levels of ketosis are often achieved with a GKI of ≤2, will help patients with cancer better tolerate chemotherapy and other standard oncology treatments [27].

More recent studies have focused on the use of the ALT/AST ratio as a predictive marker to screen for insulin resistance [58]. ALT and AST are enzymes involved in gluconeogenesis, and an increasing ALT-to-AST ratio is linked to worsening glucose regulation, metabolic impairment, and organ dysfunction, including chronic diseases such as MASLD and CVD. The KNHANES have shown that higher ALT/AST ratios were associated with impaired fasting glucose, insulin resistance (measured by HOMA-IR), undiagnosed type 2 diabetes, and declining metabolic health [58,59,60]. A higher ALT/AST ratio may offer an early identification of insulin resistance, even in those with a normal BMI [61,62]. The optimal cut-off point for the ALT/AST ratio for identifying insulin resistance was ≥0.82 in non-obese individuals and ≥1.02 in obese individuals [61]. In non-obese subjects, the positive likelihood ratio was greatest for the ALT/AST ratio. In this open-labelled cross-over trial, we observed an increase in the ratio during the suppression of ketosis (hypoketonaemia) from an average of 0.74 to 0.96 during SuK, which returned to baseline levels after sustaining euketonaemia in P3. This indicates that a shift in the ratio is evident even in this healthy population of habitually keto-adapted premenopausal lean women after a 3-week suppression of ketosis intervention and may be a more sensitive method for the detection of metabolic health decline, insulin resistance, and individual-specific hyperinsulinaemia (subclinical hyperinsulinaemia within average population reference ranges) with the absence of an elevated HbA1c and HOMA-IR.

During non-alcoholic steatohepatitis (NASH), a highly inflammatory stage of liver disease, insulin resistance has been found to be a predominant contributing factor for the lack of hepatic ketogenesis [63] which is evidenced by reduced circulating ketone body levels. It has also been shown that a 6-day ketogenic diet significantly reduced liver fat content and hepatic insulin resistance via decreased endogenous glucose production and lower serum insulin levels. This study found that a ketogenic diet could produce potential strategies for treating NAFLD/MASLD and other health conditions, where liver markers are monitored in order to determine treatment tolerance and responses [64].

Levels of ALP, albumin, total protein, total and direct bilirubin, CK, and iron did not change across the three phases of this study. Although the changes in ALP did not reach statistical significance, the observed trend towards elevation may indicate early cholestatic responses during the suppression of ketosis. This warrants further investigation to determine if sustained suppression could lead to subclinical bile duct stress. Studies have indicated a link between CK levels and BMI, waist circumference, and waist/hip ratio, suggesting that CK could serve as a marker of obesity and may also identify individuals at risk of developing obesity [65,66]. CK is a proxy marker for HbA1c, and in a non-diabetic population, CK was linked to insulin resistance and is recognised to have a positive and independent association between CK and HbA1c [67]. However, it is unlikely that levels of CK would change significantly over the 21-day period in this trial due to the effect of blood collection during the trial, which can reduce levels of HbA1c for up to a 2-month period [68], and therefore, this may have affected the trial results. Levels of iron may also have been affected by the indirect blood donations at each phase of the trial. Recent studies have linked higher serum ALP with increased mortality and an association with a high risk of developing metabolic syndrome [69,70]. Although ALP levels significantly increased from P1 to P2, they did not return to baseline levels in P3. In our mixed-effects models, ALP did not show a statistically significant relationship with both insulin and HOMA-IR; however, ALP did show a significant effect with changes in GKI levels across the study phases. ALP plays a role in various bodily functions, including the breaking down of proteins and bile duct function. Cholestasis has been shown to increase levels of ALP [71]. Further investigation on bile synthesis and transport during changes in metabolic phenotypes is warranted.

It can be seen that there is a significant positive relationship between the GKI and liver biomarkers. The relationship is even more pronounced when looking at the relationship between the 21-Day Average GKI and change in liver biomarkers, where all markers were significantly positively associated (Figure 12, Figure 13, Figure 14 and Figure 15). Meaning, the higher the average GKI is over 21 days, the higher the liver enzymes, indicating a negative trajectory with regard to liver health. When participants returned to maintaining a consistently low GKI, those who maintained the lowest had the greatest decrease in liver inflammation markers, indicating that it was the intervention that affected liver inflammation markers. The GKI was initially developed as a tool to monitor nutritional ketosis in patients with cancer [27,72] to enable improved tolerance to standard of care or metabolism-based treatments; patients with cancer may benefit from maintaining a consistently low GKI [27]. It is noteworthy to understand that increased liver inflammation markers are associated with an increase in extra-hepatic cancers [26]. The knowledge that a sustained low GKI is generally associated with healthy lower liver biomarkers would predict a decrease in risk of many chronic diseases.

Liver markers appear to be reliable early indicators for changes in metabolic health, even within a lean healthy cohort. While ALT and AST are indeed well-established clinical markers for hepatic function, this study contributes to the literature by examining how these markers respond dynamically within a controlled ketogenic intervention and reintroduction phase in a real-world cohort. This study further builds on this by exploring the relationship between liver markers and the GKI, a marker of metabolic status that integrates hepatic glucose and ketone regulation, an area that, to our knowledge, has not been previously explored in a healthy human cohort, especially in females, against liver markers. The inclusion of GKI provides a novel area to interpret shifts in liver markers in the context of sub/clinical hyperinsulinaemia, insulin resistance, and metabolic flexibility.

We included GKI in order to explore whether changes in this index correspond with shifts in liver-related biomarkers. Given the liver’s central role in regulating both glucose and ketone metabolism, GKI provides an indirect insight into hepatic metabolic function and also may be of functional utility in an at-home testing setting, as a cost-effective, easy surrogate index, simplified for general use for the public, which could help reduce burden on the healthcare system. This is the first study to investigate these relationships in a female cohort following a ketogenic diet, particularly using a mixed-effects model.

Liver markers are inexpensive and commonly measured. The liver acts as an early expression of sub/clinical hyperinsulinaemia, where chronically elevated insulin signalling (indicated as chronic hypoketonaemia) simultaneously affects other organ systems, which are less accessible for monitoring until symptomatic pathology presents. Elevated liver enzymes in the context of metabolic dysfunction are a proxy indicator of excessive insulin signalling (sub/clinical hyperinsulinaemia) and possible damage to extra-hepatic tissues. This study reinforces the clinical value of routine liver marker monitoring not just for disease detection but also for the real-time assessment of metabolic stress and dysfunction. Regular screening of ALT, AST, and GGT in at-risk populations could pre-emptively adopt early dietary interventions, potentially reverse hepatic insulin resistance, and prevent the onset of metabolic disease. Tracking biomarkers of liver function would enable patients and healthcare providers to take earlier action to prevent the development of overt pathologies, such as cancers, CVD, neurodegenerative diseases, gastrointestinal disorders, and renal system diseases, or understand when there may be a need to initiate a wider health investigation.

5. Strengths and Limitations

Our study provides a novel investigation specifically on liver biomarkers in chronic diseases of ageing in a healthy premenopausal female population who are living in sustained long-term ketosis. Collectively, these biomarkers have not been measured in this cohort before using an open-labelled, non-randomised cross-over trial. Further work assessing these biomarkers in larger male and female cohorts, both keto-adapted and non-keto-adapted individuals across different ages and in pathologies, should be considered to enhance the generalisability of this study. In a future publication and follow-up study, we plan to investigate the influence of menstrual cycle phases on all blood biomarkers measured in this cohort. This study will provide a more comprehensive assessment of the menstrual cycle and iron status, in the context of many biomarkers, whereby we aim to evaluate and quantify the effects of menstrual timing on biomarker variability in more depth, highlighting this in a broader context than only female sex hormones. Further details on strengths and limitations have been previously discussed in detail in our earlier publications, including study power and sample size calculations [1,2,28].

A strength of our study was the cultural and ethnic diversity of the participants, with a wide age range spread, including women who had not had children and had children. By implementing rigorous dietary monitoring across all phases, this ensured adherence, and each participant was therefore their own control, ensuring internal consistency and reliable cross-phase comparisons. Participants maintained their usual lifestyle and dietary preferences, and all meals were recorded daily for each meal, ensuring controlled dietary monitoring throughout the 9-week experimental period. This controlled monitoring, combined with the ethnic diversity of the participants, enhances the robustness of the findings by capturing a broader spectrum of potential metabolic responses across different genetic and cultural backgrounds. It is precisely due to the diverse ethnic origins in this cohort that it is impossible to adjust for ethnicity in our statistical models (because with only one degree of freedom for every ethnicity, the models would be automatically over-fitted). The ethnic backgrounds of participants included Colombian, Palestinian, American, Brazilian, Italian, Chinese, Thai, French, and British. This diversity adds to the generalisability of our findings and enhances the robustness of the results by capturing a broader range of metabolic responses across different genetic and cultural backgrounds.

6. Conclusions

Long-term nutritional and therapeutic ketosis (euketonaemia) maintained healthier lower levels of liver enzymes, with very clear healthy liver function parameters. Conversely, suppression of ketosis (hypoketonaemia) led to significantly elevated levels of liver enzymes ALT and GGT, as well as an increased ALT-to-AST ratio. 21 days of hypoketonaemia also significantly decreased the AST-to-ALT ratio. Our trial data supports that long-standing euketonaemia maintains health benefits to markers of liver and metabolic–endocrine health and establishes a detailed liver function profile in a healthy female cohort in long-standing nutritional and therapeutic ketosis. Our study also demonstrates that a shift in the ALT-to-AST ratio is evident, even in a long-standing healthy ketosis cohort after 21 days of hypoketonaemia via an increasing insulin demand and secretion intervention. This ratio may be used as a more sensitive method to detect changes in metabolic–endocrine health, sub/clinical hyperinsulinaemia, and insulin resistance, thereby enabling preventive healthcare. Carbohydrate consumption that results in the chronic suppression of ketosis (hypoketonaemia) for the majority of the awake hours and for many consecutive days had a significant impact on GGT and aminotransferase levels. We recommend restricting carbohydrates to within personal tolerance levels, resulting in long-standing euketonaemia (detectable at the pre-evening meal) for healthy metabolic–endocrine ageing, and to support medical interventions in order to reduce toxicity and improve tolerance and efficacy of pharmacological treatments, such as chemotherapy, enabling better patient outcomes.