The Antioxidant Potential of the Mediterranean Diet as a Predictor of Weight Loss after a Very Low-Calorie Ketogenic Diet (VLCKD) in Women with Overweight and Obesity

Obesity involves a chronic state of low-grade inflammation, which is linked to the development of several comorbidities. Recently, the very low-calorie ketogenic diet (VLCKD) has gained great interest in the treatment of obesity, almost ousting the ancient and healthy Mediterranean diet (MD). However, because these dietary regimens exploit different pathophysiological mechanisms, we hypothesize that adherence to the MD may play a role in determining the efficacy of the VLCKD. We enrolled 318 women (age 38.84 ± 14.37 years; BMI 35.75 ± 5.18 kg/m²) and assessed their anthropometric parameters, body compositions, and adherence to the MD (with the PREvención con DIetaMEDiterránea (PREDIMED) questionnaire) at baseline. The anthropometric parameters and body composition were repeated at the end of the VLCKD. At the end of the VLCKD, the women with high adherence to the MD achieved the best results in terms of weight loss and improved body composition. Specifically, the women who were above the median of fat mass (FM)% reduction had the best MD pattern, characterized by a higher consumption of extra virgin olive oil (EVOO), fruits, vegetables, and red wine, as well as a higher adherence to the MD than the women who were below the same median. In a multiple regression analysis, the PREDIMED score was the main predictor of the FM% reduction score and came in first, followed by fruit, EVOO, and glasses of wine, in predicting the percentage reduction in FM. A PREDIMED score value of > 5 could serve as a threshold to identify patients who are more likely to lose FM at the end of the VLCKD. In conclusion, high adherence to the MD resulted in higher VLCKD efficacy. This could be due to the antioxidant and anti-inflammatory properties of the MD, which are capable of establishing a metabolic set-up that is favorable to the onset of more effective ketosis.

hypothesize that previous adherence to the MD may play a role in determining the efficacy of the active phase of the VLCKD.
Thus, the aim of this study was to evaluate the differences in weight loss and body composition improvements based on adherence to the MD prior to nutritional intervention in a cohort of women with overweight and obesity undergoing the VLCKD.

Design and Setting
We enrolled 318 women with overweight or obesity who were referred clinically for weight loss at the Centro Italiano per la cura e il Benessere del paziente con Obesità (C.I.B.O), Division of Endocrinology, Department of Clinical Medicine and Surgery, University Federico II in Naples, Italy. The population was selected based on the following criteria: a) Inclusion criteria: 18 years of age or older; overweight or obesity in accordance with a BMI of ≥ 27.0 kg/m 2 ; naïve subjects (no previous treatment with anti-obesity drugs or bariatric surgery); b) Exclusion criteria: type 1 diabetes mellitus or T2DM on insulin therapy or latent autoimmune diabetes in adults; renal, hepatic, cardiac (NYHA III-IV), or respiratory insufficiency; unstable angina; stroke or myocardial infarction in the last 12 months; cardiac arrhythmias; psychiatric and eating disorders; alcohol and drug addiction; active and/or severe infections; rare metabolic diseases; pregnancy; lactation.

Anthropometric Measurements and Physical Activity
Anthropometric measurements were performed at the beginning (5 days before the start of the VLCKD) and at the end of the active phase of the VLCKD (after 45 days) by the same qualified nutritionist. The body weight was determined with a calibrated beam scale (0.1 kg approximation, Seca 711; Seca, Hamburg, Germany), and the height was determined with a wall stadiometer (0.5 cm approximation, Seca 711; Seca, Hamburg, Germany). BMI is a commonly used quantitative measure of adiposity [19]. Thus, the weight and height were combined in the following formula: Weight (kg)/Height (m 2 ) The result was then used to categorize the BMI (overweight, obesity I, II, and III classes) according to WHO criteria. Finally, the waist circumference (WC) was measured with a non-elastic tape measure at the natural indentation or halfway between the lower edge of the rib cage and the iliac crest if no natural indentation was visible (0.1 cm approximation), as reported by the National Center for Health Statistics [20]. Thirty minutes of physical activity per day was the threshold for defining physically active women, as previously reported [21].

Bioelectrical Impedance Analysis
A bioelectrical impedance analysis (BIA) was performed at the beginning and at the end of the active phase of the VLCKD by the same qualified nutritionist to assess body composition. The assessment was performed using a phase-sensitive BIA system with an 800-µA current at a frequency of 50 kHz (BIA 101, Akern Bioresearch, Florence, Italy), according to the recommendations of the European Society of Parental and Enteral Nutrition [22]. The phase angle (PhA) is calculated as the relationship between the resistance (R) of tissues, which depends mainly on tissue hydration, and their reactance (Xc), which is related to cellularity, cell size, and cell membrane integrity, according to the following formula: PhA ( • , degrees) = Xc/R* (180/π), as reported in previous evidence [23]. The same operator and instrument performed the BIA determinations under strictly standardized conditions to avoid inter-observer and inter-instrument variability.

Assessment of the Adherence to MD
Prior to the start of the VLCKD, a qualified nutritionist assessed the adherence to the MD using the PREvención con DIetaMEDiterránea (PREDIMED) questionnaire, which consists of 14 items that investigate the consumption of EVOO, fruit, vegetables, nuts, legumes, red meat, poultry, fish, animal fats, sugary drinks, sweets, and sofrito [24]. As already widely reported [25], the PREDIMED score calculation provided the following MD adherence categories: 0-5, low adherence; 6-9, average adherence; ≥10, high adherence [24].

VLCKD Intervention
Women who fulfilled the inclusion criteria were administered the VLCKD protocol with a meal replacement, consisting of active, re-education, and maintenance phases. For the VLCKD, a commercial weight-loss program (New Penta, Cuneo, Italy) was used. After a nutritional status assessment, a qualified nutritionist prepared the active-phase diet, which was prescribed by an endocrinologist, as already reported [26]. The protocol was carried out according to the European Association for the Study of Obesity (EASO) guidelines [14]. As stated in the guidelines, the VLCKD was characterized as follows: a daily energy intake of approximately <800 kcal with 13% of carbohydrates (<30 g/day), 43% of protein (1.2-1.5 g/kg of ideal body weight), and 44% of lipids (including 10 g/day of EVOO as the only fat seasoning). This composition was achieved by using high biological value protein preparations (formulated on the basis of peas, eggs, soy, and whey), containing between 15 and 18 g of high biological value protein. Four or five meals a day were provided, according to the individuals' needs. Guidance was also given on the vegetables to be consumed with regard to type, quantity (unlimited or grammed), and method of preparation/cooking. No other food was allowed beyond the meal replacements and the proposed vegetables. In addition, each woman was recommended to be well hydrated daily by consuming only water. Given the high caloric restrictions, all of the women were prescribed vitamin and mineral supplements to maintain the physiological acid/base balance (PentaCal, Penta, s.r.l., Cuneo, Italy) (B-complex vitamins, vitamins C and E, and minerals, including potassium, sodium, magnesium, calcium, and omega-3 fatty acids), in accordance with international recommendations [14]. A qualified nutritionist ascertained compliance with the VLCKD and physical activity recommendations through individual telephone counselling and also instructed the women to measure their β-hydroxybutyrate from capillary blood using test strips once a week (Optium Xceed Blood Glucose and Ketone Monitoring System; Abbott Laboratories, Chicago, IL, USA) in the morning on an empty stomach, preferably at the same time.

Statistical Analysis
The Kolmogorov-Smirnov test was used to test the data distributions. The continuous variables were expressed as mean ± standard deviation (SD) when normally distributed. The categorical variables were expressed as numbers and percentages (%). The variations were analyzed using a paired t-test for the normally distributed variables. A chi-square (χ 2 ) test was used to test for differences in the frequency distributions. The differences among the categories of adherence to the MD were analyzed by a between-groups ANOVA test followed by a Bonferroni post-hoc test. A Pearson's correlation was used to assess the association between the PREDIMED score at baseline, weight loss, and changes in body composition pre/post-intervention. A partial correlation was performed to adjust the associations for the confounding factors (BMI, fat mass (FM), and WC). A multinomial logistic regression model, odds ratio (OR), p-value, 95% interval confidence (IC), and R 2 were performed to assess the associations among the ∆% FM kg with single items of the PREDIMED questionnaire and adherence to the MD categories. In addition, two multiple linear regression analysis models (the stepwise method), expressed as R 2 , beta (β), and t, with the PREDIMED score as a dependent variable, were used to estimate the predictive value of the weight and body composition parameters (Model 1), and with the ∆% FM kg as a dependent variable, were used to estimate the predictive value of the PREDIMED score and single items of the PREDIMED questionnaire (Model 2). A receiver operator characteristic (ROC) curve analysis was performed to determine the area under the curve (AUC), criterion, sensitivity, specificity, standard error, 95% IC, and the cut-off values for the PREDIMED score in detecting patients who were most likely to lose FM at the end of the active phase of the VLCKD. A statistical analysis was performed according to standard methods using the statistical package for the social sciences software 26.0 (SPSS/PC; SPSS, Chicago, IL, USA). A p value of < 0.05 was considered statistically significant. Since this was a pilot study, no power calculations were performed. Therefore, all of the findings need to be confirmed by larger clinical trials.

Results
A total of 318 women with overweight and obesity (BMI 35.75 ± 5.18 kg/m 2 ), aged 38.84 ± 14.37 years, met the inclusion/exclusion criteria and were included in this study. All of the women were evaluated at baseline and at the end of the active phase of the VLCKD. Once a week, a nutritionist conducted a telephone interview to confirm compliance with the VLCKD, ketosis status, and any changes in the physical activity levels of the women.
Before the start of the VLCKD, the PREDIMED score was 6.38 ± 2.36. Most of the women (172, 54.2%) had average adherence to the MD, followed by low (117, 36.8%) and high adherences to the MD (29, 9.1%); Figure 1. Compared to the baseline, no women changed their physical activity levels during the 45 days of the VLCKD (101, 31.8% vs 101, 31.8%; χ 2 = 0.01, p = 0.932). Table 1 reports the anthropometric measurements and body composition parameters of the study population at baseline and at the end of the active phase of the VLCKD, respectively. At Compared to the baseline, no women changed their physical activity levels during the 45 days of the VLCKD (101, 31.8% vs. 101, 31.8%; χ 2 = 0.01, p = 0.932). Table 1 reports the anthropometric measurements and body composition parameters of the study population at baseline and at the end of the active phase of the VLCKD, respectively. At the end of the active phase of the VLCKD, the mean percentage weight loss was −7.35 ± 2.72%, with a statistically significant decrease in WC (p < 0.001). Of interest, the distribution of the women across the BMI categories was significantly modified, with an increase in the prevalence of women with normal weight (+2.2%), overweight (+15.4%), and grade I obesity (+0.3%), while the prevalence of women with grade II and grade III obesity dropped to −5.9% and −12.0%, respectively. As for the body composition parameters, all of the parameters statistically changed at the end of the VLCKD. The FM (kg and %) was statistically reduced (p < 0.001), and the fat-free mass (FFM) (kg and %) (p < 0.001) and PhA (p < 0.001) were statistically increased. Data are expressed as mean ± standard deviation (SD) or as numbers (n) and percentages (%). VLCKD, very lowcalorie ketogenic diet; ∆%, percentage change; BMI, body mass index; WC, waist circumference; BIA, bioelectrical impedance analysis; R, resistance; Xc, reactance; FM, fat mass; FFM, fat-free mass; PhA, phase angle. A p-value in bold denotes a significant difference (p < 0.05). Table 2 reports the anthropometric measurements and body composition parameters of the study population according to adherence to the MD. At the baseline, the subgroups of women did not differ in age (p = 0.264), R (p = 0.291), FFM (kg) (p = 0.136), and PhA (p = 0.091). The women with high adherence to the MD had significantly lower BMIs (p < 0.001), WCs (p < 0.001), and FMs (kg and %) (p < 0.001) than the women with low adherence to the MD. At the end of the active phase of the VLCKD, the women with high adherence to the MD had lost more body weight (p < 0.001), WC (p < 0.001), and FM (kg and %) (p < 0.001) and had increased FFM (%) (p < 0.001) and PhA (p = 0.010) compared to the women with low adherence to the MD; Table 2. Table 3 reports the physical activity levels and BMI categories of the study population according to adherence to the MD. Both at baseline and at the end of the active phase of the VLCKD, the women had not changed their physical activity levels (p = 0.059). At baseline, the lowest percentage of women had grade III obesity (p < 0.001). At the end of the active phase of the VLCKD, all of the BMI categories changed, showing a reduction in grade III obesity (p < 0.001) and an increase in normal-weight women (p < 0.001).  Table 4 shows the response frequency of the single items included in the PREDIMED questionnaire of the study population. EVOO, which was the main culinary lipid used, was the main food of the MD consumed by the study population (84.3%), followed by low consumption of red/processed meats (51.3%).
The single items included in the PREDIMED questionnaire and the categories of adherence to the MD were examined by stratifying the study population above and below the median of the ∆% FM kg (−14.84%); Table 5. The women that were above the median of the ∆% FM kg had the best Mediterranean dietary pattern, characterized by the highest consumption of EVOO (p < 0.001), fruit (p < 0.001), vegetables (p < 0.001), and red wine (p < 0.001), and high adherence to the MD (p = 0.002) compared to the women that were below the median of the ∆% FM kg.   Figure 2 shows the PREDIMED scores above and below the median of the ∆% FM kg (−14.84%). The women that were above the median of the ∆% FM kg had the highest PREDIMED score than the women that were below the median of the ∆% FM kg (p < 0.001).  Figure 2. PREDIMED scores of the study population above and below the median of the Δ% FM kg. PREDIMED, PREvención con DIetaMEDiterránea. A p-value in bold denotes a significant difference (p < 0.05). Table 6 reports the simple and after-adjustment correlations for the confounding variables among the PREDIMED scores and changes in the anthropometric measurements and body composition parameters at the end of the active phase of the VLCKD. The Figure 2. PREDIMED scores of the study population above and below the median of the ∆% FM kg. PREDIMED, PREvención con DIetaMEDiterránea. A p-value in bold denotes a significant difference (p < 0.05). Table 6 reports the simple and after-adjustment correlations for the confounding variables among the PREDIMED scores and changes in the anthropometric measurements and body composition parameters at the end of the active phase of the VLCKD. The PREDIMED score correlates with all of the parameters except for the ∆% WC (p = 0.093) and ∆% PhA (p = 0.430). Table 6. Simple and after-adjustment correlations for the confounding variables among the PRED-IMED score and changes in the anthropometric measurements and body composition parameters at the end of the active phase of the VLCKD. The results of the bivariate proportional odds ratio (OR) model, performed to assess the association of the ∆% FM kg with the quantitative variables of the PREDIMED questionnaire and adherence to the MD categories, are reported in Table 7. The highest values of the ∆% FM kg were significantly associated with the consumption of all foods of the MD included in the PREDIMED questionnaire, except for the intake of poultry more than red meats (p = 0.276) and the use of sofrito sauce ≥ 2/week (p = 0.195). Of interest, the highest values of the ∆% FM kg were also significantly associated with all categories of adherence to the MD (p < 0.001 for all).

Parameters
To compare the relative predictive power of the PREDIMED score associated with the weight loss and body composition changes at the end of the active phase of the VLCKD, we performed a multiple linear regression analysis model. In this model, the ∆% FM kg represents the most predictive parameter of adherence to the MD, which was entered in the first step (p < 0.001). The results are reported in Table 8.
To compare the relative predictive power of the PREDIMED score and the single items of the PREDIMED questionnaire associated with the ∆% FM kg, we performed a second multiple linear regression analysis. In this second model, the PREDIMED score was entered in the first step (p < 0.001), followed by fruits ≥ 3 servings/day (p = 0.010), the use of EVOO as the main culinary lipid (p = 0.001), and wine glasses ≥ 7/week (p = 0.002); Table 9.   A ROC analysis was performed to determine the cut-off value of the PREDIMED score, which was predictive of the greatest FM loss at the end of the active phase of the VLCKD (above the median value of the ∆% FM kg, −14.84%), as shown in Figure 3. A value of the PREDIMED score of >5 (p < 0.001; AUC 0.825; standard error 0.024; 95% IC 0.778 to 0.872) could serve as a threshold for identifying patients who are most likely to lose FM at the end of the active phase of the VLCKD.
1 Figure 3. ROC for the predictive values of the PREDIMED score in detecting patients who are most likely to lose FM at the end of the active phase of the VLCKD. A p-value in bold denotes a significant difference (p < 0.05).

Discussion
The aim of this prospective observational study was to determine whether adherence to the MD could predict the efficacy of the active phase of the VLCKD in terms of weight loss and body composition changes in women with obesity.
Only a small percentage of the women had high adherence to the MD (9.1%), and more than 80% of the women reported consuming EVOO as their main condiment at the expense of using other, less healthy condiments. As expected, the women with high adherence to the MD had the best anthropometric and body composition parameters. In fact, there is consistent scientific evidence that adherence to the MD is inversely associated with adiposity, in particular, abdominal adiposity [27,28].
Our study revalidated the efficacy of the VLCKD in the treatment of obesity by allowing the enrolled women to achieve significant weight and FM losses and an increase in PhA after 45 days of the active phase. PhA is a measure derived from BIA analyses and, being an indicator of cellular health and integrity [29], is, therefore, a valuable tool for determining the inflammatory state associated with many diseases, such as obesity [30]. Thus, along with the reduction in anthropometric parameters and improvements in body composition, the VLCKD also shows promising effects in reducing the low-grade inflamma-tion associated with obesity. In this regard, in a pilot study of 260 women with obesity, we recently reported changes in PhA during the active phase of the VLCKD that occurred very early and independently of weight loss and were negatively associated with high-sensitive C-reactive protein levels [26]. The efficacy of the VLCKD on anthropometric parameters, body composition, and, more recently, on inflammatory status is mainly mediated by ketone bodies (acetone, acetoacetate, and β-hydroxybutyrate) [31][32][33], which are produced at the mitochondrial level of different tissues, mainly from the oxidation of fatty acids, and are exported to peripheral tissues to be used as an energy source [34]. These new and interesting effects of the VLCKD are underpinned by several mechanisms exerted by ketone bodies, including activation of the peroxisome proliferator-activated receptor gamma and hydroxycarboxylic acid receptor 2, inhibition of NF-κB activation, and inhibition of the NLRP3 inflammasome [35].
As novel findings, we report that the women who had higher adherence to the MD at baseline showed greater weight and FM losses with a greater PhA increase. The high adherence to the MD was statistically associated with percentage decreases in BMI, WC, and FM, even after adjustments for the main confounding factors, including BMI, WC, Xc, FM, and PhA, at baseline. Interestingly, at the end of the active phase of the VLCKD, 13.8% of the women with high adherence to the MD became normal weight, while none of the women with low adherence to the MD achieved the same results, with no difference in their physical activity levels. Analyzing the individual foods from the PREDIMED questionnaire, the participants above the median value of FM loss showed a higher consumption of typical Mediterranean pattern foods, such as EVOO, vegetables, fruit, red wine, fish, nuts, and white meat instead of red meat, and a lower consumption of butter, soda drinks, commercial sweets, and confectionery. These foods were also statistically associated with FM loss. In fact, the participants with the highest PREDIMED scores were located above the median FM loss value.
Studies have shown that obesity causes mitochondrial dysfunction, or the inability of mitochondria to produce and maintain sufficient levels of adenosine triphosphate (ATP), which is the result of an imbalance in nutrient signal input, energy production, and oxidative respiration [36]. Indeed, low-grade inflammation and excessive nutrient intake (as is often the case in obesity) can lead to an increased production of ROSs and cause oxidative stress. Oxidative stress, in turn, can trigger mitochondrial changes, called mitochondrial dysfunction [36]. Mitochondria are essential for cellular energy metabolism, as they generate ATP by oxidizing carbohydrates, lipids, and proteins [37]. Of note, the production of ketone bodies during the VLCKD occurs at the hepatic mitochondrial level [37], which allows us to speculate that weight excess could negatively affect these processes as well.
Recently, the so-called "mitochondrial nutrients", which include polyphenols, plantderived compounds, and polyunsaturated fatty acids (PUFAs), have attracted the attention of researchers with the idea that improving mitochondrial structure and function is a plausible strategy to correct mitochondrial dysfunction [38]. In this respect, scientific evidence has demonstrated the beneficial effects of various bioactive MD compounds on mitochondrial dysfunction [38]. Such valuable effects might explain why the high adherence to the MD and the parallel consumption of certain specific nutrients might enhance the efficacy of the VLCKD.
In our study, EVOO, the key condiment in the MD, was associated with better outcomes in terms of weight loss and improvements in body composition following the VLCKD. Hydroxytyrosol, a polyphenol from olive oil, has been shown to be effective in regulating mitochondrial dysfunction in animal models by exploiting its ability to modulate the mitochondrial apoptotic pathway in skeletal muscle and the liver [39]. Hydroxytyrosol can also enhance mitochondrial biogenesis (specifically by increasing mtDNA and the number of mitochondria) by increasing the expression of the genes involved (PGC-1α, NRF-1, and TFAM) via the AMPK pathway [40].
Similar results were seen for some compounds found in abundance in red wine and plant products, such as fruit and vegetables, which we also found to be associated with improved efficacy of the VLCKD. In this regard, a study in mice reported that resveratrol, which is found in foods such as grapes and berries, improved mitochondrial activity by competing with NAD+ in a solubilized complex [41]. Additionally, in mice, resveratrol appears to be able to prevent metabolic diseases (such as insulin resistance and obesity) by improving mitochondrial function through the activation of genes such as PGC-1α and SIRT-1 [42]; this result was later corroborated in another very similar study [43]. Among the antioxidants, lycopene, which belongs to the phytochemical family of carotenes and is abundant in foods such as tomatoes and grapefruit, also improved mitochondrial dysfunction by exerting an anti-inflammatory effect on mice [44]. In particular, lycopene regulated the expression of certain genes, such as SIRT1, PGC1α, Cox5b, Cox7a1, Cox8b, and Cycs, as well as many respiratory chain complexes [44]. Finally, ellagic acid, which is found in strawberries and walnuts, prevented metabolic disorders by acting on mitochondria in the following two ways: directly, by decreasing the amount of ROSs and mitochondrial damage, and indirectly, by restoring the total dehydrogenase activity in the mitochondria [45].
We cannot omit to mention PUFAs, which are precursors of many other substances in the body (particularly some of these involved in blood pressure regulation and inflammatory responses) and fundamental components of cell membranes [46]. It is now known that excessive amounts of n-6 PUFAs (and a high n-6/n-3 ratio), as found in Western diets, promote the pathogenesis of many diseases, including cardiovascular diseases, cancer, and inflammatory and autoimmune diseases, while higher levels of n-3 PUFAs (and a low n-6/n-3 ratio) exert suppressive effects. Interestingly, it has been observed that a diet with a high n-3/n-6 ratio, as is the case with MD, shows an enhancement of the activities of mitochondrial complexes, accompanied by an attenuation of fumaric acid and oxidative stress [46]. In particular, walnuts, which are part of the MD and are the only tree nuts to provide n-3 PUFAs (with the n-6/n-3 ratio being around 4/1), have antioxidant and anti-inflammatory activities via the activation of the Nrf2/ARE and down-regulation of the NF-κB pathways [47].
However, we showed that the main predictor of efficacy of the VLCKD was the PREDIMED score, i.e., the whole Mediterranean nutritional model. In particular, a cut-off value above 5 was the best predictor of FM loss after the active phase of the VLCKD. This result is of fundamental importance, more than the individual foods, since it is the entire dietary pattern that probably sets up the body in a more favorable condition for weight loss and improved body composition through the synergy of the different nutrients. In addition to the individual mechanisms discussed above, which are limited to animal models, this can be partly explained by the intake of polyphenols and monounsaturated fatty acids in the Mediterranean dietary pattern, which, in addition to improving mitochondrial dysfunction, have been shown to inhibit de novo lipogenesis, improve peripheral insulin sensitivity [48], activate β-oxidation, and increase lipolysis [49]. It is probably also for this reason that the VLCKD in subjects with high adherence to the MD finds a more supportive ground for implementing its weight loss mechanisms.
In brief, it is possible to hypothesize that high adherence to the MD may, in part, attenuate mitochondrial dysfunction in women with obesity, resulting in a more efficient production of ketone bodies, and, in part, support those metabolic pathways directed toward fat oxidation, ultimately resulting in a basal metabolic set-up that is more conducive to weight loss.
We are aware that there are some limitations in our study, but also several strengths. First, the cross-sectional nature of this study did not allow for any statements on a causal relationship between adherence to the MD and weight loss and body composition improvements associated with the active phase of the VLCKD. Second, because we did not assess KB levels in the blood, we could not assess any differences in KB production. However, ketur capillary tests, which were performed independently by our patients, confirmed the ketosis status of the study population. Third, the PREDIMED score, although easy to perform by the participants, only allows for relative but not absolute statements regarding the degree of adherence to the MD. Nevertheless, the PREDIMED questionnaire was per-formed face-to-face and was not self-reported in order to minimize any bias related to filling out the questionnaire. Of interest, the PREDIMED questionnaire allows one to provide feedback to the subjects immediately after the interview is completed. In addition, to avoid inter-operator variability, only one expert nutritionist evaluated the anthropometric measures, administered the PREDIMED questionnaire, and assessed, executed, and interpreted the BIA measurements. Moreover, we increased the homogeneity of the patients' samples by including only women with overweight or obesity in order to improve the power of the study. Additionally, we did not perform a quantitative assessment of inflammation markers and oxidative stress; however, we have already widely reported that PhA may be a marker of inflammation and is associated with the MD [26,30]. In addition, with regard to logistic regression analyses, when the incidence of events/number of patients is high (as in this study), then the odds ratio can be misleading as it exaggerates the size of the effects [50]. Finally, although the gold standard for body composition assessments is dual energy X-ray absorptiometry (DXA), we used BIA in our study. However, BIA is also a validated and widely used method for assessing body compartments in subjects with obesity, with close concordance with DXA results [51,52].

Conclusions
In conclusion, this is the first study to report that the efficacy of the VLCKD is greater in women with high adherence to the MD before starting the ketogenic diet. This could be due to the presence of interesting bioactive compounds in the Mediterranean diet, which together and in the context of the MD could lead to a favorable metabolic set-up for the onset of a more effective ketosis. Furthermore, as a possible translational application, these results suggest that a specific cut-off value for the PREDIMED score (<5) can help to identify women in whom the VLCKD would achieve lower efficacy results. It is, therefore, possible to hypothesize that these women could boost the efficacy of the VLCKD with an earlier, brief nutritional intervention aimed at improving adherence to the MD, particularly by increasing foods such as EVOO and fruits before starting the ketogenic protocol.