Recurrent circadian fasting (RCF) improves blood pressure, biomarkers of cardiometabolic risk and regulates inflammation in men

The study employed an open label longitudinal follow-up design with convenience sampling to obtain three measurement points: baseline (before fasting), end of fasting (4 weeks) and post-fasting (1 month) (Fig. 1a). The target population was adults aged 20 years and above, belonging to eight Union Councils within Peshawar, Charsadda (central), Malakand and Swat (northern) districts of Khyber Pakhtunkhwa (KPK), Pakistan. The subjects were contacted through NEAT (nutrition, education, awareness and training), a health organization registered with the Government of Pakistan, Department of Social Welfare, KPK, Pakistan. NEAT advocates nutrition awareness and counselling in KPK, Pakistan. Addresses and mobile/phone numbers of participants were obtained from the address directory of NEAT. A SMS was sent to subjects who fulfilled the inclusion/exclusion criteria to invite them to participate in this study. A 1-day workshop was arranged for potential participants to educate them about the study design and purpose. Written informed consents were obtained from the participants during the workshop. The inclusion criteria were to include otherwise medically healthy subjects with no recent past history of infectious or uncontrolled non-communicable chronic diseases. The study was approved by the Research Ethics and Review Board of NEAT (Social Welfare Department, Govt. of Pakistan). While both genders were targeted for the study, only few females (n = 5) participated which did not allow sufficient numbers for inclusion in the analysis. By tradition, females rarely contribute to such studies in presence of males. In 2016, Ramadan fasting began on the 6th of June and ended on the 4th of July, lasting for 29 days. Humidity (40%) and temperatures was consistent throughout the duration and averaged 38/30 °C, with the minimum and maximum at 35/29 °C and 41/38 °C respectively. The average fasting duration was 16 h, lasting between dawn and sunset. The baseline survey was conducted 1 week before the start of fasting and an ongoing survey was conducted during the last week (week 4) of fasting. The final survey (post-fasting survey) was conducted 1 month after the end of fasting (1st week of August).

We calculated sample size based on weight loss, our primary outcome, and the necessary physiological adaptation when RCF is conducted appropriately. In average, a loss of 0.8 kg per week is achieved during RCF based on our observation from previous RCF analysis. This open label longitudinal follow-up study, was powered to detect a mean change of 0.1 kg/day between the fasting and post-fasting periods. With α = 0.05 and β = 0.8, we required n = 34 participants to detect a change of 3.0 kg for the whole RCF period. As we were also interested in older adults [22], we therefore aimed at recruiting a final sample of n > 34 individuals < 35 years and n > 34 individuals > 50 years old. The final number of participants to this study was n = 100 and we could collect samples and analyse the respective blood from n = 78 adult males, including n = 37 young and n = 41 older adults.

Data were obtained through pre-tested, semi-structured questionnaires. One questionnaire elicited information related to the socio-demographic, fasting and dietary behaviour of the participant. The same set of questionnaires was used for all follow-up surveys. Subjects were also instructed to communicate the consumption of any special food items during the fasting period. Data collection was performed by student members of NEAT who received special training on questionnaire administration and anthropometric measurements before each assessment. Field supervisors were also deployed to oversee and perform on-site questionnaire checks to ensure completeness and same-day rectification of incomplete responses and measurements. Electronic weighing scales were used for all weight measurements (± 0.1 kg). Body mass index (BMI) was calculated as weight/height² (kg/m²), and body fat was measured using the Bio-impedance Analysis (BIA) technique (Bodystat 1500, Bodystat, Isle of Man, UK).

Dietary data were collected using semi-quantitative Food Frequency Questionnaire (FFQ) by conducting face-to-face interviews as well as the 24 h dietary recall. In these sessions, subjects were asked to recall their nutritional intake—including breakfast and their final meal of the day. Information on the cooking method (fried, boiled, steamed, roasted) and source of food (home-cooked vs. outsourced) were also collected. During the interview, an adult of the household would be present to corroborate the nutritional report so as to ensure accurate verbal reporting. Serving size was quantified in terms of cups, bowls, and spoons using a specified reference (for instance, half of the small bowl). When interviewees gave unspecific answers (e.g. ‘I used a lot of milk in my tea’), they were prompted to convert their estimation to cup equivalents. The average food intake over these 3 days was calculated and used to estimate consumption. Nutritional content was computed based on food composition tables specific for the Pakistani context and extensively used in our previous studies [23, 24]; the analysed categories included total energy, fat, protein, carbohydrate and selected vitamins and minerals. The percentage of energy derived from fat, protein, and carbohydrate were also calculated. Given the distinction in energy distribution between liquid and solid food [25], we also stratified the energy contribution (%) obtained from both solid (solid calories) and liquid sources (liquid calories) in the estimation of caloric intake. Participants were asked to disclose any consumption of prescribed medication or dietary supplementation.

Blood samples were tested for blood count (white blood cells, haemoglobin, platelets) at the Clinical and Research laboratory in Charsadda, KPK, Pakistan. Plasma samples were analysed for the levels of glucose, triglycerides, C-reactive protein, creatinine, ferritin and blood urea nitrogen (BUN) using a biochemical blood analyzer (Hitachi 7180, Hitachi, Japan). Cell counts for WBC (white blood cells), RBC (red blood cells), neutrophils, lymphocytes, monocytes, eosinophils, basophils, as well as haemoglobin concentration, HCT (haematocrit), MCV (mean corpuscular volume) and MPV (mean platelet volume) were obtained using a blood cell counter (Hemavet 0950, CDC Technology, USA). Serum from participants were thawed and adiponectin (Abcam, Cambridge, UK), ketone bodies (Cayman Chemical, MI, USA), orexin A (Wuhan USCN Business Co., Ltd., Hubei, China), PPARβ (MyBioSource, CA, USA) and sCD14 (Elabscience, Hubei, China) levels were measured by Enzyme-Linked Immunosorbent Assay (ELISA). Serum samples were added to 96-well plates that have been pre-coated with antibodies specific to the marker of interest. The bound markers were then incubated with enzyme-linked antibody conjugates. Wells were washed to remove any unbound antibodies. Substrate solution was added to the wells and the reaction was stopped when colour development was observed to be optimal, as per manufacturer’s recommendation. The optical densities were measured at wavelengths indicated in the manufacturers’ protocol on the EnVision^® 2104 multimode microplate reader (Perkin Elmer, MA, USA). All other biomarkers were measured using multiplex immunoassays, MILLIPLEX© Multiplex Assays (Merck Millipore, MA, USA) and ProcartaPlex multiplex panel (Thermo Fisher Scientific, MA, USA), based on Luminex^® xMAP^® technology. Serum samples were incubated overnight in 96-well plates with fluorescent-coded magnetic beads; each conjugated with capture antibodies against the marker of interest. The plates were washed and biotinylated detection antibodies were incubated with the complex for 1 h. Streptavidin-PE was then added and incubated for 30 min. The plates were washed and the beads were resuspended in sheath fluid for measurement on the Luminex™ FLEXMAP 3D^® (Luminex, TX, USA). Data was acquired using xPONENT^® 4.0 (Luminex, TX, USA) software and analyzed with the Bioplex Manager™ 6.0 software (Bio-Rad Laboratories, Hercules, CA, USA).

The analysis of specific biomarkers was performed using Prism V8.0 (GraphPad). Violin plots reveal the medians, quartiles and the distribution of individual data and one-way ANOVA with repeated measures was used for group comparisons. Log- or square root transformation (depending on the obtained best fit to the Gaussian model) was employed when data were not normally distributed. Chi squared test was used in order to the categorical variables’ distribution. Regression analyses were performed using Stata 14 (Stata Corporation, College Station, Texas, 2015), α = 0.05. Results are presented as estimates with associated standard errors or 95% confidence intervals. All quantitative biological variables related to inflammation with left censored data were converted to categorical variables: (i) by using the limit of detection or LOD (0.64 pg/ml); or (ii) by using the mean or the median to obtain binary variables. The selected cut-off point depended on the variable distribution: (i) the detection limit was used if more than 50% of the observations fell below it; (ii) the median was used where the frequency of observations below the detection limit was > 25% and < 50%.

Finally, in order to appreciate shifting trends in biomarker concentrations during fasting, we used identical cut-off points to categorize data into three study phases. Specifically, the median computed during the pre-fasting phase were used to categorize data during the fasting and post-fasting phases. Only CRP, platelets, WBC, APRIL, MIG and calcitonin variables, with fully available observations, were considered as quantitative variables.

In this prospective study, we followed the progress of adult men, who completed a daily average of 16 h fasting for 29 consecutive days. Women participants (n = 5) were not included in this analysis due to their paucity. Participants began fasting at sunrise and ended their fast at sunset, leaving a maximum of 8 h for food intake (Fig. 1a). The nutritional intake of participants was not restricted but the study area was selected based on trends and dietary behaviour from previous fasting seasons during which an average of 3.0 kg weight loss was observed. Questionnaires were collected in the presence of a household relative to corroborate the dietary report and participants were requested to resume their daily routine. All participants fasted simultaneously, starting and ending at a similar time of day; the elderly were not precluded so that we could study the effect of age. Blood was collected before the commencement of fasting, on Week 4 of fasting and 1 month after the end of fasting, for the measurement of cardiometabolic and inflammatory markers.

The basic baseline demographic characteristics of participants are shown in Table 1. All participants were recruited from semi-urban areas with few displaying chronic diseases, as this is often a medical condition preventing from fasting. For instance, only 7.6% have reported diabetes as a condition and while hypertension was not reported we assume a similar limited fraction of participants exhibit such condition. Most participants (80%) were free from adverse health events during the fasting period, while some suffered from diarrhoea (5.0%), fever (5.0%) or malaria (10%). Mean age was 47.1 ± 21.8 years—young (n = 37, 20 to 30 years old), middle-aged (n = 18, 52 to 64 years old) and elderly (n = 23, 65 to 83 years old). Although the educational level of participants was generally low, as more than 30% had not received formal education, a majority of participants (90%) had average monthly incomes (mean USD 340.8 ± 212.8). The average age when participants started annual long-term fasting was 11.5 ± 2.5 years. Five participants missed fasting for 2–3 days as they were travelling (n = 2) or ill (n = 3). Approximately one-third of participants (30.7%) fasted for several days (5.4 ± 1.6 days) in other periods of the year.

Aside from small quantities of cooked food items that were purchased from the market, the bulk of consumed food was prepared at home (84%) thus mainly excluding the possibility of results affected by foods of non-home origin. The descriptions of consumed food and beverage, as well as average consumption during the fasting period are provided in Additional file 1: Tables S1, S2. Any significant deviation from the average dietary pattern was recorded, including information on the frequency and type of food. For example, participants had 3.4 ± 1.6 different types of dishes a day during non-fasting periods but this expanded to 12.5 ± 2.8 different dishes during the 29-day fasting period. The frequency of food intake, however, was higher (Fig. 1b) during non-fasting days as compared to fasting days (7.2 ± 1.2 vs. 4.2 ± 2.3 times per day, respectively). With less frequent but more diverse nutritional intake, fasting resulted in a significant reduction of caloric (p < 0.0001) and carbohydrate intake (p = 0.003), while the consumption of fat, proteins and fibre was maintained (Table 2).

While TRF and FMD are not usually accompanied by weight loss, we observed a significant reduction in the weight of participants at the end of fasting (mean difference = − 3.69 kg ± 0.15), but weight recovery ensued in the following weeks (mean difference = + 2.60 kg ± 0.03). This matched our previous observations and also confirms the power of the study to be sufficient for the observation of > 3.0 kg weight loss. Although the extent of weight loss was heterogenous among subjects (in terms of absolute values), post-fasting weight gain was more consistent throughout the cohort (Fig. 1c). During the fasting period, we analysed metabolic markers (Fig. 1d) associated with nutrient sensing. Leptin, which promotes satiety, was significantly reduced during fasting (p = 0.0177) and post-fasting leptin levels remained significantly lower than baseline (p = 0.0102). Peroxisome proliferator-activated receptors (PPAR)-β, which is involved in lipid and glucose homeostasis and is the most abundant member of the PPAR family, was also reduced during (p = 0.0002) and post-fasting (p = 0.0052). Similarly, fasting glucose levels were significantly reduced during the fasting period (p < 0.0001) and the difference was sustained during the post-fasting period (p < 0.0001). A series of nutritional, clinical, and inflammatory markers are presented as a heat map in Additional file 2: Figure S1 to provide an overview of the shifts induced by fasting and any evidence of post-fasting recovery.

As shown in Fig. 2a, systolic blood pressure was significantly reduced by fasting (140.6 mmHg ± 25.9 vs. 124.2 mmHg ± 22.9). This effect persisted in the post-fasting period as average systolic pressure remained at 130.0 mmHg ± 22.6 1 month after fasting (Table 2, p < 0.0001). Similar observations were made for diastolic blood pressure (Table 2) and heart rate (Fig. 2b, p < 0.0001). In the studied population, 33/78 (42.3%) individuals had high systolic BP but this was reduced to 19/78 during fasting. Triglyceride levels, which are positively associated with cardiovascular risks, were reduced during and after fasting (Fig. 2c).

We observed significant shifts in the levels of circulating biomarkers during and after fasting, as highlighted in dotted boxes in the heat map shown in Fig. 2d, which represents all the circulating biomarkers measured in the serum by ELISA/Multiplex. Specific biomarkers of interest related to cardiovascular and metabolic health are described in further detail in Fig. 2e, f, respectively. d-Dimer, the by-product of fibrin, reflects thrombus formation and lysis while hepatocyte growth factor (HGF), initially thought to be liver-specific, promotes angiogenesis and may play an important role in cardio-protection and post-injury cardiomyocyte regeneration [29]. Resistin acts in the pathogenesis of atherosclerosis by promoting endothelial dysfunction, vascular smooth muscle cell proliferation, arterial inflammation, and the formation of foam cells. Resistin is also an adipokine that is linked to insulin resistance. All three molecules are independent risk factors for cardiovascular diseases [30]. While not significant (p = 0.075), likely due to heterogeneous readings, a 30% reduction in average d-dimer levels was observed during the fasting period. Nevertheless, fasting induced a significant reduction in HGF (p = 0.0009) and resistin (p < 0.0001) levels, although both were restored to baseline levels after fasting.

As liquid intake was prohibited during the 16 h fasting period, we also analysed kidney function by measuring blood urea nitrogen (BUN) and creatinine levels. As expected, both metabolites were significantly lowered by fasting (Additional file 2: Figure S2A, p < 0.0001) and the BUN-to-creatinine ratio before, during and after fasting was 14.55 ± 2.33, 16.25 ± 4.11 and 14.07 ± 1.63 respectively (p < 0.0001). As for the assessment of liver function, we observed significant changes in l-alanine aminotransferase (ALAT) and l-aspartate aminotransferase (ASAT) levels during fasting (p < 0.0001, data not shown), in addition to previously described changes in HGF levels.

As our initial analysis revealed prolonged glycemia during and after fasting, we were interested in the levels of β-hydroxybutyrate, which could be triggered as a compensatory mechanism to promote ketosis. On the contrary, we observed that fasting contributed to reduced β-hydroxybutyrate levels (Fig. 2f). In investigating the potential benefits of fasting, we tested metabolites associated with bone health. Levels of osteoprotegerin (OPG), which inhibits osteoclastogenesis and bone resorption, were increased but this difference did not reach statistical significance. Nevertheless, we observed a robust elevation in the levels of orexin-A (p = 0.0183), a hypothalamus-derived hormone which promotes muscle toning and wakefulness. Finally, we were interested in whether there were any fasting-induced changes in gut-permeability and measured levels of sCD14—a surrogate marker of microbial translocation. sCD14 levels were significantly increased during fasting (Additional file 2: Figure S2, p = 0.0422), which suggests reduced barrier function and more bacterial translocation.

Figure 3a shows a comparison of the levels of inflammatory cytokines at all three time points. All participants had CRP levels below 3 mg/ml during the fasting period (1.89 ± 0.25 mg/ml), but post-fasting levels of CRP were slightly elevated above baseline (3.14 ± 0.84 mg/ml); the number of individuals with elevated post-fasting CRP levels was similar to the number observed before fasting. While a similar profile was observed for TNFα, IL-6 levels were elevated (p = 0.0281) during fasting and remained elevated thereafter (p = 0.0056). Using the Chi squared test, multinomial regression and linear regression, we further probed the evolution of CRP, IL-6 and TNFα during the different stages of fasting (Table 3). Plasma TNFα concentration was significantly reduced during (− 26.3%, p = 0.001) and after fasting (− 9.6%, p = 0.003). Similarly, CRP levels were significantly altered across the three study points (p < 0.0001). Data from regression (GLS and RMRL) models indicate that fasting promotes a reduction of TNF-α levels (coef. = − 1.16; 95% CI [− 1.91 to − 0.42]) but increases IL-6 levels (coef. = + 0.77; 95% CI [0.06 to 1.49]). After eliminating confounding factors, only older adults (aged ≥ 52 year) were resistant to elevations in TNFα levels (coef. = − 1.56; 95% CI [− 2.61 to − 0.52]) during fasting. Younger adults were more resilient to increased plasma TNFα levels in the post-fasting phase (adjusted coef. = − 1.35; 95% CI [− 2.52 to − 0.18]). While fasting was associated with a significant decrease in CRP levels (− 1.14 mg/L) across all participants, age was associated with greater CRP decrease (− 1.61 vs. − 0.65 mg/L; p < 0.0001).

In consideration of our observations on inflammatory markers, we further tested whether molecules composite to the senescence-associated secretory phenotype (SASP) were also similarly regulated by fasting (Fig. 3b). Here, we show that IP-10 (interferon gamma-induced protein 10) and platelet-derived growth factor AA (PDGF-AA) were significantly downregulated following recurrent fasting (p = 0.0218 and p = 0.0126, respectively). However, IL-8 levels were increased during fasting (p = 0.0025) and remained high after the fasting period, similar to IL-6. Altogether, the data suggest an complex regulation of inflammatory processes during RCF. Investigating deeper, we tested for any modulation in the circulating levels of chemokines related to three immune cell populations: T helper cells, B cells and monocytes (Fig. 3c). The data suggest an overall diminution in the expression of chemokines during fasting. Based on this data we finally intended to show a potential involvement of a key regulator of inflammation, the inflammasome (Fig. 3d). IL-1β, a core molecule of the inflammasome, was significantly increased by fasting (263.3 ± 1383 vs. 674.9 ± 2079, p < 0.0001). Platelets, a significant source of inflammation showed a concomitant increase in their numbers (p < 0.0001). IL-1β signalling is known to be counteracted by circulating IL1-RA, which binds competitively to IL-1R. Nevertheless, we did not observe a significant increase in IL1-RA levels in our paired analysis.

For a more holistic approach, we performed two types of statistical analysis on participant data during the three fasting phases. Table 4 shows the different associations between the levels of biomarkers with fasting status.

In light of the differential resilience to RCF in older adults (Tables 3, 4) we furthered our exploration in this context. Aging studies would usually classify individuals as elderly when they reach 65 years old despite life expectancy is > 80 years. Since life expectancy in Pakistan is 66.5 years, we were more liberal in our classification of elderly subjects. We compared young individuals (< 35 years) with older adults in general (> 50 years) and alternatively further separated the older adults into elderly individuals (> 65 years).

We provide the results of the principal component analysis (PCA) analysis of biomarkers for all participants indistinctively of age (Fig. 4a). The principal component analysis (PCA) was applied so as to identify components reflecting young or old adults associated profile or signature. We observed an obvious shift in PC1 (29.3%) with fasting status, as depicted by the centroids of pre-fasting (green), fasting (blue) and post-fasting (red) data points from baseline. The top 20 molecules that contribute to each of the components are reported in Table 5. While PC1 consisted mainly of inflammatory and metabolic molecules, explanatory variables for PC2 are more closely related to the inflammasome. When both age groups were individually tested, young (n = 37, 26.5 ± 3.2 years, 20–30) and older adults (n = 41, 68.1 ± 8.6 years, 52–85), we observed a larger separation of biomarkers between post-fasting data points and baseline data points in younger adult, while data points for older adults in the post-fasting period clustered more closely with baseline values. These observations prompt us to conclude that older participants were more resilient to changes induced by fasting, i.e. that biomarkers from older adults equilibrate more rapidly than younger individuals in the post-fasting period. We also identified that weight loss (in percentage) was less significant in older individuals (Fig. 4b, p < 0.0001)—older adults lost 4.75 ± 1.52% (2.57–10.44) as compared to 6.62 ± 2.07% (3.09–10.44) for younger participants. Nevertheless, we did not observe any differential regulation for glucose or PPARβ. We further stratified the older adults into middle-aged (n = 18, 52–64 years) and elderly (n = 23, 65–85 years) as the oldest display higher levels and spread of biomarkers related to cardio-metabolic risks (Fig. 4c). As observed from the dotted lines within graphs, which indicate the median, levels of CRP and systolic BP display a stepwise increase with age. In our analysis of IL-6, d-dimer and OPG levels (Fig. 4d), which correspond to inflammatory, cardiovascular and metabolic domains respectively, we observed that IL-6 levels were more significantly up-regulated in the older age groups as compared to the youngest group. For d-dimer levels, the middle age group revealed the least pronounced reduction during fasting. Although OPG levels appeared unaltered by fasting in our global collective analysis (Fig. 2c, significant regulation was observed in the oldest group. Plasma OPG levels increased most significantly (27.3 ± 15.0 vs. 44.7 ± 23.4 pg/ml, p = 0.005) but post-fasting OPG levels were similar to the baseline of the youngest group (p = 0.4922, median: 33.1 vs. 33.6 pg/ml, respectively), showing a beneficial long-lasting effect.

Finally, we further characterised two important markers associated with inflammation (CRP) and metabolism (glucose) (Fig. 5). We separately plotted donors with the lowest and highest baseline fasting glucose levels (n = 15 each) over the three time-points and found that fasting was effective in reducing glucose levels in both participants with high (by 18.6 ± 7.6%) and low (by 19.6 ± 0.9%) baseline glucose levels. In the fasting period, glucose levels were still significantly higher (p < 0.0001) in the baseline^HIGH group (122.0 mg/ml ± 3.4; 117.6–129.2) as compared to the baseline^LOW group (82.8 mg/ml ± 2.9; 77.7–86.9). In contrast, while plasma CRP levels were significantly reduced in both CRP^high and CRP^low (4.17 ± 1.0 and 2.03 ± 0.21), we did not observe a significant difference in absolute CRP levels during fasting (1.85 ± 0.27 and 1.87 ± 0.21). This suggested a differential regulation of CRP and glucose levels in fasting versus non-fasting periods. When directly compared for each fasting period (Fig. 5b) we identified that both molecules were strongly positively correlated only during RCF (r = 0.6109, p < 0.0001). As changes in nutrient availability and sensing may affect the inflammatory profile, we examined the relationship between CRP levels and nutritional markers (heat map of correlation values, Fig. 5c). Nutritional data that display significant correlation with CRP levels at baseline (p < 0.05) was followed at the different stages of the study. As proof of concept, the data demonstrate how CRP levels are positively correlated with age in the pre-fasting phase, but this relationship is lost during fasting. Likewise, this trend was observed in the associations between CRP levels and the consumption of energy, protein, fat and most vitamins and minerals. Nevertheless, the relationships between CRP levels and phytochemical index (PI), net endogenous acid production (NEAP), and DII levels appeared to be stable during fasting. Altogether, this suggests interactions between physiological systems to be regulated by RCF and nutritional intake to differentially disturb these.

Many physiological and psychological changes take place during fasting in response to changes in eating behaviour, frequency and sleeping pattern. Generally, the diet of participants in our study consisted of a combination of both healthy (high in fruits, vegetables, poultry, cruciferous or green leafy vegetables, tea, fruit juices, and whole grains), and processed (high in refined grains, red meat, butter, commercial meat products, high and low-fat dairy products, sweets, desserts, hydrogenated fats, soft drinks) food. In this cohort of fasting practitioners, we identified cardiometabolic benefits, differential regulation of inflammation, weight loss, changes in metabolic regulation (leptin, PPARβ) [32]. Moreover, in our characterisation of biological biomarkers (blood pressure, CRP, glucose levels), we observed that the elderly may receive more benefits from fasting than younger men, due either to the fact elderly are more susceptible to the extreme variations in nutriments intakes, to the longer experience and better management of fasting or simply to the fact they are the at-risk population [33]. Our study also describes the novel effect of fasting on biomarkers associated with physiological functions—such as orexin-A (wakefulness) [34], β-hydroxybutyrate (non-conventional metabolic pathways) [35] and PDGF-AA (aging/senescence) [36].

Specific to inflammation, the modulation of inflammatory markers (IL-6, TNFα, IL-8, CRP) is an important outcome of fasting as the current literature continues to support their association with morbidity and mortality. Pro-inflammatory molecules such as C-reactive protein (CRP), interleukin-6 (IL-6) and tumor-necrosis factor alpha (TNFα) are associated with major chronic conditions, including diabetes, cardiovascular diseases, and age-associated conditions such as frailty and dementia [37, 38]. Importantly, these molecules are differentially regulated during fasting, as CRP and TNFα were strongly downregulated but IL-6 and IL-8 reveal an inverse trend—dissociating the interactions between these signalling pathways could be important in discerning whether this is an effective homeostatic mechanism to maintain immune competency. Although the sustained or disproportionately high release of inflammatory molecules is often detrimental to cellular and tissue integrity [39], inflammation is a necessary tool that allows the immune system to facilitate the clearance of pathogen, restore tissue homeostasis and promote wound healing.

This study benefits from access to a state-funded population survey and a high compliance rate, as recruitment was specifically in selected regions where nutritional intake during fasting periods was predictable and least divergent from their non-fasting periods [23, 24]. The latter behaviour is linked to multiple socioeconomic factors, including rurality, food availability, income and religiosity. Accordingly, none of the participants gained weight (from compensatory eating behaviour) and weight loss remained within an acceptable range during fasting. The low-income status of most participants, however, may be associated with a mild medical history, such as a higher frequency of both diagnosed undiagnosed cases of pre-diabetes/diabetes and/or hypertension. We also report significant numbers of participants with high systolic and diastolic blood pressures but who do not receive medication. A clear benefit of our study is the inclusion of a wide age-range, which allows us to stratify participants into multiple age groups: young (representing healthy non-at-risk individuals), middle-aged, and elderly (representing at-risk and potentially frail) individuals. Since we included biomarkers that spanned a diverse range of physiological systems, we were able to identify domains where fasting effects differed due to age [40, 41].

There are several caveats in our study, for example, only males were included in this study and our findings may not be extrapolated to women. This mainly results from difficulties in recruiting females concomitantly with males in a single study. In light of the beneficial effects of RCF at the cardiometabolic level and the higher prevalence of associated diseases in women, we have planned a dedicated study designed to investigate the beneficial impact of fasting on cardiometabolism, inflammation and hormonal regulation. Self-reporting on dietary intake may result into over- or underestimation of dietary intake. Nevertheless, we have restricted our study to a relatively small but sufficiently powered group of individuals (n = 78) to assess weight loss, our primary outcome. The number of participants also enabled to sustainably ensure compliance and supervision to facilitate accurate reporting and questionnaire administration. A subsequent study in other regions of the work with higher sample size would be useful in validating our main findings. Some biomarkers were not measured or did not reach sufficient quality, especially lipids and IGF-1/insulin and must be investigated in subsequent studies. A separate limitation is the absence of a separate non-fasting, age-matched, control population within the same study period. Because of the design of the study, it would be unethical to interfere with the fasting practices to obtain a comparable control group. The non-fasting individuals are usually those with chronic conditions (diabetes), acute infections or young adults, pregnant women and the frail elderly. Finally, we did not record changes in physical activity levels during the fasting period, which may be important due to associations between physical activity and the levels of investigated biomarkers, such as CRP and IL-6 [42].

Only a few recurrent fasting protocols have been investigated and the largest and longest report includes the monitoring of 1422 subjects who fasted for 4–21 days in accordance to the Buchinger fasting method. This extreme fasting method contributed to the loss of > 10 kg in males who fasted for more than 20 days. However, among 1422 subjects, only 37 (2.6%) were able to achieve this 20 days target. Moreover, we note that most metabolic changes did not persist for more than 15 days (e.g. glucose, triglycerides), possibly due to the extent of nutritional depletion. The dietary changes observed in our setting are more similar to other reports on RCF [43‐45]. A decrease in daily calorie intake has been noted as one of the desired effects of RCF fasting [46]. Similar to what we observed, Salahuddi et al. [47] reported a significant drop in systolic and diastolic BP during fasting. However, conflicting reports suggested RCF fasting to have no effect on mean systolic and diastolic blood pressure in patients with hypertension, despite the reduced hydratation [48]. There is a lack of consensus on whether recurrent circadian fasting improves [47] or worsens lipid profiles [49] probably due to health heterogeneity in the studied populations [17]. The same applies for HDL [19, 50]. Adlouni et al. [51] conducted a study where a significant reduction in total cholesterol and triglyceride levels was seen during RCF fasting, but suggested that these effects may be specific to or could be explained by the Mediterranean dietary habits in Morocco.

Several studies on intermittent fasting have revealed a positive effect on inflammatory state and relieves the inflammatory burden in healthy individuals and in the contexts of pregnancy, cardiovascular diseases, asthma, brain disorders and as part of intense physical training [52]. However, only a few studies demonstrated reduced IL-6, and TNF-α levels during RCF (p < 0.05) as most studies only focused on CRP [53, 54]. The role of the inflammasome in the RCF-induced regulation of inflammation is still to be proven. While IL-1β is up-regulated following RCF this does not preclude that overall inflammation is reduced, due to the up-regulation of anti-inflammatory molecules such as IL-10, TGF-β, and receptors to IL-6, TNF-α, IL-18 among others. For instance, we show IL-6 up-regulation which aligns with IL-1β up-regulation and platelets elevation. Studies have shown the release of IL-6R by platelets [55] and their ability to mediate inflammatory signalling via IL-1β synthesis [56]. The emerging role of platelets in controlling inflammation [57] and related cardiometabolic diseases [58, 59] deserves more focus in the context of fasting [60]. Experimental studies by Sack et al. suggested that timing of measurement is extremely important to study inflammasome activation [61, 62]. It is likely that the timing for blood collection and the fasting/feeding period difference between studies accounts for discrepancies. Taken together, these results support the perception of reduced inflammation during RCF fasting. In addition to the modulation of these molecules, we also observed a significant downregulation in IP-10 and PDGF-AA after fasting; to the best of our knowledge, no previous study on fasting has evaluated the impact of fasting on these two chemokines. IP-10, initially identified as an IFN-γ-inducible chemokine and functionally categorized as a pro-inflammatory chemokine, is known to regulate immune responses by activating and recruiting leukocytes including T cells, eosinophils, monocytes and NK cells [63]. Moreover, IP-10 is secreted by a variety of cells including monocytes, neutrophils, endothelial cells, keratinocytes, mesenchymal cells, dendritic cells, astrocytes and hepatocytes. Other research has revealed that intrahepatic and circulating IP-10 is associated with obesity and insulin resistance in patients with chronic hepatitis C virus (HCV) infection and in patients with HCV/human immunodeficiency virus (HIV) co-infection [64]. A major issue in such biomarker analysis is the identification of the source of inflammation. In the context of fasting, we envision immune cells and organs to adapt differently at the inflammatory level. Including exosome-derived inflammatory measures would enable to identify the origin of the circulating biomarkers [65, 66] and better understand the impact of fasting on the several physiological systems.

Recurrent circadian fasting may alter the circadian consumption and distribution of energy and other nutrients for the fulfilment of daily requirements. For example, breakfast accounts for up to 30% of the daily calorie intake and the post-sunset meal constitutes 60% of the daily energy intake [13] but larger datasets should be gathered on the physiological consequences [14]. A recent meta-analysis suggested that breakfast may not be recommended for individuals whose intention is to lose weight [15]. On a related note, recent studies have demonstrated a negative impact of forgoing breakfast [16], and late night eating habits on cardiometabolic health [18]. Alternatively, the American Heart Association recommends that dietary habits which pay mindful attention to the factors of timing and frequency (which also implies quantity and quality) promote the better management of cardiometabolic risks. As RCF adheres to a strict schedule, the beneficial effects observed in our study supports the latter conclusion.

A main concern of RCF fasting studies relates to the radical dietary readjustments that occur during this period, particularly concerning carbohydrate and protein intake [44]. This may be due to higher consumption of dates, honey, sweets and soft drinks as often noted in middle-eastern countries. The rapid elevation of socio-economic status and mal-adaptation to western diets also contribute to a lower compliance to nutritional intake recommendations during fasting. Differences in physical activity levels may also account for some of the discrepancies in observations between studies. While some societies reveal diminished physical activity levels during RCF, some groups of individuals, especially the elderly, participate in longer periods of moderate physical activity due to their active participation in mindfulness-related activities that centre on prayer [67]. Related to the latter, sleeping patterns may also be interrupted by fasting, as suggested by the heterogeneity of orexin-A levels in our study. Nevertheless, RCF compliance with recommended lifestyle (maintaining daily activities and unchanged/reduced carbohydrate and protein intake) in our study, may lead to plausible health benefits that relate to weight loss and inter-systems regulation. The extent of RCF-induced benefits may differ in other studies due to factors associated with geographical and cultural predisposition (Middle-East, Eurasian, South-East Asia, immigrants in western countries). It has been proposed that health outcomes resulting from RCF are affected by various factors, including variations in the length of fasting (e.g. 16–20 h in the same year), daytime temperature [31, 68], dietary habits/customs during RCF, fluid consumption (e.g. sweetened drinks or water/juices), sleep duration and pattern [69, 70], working hours, job conditions, physical activity levels [71] and ritualistic as well as social behaviours [72].

A possible mechanism for the lowering of inflammation and oxidative stress markers with RCF may relate to reduced activation of the insulin/insulin-like growth factor 1 (IGF-1) axis, which predisposes individuals—particularly the obese—to inflammatory and oxidative stress [73]. Behaviours associated with obesity may also sustain chronic diseases through the induction of pro-inflammatory cytokines and insulin and insulin/IGF-1 resistance—both processes further contributes to the increased production of reactive oxygen species and residual inflammation. Although we could not measure insulin resistance and IGF-1 levels in our study, other studies have associated RCF with improved insulin levels and sensitivity in obese individuals with metabolic syndrome [74]; the sustained reduction in glucose levels observed in our study may act as supporting mechanism for the latter findings.

Wegman and colleagues [75] reported an increase in the expression of SIRT3, which encodes for NAD-dependent deacetylase Sirtuin-3 (reduced during aging) during fasting, which aligns with our observed down-regulation of PDGF-AA (associated with SASP). The significant reduction of total body and visceral fat reported at the end of RCF in our study and others [76], may further lead to adipocyte dormancy and reduced secretion of adipokines [77]. During RCF, drastic changes in the timing and frequency of fluid and food intake were also observed in our study (Fig. 1a)—this contributes to chronobiological changes in the circadian distribution of body temperature, melatonin, cortisol, blood glucose levels, as well as day- and night-time wakefulness [68]. These physiological changes have been reported to benefit several systems and reduce overall physiological stress [78].