Metabolomics of Ramadan fasting and associated risk of chronic diseases

Background The dramatic change in lifestyle associated with Ramadan fasting raises questions about its effect on metabolism and health. Metabolites, as the end product of metabolism, are excellent candidates to be studied in this regard. Objective This study aims to investigate the effect of Ramadan fasting on the metabolic profile and risk of chronic diseases. Methods The London Ramadan study (LORANS) is an observational study in which 2 blood samples were collected from 72 participants a few days before and after the fasting month of Ramadan. We conducted metabolomic profiling using nuclear magnetic resonance spectroscopy to assess the change in individual metabolites from before to after Ramadan. Also, we generated metabolic scores (scaled from 0 to 100) for 7 chronic diseases in the UK Biobank and assessed the association of Ramadan fasting with these scores in LORANS. Results Of the 72 participants, 35 were male (48.6%); the mean (± standard deviation) age was 45.7 (±16) y. Ramadan fasting was associated with changes in 14 metabolites (1 inflammation marker, 1 amino acid, 2 glycolysis-related metabolites, 2 ketone bodies, 2 triglyceride, and 6 lipoprotein subclasses), independent of changes in body composition. Using data from 117,981 participants in the UK Biobank, we generated metabolic scores for diabetes, hypertension, coronary artery disease, renal failure, colorectal cancer, breast cancer, and lung cancer. The metabolic scores for lung cancer, colorectal cancer, and breast cancer were lower after Ramadan in LORANS (−4.74, 9.6%, 95% confidence interval −6.56, −2.91, P < 0.001), (−1.09, −2.4%, −1.69, −0. 50, P < 0.001), and (−0.48, −1.1%, −0. 81, −0.15, P = 0.006), respectively. Conclusions Ramadan fasting is associated with short-term favorable changes in the metabolic profile concerning risk of some chronic diseases. These findings should be further investigated in future, larger studies of longer follow-up with clinical outcomes.


Introduction
Fasting is practiced by a large number of people for a variety of reasons, including weight loss and improving general health.Time-restricted fasting in particular has become popular, but comprehensive studies on the effects of such diets are lacking.Ramadan (religious) fasting is similar to time-restricted fasting, practiced by hundreds of millions of Muslims around the world [1], providing a "natural experiment" to study the potential metabolic effects of time-restricted fasting.
Metabolic changes that follow fasting in the first couple of hours include glycogenolysis (consuming the glycogen stores), followed by an adaptive phase after a more extended period of fasting (>8 h), where the insufficiency/absence of the primary fuel (glucose) is compensated for by converting glucogenic amino acids into glucose (gluconeogenesis) and converting ketogenic amino acids and fatty acids into ketones (ketogenesis) [2,3].However, the longer-term metabolic effects of fasting and its subsequent effects on health remain to be elucidated.
The metabolic effects of fasting on different systems can be studied using various methods within the global -omics umbrella.Metabolomics is an approach where a wide range of small molecules in biological specimens are analyzed [4].A previous study found that prolonged fasting significantly altered 166 metabolites [5], with the metabolic profile being associated with some chronic diseases [6,7].
We hypothesize that Ramadan fasting is modulating the metabolic profile that may have an effect on risk of chronic disease in the long term.
So far, only 2 small studies were conducted to explore the metabolic effects of Ramadan fasting.The first is a pilot study of Ramadan fasting using metabolomics involving measurements of 186 metabolites collected on the 7th and 26th day of Ramadan in 11 healthy males, with significant changes observed for 20 metabolites [8].The second study used mass spectrometry to investigate the metabolomics changes in 25 individuals with overweight and obesity after Ramadan fasting [9].
Here, we report the metabolomics changes after Ramadan fasting using data and samples from the London Ramadan study (LORANS).Moreover, to speculate on the clinical impact of fasting, we built metabolite risk scores for several chronic diseases using the UK Biobank data and evaluated whether the metabolite risk score of any of these diseases was significantly modified after 1 mo of Ramadan fasting.

London Ramadan study
LORANS is an observational study conducted in 2019 in London, United Kingdom.Those aged 18 y and above who were planning to fast for >20 d of Ramadan were eligible for the study, with data collection carried out in clinics set up in 5 mosques in London.The study included 2 visits: one before the fasting month of Ramadan and the second within 8-12 d after the end of Ramadan.We excluded pregnant females and those who were unwilling to attend the second visit.Before Ramadan, we collected blood samples from 140 individuals of whom 78 (55.7%) participants attended the second visit and gave blood samples after Ramadan.We collected 2 EDTA blood samples (9 mL) from each participant per visit, which were transported in a thermoporter to a bio-repository at Charing Cross Hospital for processing and long-term storage at À80 C. We excluded 6 individuals whom the first visit sample was not processed because of a delay in transporting the samples for processing, leaving 72 participants with blood samples before and after Ramadan (Supplemental Figure 1).Supplemental Table 1 compares the study sample's demographics (n ¼ 72) to those of people not attending the second visit (n ¼ 68); the 2 groups were similar except for age.In addition to blood samples, we collected socioeconomic and lifestyle data using online questionnaires and measured blood pressure using an automatic blood pressure monitor and body composition using a bioelectrical impedance analyzer at each visit, more details reported elsewhere [10,11].
All participants provided written informed consent before data collection.The study was approved by the Ethics Committee at Imperial College London (reference: 19IC5138, dated April 17, 2019).

UK Biobank
The UK Biobank is a large-scale prospective study with over 500,000 participants from across the United Kingdom.The study includes hospital inpatient data on diagnosis in the form of codes (ICD-9 and ICD-10) [13] with NMR metabolomic biomarkers data (Nightingale Health Ltd.) available on 117,981 UK Biobank participants [14].

Statistical analysis
Change of metabolite levels in LORANS.Metabolite levels were first normalized between 0 and 1 because normalizing data of different ranges is essential to compare results; this process transformed metabolites' values to be between 0 and 1, reserving their unique variances.Next, we used a linear mixed-effects model for each metabolite to investigate the difference in its levels before and after Ramadan fasting.This model had age and sex as fixed effects and participant IDs, mosques, and day of the second measurement as random effects.To investigate whether the observed changes in metabolites were a result of changes in body composition, we constructed 2 additional models, further adjusting for: 1) changes in waist circumference, free-fat mass, and BMI; 2) changes in total body water and fat percentage.We applied Benjamini-Hochberg for false discovery rate (FDR) with a q value of 0.05 to correct for multiple testing [15].
The metabolic risk score for complex diseases.First, using data from the Nightingale platform in the UK Biobank, we used metabolite levels to construct metabolic risk scores for cardiometabolic disorders and incidence of a number of common cancers, namely coronary artery disease, diabetes, hypertension, renal failure, colorectal cancer, breast cancer, and lung cancer.These 7 diseases were selected because of their commonness and being the most likely to be affected by fasting.To this end, ICD-9 and ICD-10 codes from the hospital inpatient data available in the UK Biobank were used to define diseases within the subset of participants who have metabolomics data (n ¼ 117,981) (Supplemental Table 2).For each disease of interest, the following steps were applied.Prevalent cases were excluded at baseline (2006).We examined the individual association of each metabolite (adjusted for age and sex) using a Cox proportional hazards model, adjusting P values for multiple testing by applying the FDR method.Metabolites associated with incident disease outcome at P < 0.05 were then included into a multivariate Cox proportional hazards model.We used the β coefficients from the models (1/metabolite) to produce a metabolic score per outcome per participant in LORANS.In cases of a high level of multicollinearity between variables in a model, we first applied a Cox Lasso regression to select variables that were then used in a multivariate Cox regression.Multicollinearity was assumed whenever the condition number (ratio of the largest singular value over the smallest singular value which is produced by the decomposition of the sample correlation matrix) was !30 [16].Finally, from the multivariate Cox regression model output, we used the coefficients of the metabolites with P < 0.05 to generate a metabolic score for each participant in LORANS (Supplemental Figure 2).The following example clarifies this process further: to calculate the metabolic score of colorectal cancer for a participant in LORANS, we multiplied the regression coefficient (from the colorectal cancer multivariate Cox regression model) of glycoprotein acetyls (1 of the 2 metabolites contributing to this metabolic score) in the glycoprotein acetyls value assessed before Ramadan and added the result to that of free cholesterol in intermediate-density lipoprotein (IDL).This sum is considered the colorectal cancer metabolic score for that participant before Ramadan.Then, similarly, we calculate the score after Ramadan, from which the score before Ramadan was substracted to explore the difference (if any).The metabolic scores were rescaled so each participant would have a value between 0 and 100 using the "scales" package in R. The change in the estimated metabolic scores after Ramadan was examined using a mixed-effects model.
For lung cancer, we built a second metabolic score after further adjustment for smoking intensity in the initial univariate Cox regression model to adjust for the effect of smoking on metabolites.
Also, we built a metabolic score for smoking to investigate whether exposure to smoking changed during Ramadan.This was done by applying the same steps outlined above using a linear regression model.
We also assessed the correlation between metabolic scores and incidence rates of the 7 chronic diseases among the UK Biobank participants using a Cox proportional hazards regression model adjusted for age and sex.Moreover, we divided the metabolic scores into 4 groups (very low, low, high, and very high) and compared the frequency of incident cases across the 4 levels of the metabolic score.We used "base," "survival," and "glmnet" packages in R (version 1.1-21) to perform statistical analyses.Also, we used complete-case analysis.
Selecting the metabolites associated with lung cancer in the UK Biobank adjusted for smoking intensity showed that the effect on lung cancer score was independent of smoking (À5.69%;À2.95; and À4.74, À1.18) (Supplemental Table 7).Also, the smoking metabolic score we created did not change significantly, meaning there was no change in smoking during Ramadan (Supplemental Table 8).

Discussion
In this study, we investigated the metabolic changes after fasting in the month of Ramadan.We found changes in serum levels of 14 metabolites.Using 7 metabolic risk scores that were built using data from the UK Biobank, we showed that the metabolic risk scores were lower for lung cancer and, to a lesser extent, for colorectal cancer and breast cancer after 1 mo of fasting in Ramadan.However, this study does not provide evidence for an association between Ramadan  fasting and a lower risk of these diseases.The findings merely showed changes in the metabolic profile, similar to that of individuals at lower risk of the diseases.Mathew et al. [8] measured 202 metabolite levels by applying extended targeted metabolomics assays to dried blood spots for 11 men during Ramadan.Twenty metabolites, mostly phosphatidylcholines, were significantly changed from the first to the fourth week of Ramadan.However, phosphatidylcholines were measured in LORANS as one broad class and did not change after Ramadan fasting.Only 2 (lactate and tyrosine) of the 14 metabolites that changed significantly in LORANS were measured in Mathew et al.'s [8] study, and none was reported to be changed.Notably, the 2 studies (LORANS and Mathew) collected the first and second blood samples at different times and used different metabolomics profiling methods, which might have contributed to the differences observed in the metabolite changes.
A systematic review and meta-analysis of data on healthy individuals showed that Ramadan fasting is associated with reduced HDL but no changes in TGs, total cholesterol, or LDL [17].In contrast, we do not report reductions in HDL, which may partly be explained by LORANS participants being recruited from a community-based sample including healthy and unhealthy individuals, whereas the review only involved studies on apparently healthy individuals.

Implication on public health and potential consequences
Ketogenesis has been mostly characterized in the context of ketogenic diets, showing positive outcomes for the microbiome, lowering hemoglobin A1c (HbA1c) and the need to reduce insulin in diabetics, helping weight loss, reducing visceral adiposity, positively regulating lipoproteins, lowering TGs, and reducing inflammation [18,19].
In addition, the observed metabolic changes have potential consequences.The increase in acetone and decrease of acetate, together with decreases in both lactate and pyruvate as well as decreasing lipoprotein TGs point to an energy pathway shift in the liver after Ramadan fasting, from glucose to fatty acid metabolism.This can be seen at 3 levels: 1) In ketogenesis, the limited availability of oxaloacetate (because of its consumption in parallel with gluconeogenesis) leads to the production of ketone bodies from acetyl coenzyme A resulting from the breakdown of fatty acids.2) Pyruvate is the product of glycolysis and a precursor to the Krebs cycle for energy production, whereas lactate is its by-product.The observed reduction in both pyruvate and lactate can thus be explained by the shift from glycolysis to ketogenesis.3) There is a reduction of fatty acid-rich TGs in multiple lipoprotein fractions, which reflects lipolysis and breakdown of fatty acids during ketogenesis.

Fasting and cancer
In this study, we found that a metabolic risk score associated with the risk of lung cancer and colorectal cancer is reduced after Ramadan fasting.A number of experimental studies on animal models have provided evidence for the preventive effect of fasting on developing cancer [20][21][22].
Previous studies have reported associations between some of the metabolites that we have used in the metabolic risk score and lung cancer, such as alanine and β-hydroxybutyrate in Zhang et al.'s [23] study and citrate, alanine, and lactate in Rocha et al.'s [24] study.

Glycoprotein acetyls
Glycoprotein acetyls was 1 of 9 metabolite measures contributing to the metabolic score of lung cancer and 1 of 2 variables that constituted the colorectal cancer metabolic score.Glycoprotein acetyls signal is mainly influenced by concentrations of 5 circulating acute-phase inflammatory glycoproteins [25].Previous studies have identified glycoprotein acetyls as a predictive marker for diabetes [26], cardiovascular diseases [27], all-cause mortality [28], total cancers [29], colorectal cancer [30], and cancer deaths [28].In our study, glycoprotein   acetyls contributed to the metabolic scores of diabetes, hypertension, coronary artery disease, renal failure, colorectal cancer, and lung cancer.
In addition, glycoprotein acetyls appeared to be the main driver of reductions in colorectal cancer and lung cancer metabolic scores after Ramadan fasting (Supplemental Table 6).

Impact of smoking on metabolic profiles
Because smoking is a major risk factor for lung cancer and is known to affect the metabolic profile [31], we conducted a number of sensitivity analyses to minimize any bias because of confounding by smoking.Unfortunately, the small number of smokers in LORANS did not allow adjusting for smoking.The small number of smokers might be because of misclassification in the collected data on smoking, because it is prohibited in Islam and is not socially accepted.Also, some Muslims find Ramadan an excellent opportunity to quit smoking; thus, some former smokers may have stopped smoking a few days before taking part in the study.However, the effect of fasting on lung cancer metabolic risk remained significant after constructing another metabolic score adjusted for smoking in the UK Biobank, and the metabolic score for smoking did not change after Ramadan fasting (Supplemental Tables 7  and 8), suggesting that smoking is not likely to be an important confounder in the fasting/metabolite relationships.

Strength and weaknesses
This study has several strengths.First, LORANS has recruited a sample from the general population with different ethnic backgrounds.Studies that have been done on each of these ethnic groups in their home countries recruited participants from different socioeconomic statuses that might affect the lifestyle changes during Ramadan and cause discrepancies across studies.In LORANS, we have a sample with similar socioeconomic status because the participants were recruited from similar neighborhoods of the same city.This makes the results of the study more generalizable compared with previous studies.Second, we used the Nightingale NMR platform that covers a range of metabolites [8].We also used the Nightingale platform to estimate metabolic risk scores in the UK Biobank study with ~120,000 participants, enabling us to create robust metabolic scores and extrapolate the metabolic changes to the risk of a range of clinical outcomes.
This study also has some limitations.First, blood samples from 68 individuals were not available for follow-up.Another limitation is that we did not apply cross-validation to the Cox Lasso regression model.
In conclusion, Ramadan fasting is associated with short-term favorable changes in the metabolic profile concerning the risk of some chronic diseases.These findings should be further investigated in future, larger studies of longer follow-up with clinical outcomes.

FIGURE 2 .
FIGURE 2. Correlation between metabolic scores and incidence rates of chronic diseases in the UK Biobank.

TABLE 4
Number of metabolites correlated with/contributing to the metabolic scores