Previously we discovered that routine periodic fasting was associated with a lower prevalence of coronary artery disease (CAD). Other studies have shown that fasting increases longevity in animals. A hypothesis-generating analysis suggested that fasting may also associate with diabetes. This study prospectively tested whether routine periodic fasting is associated with diabetes mellitus (DM). Patients (n = 200) undergoing coronary angiography were surveyed for routine fasting behavior before their procedure. DM diagnosis was based on physician reports of current and historical clinical and medication data. Secondary end points included CAD (physician reported for ≥1 lesion of ≥70% stenosis), glucose, and body mass index (BMI). Meta-analyses were performed by evaluation of these patients and 448 patients from a previous study. DM was present in 10.3% of patients who fasted routinely and 22.0% of those who do not fast (odds ratio [OR] 0.41, 95% confidence interval [CI] 0.17 to 0.99, p = 0.042). CAD was found in 63.2% of fasting and 75.0% of nonfasting patients (OR 0.42, CI 0.21 to 0.84, p = 0.014), and in nondiabetics this CAD association was similar (OR 0.38, CI 0.16 to 0.89, p = 0.025). Meta-analysis showed modest differences for fasters versus nonfasters in glucose concentrations (108 ± 36 vs 115 ± 46 mg/dl, p = 0.047) and BMI (27.9 ± 5.3 vs 29.0 ± 5.8 kg/m 2 , p = 0.044). In conclusion, prospective hypothesis testing showed that routine periodic fasting was associated with a lower prevalence of DM in patients undergoing coronary angiography. A reported fasting association with a lower CAD risk was also validated and fasting associations with lower glucose and BMI were found.
This study of a newly ascertained patient population was undertaken to determine whether the hypothesis-generating diabetes mellitus (DM) findings of our original fasting study were a chance occurrence or whether fasting has a consistent association with a lower DM risk. The secondary objective of the study was to validate the association of fasting with coronary artery disease (CAD).
Methods
Patients enrolled in the Intermountain Heart Collaborative Study (IHCS) during 2007 to 2008 were approached by a study coordinator to obtain informed consent before patients’ coronary angiographic procedure. This study was approved by the Intermountain Healthcare institutional review board and all participants provided written informed consent ( http://clinicaltrials.gov , identifier NCT00406185 , database registry of the IHCS). After providing consent, patients were surveyed for fasting behavior and other lifestyle factors using a validated, previously described survey. Fasting was defined in the survey as routine abstention from food and drink for extended periods. This definition was based on validated survey language that was crafted in general terms but was shown before implementation to primarily represent 1 fasting definition: 1 time/month for 24 hours. This study was possible because 67% of the local cardiac patient population reports a Latter-Day Saint (LDS; or Mormon) religious preference and a large proportion of LDSs routinely fast (1 time/month for 24 hours). Previously, it was shown that 92% of local cardiac patients who fast routinely reported an LDS religious preference.
Other independent study variables included age, gender, and body mass index (BMI; kilograms per meter squared). Physician-reported hyperlipidemia was defined as total cholesterol ≥200 mg/dl or low-density lipoprotein cholesterol ≥130 mg/dl or use of lipid-lowering medication. Hypertension was defined as systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or use of antihypertensives. Smoking was defined as active smoking or a >10 pack-year history. Family history of early CAD was present when a first-order relative had developed cardiac death, myocardial infarction, or coronary revascularization at <65 years of age for women and <55 years of age for men.
Physical activity history was self-reported by patients through 3 survey questions that inquired about the amount of time each day and number of days per week that patients engaged in moderate activities and in vigorous activities and the number of days per week engaged in strength training. Socioeconomic measurements of income, educational attainment, and ancestry (race/ethnicity) were also obtained by survey. These physical activity and socioeconomic data were available in only 94% of patients (n = 187) because of self-selection of patients in completing these survey questions. Education was categorized into 6 groups based on level of attainment: some high school education or less, graduation from high school, some college or vocational school, college graduation or some graduate school, a master’s degree, or a doctoral degree. Income was categorized into 3 groups: low (<$30,000), intermediate ($30,000 to $69,999), and high (≥$70,000).
DM diagnosis was based on a physician report of current and historical, clinical, and medication data. For clinical laboratory findings, a 12-hour fasting blood glucose level ≥126 mg/dl was considered DM. Prescription of an anti-DM medication at the time of enrollment identified patients with a previous DM diagnosis.
Serum glucose levels were queried from electronic clinical records for the present study’s population and for the 448 patients from the previous fasting study. Glucose concentrations were measured after an 8-hour fast for clinical purposes during the index hospitalization.
For CAD, lesion information was provided by each patient’s attending cardiologist. The presence of ≥1 flow-limiting lesion of ≥70% stenosis was defined as a CAD. No-CAD controls were patients free from any lesion or with luminal irregularities of <10% stenosis. For clarity of diagnosis, moderate CAD (most severe lesion 10% to 69% stenosis) was excluded as indeterminate.
For baseline characteristics, differences between patients who reported fasting behavior and those who reported that they did not fast routinely were compared by chi-square test for discrete variables and Student’s t test for continuous variables. Associations of fasting with DM and with CAD were evaluated by chi-square test. Logistic regression was used to adjust the associations for covariables including age and gender for the DM end point and in the 94% of patients (n = 187) who had the data, statistical modeling added physical activity, education, income, and ancestry variables.
For the CAD end point, adjustments in regression were made for age, gender, diabetes, BMI, smoking, hyperlipidemia, hypertension, and family history of early CAD. Further modeling added physical activity, education, income, and ancestry variables, although these models included only 187 patients (94%) because of missing data. The association of fasting with CAD was also tested in the nondiabetic substratum to further evaluate whether the CAD association depended on the DM association.
Meta-analysis was used to evaluate the combined populations from the present and the original fasting studies. The combined sample size from the 2 studies was 648 patients (200 from the present study, 448 from the previous study). The Mantel–Haenszel method was used for the initial univariable meta-analysis of the fasting associations with DM and with CAD, and logistic regressions that included a study-identifying covariable were used for multivariable meta-analyses. Because of their connection to metabolic disease, BMI and glucose level were also evaluated by meta-analysis for an association with fasting. Meta-analysis used analysis of variance to test differences in BMI and glucose by fasting status and entered the variable identifying whether the patient was included from the present or previous study.
Two-sided statistical tests required a p value ≤0.05 for the DM end point, which was defined as the primary test of hypothesis. For the secondary CAD end point, statistical significance was corrected by the Bonferroni test and required a p value ≤0.025. Analyses were performed using SPSS 15.0 (SPSS, Inc., Chicago, Illinois).
Results
In patients newly surveyed for routine, periodic fasting behavior (n = 200), ages ranged from 26 to 90 years and 34.0% reported routine fasting. Other population characteristics are listed in Table 1 . These patients were mutually exclusive of the 448 who were studied in a previous report.
| Characteristic | Overall | Routine Faster | Nonfaster | p Value |
|---|---|---|---|---|
| (n = 200) | (n = 68) | (n = 132) | ||
| Age (years) ⁎ | 66.1 ± 10.6 | 69.0 ± 11.0 | 64.6 ± 10.2 | 0.006 |
| Women | 35.5% | 32.4% | 37.1% | 0.50 |
| Body mass index (kg/m 2 ) ⁎ | 28.8 ± 6.0 | 27.4 ± 5.5 | 29.4 ± 6.1 | 0.038 |
| Race/ethnicity | ||||
| African | 1% | 1% | 1% | 0.84 |
| Asian | 1% | 0% | 1% | |
| Caucasian | 85% | 87% | 83% | |
| Hispanic | 1% | 2% | 1% | |
| Native-American | 2% | 0% | 2% | |
| Other/multiple races | 4% | 4% | 4% | |
| Unknown ancestry | 8% | 6% | 9% | |
| Hypertension † | 63.0% | 58.8% | 65.2% | 0.38 |
| Hyperlipidemia ‡ | 58.5% | 51.5% | 62.1% | 0.15 |
| Smoker | 9.0% | 2.9% | 12.1% | 0.032 |
| Physical activity (n = 187) | ||||
| Moderate activities (hours/week) | ||||
| 0 | 14% | 8% | 18% | 0.016 |
| 1–2 | 14% | 6% | 18% | |
| 3–16 | 52% | 63% | 47% | |
| ≥17 | 19% | 23% | 17% | |
| Vigorous activities (hours/week) | ||||
| 0 | 56% | 46% | 62% | 0.24 |
| 1–2 | 9% | 12% | 7% | |
| 3–16 | 26% | 31% | 23% | |
| ≥17 | 9% | 11% | 8% | |
| Strength training (days/week) | ||||
| 0 | 74% | 74% | 75% | 0.99 |
| 1–3 | 12% | 12% | 11% | |
| 4–7 | 14% | 14% | 14% | |
| Income (n = 187) (dollars/year) | ||||
| <$30,000 | 25% | 26% | 25% | 0.46 |
| $30,000–$69,999 | 31% | 37% | 28% | |
| ≥$70,000 | 15% | 11% | 17% | |
| Education (n = 187) | ||||
| <12 years | 9% | 12% | 7% | 0.016 |
| High school graduate | 24% | 11% | 30% | |
| Some college | 33% | 32% | 34% | |
| Bachelor’s degree | 20% | 22% | 19% | |
| Master’s degree | 9% | 12% | 7% | |
| Doctoral degree | 6% | 11% | 3% |
⁎ Values are presented as mean ± SD.
† Hypertension was defined as systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or use of antihypertensive medication.
‡ Hyperlipidemia was defined as total cholesterol ≥200 mg/dl or low-density lipoprotein cholesterol ≥130 mg/dl, or use of lipid-lowering medication.
A DM diagnosis was carried by 10.3% of patients who routinely fasted compared to 22.0% of nonfasters (odds ratio [OR] 0.41, 95% confidence interval [CI] 0.17 to 0.99, p = 0.042). This remained significant with adjustment for age and gender (p = 0.044; Figure 1 ). Adjustment for physical activity and socioeconomic measurements did not confound the association of fasting with DM, with changes in the regression coefficient of fasting by <10%. Meta-analysis of the 200 patients with the previously studied 448 confirmed a fasting association with DM (OR 0.56, CI 0.36 to 0.88, p = 0.012, adjustment for age and gender p = 0.014; Figure 1 ).
