Association of periodic fasting lifestyles with survival and incident major adverse cardiovascular events in patients undergoing cardiac catheterization

Abstract

Aims

Animal models repeatedly show fasting increases longevity. Human data, though, are limited to anecdotal claims. This study evaluated the association of routine fasting with survival and, secondarily, with incident major adverse cardiovascular events.

Methods and results

Cardiac catheterization patients enrolled in the Intermountain INSPIRE longitudinal cohort (n = 2785) during 2013–2015 were followed through March 2019. A fasting survey was completed in n = 2025 (73%) of this cohort and 1957 were included in the final data analysis after 68 participants were removed (24 for data issues and 44 for fasting less than 5 years). Self-reported routine fasting behaviour, years of participation in fasting, and other fasting characteristics were surveyed. Mortality was the primary outcome and incident myocardial infarction (MI), stroke, and heart failure (HF) were secondary. Routine fasters (n = 389, mean age 64 ± 14 years, 34% female) averaged 42 ± 18 years of routine fasting (minimum 5 years). Non-fasters (n = 1568, aged 63 ± 14 years, 36% female) included never fasters (n = 1120 with 0 years of fasting) and previous fasters (n = 448 who averaged 32 ± 21 years of prior fasting but had stopped prior to enrolment). Routine fasters had greater survival vs. non-fasters [adjusted hazard ratio (HR) = 0.54, 95% confidence interval (CI) = 0.36–0.80; P = 0.002] and lower incidence of HF (adjusted HR = 0.31, CI = 0.12–0.78; P = 0.013), but not MI or stroke after adjustment.

Conclusions

Routine fasting followed during two-thirds of the lifespan was associated with higher survival after cardiac catheterization. This may in part be explained by an association of routine fasting with a lower incidence of HF.

Clinical study registration

The Intermountain INSPIRE registry https://clinicaltrials.gov/, NCT02450006.

Periodic fasting, Routine fasting, Therapeutic fasting, Life span, Longevity

Introduction

Remarkable advances have enhanced the treatment and prevention of cardiovascular events and metabolic diseases, with an 18.6% decrease in cardiovascular death from 2006 to 2016.¹ Unfortunately, heart disease continues to be the leading killer of adults² and heart failure (HF) prevalence is increasing.³ Other preventive approaches are needed to decrease mortality and HF risk.⁴ Unhealthy diet is one key lifestyle factor that is prevalent,⁵ but is also a modifiable risk contributor that, when replaced with healthy dietary practices, is connected with improved cardiovascular health and lower risk of mortality.^6,7

Decades ago, studies in California of observant members of the Church of Jesus Christ of Latter-day Saints revealed a substantially longer life expectancy,⁸ and lower standardized mortality ratios.⁹ For many decades, Utah—where approximately two-thirds of the population are Latter-day Saints—has had among the lowest coronary heart disease mortality rates in the USA.¹ Fasting for religious purposes is a common practice among these populations and is associated with lower clinical risks, regardless of smoking.^10,11 Using fasting as a means of promoting health has been advocated throughout history.¹² While the purpose of religious fasting differs from rapid weight loss fasting diets (whose goals are both cosmetic and health-associated),^13–15 one should not confuse the purpose with the act of fasting. The benefits of fasting are evoked during the fast and, while some physiological carryover into the refeeding period occurs,¹⁶ a higher fasting frequency will simply accumulate those benefits more rapidly.

Scientific study of fasting and longevity began among rodents in 1946.¹⁷ Fasting has repeatedly resulted in increased animal survival.¹⁸ While no direct evidence exists regarding fasting and survival, humans may benefit from fasting through improved body weight,^13–15glycaemic control,¹⁹ and favourable effects on genes related to autophagy.¹⁹ Fasting was used before 1975 for weight loss,²⁰ and fasting twice or more per week involves substantial drop-out rates with weight lost similar but not superior to caloric restriction.^13–15 Fasting, like caloric restriction,²¹ produces benefits regardless of weight loss.²² Similarly, a review of the literature on novel low-risk lifestyle factors suggests that their, ‘benefit occurs even when the impact on conventional risk markers is discouragingly minimal or not present’.²³ Fasting associations with clinical endpoints are reported, including coronary artery disease (CAD)¹⁰ and diabetes mellitus.¹¹

Despite this, the understanding of how fasting affects humans remains limited.²⁴ The primary purpose of this study was to assess the association of long-term routine fasting with survival among patients referred for CAD evaluation. Secondary objectives evaluated fasting associations with incident major adverse cardiovascular events (MACE) and prevalent cross-sectional endpoints.

Methods

The primary objective was to evaluate the association of fasting with survival after cardiac catheterization, with secondary objectives to evaluate fasting associations with: (i) the incidence of MACE, including myocardial infarction (MI), HF, and stroke in patients free of these events at baseline and (ii) the cross-sectional prevalence of CAD, diabetes, and other baseline outcomes. This was an observational, non-randomized longitudinal cohort study of patients enrolled in a long-term project (the INSPIRE registry) collecting biological samples, clinical data, epidemiologic data, and longitudinal outcomes among patients presenting to Intermountain Healthcare. Patients included adult men and women of any age, race, or ethnicity who were referred by their physician to cardiac catheterization based on symptoms and clinical judgement. Coronary artery disease presence was determined from a review of angiograms by attending cardiologists. This study was approved by the Intermountain Healthcare institutional review board and all patients provided consent (https://clinicaltrials.gov/, identifier NCT02450006).

Adult patients (N = 1957) enrolled in the INSPIRE registry from February 2013 to March 2015 completed an in-hospital survey (see the Supplementary material online, Appendix) in an acute care environment, reporting their fasting behaviours and information regarding income, education, physical activity, religion, race, and ethnicity. Surveys were administered prior to cardiac catheterization by a trained researcher. The survey utilized validated questions from prior Intermountain studies,^10,11 national surveys (e.g. the National Health and Nutrition Examination Survey), US Federal standards (e.g. standardized race and ethnicity questions), and the British National Health Service General Practice Physical Activity Questionnaire. Survey questions 2–8 were newly developed for this study. A fasting definition was not imposed on patients, but the questions characterized fasting behaviour (Supplementary material online, Appendix).

Fasting behaviour categories consisted of those who were ‘non-fasters’ (either never fasters or previous fasters who had historically fasted but stopped doing so some time prior to enrolment) and ‘routine fasters’ who had routinely fasted for 5 years or more (primarily for religious purposes). Other data elements were collected from electronic health records (Supplementary material online, Methods).

Fasting variables included whether patients routinely abstained from food and drink for extended periods of time, whether the patients drank water during fasts, the maximum consecutive length of time they abstained from food, their primary reason for fasting, how often they fasted, and how many years patients had fasted once per month, how many years at least once per year, and how many times in the last year the patient had fasted for 20 h or more (see Supplementary material online, Appendix).

Outcome variables included incident MI, incident HF, incident stroke, and the primary outcome of mortality. Mortality data were collected from the Social Security death master file, Utah death certificates, and electronic hospital records, providing thorough follow-up for mortality. Myocardial infarction, HF, and stroke events were queried from hospital and clinic encounters in the Intermountain electronic data warehouse that contains all electronic health record data from the 22 hospitals and >180 clinics in the integrated health system. Since Intermountain serves the healthcare needs of about two-thirds of people in Utah and southern Idaho, >90% of cardiovascular events and diagnoses are identified by this electronic method.²⁵ While this is not 100% complete, the small proportion of patients who may change to a different health system for subsequent care are unlikely to systematically do so based on fasting status.

Statistical considerations

Comparison of fasting status with demographics, clinical factors, medications, survey variables, and other study covariables was performed using the Student’s t-test or the χ² test, as appropriate. The primary study comparison was between routine fasters (i.e. fasting for ≥5 years) with non-fasters.

Survival analyses comparing the association of fasting status with mortality and the other longitudinal events were performed using Cox proportional hazards regression. They were also visually evaluated using Kaplan–Meier methods. The hazard ratio (HR) and 95% confidence intervals (CIs) for univariable and multivariable Cox regression analysis were used to assess differences in mortality and other outcomes between routine fasters and non-fasters. Multivariable Cox regression evaluated 61 covariables that included age, sex, education, physical activity, smoking, alcohol consumption, religious preference, depression history, renal failure, and more (all covariables for survival analyses are listed in Tables 1 and 2 and Supplementary material online, Table S1). Modelling was performed by examining univariable models of each covariable and bivariable models also entering fasting. Further backward and forward multivariable selection evaluated covariables that had associations (P < 0.15) or potential confounding effects (bivariable change of the beta-coefficient of fasting >10%). Stepwise evaluations of variable groupings were performed with forced entry of sets of covariables in a stepwise procedure which was repeated for each subsequent group of covariables that were significantly associated with the outcome or confounders of fasting. Final models used forced variable entry of all significant or confounding variables. Given the number of deaths, up to 25 variables could be entered simultaneously in regression models (the final model had 18). All statistical analyses were performed using SPSS v.23.0 (IBM SPSS, Inc., Armonk, NY, USA).

Table 1

Baseline characteristics of the study population (see also Supplementary material online, Table S1A).

Characteristics	Overall	Non-faster	Routine faster
Sample size	N = 1957	n = 1568	n = 389
Age (years)	63.1 ± 14.1	63.0 ± 14.2	63.8 ± 13.8
Sex (female)	697 (35.6%)	566 (36.1%)	131 (33.7%)
Race
American Indian/Alaska Native	24 (1.2%)	19 (1.2%)	5 (1.3%)
Asian	12 (0.6%)	11 (0.7%)	1 (0.3%)
Black/African American	12 (0.6%)	10 (0.6%)	2 (0.5%)
Native Hawaiian/Pacific Islander	10 (0.6%)	7 (0.4%)	3 (0.8%)
White	1741 (88.0%)	1384 (88.3%)	357 (91.8%)
Other race	22 (1.1%)	19 (1.2%)	3 (0.8%)
Decline to answer	136 (6.9%)	118 (7.5%)	18 (4.6%)
Ethnicity
Hispanic or Latino	56 (2.9%)	51 (3.3%)	5 (1.3%)
Not Hispanic or Latino	1496 (76.4%)	1188 (75.8%)	308 (79.2%)
Unknown	22 (1.1%)	16 (1.0%)	6 (1.5%)
Decline to answer	383 (19.6%)	313 (20.0%)	70 (18.0%)
Smoking history
Never	1748 (89.3%)	1384 (88.3%)	364 (93.6%)
Past	154 (7.6%)	130 (8.3%)	19 (4.9%)a
Current	60 (3.1%)	54 (3.4%)	6 (1.5%)a
Alcohol consumption (drinks)
None	1267 (64.7%)	919 (58.6%)	348 (89.5%)b
<1 per weekc	271 (13.8%)	258 (16.5%)	13 (3.3%)
1–7 per week	222 (11.3%)	207 (13.2%)	15 (3.9%)
>7 per week	92 (4.7%)	89 (5.7%)	3 (0.8%)
Decline to answer	105 (5.4%)	95 (6.1%)	10 (2.6%)
Physical exercise
None	1275 (65.0%)	1044 (66.6%)	229 (58.9%)
Some (<1 h)	135 (6.9%)	105 (6.7%)	30 (7.7%)
1–2 h	137 (7.0%)	106 (6.8%)	31 (8.0%)
3 h or more	157 (8.0%)	113 (7.2%)	44 (11.3%)d
Did not respond	255 (13.0%)	200 (12.8%)	55 (14.1%)
Yearly gross household income
<$30 000/year	317 (16.2%)	274 (17.5%)	43 (11.1%)
$30 000–$49 999/year	348 (17.8%)	279 (17.8%)	69 (17.7%)
$50 000–$69 999/year	263 (13.4%)	201 (12.8%)	62 (15.9%)d
$70 000–$99 999/year	270 (13.8%)	213 (13.6%)	57 (14.7%)d
≥$100 000/year	316 (16.1%)	231 (14.7%)	85 (21.9%)b
Decline to answer	443 (22.6%)	370 (23.6%)	73 (18.8%)
Religious preference
Atheist	26 (1.3%)	23 (1.5%)	2 (0.5%)a
Buddhist/Hindu	6 (0.3%)	5 (0.3%)	1 (0.3%)
Jewish	5 (0.2%)	5 (0.3%)	0 (0%)
Latter-day Saintc	1100 (56.2%)	755 (48.2%)	345 (88.7%)
None	210 (10.7%)	205 (13.1%)	5 (1.3%)b
Protestant	175 (8.9%)	169 (10.8%)	6 (1.5%)b
Roman Catholic	127 (6.5%)	123 (7.8%)	4 (1.0%)b
Orthodox Christian	15 (0.8%)	15 (1.0%)	0 (0%)
Other	95 (4.9%)	90 (5.7%)	5 (1.3%)b
Decline to answer	199 (10.2%)	178 (11.4%)	21 (5.4%)b
Presenting symptoms at index hospitalization
Stable angina	1508 (77.1%)	1212 (77.3%)	296 (76.1%)
Unstable angina	309 (15.8%)	240 (15.3%)	69 (17.7%)
Acute myocardial infarction	140 (7.2%)	116 (7.4%)	24 (6.2%)

a

P ≤ 0.05 (P > 0.01) vs. non-fasters.

b

P ≤ 0.001 vs. non-fasters.

c

Referent category in this variable (where unspecified for categorical variables, the referent is the lowest value or first category in the variable).

d

P ≤ 0.01 (P > 0.001) vs. non-fasters.

Table 2

Selected baseline characteristics as cross-sectional outcomes (see also Supplementary material online, Table S1B).

Characteristics	Overall (N = 1957)	Non-faster	Routine faster
Weight (kg) (n = 1926)	90.8 ± 23.3	90.6 ± 23.5	91.6 ± 22.7
Body mass index (kg/m2) (n = 1925)	30.1 ± 7.2	30.1 ± 7.2	29.9 ± 7.0
SBP (mmHg) (n = 1255)	141 ± 23	142 ± 23	139 ± 23
DBP (mmHg) (n = 1255)	82.6 ± 12.9	82.7 ± 13.0	81.9 ± 12.6
Total cholesterol (mg/dL) (n = 1445)	161 ± 44	161 ± 44	161 ± 44
LDL-C (mg/dL) (n = 1415)	89.3 ± 34.9	88.3 ± 34.3	93.2 ± 37.0
HDL-C (mg/dL) (n = 1443)	43.4 ± 14.2	43.7 ± 14.6	42.4 ± 12.4
Triglycerides (mg/dL) (n = 1445)	145 ± 108	149 ± 114	129 ± 81a
Diabetes diagnosis history	777 (39.7%)	646 (41.2%)	131 (33.7%)b
Glucose (mg/dL) (n = 1925)	114 ± 39	114 ± 40	111 ± 36
HbA1c (%) (n = 728)	6.6% ± 1.7%	6.6% ± 1.6%	6.5% ± 1.8%
Coronary disease diagnosis	776 (43.1%)	637 (44.2%)	139 (38.6%)a
Myocardial infarction history	308 (15.7%)	250 (15.9%)	58 (14.9%)
Heart failure history	478 (24.4%)	404 (25.8%)	74 (19.0%)b
BNP (pg/mL) (n = 1336), median (IQR)	131 (47–341)	136 (47–359)	110 (44–255)a
Depression diagnosis history	448 (23.3%)	382 (24.7%)	66 (17.6%)b

a

P ≤ 0.05 (P > 0.01) vs. non-fasters.

b

P ≤ 0.01 (P > 0.001) vs. non-fasters.

BNP, B-type natriuretic peptide; DBP, diastolic blood pressure; HbA1c, haemoglobin A1c; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol; SBP, systolic blood pressure.

Results

Overall, 2025 patients completed surveys, constituting 73% of 2785 patients approached for INSPIRE enrolment (7% declined INSPIRE and 20% enrolled in INSPIRE but declined the survey). Exclusions included 3 patients with unknown death dates, 19 with unknown fasting status, 2 who declined to answer the fasting questions, and 44 who had engaged in fasting for <5 years. Baseline characteristics for included subjects (N = 1957) are shown in Tables 1 and 2 and Supplementary material online, Table S1. Thirty-six percent were female and age averaged 63 ± 14 years. Body mass index (BMI) averaged 30.1 ± 7.2 kg/m² and 51.1% had a history of hypertension. The average total cholesterol was 161 ± 44 mg/dL and 39.7% were diagnosed with diabetes. A HF history was found in 24.4% and the average B-type natriuretic peptide (BNP) was 131 pg/mL.

Survey results (Supplementary material online, Table S2) showed that at baseline routine fasters (n = 389) had fasted routinely ∼1 day per month for 42.2 years (±18.1 years). Routine fasters included 54.0% who fasted for more than 19 h at a time. Furthermore, 51.2% of routine fasters drank water when fasting (44.0% did not, 4.9% declined). The primary reason for fasting among routine fasters was religious/spiritual for 84.1%. Among non-fasters (n = 1568), never fasters (n = 1120) had 0.0 ± 0.0 years of fasting during life and previous fasters (n = 448) had 31.9 ± 20.7 years of fasting that ended prior to baseline. The percentage of life that patients had experience with periodic fasting is shown in Supplementary material online, Table S3. Of note for the two subgroups of non-fasters, never fasters vs. previous fasters averaged 62 vs. 65 years of age, 36% vs. 78% were Latter-day Saints, 14% vs. 7% were current or past smokers, 66% vs. 68% were physically inactive, and 36% vs. 14% were mild/moderate alcohol consumers.

Mortality was 8.2% in routine fasters, 14.7% for previous fasters, and 14.5% in never fasters. Mean follow-up was 4.43 years (±1.12 years, maximum: 6.08 years). In univariable Cox regression, routine fasters had HR = 0.51 [CI = 0.35–0.75; P < 0.001; 91.8% survivors (n = 357/389)] compared to non-fasters [85.5% survivors (n = 1340/1568)]. Kaplan–Meier estimates are shown in Figure 1. Adjusted Cox results are shown in Table 3, including for a fully adjusted model that had HR = 0.54 (CI = 0.36–0.80; P = 0.002). Interestingly, the association of routine fasting with lower mortality risk improved slightly in a bivariable model entering both fasting and alcohol consumption (and was unchanged by alcohol consumption after full adjustment, data not shown). However, 7.2% of routine fasters vs. 29.7% of non-fasters had mild/moderate alcohol consumption (up to seven drinks per week) and elevated mortality risk was found (fully adjusted compared to mild/moderate) for no alcohol consumption (HR = 1.24, CI = 0.86–1.79), heavy consumption (HR = 1.65, CI = 0.83–3.29), and those who declined to answer (HR = 2.03, CI = 0.97–4.26). Supplementary material online, Figure S1 shows associations for routine fasters with lower mortality risk in patients with or without CAD.

Figure 1

Kaplan–Meier survival curves following index catheterization for routine fasters vs. non-fasters (log-rank: P < 0.001).

Table 3

Survival analyses

Cox regression model	Hazard ratio	95% confidence interval	P-value
Model 1a	0.51	0.35–0.75	<0.001
Model 2b	0.49	0.33–0.71	<0.001
Model 3c	0.51	0.35–0.75	<0.001
Model 4d	0.55	0.37–0.82	0.003
Model 5e	0.54	0.36–0.80	0.002

Association of routine fasting with mortality.

a

Univariable Cox regression model for routine fasters compared to non-fasters.

b

Cox regression Model 2 entered age, race, religious preference, and income in addition to fasting status.

c

Model 3 entered the factors listed for Models 1 and 2, plus physical exercise, walking, and physical activity at work.

d

Model 4 adjusted for the factors listed for Models 1–3, plus diabetes diagnosis, heart failure diagnosis, prior atrial fibrillation ablation, cancer diagnosis, renal failure, and peripheral vascular disease.

e

Model 5 adjusted for the factors listed for Models 1–4, plus in-hospital treatment with coronary artery bypass grafting or percutaneous coronary intervention, and discharge prescriptions for anticoagulants, antiplatelet agents, and angiotensin receptor blockers. See the Supplementary material online, Appendix for survey variable definitions (all study covariables were evaluated during initial multivariable modelling; however, because some did not remain as significant independent predictors of mortality they were not retained in the final models presented here; please see the Methods section for a full description of the model-building approach).

Among patients with no HF history (n = 1472), routine fasters had lower incidence of HF (Table 4): adjusted HR = 0.31 (CI = 0.12–0.78; P = 0.013). Similarly, in patients with no prior MI (n = 1647), routine fasters had a lower incidence of MI: HR = 0.57 (CI = 0.37–0.88; P = 0.011) (Table 4), but after multivariable adjustment for the number of diseased coronary vessels, HF history, and alcohol consumption this association was reduced (HR = 0.67, CI = 0.43–1.05; P = 0.08). Incidence of stroke for routine fasters among those free of a stroke history (Table 4, n = 1841) had an adjusted HR = 0.77 (CI = 0.43–1.37; P = 0.37).

Table 4

Incident major adverse cardiovascular events

Cox regression model	Hazard ratio	95% confidence interval	P-value
Incident heart failurea
Univariable	0.31	0.12–0.78	0.013
Full adjustment	0.31	0.12–0.78	0.013
Incident myocardial infarctiona
Univariable	0.57	0.37–0.88	0.011
Full adjustment	0.67b	0.43–1.05	0.08b
Incident strokea
Univariable	0.74	0.41–1.31	0.30
Full adjustment	0.77	0.43–1.37	0.37

Association of routine fasting with incident cardiovascular events.

a

Analyses were performed among study participants free of this diagnosis at baseline.

b

The association of routine fasting with myocardial infarction was reduced by adjustment for the number of diseased coronary vessels, heart failure history, and alcohol consumption.

Discussion

Among people referred to coronary angiography, routine fasting ∼1 day per month during two-thirds of the patient’s life was associated with greater survival. Fasting was also associated with lower incidence of HF but not stroke during prospective longitudinal follow-up, and an association with incident MI was reduced by multivariable adjustment. At baseline, fasters had a lower prevalence of CAD and diabetes, as previously reported; furthermore, baseline triglycerides, prior HF diagnosis and baseline BNP concentration, and prior depression diagnosis were lower among fasters, but no fasting-associated differences in weight or BMI were found.

Scientists have been studying the relationship between fasting and lifespan since at least the 1940s,¹⁷ and fasting was evaluated as a weight loss method prior to 1975.²⁰ Many animal studies demonstrated a positive effect of fasting on lifespan.^{17,18,26–28} For example, Mitchell et al.¹⁸ found that long periods of fasting regardless of caloric intake or diet composition increased survival of mice. Pijpe et al.²⁸ reported that selecting butterflies for those resistant to starvation increased life span 50–100% compared to unselected adult butterflies. Beauchene et al.²⁶ found a 97% survival rate in rats that had limited feeding times. Anson et al.²⁷ noted that animals experienced extended longevity and stress resistance with fasting compared to ad libitum feeding.

Given these animal data, the results for survival of humans in the present study are not surprising. One may question, though, whether the average regimen in this study of fasting ∼1 day once per month for more than four decades could have a direct impact on mortality or MACE. Physiological changes leading to death or MACE can occur over just weeks or months rather than the decades it takes for chronic diseases like CAD and diabetes to develop.^10,11 Mechanisms exist, though, in which fasting may lead to the benefits observed here, especially HF-related risk reductions (see below), and outcomes were consistent with those benefits. While 1 day of fasting can evoke the benefits,^16,29 routine fasters in this study had followed their fasting practices for an average of 42 years and were continuing to fast. It may be that long-term routine fasting preconditions the body to expect food scarcity and activate physiological benefits of fasting earlier than usual in the fasting period such that shorter duration fasting also activates those mechanisms. Health benefits have been noted of time-restricted feeding—a 16–18 h daily fasting period,²² which is not much longer than a typical 12–14 h evening/overnight fast. It is plausible that a periodic longer fast (e.g. 24 h once per month) may condition the body to evoke fasting’s benefits more quickly (e.g. after just 8–10 h of fasting rather than the usual 12–14), so they occur each day and accumulate through repeated activation of fasting-related biological mechanisms.^{16,19,20,29,30} Genetically, such conditioning may have been conserved historically to aid in daily survival during periods of moderate food scarcity.

Development and progression of HF involves the human growth hormone/insulin-like growth factor-1 axis,³¹ as well as anaemia, excess sodium, and other factors that weaken the heart.³ Fasting is known to markedly increase human growth hormone during a calorie-free fast,²⁹ and to otherwise affect the human growth hormone/insulin-like growth factor-1 axis.³⁰ Fasting switches the body into ketosis and uses free fatty acids as the primary energy source.¹⁶ Furthermore, fasting induces natriuresis in which circulating levels of sodium are reduced due to its preferential excretion,^20,29,32 potentially lowering blood pressure, reducing cardiac load, and reducing HF risk. Interestingly, the natriuretic peptide BNP was significantly lower in routine fasters in this study, supporting that a tangible effect on HF risk existed. Finally, fasting increases haematocrit and haemoglobin,²⁹ which could decrease HF symptoms related to anaemia. Interestingly, such effects of fasting on HF-associated risk mechanisms closely mirror proposed effects of sodium-glucose cotransporter-2 inhibitors on HF, which mechanistic connections likely involve the human growth hormone/insulin-like growth factor-1 axis, ketosis, natriuresis, and reduced anaemia.

Furthermore, fasting was shown in a mouse model of HF to activate the autophagy-regulating transcription factor EB in the myocardium.³³ Fasting treatment of mice in that study reversed a form of cardiomyopathy by inhibiting mammalian target of rapamycin, increasing autophagy, and normalizing desmin localization.³³ Such pathways precondition the myocardium and may aid in inhibiting the development of HF. Some of these mechanisms and those mentioned above are among those proposed in reviews of various novel low-risk lifestyle factors.²³

This study’s results also suggest a connection of fasting with lower MI-associated risk.^34,35 Fasting may have a cardioprotective effect in reducing the severity of cardiovascular events.^36,37 Ahmet et al.³⁷ found that fasting rats had two-fold smaller MI, four times lower apoptotic myocytes in the at-risk area, and a smaller inflammatory response. This cardioprotective effect may be mediated by adiponectin, a hormone known to protect the heart from ischaemic injury by activating the cyclic adenosine monophosphate-dependent protein kinase—Akt pathway and inducing the expression of Hif-1.³⁶ These effects may also reduce the risk of HF after acute MI, but require additional investigation given the non-significant multivariable result for MI in this study.

Limitations and strengths

This was a non-randomized, observational study, making the determination of cause and effect challenging. Some confounding variables may be unmeasured (e.g. other dietary factors) and statistical analysis may not have completely controlled for some measured confounders. The study did evaluate 61 potential confounding covariables in survival analyses, while these findings generalize directly to patients referred for cardiac catheterization. Some variables such as laboratory results for cholesterol levels, haemoglobin A1c, and BNP and diagnostic results from echocardiogram [i.e. left ventricular ejection fraction (LVEF)] were available only in limited subsets of patients, thus adjustment for these factors is constrained to less than the full population and limits the interpretation of the statistical modelling, which future studies should address. Additionally, only a few (15.7%) of the patients received an echocardiogram during longitudinal follow-up and those who did had the testing at varying times, making it difficult to assess changes in LVEF and other measures in association with fasting.

Questions may arise because of unique characteristics of patients who have a Latter-day Saint religious preference, including lower smoking and alcohol consumption, and higher physical activity levels. Most Latter-day Saints, however, reported that they did not fast routinely (i.e. n = 755 of 1123 Latter-day Saints were non-fasters, similar to published percentages^10,11) and relatively small differences in smoking and physical activity existed between routine fasters and non-fasters. A substantial difference in alcohol consumption did exist for routine fasters vs. non-fasters (7.2% vs. 29.7% mild/moderate alcohol consumption). Studies suggest that no consumption and heavy consumption are associated with higher risk compared to moderate alcohol consumption,³⁸ and this study supported those associations. Consequently, any confounding from no alcohol consumption is expected to reduce the association of fasting with survival, but it did not affect associations with mortality or incident HF.

For previous fasters, when they stopped fasting routinely was unknown, but because their characteristics were similar to routine fasters, it is reasonable to assume that they started fasting about the same time in life as routine fasters which would place the ending of their fasting regimen ∼10 years prior to study entry. While some previous fasters may have stopped due to health problems, comorbidities were generally similar between routine and previous fasters.

A strength of this article is that it reports the first prospective longitudinal cohort study of fasting behaviour among thousands of individuals in association with mortality and MACE.

Conclusions

Participation in routine fasting across two-thirds of the lifespan was associated with greater survival after cardiac catheterization among a population evaluated angiographically for CAD. Furthermore, an association of fasting with lower incidence of HF was found among patients free of HF at baseline and this was robust to multivariable adjustment. A similar association with incident MI was found but this was weakened by adjustment. Fasting was also associated with lower odds of prevalent HF but not MI. While previously reported associations of fasting with prevalent CAD and diabetes were replicated, this study primarily reveals that progression of disease to HF and death may be prevented by fasting. Known and proposed biological mechanisms support the patient outcomes observed in the study, especially HF-associated mechanisms, and uniquely this epidemiologic study indicates that other undiscovered mechanisms may exist. Because this was an observational study, further mechanistic investigations and prospective randomized clinical trials of fasting are required to demonstrate that fasting causes these health benefits.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology online.

Data availability

Due to ethical restrictions related to protecting patient confidentiality, data cannot be made publicly available. Data are available through collaborative contract with Intermountain Healthcare by contacting Dr Benjamin D. Horne ([email protected]).

Funding

This study was funded by the Intermountain Research and Medical Foundation [award 614] and by internal institutional funds. The funding sources had no role in the design of the study, the data analysis, the interpretation of the findings, or the writing or publication of the study manuscript.

Conflict of interest: B.D.H. is the PI of other fasting-related research grants from the Intermountain Research and Medical Foundation. All other authors declared no conflict of interest.