Does intermittent fasting raise the risk of early death?

Executive summary

* A study examined the associations of meal frequency, skipping meals, and intervals between meals with all-cause and cardiovascular disease (CVD) mortality.

* The study involved 24,000 US adults who were followed for an average of approximately 7.7 years. Almost one in five people died during the study, which is high.

* The study claimed that, compared with people consuming three meals a day, those consuming one meal a day had a 30% higher (relative) risk of dying from anything and an 83% higher (relative) risk of dying from CVD.

* There were other findings related to skipping breakfast, lunch and/or dinner and one finding related to the average number of hours between meals.

* There were many non-findings, and these showed inconsistencies with the headline findings.

* As usual, the characteristics table showed that the ‘meal skipper’ was a totally different person to the ‘three meals a day’ person. Differences were adjusted for, but you can’t adjust for a whole lifestyle.

* The plausible mechanism discussion in the paper blew the paper apart. It explored why skipping meals might be associated with higher mortality. Association was the key word. The meal patterns were markers of other issues (e.g., poverty/lack of access to food). None of them explained why skipping meals per se would cause any early deaths.

Introduction

I was browsing online, when a news site popped up with the headline “Intermittent fasting may raise your risk of an early death” (Ref 1). I’m not the biggest fan of intermittent fasting, but I don’t think it’s a deadly activity to adopt (I summarise my view in Ref 2). We’ve looked at intermittent fasting a number of times in Monday notes (Ref 3). I located the academic article behind the headlines. It was called “Meal Skipping and Shorter Meal Intervals Are Associated with Increased Risk of All-Cause and Cardiovascular Disease Mortality among US Adults” and it was by Sun et al (Ref 4).

The objective of the study was “to examine the associations of meal frequency, skipping, and intervals with all-cause and cardiovascular disease (CVD) mortality.”

The study

The study was epidemiological i.e., a population study. Hence we know immediately that we’re looking at association and not causation. The study involved US adults and thus may not be generalisable beyond this group. A total of 24,011 people were examined. They were aged 40 years or older, with an average age of approximately 56 years. Approximately 10% of participants were Hispanic, which led to a number of references to “Spanish” in the paper. The participants were part of the National Health and Nutrition Examination Survey 1999-2014. Eating patterns were assessed using 24-hour recall (so we also have the usual limitations of Food Frequency Questionnaires).

Remember that the characteristics table (Table 1) is the place to start in any population study. This groups people by the factor of interest (meal frequency in this case) and then the characteristics of people who consume different numbers of meals are easy to compare.

In this study, compared with people consuming three meals a day, those who consumed one or two meals a day (especially one) were more likely to be male, more likely to be younger, with far lower energy intake, far lower education, far greater incidence of poverty, far more likely to be a current smoker, doing far less activity, less likely to be a moderate drinker and more likely to have food insecurity.

Food insecurity was defined according to the Adult Food Security Survey Module. This is a standardized 10-item questionnaire developed by the US Department of Agriculture, which is used in many national surveys (Ref 5). Food insecurity is “the limited or uncertain availability of nutritionally adequate and safe foods, or limited or uncertain ability to acquire acceptable foods in socially acceptable ways.” Not having access to sufficient food is serious and highly likely to be associated with dying early.

The only characteristic acting in favour of those consuming fewer than three meals a day was age (they were younger on average). All other lifestyle characteristics acted against the fewer meals people. The fewer meals people had some advantages related to baseline conditions. Compared with people consuming three meals a day, those who consumed one or two meals a day were less likely to have diabetes or cancer. Pre-existing cardiovascular disease was mixed – being highest in those consuming two meals a day and lowest in those consuming four.

The characteristics table usually paints a picture of a different type of person. The differences were stark in this study. The three meal a day people were completely different people to those consuming fewer meals than this. The ‘meal skipper’ was male, young, poor, less educated, less active, struggling to get food and a smoker.

One line in the characteristics table caught my eye – “Snacks frequency (times/day).” It was reported that those eating three meals a day snacked 2.1 times a day and those eating one meal a day snacked 2.8 times a day. When is a meal a snack and vice versa?

The answer to this could be found on page 2 of the paper: “Reports of breakfast, lunch, and dinner or their equivalents in Spanish were considered as meals in this study. Breakfast consumption was defined based on self-reported consumption of morning meal, including ‘breakfast’ and the Spanish equivalents. Lunch consumption was defined based on self-reported consumption of noon meal, including ‘lunch,’ ‘brunch,’ and the Spanish equivalents. Dinner consumption was defined based on self-reported consumption of afternoon meal, including ‘dinner,’ ‘supper,’ and the Spanish equivalents. All other eating events were considered as snack.”

Hence meal frequency in this paper came down to self-reported timing. The three meals a day people consumed their three meals at times that we associate with breakfast, lunch and dinner. Plus, they snacked an average of 2.1 times a day outside these times. The one meal a day people consumed one meal at a time that we associate with breakfast, lunch or dinner. Plus, they snacked an average of 2.8 times a day outside these times. That’s 5.1 occasions of eating vs 3.8. That’s not so different.

Adjustments were made for “age, gender, race and ethnicity, socioeconomic status, lifestyle factors, and baseline history of diseases.” However, as we must reiterate – you can’t adjust for a whole person/lifestyle.

Results

The main outcome of interest was death – cardiovascular disease (CVD) mortality and all-cause mortality. This is ideal. There’s no point dying less of one thing to die more of another. During 185,398 person years of follow-up, 4,175 deaths occurred including 878 CVD deaths. That’s high. There was an average 7.7 years of follow-up. Of the 24,011 people in the study, 4,175 died during the follow-up period. That’s almost 1 in 5 people who died (17%), having been followed for an average of approximately 7.7 years from a starting average age of approximately 56 years. This reflects the fact that the study included people with pre-existing conditions and clearly the poverty/food access issues were key.

There were a number of results. The following passage is the summary presented in the abstract. I’ll leave the confidence intervals in this passage (for those who like them) and then we’ll break down these results – and include the non-results – afterwards:

“Compared with participants eating three meals per day, the multivariable-adjusted HRs for participants eating one meal per day were 1.30 (95% CI 1.03 to 1.64) for all-cause mortality, and 1.83 (95% CI 1.26 to 2.65) for CVD mortality. Participants who skipped breakfast have multivariable-adjusted HRs 1.40 (95% CI 1.09 to 1.78) for CVD mortality compared with those who did not. The multivariable-adjusted HRs for all-cause mortality were 1.12 (95% CI 1.01 to 1.24) for skipping lunch and 1.16 (95% CI 1.02 to 1.32) for skipping dinner compared with those who did not. Among participants eating three meals per day, the multivariable-adjusted HR for participants with an average interval of <=4.5 hours in two adjacent meals was 1.17 (95% CI 1.04 to 1.32) for all-cause mortality, comparing with those having a meal interval of 4.6 to 5.5 hours.”

The results can be categorised as follows:

1) Meal frequency

Compared with participants eating three meals per day, the (adjusted) risk ratios for participants eating one meal per day were 1.30 for all-cause mortality, and 1.83 for CVD mortality. i.e., one meal a day people had a 30% higher (relative) risk of dying from anything and an 83% higher (relative) risk of dying from CVD.

The abstract didn’t report that there were no significant differences in risk for people who consumed two meals a day or four or more meals per day when compared with those consuming three meals a day. The ‘fewer meals = more deaths’ hypothesis only held for one vs three meal eaters.

2) Missing meals

Participants who skipped breakfast had a 40% higher (relative) risk of dying from CVD compared with those who didn’t skip breakfast.

The abstract didn’t report that they had no greater risk of dying from any cause. If they were dying more from CVD and not differently overall, they were likely dying less of something else.

Skipping lunch or dinner worked the other way round – the differences were significant for all-cause mortality, but not for CVD mortality. People who skipped lunch had a 12% higher (relative) risk of dying from anything, but no greater risk of dying from CVD. People who skipped dinner had a 16% higher (relative) risk of dying from anything, but no greater risk of dying from CVD.

3) Intervals between meals

This one was interesting. Having claimed that having fewer meals was not good and skipping specific meals was not good, the final area of analysis was the average number of hours between meals. This found that shorter intervals were not good. Table 4 in the paper reviewed only those people consuming three meals a day. It grouped them into those having meals 4.5 hours or fewer apart, those having meals 4.6-5.5 hours apart and those having meals more than 5.5 hours apart.

The middle group was taken as the reference point. Relative to those averaging meals 4.6-5.5 hours apart, there was a 17% higher (relative) risk of dying from any cause for those having meals 4.5 hours or fewer apart. The differences between the 4.6-5.5 hours interval group and the more than 5.5 hours were not significant. (There were no significant results for CVD mortality).

This could be argued either way. It could be argued that not leaving a longer interval, i.e., not ‘fasting’ for longer between meals was riskier. It could be argued that having meals closer together left a longer period outside the three meals when meals were not being eaten and this longer fast was riskier. It wasn’t clear what was being implied as a finding here.

The non-results were just as important as the results and they showed that the results were not consistent. One meal relative to three showed a difference, but not two or four. Skipping breakfast was riskier for dying from CVD but not all causes. Skipping lunch or dinner was the other way round. Closer meals seemed ‘riskier’, but this left more ‘fasting’ time outside of meals.

Some of the non-results were mentioned in the “Discussion” section of the paper but they weren’t highlighted in the paper abstract (which is all that most people read). The Discussion section noted that most research has been conducted into skipping breakfast, not lunch or dinner.

Plausible mechanism?

The Discussion section is always supposed to offer a plausible mechanism for findings. This paper suggested four reasons for why skipping meals might increase the risk of death:

1) skipping meals might indicate the INability to acquire adequate food, which has been shown to be a CVD risk among adults.

2) skipping meals might be a marker for an unhealthy lifestyle.

3) skipping meals usually means eating meals with a larger energy load (i.e., fewer, bigger meals), which could adversely impact metabolism and

4) skipping meals may reflect psychosocial conditions, which are associated with higher CVD risk e.g., anorexia.

Three of these (1, 2 and 4) offer explanations for why skipping meals might be a marker of something that could impact mortality. They don’t offer an explanation for why skipping meals per se can impact mortality.

We know from the characteristics table that (1) was the case. The income/poverty and food security information confirmed that the one meal a day person was poor and disadvantaged. Being poor and disadvantaged would increase mortality risk. The number of meals per day was just the marker of this.

We know from the characteristics table that (2) was the case. The one meal a day person had a bad lifestyle. Having a bad lifestyle would increase mortality risk. The number of meals per day was just the marker of this.

If point (4) were correct, (i.e., skipping meals may reflect psychosocial conditions), then those conditions would be the factors that increased mortality risk (especially anorexia if applicable). Again, the number of meals would just be a marker of this.

The third point (fewer, bigger meals) didn’t withstand scrutiny. According to the characteristics table, the one meal a day person was consuming 83% of the calories of the three meals a day person. I calculated the average calories consumed at each eating occasion (i.e., meals and snacks). For the one meal a day person it averaged 456 calories at each eating occasion. For the three meals a day person it averaged 410 calories at each eating occasion. Two small carrots would account for that difference. It’s not going to cause early death.

The bottom line

The plausible mechanism analysis was the one that got to the crux of this paper. I was trying to think how skipping a meal might increase the risk of death. It was the paper’s job to answer this. In the Discussion section it did. However, this showed that any plausible mechanism simply confirmed that meal skipping could be a marker of an issue. The issue itself would be the factor that could impact mortality.

The Daily Mail, UK, ran with the headline “Intermittent fasting – the celeb-favorite diet followed by Kourtney Kardashian and Mark Wahlberg – may RAISE your risk of an early death by 30%, study suggests” (Ref 6). This is not true. The number of meals consumed by people in this particular US study was associated with a number of factors that could increase the risk of mortality (being poor, insecure access to food, having a bad lifestyle etc.). The number of meals a day did not cause anything. Intermittent fasting did not cause anything. The number of eating occasions simply told us something about the people with a higher risk of mortality.

References

Ref 1: https://newsachieve.com/2022/11/24/intermittent-fasting-may-raise-your-risk-of-an-early-death-by-30-study-suggests/
Ref 2: The main reasons for this are: i) I don’t think that the evidence is compelling that the gain (especially relative to ‘pain’) is significant in the normal use of the word; ii) I don’t think that restricting eating to certain times helps with eating disorders; iii) I think that the body gets used to restrictions if adopted and then there are consequences if normal eating is resumed; and iv) if people follow very low carb diets and intermittent fasting and are still struggling with weight, there’s not much left for them to try.
Ref 3: https://www.zoeharcombe.com/?s=intermittent+fasting
Ref 4: Sun et al. Meal Skipping and Shorter Meal Intervals Are Associated with Increased Risk of All-Cause and Cardiovascular Disease Mortality among US Adults. Journal of Academy of Nutrition and Diabetics. August 2022.
https://www.sciencedirect.com/science/article/pii/S2212267222008747
Ref 5: Their refs:
US Dept of Agriculture, Economic Research Service. The U.S. Adult Food Security Survey Module. Accessed June 22, 2022. https://www.ers.usda.gov/media/8279/ad2012.pdf and
Opsomer et al. An evaluation of the U.S. Department of Agriculture food security measure with generalized linear mixed models. J Nutr. 2003.
Ref 6: https://www.dailymail.co.uk/health/article-11462201/Intermittent-fasting-RAISE-risk-early-death-30-study-suggests.html

2 thoughts on “Does intermittent fasting raise the risk of early death?

  • avatar
    December 15, 2022 at 9:05 am
    Permalink

    This looks like a study fudged, sorry, ‘powered’ to prove that 3 meals a day and 2 snacks is the healthiest option, including calling a meal a snack when it suits the preferred outcome.

    Reply
  • avatar
    December 12, 2022 at 9:45 pm
    Permalink

    Hi Zoë
    You really have to work hard at times. I mean, reading and analysing stuff this quality.

    Snack frequency
    omad twmad thmad f+mad p
    2.8 (0.1) 2.4 (0.03) 2.1 (0.02) 2.0 (0.04) < 0.001
    1 meal 2 meals 3 meals 4+ meals
    3,8 4,4 5,1 6+ meals in reality

    This is strange. You only count "main meals", and the snacks are like "free meals".
    However, the glucose and insulin response are a bit overdriven, with small "snacks", as I recall it.
    Especially if one follows the "healthy" advice.
    "Meal frequency was calculated as the sum of frequency of breakfast, lunch, and dinner consumption, and further categorized as one, two, three and four or more meals per day." (some reported multiple breakfast etc.).

    What has this actually to do with intermittend anything? I thought the idea was not-to-eat while waiting for the next meal. Since they measure the time only inbetween the main meals, omitting the snacks, could it become to worse self deception?

    The whole idea with "intermittend" is to change metabolism by not-eating, enabling burning the fat, that the extra of it, nearly everybody wants to get rid of.

    Prof. Brian Wansink, it is all forgiven (almost 20 studies retracted). An intelligent fraud is much more interesting than a stupid study. Come back, and do I have a theme for you. If you sit a) at entrance b) near window c) near corner, at restaurants you frequently visit, what are your chances to get CVD? Possible mechanism; too much hassle at entrance, too lonely at corner, window inspires your metabolism. Boy what amount of clicks and nutrition chics to report on this.
    JR

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.