Why fasting matters: IGF-1 and inflammation

A very interesting and well designed intermittent fasting (IF) study published last year (hereafter denoted as “this study”) has been widely discussed because of its results on body composition. However, the data in this study shows a lot more if you are interested in health and longevity.

There is some debate in the calorie restriction (CR) community about the effects of IF in humans (as rodent data is mostly supportive of an independent effect). Overall, and currently, there are some major points that are thought responsible for the beneficial effects of CR, besides energy restriction per se. One of the most important ones is protein intake. There was a shift in the average thinking on dietary protein for some long-term CRONers over the years as to what the optimal intake of dietary protein is.

This is based mainly on the effect of protein (specially from animal origin) on IGF-1 levels, as high IGF-1 levels have been associated with increased risk of cancer. This was particularly troubling considering that protein intake directly regulates IGF-1 levels: in long term CRONers, a reduction of protein intake was necessary to lower serum IGF-1 levels.

This means that, in contrast to rodents, energy restriction was not sufficient to lower IGF-1. Further evidence of the effects of dietary protein come from studies done on Drosophila, which show that protein (but not energy) restriction was critical for the beneficial effects on longevity. Similar findings in rodents have also been published. I think there are some caveats and nuances to take into account, but that is a topic for another day.

So, in humans, long term CR appears to show most of the health benefits observed in other species, with the exception of lower IGF-1 levels. Thus, the current trend is to restrict protein intake to up to 0.8g/kg of body weight.

To quote Fontana et al., 2008 (my emphasis):

(…) our findings demonstrate that, unlike in rodents, long-term severe CR does not reduce total and free IGF-1 levels in healthy humans if protein intake is high. In addition, our data suggest that chronic protein intake is more powerful than calorie intake in modulating circulating IGF-1 concentration in humans. (…) these findings underscore the importance of dietary macronutrient intake in regulating metabolic events, and suggest that reduced protein intake may become an important component of anti-aging and anticancer dietary interventions, due to the importance of IGF-1 in the biology of aging (…) and in the pathogenesis of many human tumors.

Moro et al. studied the effects of an IF 16/8 schedule (16h fast, 8h eating window) or 3-meals-per-day (control) in resistance-trained subjects for 8 weeks. Although the IF group lost body fat and gained fat-free mass, the most interesting bit is in the blood markers measured. I will focus mainly on IGF-1, but it is worth mentioning that overall, markers improved dramatically in the IF group.

As the IF group consumed 1.93g of protein per kg of body weight and total calories were not restricted (they were around maintenance with approximately 14.97 kcal/lb of body weight and were not statistically different than in the 3-meals-per-day group), one can assume that the changes in several parameters are the result of the restriction of calories to a shorter window of time during the day. In other words, the design of this study helps us dissociate the effects of CR and protein intake from that of fasting or time-restricted feeding.

What happened with IGF-1 levels? They went down significantly, despite consuming a high protein and a normocaloric diet. In the IF group, IGF-1 levels (ng/mL) fell from 216.94 ± 49.55 to 188.90 ± 31.48, while in the control group (shown as “No IF” below) it didn’t change much (from 215.59 ± 56.25 to 218.41 ± 42,24). Just restricting calories to an 8 hour window reduced serum IGF-1 levels by 13% in 8 weeks despite consuming maintenance calories and high protein.

IGF-1 levels pre and post intervention in the 16/8 group (IF) and normal 3-meal per day pattern (No IF).

If we consider the proposed cancer risk that carries consuming a high protein diet (which in itself is controversial but not far-fetched in certain contexts), IF seems to promote “healthier” IGF-1 levels than spreading calories throughout the day.

We can also compare the effect of IF + resistance exercise + high protein/normocaloric diet (IF+EX) to the effect of CR without IF* on IGF-1 levels. After 8 weeks of IF + EX, IGF-1 levels were around 188.90 ng/mL, while after 6 years of CR, levels are around 194 ng/mL. I would argue that the difference is neither significant nor meaningful. But it shows that in humans, IF might be a more viable way of reducing high serum IGF-1 levels than CR (without protein restriction).

IGF-1 levels have also been shown to be responsive to a low calorie, low protein fasting-mimicking diet (FMD). The last study showed a 13% reduction after 3 months (or 3 cycles of 5 days of FMD), whereas a previous pilot studyshowed a reduction of 15%. Overall, reduction either by a 5-day per month FMD or by a 16/8 IF-high protein protocol seem to be comparable. Whether a larger fasting window (or a shorter eating window) promotes additional reductions is unknown. Probably combining IF + CR is synergistic.

An important point to mention is that its not a matter of “how low” you can go: low and high IGF-1 levels are both detrimental. One should look for the optimal serum level. Unfortunately, there is no consensus and the “optimal” level is unknown. Sufficient to say that IGF-1 levels per se are meaningless if not coupled to optimization of other markers (ie. insulin), but that is beyond the scope of this post.

Finally, concerns about lower serum IGF-1 levels being detrimental for muscle mass gains are based on a misunderstanding of physiology (which goes in line with the “more is better” mentality of some). Serum IGF-1 levels are not correlated with skeletal muscle hypertrophy (see also here), which is also exemplified in the study discussed here (no difference in muscle mass gains between groups despite lower serum IGF-1 levels).

This is because of differential expression and protein content/signaling in muscle vs liver (the latter is reflected in serum). Simply put: what matters for muscle hypertrophy is the IGF-1 available in muscle (not correlated to serum IGF-1), which is mostly produced locally and acting in an autocrine/paracrine manner.

*It is not stated whether CRONers in the Fontana, et al. study also fasted regularly, but based on most plans I’ve seen (and guides by the CR Society), food is spread during the day, in contrast to what happens in mice undergoing CR.


Inflammation

Initially I was going to make a different post about it, but it made sense to just add it to this one. Besides significant reductions in glucose and insulin levels, all markers of inflammation (IL-6, IL-1b, TNF-a, leptin) went down and adiponectin (anti-inflammatory) went up. You can see the values compared to the 3-meals-per-day group below:

Levels before the intervention are shown as black bars, while post-intervention, in white bars.

As these are markers associated with inflammation but also adiposity, can’t these results be explained by the loss of bodyfat in the IF group?

Unfortunately there is not much information on adipocytokines in lean populations in the absence of weight loss. But we can compare the magnitude of change with respect to body fat loss in obese subjects after a weight loss diet. In obese women, 3 weeks of a very-low calorie diet which reduced bodyfat by 3 kg, decreased IL-6 and leptin but no change was seen for TNF-a (and there are mixed results on the effect of weight loss on the latter). A reduction of TNF-a has been seen in obese subjects with a calorie restricted diet (-500 kcal), sibutramine and exercise. Achieving 10% of weight loss through an aggressive weight loss program (500–1000 kcal of restriction) in obese women decreased IL-6 by 27%, compared to around 19% observed for IF subjects. Overall, it looks like the reduction in inflammatory markers is not proportional to the amount of body fat lost (1.6 kg) in this study. This suggests a different effect of IF per se.

As before, the most relevant comparison could be to long-term CRONers. Adiponectin levels in people practicing sever CR for approximately 7 years have been shown to be around 15.7 ug/mL, compared to 13.9 ug/mL seen after 8 weeks of IF. In comparison, long-term endurance runners had lower levels (11.1 ug/mL) which is more or less the values observed for the No IF group in this study (10.9 ug/mL) (which can also serve as a resistance-trained control group, as they were experienced lifters). Importantly, these levels are comparable to those of matched sedentary controls consuming a typical Western diet (WD) (9.5 ug/mL). This suggests that fasting and CR are better for increasing serum adiponectin levels than exercise (endurance or resistance), being CR the most efficient. The lack of effect of exercise on adiponectin levels has been shown before. There is also some evidence that moderate weight loss does not result in higher adiponectin levels in obese subjects (see also here).

On the other hand, we can also compare the effect of IF or CR on IL-6 levels: CRONers showed significantly lower levels of serum IL-6 (0.73 ng/L) compared to IF subjects in this study (1.08 ng/L). Endurance runners also showed lower IL-6 levels (0.71 ng/L) than IF subjects. Disturbingly, those sedentary on a WD had almost the same levels (1.21 ng/L) as resistance-trained subjects in this study (1.33 and 1.24 ng/L before the intervention).

Chronic low-grade inflammation has been implicated in several diseases related to obesity and insulin resistance, in which adipocytokines play a major role. As described here, levels of these pro-inflammatory cytokines are related to cardiovascular disease:

Interleukin‐6 (IL‐6), interleukin 1b (IL‐1b) and tumour necrosis factor α (TNFα) are the principal pro‐atherogenic cytokines,1 which are also produced in tissues other than the vascular wall and immune system, such as adipose tissue, myocardium, intestine, etc.1 They upregulate the expression of adhesion molecules on vascular endothelium, depress nitric oxide synthesis and promote the subendothelial migration of leucocytes. Further to their local regulatory role at a vascular level, these cytokines induce the liver‐derived synthesis of acute phase proteins, such as fibrinogen, plasminogen, C‐reactive protein (CRP) and serum amyloid α (SAA), which amplify inflammatory and pro‐coagulant responses.

The importance of the molecules described above can be further seen in this review. Some extracts below (my emphasis):

Adiponectin

Adiponectin has also been reported to have antiatherogenic effects (Funahashi et al. 1999, Ouchi et al. 1999). In addition, adiponectin exhibits cardioprotective activity in ischemic heart disease through AMPK and cyclooxygenase 2 pathways (Shibata et al. 2005). (…) Adiponectin also has anti-inflammatory effects that contribute to its protective role against metabolic stress in obesity. Adiponectin suppresses TNFα production in obese mice (Xu et al. 2003a), and adiponectin-deficient mice have high levels of TNFα in adipose tissue (Maeda et al. 2002). Low levels of plasma adiponectin are associated with C-reactive protein in humans (Ouchi et al. 2003). Adiponectin enhances the clearance of apoptotic cells by facilitating their opsonization and uptake by macrophages (Takemura et al. 2007). Some of the anti-atherogenic effects of adiponectin are also mediated by its role in the suppression of inflammatory responses. Adiponectin inhibits nuclear factor-κB (NFκB) activity and its downstream adhesion molecules leading to reduced monocyte adhesion to endothelial cells (Ouchi et al. 1999, Okamoto et al. 2002). In addition, adiponectin confers vascular-protective activities by suppressing the apoptosis of endothelial cell (Kobayashi et al. 2004).

Leptin

Leptin is structurally similar to Class I helical cytokines and shares the same JAK–STAT pathway downstream of its receptor. Leptin expression can be induced by endotoxin or cytokine TNFα (Grunfeld et al. 1996). Conversely, leptin increases thymic secretion of acute-phase reactants and TNFα and promotes T helper 1 cell differentiation (La Cava & Matarese 2004). Leptin acts on T cell, macrophages, and other immune cells to stimulate the production of a wide spectrum of cytokines (La Cava & Matarese 2004). In light of the role of several cytokines in enhancing energy expenditure and suppressing food intake (Ye & Keller 2010), this proinflammatory action of leptin might contribute to its overall effects in body weight regulation.

TNF-a

TNFα was the first cytokine identified in the adipose tissue of obese mice, marking the start of the metabolic inflammation concept (Hotamisligil et al. 1993). The direct involvement of TNFα in obesity-induced insulin resistance was confirmed by observations that TNFα treatment interferes with insulin signaling and blocks insulin actions (Hotamisligil et al. 1994). Mice lacking the functions of TNFα or its receptors are protected from obesity-induced insulin resistance and hyperglycemia (Uysal et al. 1997, 1998). It was initially thought that adipose-derived TNFα was produced mainly by adipocytes, but the parallel trend of macrophage infiltration and TNFα expression in adipose tissue of obese mice suggests that a significant portion of the adipose TNFα pool might be derived from macrophages and other immune cells. Interesting, FFA strongly stimulates TNFα production in macrophages (Nguyen et al. 2005) and in turn, TNFα stimulates lipolysis to increase fatty acid release from adipocytes (Wang et al. 2008). This FFA-cytokine cycle suggests that metabolic inflammation, once started, can use this self-perpetuating mechanism to further its inhibitory effects on insulin signaling and energy metabolism. In addition, TNFα directly stimulates hepatic lipogenesis in vivo (Feingold & Grunfeld 1987), and adipose-derived TNFα is also a major mechanistic link between obesity and cancer (Park et al. 2010).

IL-6

IL6 is one of the major pro-inflammatory cytokines whose expression level increases in the adipose tissue of obese mice and patients, but its role in glucose metabolism has not been fully resolved. (…) There are several potential explanations for the seemingly contradictory data regarding IL6 in insulin action and glucose metabolism. Effects of acute vs chronic treatments need to be differentiated and dose and site of action of IL6 need to be carefully considered. In addition, IL6 produced by different organs might also contribute to its complex effects on metabolic regulation.

From an in-depth review of IL-6 and metabolic inflammation, please read here. Although its not clear if IL-6 has a direct causative effect (it can have pro and anti-inflammatory effects), it has been associated with the T2DM, CVD and inflammation:

Low-grade chronic inflammation in obesity, reflected by a two- to threefold increase in the systemic level of cytokines including IL-6, appears to precede and is a risk factor of the subsequent development of insulin resistance and T2DM (Spranger et al., 2003; Wang et al., 2013; Lowe et al., 2014). (…) IL-6 has been identified as an independent predictor of T2DM and associated cardiovascular events (Spranger et al., 2003; Lowe et al., 2014). Adipocytes and macrophages residing in adipose tissue are the major sources for the elevated plasma IL-6 concentration up to 2–3 pg·mL−1 in patients with obesity and T2DM (Pradhan et al., 2001; Spranger et al., 2003). Nevertheless, the existing evidence is not enough to establish a causal association between IL-6 levels and the progression to metabolic and cardiovascular disorders. Due to its pleiotropic actions in various tissues and organs, the exact role of IL-6 in the pathogenesis of diabetes must be examined carefully in a cell- and tissue-specific manner, but allowing for the possibility of crosstalk between affected tissues and organs.


Overall, it seems that restricting calories daily to a short window of time might be metabolically beneficial in the absence of significant calorie or protein restriction. In combination with a diet with adequate protein and resistance exercise, it seems to promote favorable changes in body composition and improve metabolic markers related to inflammation. The combination of IF and CR might be synergistic, while the effects of longer daily fasting periods with a shorter eating window is unknown.

Let’s see at the body composition data from the study.

  • The IF group lost 0.9 kg in 8 weeks. This represents 1% of body weight. They lost 15% of body fat mass.
  • In kcal/lb (a rough measure of energy intake level), at the end of the study, the IF group was consuming around 14.97 kcal/lb, while the control group was consuming 15.47 kcal/lb. Generally, a good estimate for maintenance calories is between 14–16 kcal/lb.
  • In the IF group, subjects went from consuming 2826 to 2735 kcal/day. In the control group, it went from 3007 to 2910 kcal/day. There was no significant difference between calorie intake between groups. But for the sake of the argument, lets say that they were in calorie deficit. Energy restriction was thus 3% from baseline in both groups. This level of restriction is well within errors in estimation and in all practical purposes, not considered as “energy restriction”. However, if we assume that it indeed is considered an energy restricted diet, then no one can argue that the deficit was small and almost non-significant. This also agrees with the small amount of weight loss, which is well within normal daily variation.
  • Protein intake was increased compared to baseline and not significantly different between groups (1.93 g/kg in the IF group vs. 1.89 g/kg in the control group).

Based on the above, one can rely on the lack of statistical significance in energy intake between groups and mean calorie intake per day to assume that both groups consumed the same number of calories, which were around maintenance. However, the calorie difference might be biologically significant as suggested by the greater loss of body fat in the IF group. Given that the focus of the criticism is in the latter, I will consider the IF group to be in a slight calorie deficit and show how it doesn’t change the main argument of the original post.

What we know about IGF-1, calorie restriction (CR) and weight/fat loss

As mentioned in the other post, it is well established than in humans, dietary protein intake is the main determinant of serum IGF-1 levels. So in theory, the higher the protein intake, the higher the IGF-1 levels. CR has a modest effect. On the other hand, CR, by definition, will result in weight loss, which could also modulate IGF-1. Thus, the significant reduction of IGF-1 in the IF group, if not because of the temporal restriction of food to a short window, could be due to:

a) Lower protein intake than the control group.

b) Lower caloric intake or greater CR than the control group.

b) Greater weight/fat loss than the control group.

Protein intake was not different between groups and was high, so if anything, it should have increased IGF-1 levels.

The CR level in both groups was 3%, so the level of restriction for both was the same (the absolute level of restriction was 91 kcal for the IF group and 97 kcal for the control group). The difference in basal energy intake between groups was due to different initial mean body weight, which was not statistically significant (83.9 kg in the IF group vs. 85.3 kg in the control group), but at the end of the study was 175 kcal. Again, the difference was not statistically significant, both diets were on maintenance levels and restricted from basal levels by the same amount. But for the sake of the argument, we will assume that this difference could account for the change in IGF-1.

The IF group lost significantly more body fat than the control group. Thus, it could also be that this loss of body fat explains the difference in IGF-1 between groups.

In summary and for further comparisons, the IF group was 3% CR, lost 1% of body weight and lost 15% of body fat mass in 8 weeks. The appropriate way to calculate the calorie deficit should be to subtract the intervention calories from basal calories (which is 91 kcal), but as people have focused on the 175 kcal difference with the control group, we will use this number as the calorie deficit (-175 kcal).

Because CR and weight loss are invariably linked, I will discuss them together.

I already mentioned that long-term (6 years) CRON results in a modest decrease in IGF-1, with average levels compared to what achieved in 8 weeks in the IF group. This population is the most relevant for discussion of independent effects of CR on metabolic markers as they are weight-stable (so no confounding due to weight loss).

There is also data from the CALERIE study, at 1 and 2 years. In this intervention, normal weight subjects were calorie restricted for 2 years to a goal of 20% CR. After 1 year, subjects achieved only 12% of CR (-279.5 kcal), reduced body weight by 10.7%, fat mass by 24%, improved insulin sensitivity and some inflammatory markers, but didn’t reduce IGF-1 significantly. The same results were seen after 2 years: body weight was reduced by 10.4%, fat mass by 22.5% with a similar CR level (-216.3 kcal), but IGF-1 levels were only reduced by 8.6%. Importantly, protein intake increased in this period to 1.28 g/kg.

I have put these differences in a table to make it easier to compare:

All changes are with respect to baseline values and just a rough difference between means.

Achieving 10% of weight loss, 22.5% of fat loss and 10% CR in 2 years didn’t produce the same reduction in IGF-1 as 8 weeks with IF with 3% CR, despite higher weight/fat loss and lower protein and calorie intake. Even if we consider that the calorie reduction was 175 kcal (which was not), it still falls short compared to that in Fontana et al., 2016 (175 kcal vs 216 kcal, or 6% CR vs 10% CR).

I believe the most adequate comparison is the one above, because it compares normal weight subjects. But there is also data from obese subjects undergoing weight loss.

In overweight women, 25% of energy restriction either continuously (CER, 1500 kcal daily) or intermittently (IER, 647 kcal for 2 days per week) resulted in similar body weight loss after 6 months. Neither of the interventions reduced significantly IGF-1. However, there is a clear trend in the IER group for reducing IGF-1 levels (baseline=201.3; 6 months=191.6 ng/mL) that was not seen in the CER (baseline=202.9; 6 months= 203.7 ng/mL) and didn’t appear to correlate with weight or fat loss.

But the IER protocol is more similar to the 5:2 diet than to a 16/8 IF protocol (and it involves less overall fasting period). Nevertheless, it shows that 6 months of 25% CR that produced significant changes in body weight and fat loss didn’t reduce IGF-1 levels. Similar results have been observed by the same authors: no change in IGF-1 with 25% daily or intermittent energy restriction despite significant weight/fat loss.

In other group of obese subjects, dietary restriction (1200 kcal/day) increased IGF-1 levels after 8 weeks but returned to baseline after 16 weeks, despite 5.8 and 8.1 kg of body weight lost (8 vs. 16 weeks, respectively). A similar increase in IGF-1 after weight loss has been observed in other study with obese women, as well as in an intervention with or without orlistat.

As described, there is no clear relationship between body weight/fat loss and degree of CR on serum IGF-1 levels in normal or obese subjects. Only after a long period of time (2 years) and constant, significant CR (10–12%) a small change in IGF-1 is observed. Thus, it is highly unlikely that the change seen in the IF group is due to either CR or weight/fat loss, specially with a high protein intake, almost no CR and short duration of the intervention (8 weeks).

Even if one assumes that the IF group was in significant calorie deficit, the degree of IGF-1 reduction in such a short time and with the amount of protein is remarkable. As mentioned in the other post, comparable reductions have only been observed after 3 cycles of a Fasting-Mimicking Diet(13%) or reducing protein from 1.67 g/kg to 0.95 g/kg for 3 weeks (22%), the latter being the most effective.

Finally, I want to mention something important that was not part of the original post as the study didn’t measure it: IGFBP (IGF-binding proteins). The activity of IGF-1 depends on its bioavailability, which in turn is determined by the ratio of IGF-1 to IGFBP (IGF-1:IGFBP ratio).

Simply put, the concentration of IGFBP determines the amount of free serum IGF-1, which in the end is the available hormone in circulation. 2 years of CRincreased IGFBP-1 (one isoform of IGFBP that is regulated by metabolic status) by 20–25%, which in turn reduced the IGF-1:IGFBP-1 ratio by 42%. So even though CR didn’t reduce IGF-1 significantly, it did reduce the amount of free IGF-1.

Why it makes sense

Quoting Fontana et al., 2016 (my emphasis):

In fact, serum concentration of IGFBP-1, unlike IGFBP-3 which binds 75–90% of circulating IGF-I, is heavily influenced by the metabolic (i.e., insulin resistance, and insulin and glucagon levels) and nutritional (fasting and refeeding) state of the individual. Excessive adiposity-induced insulin resistance and compensatory hyperinsulinemia have been shown to decrease hepatic synthesis of IGFBP-1, which translates into increased concentrations of bioavailable IGF-1, without modifications in serum total IGF-1 levels (Lukanova et al., 2001; Maddux et al., 2006).

Patients with type 1 diabetes have higher serum IGFBP-1 concentrations than normoglycemic controls (Suikkari et al., 1988), and acute steady state hyperinsulinemia lowers serum IGFBP-1 levels by 40–70% in normal individuals (Yeoh & Baxter, 1988; Snyder & Clemmons, 1990). Moreover, it has been shown that circulating levels of IGFBP-1 are acutely increased by 3–4 fold in response to overnight fasting and decline rapidly after a meal (Busby et al., 1988; Smith et al., 1995).

Indeed, IGFBP-1 has been proposed as a marker of hepatic insulin sensitivity. While IGF-1 levels are primarily determined by dietary protein intake, IGFBP-1 levels are regulated by insulin secretion. It appears that fasting decreases IGF-1 and increases IGFBP-1, effectively reducing IGF-1 bioavailability. In normal subjects after 36 hours of fasting, IGF-1 levels are reduced from 249.5 to 219.4 ng/mL (-12%) and IGFBP-1 levels are increased from 27.4 to 205.2 ng/mL (+649%).


A note on adipocytokines

In the previous post I presented evidence that the relationship between weight loss and levels of adipocytokines is equivocal. However, I found information on TNF-a in non-obese subjects. From the same CALERIE study, TNF-a was reduced by 22% after 2 years, compared to only 8% in the IF group, which is almost the same reduction observed for the CR subjects in CALERIE after 1 year (8.5%) with a higher CR level. Still, absolute values are significantly lower in long-term CRONers.