Blood sugar and cortisol: the axis behind many hormone symptoms

A team around Verhulst searched PubMed and EMBASE for work published between 1980 and 2018 in which blood sugar was lowered step by step under controlled conditions. 63 papers remained, 51 of them with participants without diabetes.

In these people without diabetes the medians were as follows: adrenaline started at 3.4 mmol/L, growth hormone at 3.2 mmol/L, noradrenaline at 3.0 mmol/L and cortisol only at 2.8 mmol/L. Autonomic symptoms appeared at a median of 3.0 mmol/L.

For you this means: your body does not wait for a dramatically low value. It reacts in a range that many lab reports would not flag as abnormal. And cortisol comes later in this sequence than adrenaline, not first.

The authors name two limitations themselves, and they belong here. In this review the thresholds for autonomic and for neuroglycopenic symptoms did not differ from one another. And the methodology of the included papers varied considerably, with the high insulin doses of most clamp studies potentially affecting the counterregulation.

A group around Mitrakou lowered blood sugar in ten healthy volunteers to fixed plateaus of 90, 78, 66, 54 and 42 mg/dL and measured hormones, symptoms and cognitive performance at each step.

Hormone release began between 65 and 68 mg/dL. The autonomic symptoms began at 58 plus/minus 2 mg/dL, and the authors list them literally: anxiety, palpitations, sweating, irritability, trembling. The neuroglycopenic symptoms, meaning hunger, dizziness, tingling, blurred vision, difficulty thinking and faintness, followed at 51 plus/minus 3 mg/dL. A measurable deterioration in cognitive tests occurred at 49 plus/minus 2 mg/dL.

For you this means: this list is not imagination. It is the description of an adrenaline response, documented in the laboratory with a clock and a catheter. That the same list also appears in a panic attack does not contradict this. Both routes end in the same adrenaline response.

A group around Dinneen blocked endogenous cortisol production in healthy people with a drug and then replaced it either with a variable infusion that recreates the nocturnal rise, or with a steady infusion that prevents it. A mixed meal followed.

With the nocturnal rise recreated, glucose release from the liver was higher (2.6 plus/minus 0.2 versus 1.5 plus/minus 0.4 nmol/kg per six hours), the disappearance of glucose from the blood lower (5.9 plus/minus 0.3 versus 7.3 plus/minus 0.9 mmol/kg per six hours) and C-peptide lower. Glucagon was higher. In addition, fatty acid release, ketone bodies and alanine rose.

For you this means: the higher morning value arrives with advance notice. In the original text the authors explicitly call this state a physiological insulin resistance. So a built-in state, not a disease.

A group around Spiegel had eleven young men spend only four hours in bed for six nights and then recover over six nights with twelve hours in bed.

Under sleep restriction, glucose tolerance was lower (p less than 0.02), the concentration of thyroid stimulating hormone lower (p less than 0.01), evening cortisol elevated (p = 0.0001) and sympathetic nervous system activity increased (p less than 0.02). The authors described the effects as resembling those of normal ageing.

For you this means: lack of sleep pushes on both ends of the axis at the same time. The sample is eleven young men. Not one woman was included. That is a serious limitation, not a footnote.

A group around Broussard studied seven healthy adults, one of them a woman, after four days with 4.5 hours in bed and after four days with 8.5 hours in bed, with controlled food and controlled activity. Fat cells taken from subcutaneous adipose tissue were then treated with rising insulin concentrations.

After the short sleep phase these cells needed almost three times the insulin concentration for the half-maximal response (0.71 versus 0.24 nM, p = 0.01), and the total response was 30 percent lower (p = 0.01). Whole-body insulin sensitivity fell in parallel (p = 0.02).

For you this means: the effect of too little sleep does not end at the level of feeling. It is measurable in extracted tissue. Seven people at one centre are very few, and the authors name that as a limitation themselves.

A group around Plymate incubated a human liver cell line for 72 hours in serum-free medium and compared the influence of insulin and prolactin with that of estradiol, thyroxine and testosterone.

Insulin and prolactin lowered SHBG production from 65.0 plus/minus 0.6 to 46.8 plus/minus 1.1 and 46.8 plus/minus 1.2 nmol per one million cells respectively (p less than 0.01). Insulin also dampened the SHBG production stimulated by estradiol and thyroxine. Estradiol and testosterone increased it.

For you this means: a direction has become visible here, not a process in your body. Not a single sentence about your blood can be derived from this paper. It is the start of the chain, not its proof.

A group around Selva treated transgenic mice carrying the human SHBG gene, as well as human liver cells, with glucose and fructose and looked at which switch SHBG formation responds through.

Both sugars lowered production of human SHBG. The route ran through a downregulation of the transcription factor HNF-4alpha, which was replaced at the SHBG promoter by another factor. The fall in HNF-4alpha ran in parallel with rising palmitate in the cell, could be mimicked with palmitoyl-CoA, and did not occur when new fat formation was inhibited.

For you this means: it is not insulin alone that stands at the beginning, but the new fat formation in the liver driven by sugar. This is the point where liver fat enters the picture. The measurements were made in cells and in mice. In humans this switching route has not been demonstrated in this form.

A group around Nestler studied six women with PCOS and obesity. First, steroid production in the ovaries was suppressed with medication so that the sex hormones drop out as a confounder. Then the active substance diazoxide lowered insulin release over ten days.

The insulin area in the glucose test fell from 262 plus/minus 55 to 102 plus/minus 33 nmol per minute and litre (p less than 0.05). SHBG rose by 32 percent from 17.8 plus/minus 2.6 to 23.5 plus/minus 2.0 nmol/L (p less than 0.003). Testosterone not bound to SHBG fell by 43 percent (p = 0.05). The sex steroids themselves did not change.

For you this means: this is the cleanest human evidence I could find for this connection, and it covers six women. Both substances used are prescription medicines and stood here as a research tool, not as treatment.

Diazoxide is licensed in Germany for the treatment of hypoglycaemia and not for hormone symptoms. Its use in this study therefore lay outside the licence. In the same study glucose tolerance worsened under diazoxide. The substance can also raise blood sugar, cause the body to retain water and sodium, and change hair growth. None of this is an approach to copy, and a dose is deliberately not given here.

Nestler and Jakubowicz randomised 24 women with PCOS and obesity to metformin or placebo and gave it for four to eight weeks. Eleven women received metformin, the rest placebo. Steroids, hormone responses and glucose tolerance were measured.

In the metformin group, that is in eleven women, the insulin area fell from 9303 plus/minus 1603 to 4982 plus/minus 911 microunits per millilitre and minute (p = 0.004). Free testosterone fell from 0.34 plus/minus 0.07 to 0.19 plus/minus 0.05 ng/dL (p = 0.009), SHBG rose from 0.8 plus/minus 0.2 to 2.3 plus/minus 0.6 micrograms/dL (p less than 0.001), and androgen formation in the ovary declined. In the placebo group none of this changed meaningfully.

For you this means: the circle closes in both directions. Less insulin went along with more SHBG and less androgen formation. All the numbers given come from eleven women, not from 24.

The full frame belongs to the picture. Metformin is a prescription medicine, licensed in Germany for the treatment of type 2 diabetes. Its use in polycystic ovary syndrome, which is what this study is about, is a use outside the licence, meaning off-label. That means: the decision is made by a doctor in the individual case, it requires informed consent, and reimbursement is not a given.

Metformin can cause gastrointestinal symptoms. Taken over longer periods it can impair the absorption of vitamin B12. And it must not be given, or only in adjusted form, when kidney function is impaired, because in rare cases lactic acidosis can occur, a severe acidification of the blood. The substance appears here as a research tool, not as a recommendation. Anyone taking it changes nothing about it on their own, and every adjustment belongs under medical supervision.

A group around Ding studied postmenopausal women not on hormone therapy, 359 with newly occurring type 2 diabetes and 359 controls, and repeated the analysis in men. Two gene variants of the SHBG gene additionally allowed a Mendelian randomisation.

Compared with the lowest quartile, the odds ratios in women were 0.16 (95 percent confidence interval 0.08 to 0.33), 0.04 (0.01 to 0.12) and 0.09 (0.03 to 0.21) in the three higher quartiles, with a trend of p less than 0.001. In the genetic analysis the predicted odds ratio per standard deviation of SHBG was 0.28 (0.13 to 0.58) in women.

For you this means: SHBG is more than a transporter. It is an informative metabolic marker. That is why I measure this value alongside when hormone symptoms and blood sugar topics are on the table together. That describes how I work and says nothing about whether a treatment at this point helps.

This paper investigated a different subject, namely the association between body mass index and breast cancer risk after the menopause. It appears here exclusively because of a methodological side finding and not because of its subject.

A collaboration around Key analysed individual data from eight prospective studies in postmenopausal women; data from 624 cases and 1669 controls were available.

When the influence of free estradiol was accounted for, the relative risk per five points of body mass index fell from 1.19 (95 percent confidence interval 1.05 to 1.34) to 1.02 (0.89 to 1.17). When SHBG was accounted for instead, the excess fell only moderately, and for androgens hardly anything changed.

For you this means: the free fraction can be biologically more informative than the total value. That is the only point this paper stands for here. On the development or treatment of breast cancer this article says nothing, it is not written for that.

A group around Schliep followed 259 healthy women between 18 and 44 years old over two menstrual cycles and recorded perceived stress daily in addition to a standardised questionnaire.

High versus low daily stress was associated with 9.5 percent lower estradiol (95 percent confidence interval minus 15.6 to minus 3.0), 10.4 percent lower free estradiol, 14.8 percent lower LH and 6.2 percent higher FSH. Luteal progesterone was 10.4 percent lower, with an interval whose lower bound at minus 0.10 percent practically touches zero. The adjusted odds ratio for a cycle without ovulation was 2.2 (1.0 to 4.7).

For you this means: the effect of everyday stress on the menstrual cycle is measurable in humans and can be named in size. It is an observational study, not a causal one, and some intervals touch one. The finding should not be made bigger than that.

A review by Gerard and Brown summarises the molecular regulation of aromatase in fat tissue, largely on the basis of cell and animal data.

It describes that obesity goes along with chronic inflammation of white adipose tissue and altered fat cell biology, and that inflammatory factors considerably influence estrogen signalling, above all through the expression of aromatase.

For you this means: the general statement holds. Fat tissue forms estrogens, and inflammation intervenes in that formation. Everything more specific comes from cell culture and animal models and has not been shown in humans in this form.

A group around O'Kelly followed 32 people whose oesophagus had been surgically removed more than twelve months earlier, for seven days with continuous glucose monitoring plus a food and symptom diary. 21,504 glucose values and 1276 meals were analysed.

226 of 1276 meals, that is 17.7 percent, were followed by a reactive hypoglycaemia event. Of these, only 19 were accompanied by symptoms, which is 8.4 percent of the events. Meals followed by an event were higher in carbohydrate (35.3 versus 31.7 g; p = 0.036), fibre (4.11 versus 3.15 g; p = 0.020) and sugar (12.65 versus 10.96 g; p = 0.048). Long gaps of more than three hours between meals and alcohol with the meal also went along with more events. For these two patterns the abstract gives no figure.

For you this means three things. Most glucose drops are not noticed at all. Long gaps between meals and alcohol with a meal went along with more events. And the authors' most important sentence comes at the end: neither nutrient composition nor eating patterns distinguished the symptomatic from the asymptomatic events. From these data one cannot therefore derive which eating behaviour avoids symptoms.

Very important alongside this: these people have a surgically altered anatomy. The numbers are not transferable to an unaltered digestive tract.

A group around Hall equipped participants with different metabolic states with glucose sensors, additionally measured insulin resistance and insulin secretion, and gave standardised test meals.

People who counted as normoglycaemic by the usual measurements spent 15 percent of the time in the prediabetes range and 2 percent in the diabetes range on continuous monitoring. Responses to identical meals differed considerably between individuals.

For you this means: the sensor shows you something real, and it shows it in people without any disease as well. Both sentences appear in the same paper. An excursion is a hint at a pattern, not proof of a disorder.

A group around Shukla had 15 people with prediabetes eat the same meal on three days, each time in a different order: once carbohydrates first and protein and vegetables ten minutes later, once protein and vegetables first and carbohydrates ten minutes later, once vegetables first.

The incremental glucose peaks in the two variants with protein and vegetables first were more than 40 percent lower than with carbohydrates first. The incremental area under the glucose curve with protein and vegetables first was 38.8 percent lower. With carbohydrates first there was a pronounced fluctuation, while in the other two variants the values stayed stable.

For you this means: it was the same foods and the same amount. Only the order was changed. Fifteen people with prediabetes, three test days, no long-term result. The finding is that big and no bigger.

A group around King gave eleven men with type 2 diabetes, on three mornings before breakfast and lunch, either 15 g of intact whey protein, 15 g of hydrolysed whey protein or a placebo.

After breakfast the glucose area under intact whey protein was 13 plus/minus 3 percent lower than under control (p less than 0.05). After lunch too, glycaemia was more favourable only under intact protein, and only there did participants report more satiety.

For you this means: in this study the effect was measurable and it was small. It was measured in eleven men with type 2 diabetes, not in women with hormone symptoms. The 15 grams are a study detail and not a recommendation, and a product name does not belong here.

A group around Reynolds compared, in 41 adults with type 2 diabetes, two weeks with 30 minutes of walking per day against two weeks with 10 minutes of walking after every main meal. Both variants meet current activity recommendations.

The area under the glucose curve was lower when walking after meals, with a ratio of geometric means of 0.88 (95 percent confidence interval 0.78 to 0.99). The difference was particularly clear after the evening meal, at 0.78 (0.67 to 0.91), meaning where the most carbohydrates were eaten and the most sitting happened.

For you this means: the same total time, distributed differently. That is the most practical single finding in the whole article. It was studied in people with type 2 diabetes.

A group around Buffey searched six databases up to August 2021 for randomised crossover studies comparing frequent short breaks in prolonged sitting through standing and through light walking against uninterrupted sitting. Seven studies entered the analysis.

Standing as a break lowered post-meal glucose by minus 0.31 (95 percent confidence interval minus 0.60 to minus 0.03; p less than 0.04), with no effect on insulin. Light walking lowered glucose by minus 0.72 (minus 1.03 to minus 0.41; p less than 0.001) and insulin by minus 0.83 (minus 1.18 to minus 0.48; p less than 0.001). Walking was also superior to standing alone.

For you this means: getting up is better than staying seated, and a few steps are better than getting up. The authors themselves stress that these are acute single-day studies and that the transfer into everyday life still has to be studied.