TMAO is a small, low-molecular-weight metabolite with an outsized reputation. It begins its life in the gut, where bacteria convert dietary precursors such as choline, phosphatidylcholine, betaine, and L-carnitine, abundant in red meat, eggs, dairy, and saltwater fish, into trimethylamine, or TMA. The liver then oxidizes TMA into TMAO via the enzyme flavin monooxygenase 3. From there, TMAO circulates in the blood and is excreted almost entirely by the kidneys, without undergoing any meaningful biotransformation along the way. Elevated plasma TMAO has been associated in numerous studies with cardiovascular disease, atherosclerosis, chronic kidney disease, and type 2 diabetes itself, and in diabetic patients specifically, higher TMAO has been tied to more cardiovascular events, faster loss of kidney function, and greater all-cause mortality.

The new analysis was prompted by a puzzling observation from a secondary analysis of the EMMY trial, in which patients treated with empagliflozin after a myocardial infarction showed a more pronounced rise in serum TMAO than those on placebo. That result raised an obvious question: does the same thing happen in people with type 2 diabetes, and does it extend to other drugs in the class, such as dapagliflozin? To find out, investigators combined individual patient data and stored samples from two prospective, randomised, placebo-controlled, double-blind crossover trials conducted at the University of Erlangen-Nuremberg, registered as NCT02471963 and NCT02383238. In each trial, patients were randomised after a run-in phase to receive either 25 mg empagliflozin or 10 mg dapagliflozin, or matching placebo, once daily for six weeks, followed by a one-week washout and a switch to the alternate treatment for another six weeks.

Fasting blood samples and 24-hour urine collections were obtained at baseline and at the end of both treatment periods, then stored at minus 80 degrees Celsius until analysis. TMAO concentrations were quantified with a validated liquid chromatography tandem mass spectrometry assay, using a deuterated internal standard and protein precipitation with methanol. After excluding a small number of samples that could not be quantified, the final analytical population comprised 123 patients, 69 from the empagliflozin trial and 54 from the dapagliflozin trial. The cohort had a mean age of 61.2 years, was 59 percent male, had a mean HbA1c of 6.7 percent, and a mean estimated glomerular filtration rate of 93.8 mL/min/1.73 m², with more than 70 percent of participants above 90. Compliance, assessed by tablet counts, averaged 99 percent in both trials.

The results were consistent in direction across the pooled population. Median plasma TMAO was 3.71 μmol/L at baseline and 3.60 μmol/L after placebo, but rose to 4.19 μmol/L after active treatment, a crossover difference of 0.36 μmol/L with a p-value of 0.006. The median intraindividual percentage change was 11.5 percent, significantly different from zero. When the drugs were analysed separately, the effect was clearer for dapagliflozin, where median TMAO climbed from 3.30 to 4.44 μmol/L, a 23.4 percent median increase, than for empagliflozin, which produced a numerically similar but statistically non-significant rise of 8.4 percent. The authors suggest the pattern points toward a possible class effect of SGLT-2 inhibitors, echoing the EMMY findings and extending them to a second drug and a different patient population.

Just as intriguing was what happened in the kidneys. Because TMAO is cleared almost exclusively by renal excretion, the researchers calculated its renal clearance from paired plasma and 24-hour urine measurements. Across all 123 patients, median renal clearance of TMAO fell from 95.8 mL/min on placebo to 83.9 mL/min on active drug, a statistically significant decline. In the dapagliflozin group the drop was significant on its own, falling from 109.2 to 97.8 mL/min, while the empagliflozin subgroup showed a non-significant downward trend. Yet the total amount of TMAO excreted in 24-hour urine barely changed, rising only non-significantly. That combination, lower clearance with unchanged total excretion, is a biochemical clue that the kidneys may not be the whole story.

The most obvious candidate explanation is the well-known initial dip in estimated glomerular filtration rate that occurs when SGLT-2 inhibitor therapy begins, a reversible haemodynamic effect of tubuloglomerular feedback. The study did observe a small but significant eGFR decline overall, and TMAO levels correlated negatively with eGFR at all time points, as expected from prior literature. But the eGFR dip was marginal, particularly in the dapagliflozin group, which showed the larger TMAO changes, and crucially, there was no correlation between individual patients’ eGFR changes and their TMAO changes. A reduced filtration rate alone, the authors conclude, cannot adequately explain the rise in plasma TMAO.

That leaves transporters and production. TMAO is eliminated primarily by glomerular filtration with a substantial contribution from transporter-mediated secretion in the proximal tubule, where the organic cation transporter 2 shuttles TMAO from blood into tubular cells and the multidrug and toxin extrusion protein 1 exports it into the urine. Animal work has shown that SGLT-2 inhibition broadly downregulates renal transport proteins, and drug interactions with OCT2 or MATE1 could theoretically reduce TMAO secretion. But if transporter-mediated excretion were truly impaired, the total urinary excretion of TMAO should have fallen, and it did not. The authors therefore favour a different hypothesis: SGLT-2 inhibitors may increase the production of TMAO, most plausibly through changes in the gut microbiome, even though some prior studies had suggested these drugs shift the microbiome in ways that should lower TMAO-producing bacteria. A subgroup analysis of the 20 participants who took antibiotics during the trials found no significant difference in TMAO changes, and the crossover design, in which each patient served as their own control, makes dietary shifts an unlikely confounder, although diets were not formally documented.

The clinical implications remain deliberately open. TMAO is widely characterised as a risk marker, and possibly a risk factor, for cardiovascular and kidney disease, so a drug-induced rise of roughly 10 to 20 percent would seem, on its face, unwelcome. Yet SGLT-2 inhibitors have delivered unmistakable benefits on cardiovascular outcomes and mortality in enormous outcome trials, in diabetic and non-diabetic patients alike, benefits that clearly operate through mechanisms beyond glucose control. Whether the TMAO increase is a harmless bystander, a counterbalancing signal, or something that modulates the drugs’ net benefit is unknown, and the authors call for studies examining outcomes in relation to baseline and on-treatment TMAO levels, as well as cohorts with much higher starting concentrations.

The study’s limitations temper any strong conclusions. Six weeks per treatment period is short, so long-term trajectories of TMAO, whether the rise plateaus, persists, or reverses, cannot be determined. No second baseline was established after the washout, so carryover effects may have blunted the measured effect in some patients. All participants had preserved kidney function, with eGFR above 60, leaving the question unanswered for the many diabetic patients with renal impairment, who typically have the highest TMAO levels. Compliance was assessed by counting returned tablets, and diet went unrecorded. Still, the pooled crossover data are internally consistent, and the message is clear enough to unsettle a comfortable assumption: a drug class celebrated for protecting the heart and kidneys may simultaneously raise circulating levels of one of the most talked-about microbiome-derived risk markers in cardiovascular medicine, and the reason why remains an open and compelling scientific question.

Subject of Research: Effects of SGLT-2 inhibitors on plasma TMAO concentrations and renal clearance in patients with type 2 diabetes

Article Title: Effects of SGLT‐2 Inhibitors on Plasma Concentrations and Renal Clearance of the Risk Marker Trimethylamine‐N‐Oxide in Patients With Type‐2 Diabetes

Article References: Effects of SGLT‐2 Inhibitors on Plasma Concentrations and Renal Clearance of the Risk Marker Trimethylamine‐N‐Oxide in Patients With Type‐2 Diabetes. (n.d.). https://doi.org/10.1002/edm2.70318

Image Credits: AI Generated

DOI: 10.1002/edm2.70318

Keywords: SGLT-2 inhibitors, TMAO, type 2 diabetes, empagliflozin, dapagliflozin, gut microbiome, renal clearance, cardiovascular risk, kidney function, eGFR, clinical trial, biomarkers

Researchers at Ankara Etlik City Hospital compared 224 patients with type 2 diabetes who had been taking either empagliflozin or dapagliflozin for at least six months against 106 age- and sex-matched patients with type 2 diabetes who had never used the drug class. The study, published in BMC Endocrine Disorders, tracked two principal outcomes: thyroid-stimulating hormone, or TSH, the pituitary messenger that regulates thyroid function, and the systemic immune-inflammation index, known as SII, a simple blood-derived marker calculated from platelet, neutrophil, and lymphocyte counts that is increasingly used as a low-cost gauge of systemic inflammation. Crucially, the investigators excluded anyone with known thyroid disease, thyroid autoantibody positivity, active inflammatory or autoimmune conditions, advanced organ failure, or ongoing immunomodulatory therapy, ensuring that the comparison was as clean as retrospective data allow.

The headline numbers tell a story of remarkable similarity across the groups. Median TSH was 1.6 micro-international units per milliliter in the empagliflozin group, 1.8 in the dapagliflozin group, and 1.9 in the control group, a difference that failed to reach statistical significance with a p-value of 0.338. The SII told the same tale: medians of 481, 531, and 468.5 respectively, with a p-value of 0.509. In plain terms, people on SGLT-2 inhibitors had thyroid function markers and inflammatory indices that were statistically indistinguishable from their peers on other diabetes regimens.

What lends this result weight is the study’s statistical diligence, which goes beyond the simple group comparison. The team ran correlation analyses, multivariate linear regression, multicollinearity diagnostics using variance inflation factors, and even a post-hoc power calculation, an unusually transparent step for a retrospective design. The power analysis showed the study had 80 percent ability to detect a Cohen’s f effect size of 0.17 or larger, yet the observed effect sizes for TSH and SII were dramatically smaller, at f of 0.02 and approximately 0.00 respectively. In other words, the study was adequately powered to find a meaningful signal, and the signal simply was not there, suggesting that if any true effect of SGLT-2 inhibitors on these markers exists, it is likely too small to matter clinically.

The regression analysis did, however, surface an unexpected protagonist. When the researchers modeled log-transformed TSH as the outcome, female sex emerged as an independent predictor, with a standardized coefficient of 0.161 and a p-value of 0.015, a finding consistent with the well-documented observation that thyroid parameters differ by sex. More intriguingly, estimated glomerular filtration rate, or eGFR, the standard measure of kidney function, independently predicted TSH levels with a coefficient of minus 0.0043 and a p-value of 0.037, meaning that as kidney function declined, TSH tended to rise regardless of which diabetes medications a patient was taking. SGLT-2 inhibitor use itself, along with SII, neutrophil-to-lymphocyte ratio, age, metformin use, and diabetes duration, all fell away as non-significant in the final model.

This renal-thyroid connection is more than a statistical footnote. The hypothalamic-pituitary-thyroid axis and the kidneys are intertwined in ways that clinicians sometimes underestimate. Declining renal clearance alters the metabolism and excretion of thyroid hormones, chronic kidney disease can produce a state resembling non-thyroidal illness, and conversely, hypothyroidism can impair renal hemodynamics. The finding that eGFR independently tracks with TSH in a population deliberately screened free of thyroid disease suggests that kidney function deserves closer attention when interpreting thyroid tests in patients with type 2 diabetes, a group in which diabetic nephropathy is common. The authors argue that this axis warrants further study independent of any drug exposure.

The mechanistic backdrop explains why the question was worth asking in the first place. SGLT-2 transporters are expressed not only in the kidney’s proximal tubule but also in extrarenal tissues, and some animal and small human studies had hinted at crosstalk between glucose-lowering agents and thyroid physiology. Meanwhile, the systemic immune-inflammation index and its cousin, the neutrophil-to-lymphocyte ratio, have become fashionable proxies for the chronic low-grade inflammation that underlies both insulin resistance and diabetic complications. Because SGLT-2 inhibitors appear to dampen inflammatory pathways in some experimental models, researchers suspected the drugs might measurably shift these indices in patients. The Turkish dataset, one of the largest head-to-head comparisons to date, found no evidence of such a shift; notably, SII and NLR were themselves strongly correlated in the sample, with a correlation coefficient of 0.467 and a p-value below 0.001, confirming that the inflammatory markers were behaving as expected even if the drugs did not move them.

For the millions of people taking empagliflozin or dapagliflozin, the practical message is one of reassurance. The study population, 330 adults in total, had used the drugs for at least six months, a duration long enough for any steady-state effect on thyroid physiology or immune-cell dynamics to manifest. The absence of significant differences in TSH or SII means there is currently no basis for routine thyroid monitoring specifically triggered by SGLT-2 inhibitor initiation in patients without preexisting thyroid conditions, and no signal that the drugs are provoking systemic inflammation. At the same time, the authors are careful to frame the result within the limits of its design: it was a single-center, retrospective, observational study, and it relied on TSH rather than the full panel of free thyroxine and triiodothyronine measurements that would capture the complete thyroid picture. Subclinical shifts in peripheral hormone conversion, for example, could theoretically escape a TSH-only lens.

Those caveats point directly at the research agenda the authors propose. Prospective, multicenter studies incorporating free thyroxine and triiodothyronine measurements, longer follow-up, and perhaps more granular inflammatory biomarkers such as interleukin-6 or C-reactive protein would settle the question with greater authority. Until then, this study stands as a well-powered negative result, a category of evidence that science needs more of. It trims the list of things clinicians must worry about with one of the most widely prescribed drug classes in modern medicine, while quietly spotlighting renal function as a variable that genuinely shapes thyroid readings. In a field saturated with hyped drug effects, a rigorous demonstration that nothing is happening can be its own kind of headline news.

The choice of the systemic immune-inflammation index as an outcome reflects a broader shift in diabetes research toward hematological composite markers. Because the index is derived from routine complete blood counts rather than specialized assays, it can be computed cheaply and repeatedly in ordinary clinical settings, which explains its growing popularity as a surrogate for the inflammatory burden associated with insulin resistance, atherosclerosis, and microvascular complications. Its strong correlation with the neutrophil-to-lymphocyte ratio in this cohort, with a correlation coefficient of 0.467, indicates that both markers captured overlapping aspects of the same underlying inflammatory physiology, lending internal coherence to the null finding.

The thyroid measurements were performed using electrochemiluminescence immunoassay, a widely adopted platform for thyrotropin detection that offers high analytical sensitivity at the low concentrations relevant to suppression and subclinical disease. This matters for interpretation, because a null result obtained with a precise assay is more convincing than one obtained with older, noisier methods. The reliance on a single TSH measurement per patient, however, means that within-person biological variation, which can be substantial for thyrotropin, was not accounted for, a limitation inherent to the retrospective design.

The composition of the cohort also deserves note. Patients taking empagliflozin outnumbered those on dapagliflozin by 139 to 85, and both active groups were pooled for some analyses rather than analyzed as separate exposures, since the two agents share the same mechanism of glucose excretion through urinary inhibition of the sodium-glucose cotransporter. Metformin use was deliberately included as a covariate because the drug is nearly ubiquitous in type 2 diabetes and has itself been anecdotally linked to thyroid parameters, yet it showed no independent association with thyrotropin in the final regression model.

The sex difference in thyrotropin, with female sex independently predicting higher values, aligns with decades of epidemiological data showing that thyroid dysfunction and mildly elevated thyrotropin are more prevalent in women, particularly with advancing age. Its emergence in a cohort screened free of thyroid disease suggests that even within the normal reference range, physiological sex-based variation persists and should be considered when clinicians interpret borderline results.

Finally, the ethics approval from the Ankara Etlik City Hospital committee and the waiver of individual consent, justified by anonymized extraction from the hospital information system, illustrate the standard governance pathway for retrospective chart-based endocrine research in Türkiye. The open-access publication under a Creative Commons license ensures that this carefully powered negative result is freely available to the clinicians and researchers best positioned to build on it.

Subject of Research: Effects of long-term SGLT-2 inhibitor use on thyroid-stimulating hormone and systemic immune-inflammation index in patients with type 2 diabetes

Article Title: Effects of SGLT-2 inhibitors on thyroid-stimulating hormone and systemic immune-inflammation index in type 2 diabetes: a retrospective study

Article References: Bayram, S. M., Demirci, H., Üçgül, E., Canlar, Ş., Cinel, M., Menekşe, B., Kuşabbi, İ. A., & Çakal, E. (2026). Effects of SGLT-2 inhibitors on thyroid-stimulating hormone and systemic immune-inflammation index in type 2 diabetes: a retrospective study. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02483-9

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02483-9

Keywords: SGLT-2 inhibitors, empagliflozin, dapagliflozin, type 2 diabetes, TSH, systemic immune-inflammation index, neutrophil-to-lymphocyte ratio, thyroid function, estimated glomerular filtration rate, retrospective study, endocrinology, inflammation