Short-term Metformin Protects Against Glucocorticoid-Induced Toxicity in Healthy Individuals: A Randomized, Double-Blind, Placebo-Controlled Trial Free

Glucocorticoids (GCs) are widely prescribed immunosuppressants used for a broad range of medical conditions. Approximately 3% of the general population receives GCs, and the number of long-term prescriptions is increasing (1). Despite their strong therapeutic effects, GCs have numerous harmful metabolic side effects, such as type 2 diabetes, dyslipidemia, osteoporosis, and myopathy (1,2). These metabolic effects limit GC use and increase morbidity and mortality (1). Despite recognizing the risks of GC-induced side effects, we lack treatment strategies to overcome GC-induced metabolic side effects (1,2).

A promising candidate for managing GC-induced side effects is metformin (3). Metformin is a widely prescribed and safe medication to lower glucose levels in patients with type 2 diabetes (4). However, its exact mechanism of action remains unknown. Although the liver is considered the primary site of metformin action, recent findings suggest its effects extend to various other tissues. Consequently, metformin has attracted interest beyond diabetes treatment and is currently being investigated as an anticancer and antiaging drug (4). Therefore, understanding the full scope of metformin action is of significant interest across many medical indications.

We previously demonstrated in a proof-of-concept study that metformin can prevent glycemic deterioration during GC therapy in patients without diabetes with various diseases (5). However, the heterogeneity of the underlying conditions posed challenges in thoroughly assessing the potential of metformin during GC excess. Therefore, we conducted a randomized, placebo-controlled study of healthy individuals receiving high-dose GCs to address this limitation.

By integrating clinical data with metabolomic and genetic insights, we had three objectives: validate our previous findings on glucose control, explore the impact of metformin on tissues beyond the liver, and gain deeper insight into the underlying mechanisms of metformin action.

This randomized, placebo-controlled, double-blind, crossover study was conducted at the University Hospital Basel, Basel, Switzerland. The Ethics Commission of Northwestern and Central Switzerland approved the study, which is registered at ClinicalTrials.gov (identifier NCT04659915). The study was conducted following the ethical guidelines of the Declaration of Helsinki. Written consent was obtained from participants after fully explaining to them the purpose and nature of all study procedures.

Male adults aged 18–40 years with a normal weight (BMI range, 18.5–25 kg/m²) were eligible for the trial. Essential exclusion criteria were any significant ongoing diseases; medications, including GCs or metformin until 4 weeks before inclusion; regular alcohol intake; and intensive physical activity (>4 h/week). Participants were recruited from the general population.

According to a predefined randomization list created by an independent investigator not involved in the study, participants were assigned to receive either metformin or placebo first. The treatment sequence was 1:1. Only the independent investigator, the Hospital Pharmacy of the University Hospital, and the principal investigator had access to the code, which was kept in sealed envelopes. The randomization sequence was unknown to the participants and all study personnel involved in the study.

After a screening visit, all participants received prednisone 30 mg/day for two 7-day periods separated by a 28-day washout period. In addition, all participants received metformin during one study period and a placebo during the other. Metformin was initiated at a dose of 500 mg/day orally. Metformin was increased by 500 mg every other day until the maximum amount of 2,000 mg was reached on the seventh day of the study period. Identical placebo tablets were administered during the other study period. The compliance was calculated based on the number of returned tablets. All investigators and study participants were blinded to the treatment sequence until the completion of the study.

All participants had a screening visit followed by four study visits. The visits took place before and after every treatment period. At each visit, the participants were equipped with a portable Holter electrocardiogram monitor the evening before to measure heart rate variability. After an overnight fast, participants attended the research facility at 8 a.m. Weight and blood pressure were measured. Basal metabolic rate was measured with indirect calorimetry (Quark RMR; Cosmed) for 20 min. After fasting blood samples were taken, a standardized mixed-meal tolerance test (MMTT; 300 mL of Ensure Plus; Abbott) was conducted, and blood was frequently sampled over 120 min.

Plasma glucose level was measured with an enzymatic assay (Cobas modular analyzer, Roche Diagnostics). Serum insulin was analyzed by ELISA (Insulin ELISA; Mercodia). C-peptide concentration was measured with a C-peptide ELISA (Mercodia). The HOMA for insulin resistance (HOMA-IR) index was calculated according to Matthews et al. (6). The Matsuda index was used to calculate whole-body insulin sensitivity (7). The insulinogenic index, a marker of β-cell function, was calculated as follows: [insulin_30 min − insulin_fasting]/[glucose_30 min − glucose_fasting] (8). Incremental area under the curve (AUC) levels over 120 min during the MMTT for insulin, C-peptide, and glucose were calculated based on the trapezoidal rule.

Lipids were measured with an enzymatic assay (Cobas; Roche Diagnostics). Immunoassays were used to measure thyroid-stimulating hormone (Elecsys TSH, Cobas; Roche Diagnostics), GDF-15 (R-Plex; Mesoscale), and total glucagon-like peptide 1 (GLP-1) (GLP1 total ELISA; Mercodia). C-terminal telopeptide (CTX), intact amino terminal propeptide of type I procollagen, and osteocalcin were analyzed on the IDS-iSYS Multi-Discipline automated analyzer (IDS iSYS; Immunodiagnostics Systems, Boldon, U.K.). The assays are based on chemiluminescence technology. Adiponectin and leptin levels were assessed as exploratory end points using electrochemiluminesence immunoassays (U-Plex Human Adiponectin Assay and V-plex Human Leptin Assay; Meso Scale Discovery). Parathyroid hormone (PTH) and 25-hydroxy vitamin D₃ levels were assessed as exploratory end points using electrochemiluminescence (Elecsys PTH and Elecsys Vitamin D total III; Cobas).

Heart rate variability (HRV) was measured to determine the autonomic nervous system tone. HRV was analyzed using the cloud-based Cardiomatics software (Cracow, Poland), which provides artificial intelligence–based automatic analysis of Holter monitor electrocardiogram recordings (9). HRV data were collected overnight from 0000 h to 0500 h while all participants were asleep, followed by an additional 30-min period directly after awakening. Wake-up reactions for changes in average heart rate, high-frequency power, low-frequency power, and root mean square of successive differences (rMSSD) were calculated based on the observed differences between the awake and asleep periods.

For metabolomic analysis, blood was sampled in the fasting state and 15 min after the intake of the liquid mixed meal at the end of each treatment period. Nontargeted metabolomics identification was performed by flow injection analysis at ETH Zürich, as detailed by Fuhrer et al. (10). In brief, metabolites were extracted using the Hamilton STAR M methanol protein precipitation protocol, separated into technical duplicates, and analyzed by flow injection analysis on the Agilent Q-TOF 6550 instrument. Raw data acquired from the instrumentation were processed using in-house MATLAB software. This software is optimized to detect metabolites present in quantities close to the background through the use of wavelet transformations to accurately identify metabolite centroids while denoising the ion signal by removal of background wavelet components. Ions with a count of <500 across all scans were filtered out to avoid including rare features that were too infrequently observed to be statistically meaningful. The metabolite centroids identified in the complete set of samples are then aggregated into a single matrix. Centroid masses were binned within the tolerance limit of the instrument (∼0.002 amu at a charge to mass ratio [m/z] of 300), and the weighted average m/z was calculated for each bin. This facilitates metabolite identification by using information across all samples to derive an accurate m/z for each detected ion. Metabolite annotation of these ions was performed based on the m/z ratio, with mass accuracy of 0.001 Da, with reference to an in-house curated set of metabolites initially derived from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

Differential analyses were performed on normalized values of the curated set of metabolites, and differential fold changes in ion intensities were calculated. Significantly differentially expressed metabolites were identified using a Student t test, and the resulting P values were corrected for multiple testing across the curated metabolite set, using the Benjamini–Hochberg procedure. Pathway enrichment analysis was then performed on the differentially expressed metabolites using fiaMiner, which uses internally curated human metabolome database–based metabolomics pathways.

In a subgroup of 10 participants, subcutaneous adipose tissue biopsy samples were taken at the end of each treatment period. Biopsies were performed after the MMTT. A local anesthesia was administered, a trained surgeon removed ∼100–250 mg of subcutaneous adipose tissue from the periumbilical region. The samples were immediately frozen on dry ice and stored at −80°C.

Total RNA was extracted from adipose tissue using Trizol reagent (Geneaid, catalog no. GZR200) according to the manufacturer’s instructions, and extracted RNA was purified with columns (QIAGEN, catalog no. 74104). RNA quality was measured by Tapestation. The mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. After fragmentation, the first-strand cDNA was synthesized using random hexamer primers, followed by second-strand cDNA synthesis. Libraries were finalized by end repair, A-tailing, adapter ligation, and size selection. The purified libraries were sequenced on Illumina platforms. Raw sequencing reads of all samples were processed and analyzed using the SUSHI framework developed by Functional Genomics Center Zurich (11). Adapter sequences and low-quality bases were trimmed off from the raw reads, using fastp, version 0.20, for quality control (12). Pseudo-alignment of the filtered reads was performed against the human reference genome assembly GRCh38.p13, and gene expression counts were quantified using Kallisto, version 0.46.1 (13). Differential gene expression was then performed on the gene count matrix, using the DESeq2 R package (version 1.40.2). The patient identification and BMI were incorporated as additional covariates alongside treatment into DESeq2's negative binomial generalized linear model. The genes were then read-count thresholded based on whether, in either of the treatment groups, at least 50% of samples had >10 identifiable reads for the gene.

To control for multiple testing, the P values for the remaining genes were adjusted using the Benjamin–Hochberg method. Differentially expressed genes, determined by the log2 fold change in gene expression between treatment groups, were identified based on the significance threshold of (adjusted) P < 0.05. For gene enrichment analysis, differentially expressed genes were imported into ConsensusPathDB (www.consensuspathdb.org). Pathways defined by the KEGG, REACTOME, HumanCyc, Wikipathways, and PharmGKB databases were selected in the online tool to discover overrepresented pathways from the differential gene expression list, and the P-value threshold was set at 0.05.

The primary end point was the change in insulin sensitivity between each study period. Secondary end points included changes in plasma glucose, insulin, C-peptide, GLP-1, growth/differentiation factor 15, and lipid levels; blood pressure; weight; body composition; energy expenditure; substrate utilization; and HRV. Additional analyses included changes in insulin secretion, bone turnover markers, metabolites in blood, and RNA expression in subcutaneous adipose tissue. Further exploratory analysis included changes in adiponectin and leptin levels. In addition, 25-hydroxy vitamin D₃ and PTH levels were measured.

The statistical analysis was performed according to a prespecified statistical analysis plan, finalized before any comparative analysis. The sample size was estimated based on the observed effect of metformin on insulin sensitivity (HOMA-IR index) in GC-treated patients (5). From this observed effect, it was calculated that 18 study participants would be required to detect an expected mean difference in the HOMA-IR index of 1.2 at 80% power (probability). This assumes a two-sided test at a significance level of 0.05 and an assumed SD of the HOMA-IR index of 1.4.

Participants who dropped out of the study and who did not receive the study drug were replaced. The full analysis set included all participants who received the study drug. The complete analysis set (CAS) comprised all participants who completed the study per protocol. The primary end point analysis was conducted in the full analysis set and repeated in the CAS. Secondary end point analyses were performed in the CAS.

We used a two-sided paired t test to analyze the primary outcome: change in insulin sensitivity (HOMA-IR index). To assess any impact of the crossover design, the primary end point was further analyzed with a linear model with treatment sequence, randomization, and their interaction term as fixed effects (Supplementary Table 1).

Missing values for the primary end point were imputed using the mean value for the specific treatment group (i.e., metformin or placebo). Missing values for the secondary end points were not imputed. Secondary end points were similarly analyzed using paired t tests. Categorical data were analyzed using the exact binomial test. The Benjamini–Hochberg method was applied to correct for multiple testing. Analysis was carried out using the R statistical software (version 4.3.1). Figures were created in GraphPad Prism, version 10.2.3 (GraphPad Inc., La Jolla, CA).

From 1 February to 31 August 2021, healthy, lean men were enrolled in the study at the University Hospital Basel, Switzerland. A total of 19 participants [mean age ± SD, 27.0 (±5.2) years, lean, and metabolically healthy (Table 1)] were randomly assigned to a study arm (Supplementary Fig. 1). One participant withdrew his consent before receiving any study medication. Of the 18 participants who received the study medication, one dropped out due to a bike accident during the washout period. Seventeen participants completed both study arms.

The results of our analyses are summarized in Table 2. During GC treatment, the glucose AUC increased significantly with placebo but remained stable with metformin treatment (Fig. 1A, Supplementary Fig. 2A and B). Similarly, insulin (Fig. 1B, Supplementary Fig. 2C and D) and C-peptide AUCs increased with placebo but remained stable with metformin. Consequently, whole-body insulin sensitivity (Matsuda index) improved with metformin compared with placebo (Fig. 1C). These findings remained significant after adjusting the P values. However, we did not identify a difference between the two treatment groups in the HOMA-IR index and insulinogenic index (Fig. 1D). There was no evidence for a carryover or a sequence effect for HOMA-IR (Supplementary Table 1, Supplementary Fig. 3). The regression analysis of the Matsuda index indicated the presence of a sequence effect (placebo → metformin) but no significant interaction effect between the treatment types (Supplementary Fig. 4). Fasting lipids remained relatively unchanged across both interventions, with no significant differences in total cholesterol, LDL cholesterol, HDL, and triglyceride levels.

During the MMTT, total GLP-1 AUC increased significantly with metformin compared with placebo (Fig. 1E, Supplementary Fig. 2E and F). There was no difference in fasting GDF-15 (Fig. 1F) or thyroid-stimulating hormone levels. The fasting CTX concentration increased with placebo but remained stable with metformin. However, this finding no longer reached significance after adjusting the P value. Intact amino terminal propeptide of type I procollagen and osteocalcin decreased with both treatments. PTH and 25-hydroxy vitamin D₃ concentrations were within the normal range at baseline (Table 1). No changes in adiponectin and leptin levels were observed (Supplementary Fig. 5). Body weight did not differ between the interventions; there was no difference in basal metabolic rate and the respiratory quotient.

Systolic and diastolic blood pressures remained similar between both study arms. Additionally, upon awakening, no differences were observed in the changes of average heart rate, high-frequency power, low-frequency power, or rMSSD, indicating no relevant change in autonomic nervous system tone.

Untargeted metabolomic analysis was performed of blood samples of each study participant in the fasting state and 15 min after ingesting a liquid mixed meal, before and after metformin and placebo exposure.

During fasting, a comparative analysis between metformin and placebo revealed differential expression of 12 metabolites (Fig. 2A, Supplementary Fig. 6). Notably, citrulline and ornithine, components of the urea cycle pathway, exhibited significant downregulation with metformin treatment. Other differentially expressed metabolites were associated with pathways involving fatty acids and bile acids.

In the postprandial state, 38 metabolites displayed differential expression with metformin (Fig. 2B, Supplementary Fig. 6). Eleven of these metabolites overlapped with those identified in the fasting state. Once again, citrulline and ornithine had the most significant impact.

Pathway enrichment analysis revealed that metformin upregulated the citric acid cycle during fasting. In the postprandial state, metformin downregulated bile acid biosynthesis and concurrently upregulated the flux of free fatty acids (Supplementary Fig. 7).

RNA sequencing was performed on subcutaneous white adipose tissue samples obtained from 10 participants at the end of each treatment period. Differential gene expression analysis conducted irrespective of fold change between metformin and placebo unveiled 80 genes exhibiting variations in adipose tissue. Nineteen of these 80 genes were differentially expressed above an absolute log2 fold change of 0.5 (Fig. 2C, Supplementary Table 2). These 19 genes were mainly involved in lipid metabolism (AACS, FADS1, FADS2, ELOVL6, SREBF1) and were downregulated with metformin. Other differentially expressed genes were involved in transcription regulation and signaling (JUNB, PRRT4, NR1D1, FOXN4, THRSP), cell cycle regulation, and proliferation (CCNG2, EPHB2).

Pathway analysis of differential gene expression further confirmed that metformin primarily affected cholesterol, steroid, and fatty acid pathways (Supplementary Fig. 8). Comparative analysis with data from a 6-week metformin study by Kulkarni et al. (14), which focused on older adults without diabetes and without GC therapy, identified shared enriched pathways predominantly linked to fatty acid metabolism (Supplementary Fig. 9).

Gastrointestinal symptoms, such as nausea and mild abdominal discomfort, were reported slightly more frequently in the metformin group than the placebo group (n = 9 with metformin compared with 6 with placebo). However, this difference was not statistically significant (P = 0.45). Moreover, the reported symptoms were generally mild, self-limiting, and did not require treatment discontinuation.

In this study, metformin prevented insulin resistance induced by GCs. Metformin enhanced whole-body insulin sensitivity, as assessed by the Matsuda index, by inhibiting the postprandial increase in glucose and insulin levels. This study underscores the role of metformin in preventing GC-induced diabetes (5).

The HOMA-IR index—primarily a marker of hepatic insulin sensitivity based on fasting values—remained unaffected by metformin (7). This is likely due to the distinct properties of GCs, which mainly induce postprandial insulin resistance (1). We observed no influence of GCs on β-cell function, as assessed through the insulinogenic index, and hence, no apparent impact of metformin. Nevertheless, effects on the HOMA-IR index and insulin secretion may manifest in patients with increased metabolic risk and more pronounced glucotoxicity (2). Although pretreatment with GCs may enhance the effect of metformin on whole-body insulin sensitivity, based on the sequencing effect observed in the Matsuda index, the absence of a significant interaction effect suggests that the overall magnitude of the treatment effect is not dependent on treatment sequence.

Our second objective was to investigate the extent of metformin action beyond glucose control. GCs can cause obesity and dyslipidemia by inducing both lipolysis and lipogenesis (2). Our metabolomic analysis revealed that metformin enhances the postprandial flow of free fatty acids. This effect may be attributed to increased tissue uptake and accelerated clearance of free fatty acids from the bloodstream. Metformin also downregulated key genes involved in fatty acid synthesis in adipose tissue, such as FADS1, FADS2, and ELOVL6, which are linked to type 2 diabetes risk (15–17). Thus, our findings highlight the potential of metformin in counteracting GC-induced dyslipidemia and obesity.

Next, we investigated the impact of metformin on bone health. We found that metformin preserved CTX, a marker of bone resorption, whereas CTX level increased with placebo, indicating reduced osteoclast activity with metformin. However, the difference in CTX level did not remain statistically significant after adjustment for multiple testing and, therefore, remains hypothesis-generating. Nonetheless, the observed statistical trend and previous observations in preclinical and human studies suggest an effect of metformin in preventing GC-induced bone loss (18–20).

Myopathy is another common side effect of GCs (21), resulting from increased protein breakdown and diminished protein synthesis (21). We found that citrulline and ornithine, two metabolites of the urea cycle pathway, were strongly downregulated, suggesting that protein breakdown is slowed with metformin. Similar metabolic shifts have been reported in patients with type 2 diabetes using metformin (19). Whether metformin can prevent muscle wasting and GC-induced myopathy needs to be further evaluated. Taken together, our study shows that the effects of metformin extend beyond glucose control, providing protective benefits to muscle, adipose, and, possibly, bone tissue.

Our third objective was to identify the underlying mechanisms of metformin action. The molecular mechanisms of metformin are still not fully elucidated, and several mechanisms have been proposed (4). Activation of AMPK is considered a key mechanism, but AMPK-independent mechanisms are also under discussion (4). GCs reduce AMPK activity, and preclinical studies have shown that metformin reverses these effects by activating AMPK (22–24). In human adipose tissue, metformin-induced AMPK activation has been linked to reduced fatty acid synthesis (25). Consistent with this, our genetic analysis of adipose tissue revealed a significant downregulation of pathways associated with fatty acid synthesis after metformin treatment. Furthermore, metformin suppressed the expression of genes like AACS,FADS1, and FADS2, which impair AMPK activation (16,26). These findings support the hypothesis that metformin counteracts the metabolic effects of GCs through AMPK activation in adipose tissue.

In recent years, AMPK-independent mechanisms of metformin have gained attention. Among these, GLP-1 has been increasingly recognized as a contributor to metformin action (4,27). Here, we observed a distinct elevation of GLP-1 after metformin administration, potentially due to alterations in the bile acid pool, a mechanism under investigation (28). Indeed, metformin reduced bile acid synthesis postprandially, which supports this hypothesis.

The impact of metformin on bile acids has received considerable attention because metformin was found to exert beneficial metabolic effects by changing the gut microbiome and secondary bile acid levels (29). In our study, metformin affected both liver-derived primary bile acids and gut-derived secondary bile acids, emphasizing the role of the gut in metformin action.

Growth differentiation factor 15, which has been implicated in the weight regulatory effects of metformin, was not affected in our study (30). This may be due to the relatively short duration of the study. In summary, our study shows that metformin acts through various pathways in both AMPK-dependent and independent manners, with tissue-specific mechanisms.

A limitation of our clinical trial is the short study duration, which restricts our ability to investigate direct clinical outcomes such as the development of type 2 diabetes. Whether the observed changes in surrogate parameters translate to preventing the toxic effects of GCs remains to be seen. However, given that our study population consisted of young and healthy individuals and the treatment duration was short, it is conceivable that the effect of metformin is more potent in a population at risk for metabolic disorders.

We cannot definitively establish whether the observed changes in white adipose tissue, bone, and muscle directly result from metformin action or indirectly from reduced insulin resistance. However, compared with studies of patients with type 2 diabetes, insulin resistance was mild and limited to the postprandial state. Additionally, comparing our genetic findings with data from a 6-week metformin study of adults without diabetes who were not receiving GC therapy identified several shared enriched pathways predominantly linked to fatty acid metabolism (14). These findings collectively indicate that the observed effects are largely independent of insulin resistance.

In conclusion, metformin prevents GC-induced insulin resistance and reduces markers of dyslipidemia, myopathy, and, possibly, bone resorption. This study highlights aspects of metformin beyond glucose control and demonstrates its impact on white adipose tissue, muscle, bone, and the gut. Additionally, we found that metformin activates both AMPK-dependent and -independent pathways. Thus, metformin could prevent GC-related toxicity, including diabetes mellitus, dyslipidemia, myopathy, and possibly osteoporosis. These insights underscore the therapeutic potential of metformin not only in managing GC side effects but also in broader clinical applications.

Our study provides robust data for designing a larger randomized-controlled trial to investigate the protective effects of metformin in patients during GC treatment. Such a trial should address long-term outcomes, including diabetes mellitus, dyslipidemia, muscle wasting, osteoporosis, and bone fractures.

Acknowledgments. The authors thank the team at Cardiomatics, especially Rafal Samborski and Nikola Fajkis-Zajaczkowska, for supporting the analysis of heart rate variability. The authors thank Lucia Seeger for her support in conducting the study. The authors thank Mirjam Christ-Crain, Ulrich Keller, and Fabian Meienberg for reviewing the manuscript. We used BioRender for creating the graphical abstract. During the course of preparing this work, the author(s) used ChatGPT for the purpose of improving the readability of the manuscript. Following the use of this tool/service, the authors formally reviewed the content for its accuracy and edited it as necessary. The authors take full responsibility for all the content of this publication.

Funding. E.S. was supported by the Swiss National Science Foundation (grant 193516), SwissLife Jubilee Foundation, Swiss Diabetes Foundation, and Novartis Foundation for Medical-Biological Research. N.Z. received support from the Strategic Focus Area “Personalized Health and Related Technologies” (grant 603) of the ETH Domain (Swiss Federal Institutes of Technology).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.T., C.J.P., and V.I. collected the data; M.E. conducted the statistical analysis; S.T., S.P., C.M., T.D., C.S.Z., M.K., M.E., A.O., N.Z., I.R., A.G., C.W., and E.S. contributed to the analysis and interpretation of the data; S.T., S.P., and E.S. did the literature search; E.S. designed and supervised the study and drafted the manuscript; S.T. and P.A.T. wrote the protocol, which E.S. edited; S.T., C.J.P., V.I., C.M., T.D., C.S.Z., M.K., M.E., A.O., N.Z., I.R., A.G., and C.W. revised the manuscript. E.S., S.T., and M.E. verified the data and had access to all raw data. All authors had final responsibility for the decision to submit for publication. E.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis

Prior Presentation. This study was presented in part at the European Association for the Study of Diabetes conference, Hamburg, Germany, 2–6 October 2023; the Swiss Society of Endocrinology and Diabetology Congress, Bern, Switzerland, 14–15 November 2023; and the European Congress of Endocrinology, Istanbul, Turkey, 13–16 May 2023.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Naveed Sattar.