Hydrochlorothiazide and chlorthalidone, a benzothiadiazide and a thiazide-like diuretic, respectively were introduced during 1957-59 and the former quickly became a mainstay in the treatment of hypertension. Its popularity at the time stemmed from its perceived clinical (anti-hypertensive) effectiveness, low cost and apparently better safety profiles compared to earlier diuretics (Au and Raisz, 1960; Conway and Lauwers, 1961; Laragh, 1962; Mizgala, 1965). Indeed, hydrochlorothiazide gained widespread popularity following a controlled trial for the management of hypertension published in 1970 (Author Anonymous, 1970). Importantly, over the period spanning the 1960s and 1970s, hydrochlorothiazide underwent continuous examination in many clinical trials, predominantly focused on controlling hypertension (Collins et al., 1990; Antonietta et al., 2022). Almost 20 years after its discovery, a pharmacokinetically dissimilar thiazide-like sulfonamide derivative of hydrochlorothiazide, i.e., chlorthalidone, emerged in the clinic (Riess et al., 1977; Chen and Chiou, 1992). Surprisingly, its efficacy in managing hypertension was first evaluated in 1979 (Author Anonymous, 1979) and several studies, decades later, consistently suggested that this and other thiazide-like diuretics may have a more favorable clinical profile than hydrochlorothiazide (Harrower et al., 1985; Ernst et al., 2006; Chalmers and Arima, 2010; Dorsch et al., 2011; Tziomalos et al., 2013; Liang et al., 2017; Khenhrani et al., 2023). Yet, hydrochlorothiazide prevailed, and still stands, as one of the most frequently prescribed medications in the United States, with a staggering ∼39 million prescriptions for this drug alone in 2021 (clincalc.com/DrugStats/Drugs/Hydrochlorothiazide).

Although there is little evidence that low doses of hydrochlorothiazide (12.5–25 mg daily) reduce the risk of heart attack, stroke or death (Messerli et al., 2011), higher doses have been proven effective in lowering blood pressure and improving cardiovascular outcomes in patients with hypertension. Early trials on small number of subjects during the late 1950s showed the benefits of higher doses, though they did not consider their metabolic effects (Beyer et al., 1957; Bayliss et al., 1958; Bunn, 1958; Freis et al., 1958; Laragh et al., 1958; Rochelle et al., 1958; Wilkins, 1958; Wilkins et al., 1958; Heinemann et al., 1959; Reinhardt, 1959; Laragh, 1967). As our understanding of hypertension grew in the 1980s and 1990s, hydrochlorothiazide remained a key treatment for hypertension (Wilhelmsen et al., 1981; Hebert et al., 1993; Moser and Hebert, 1996; Savage et al., 1998), even as newer drugs with fewer, if any, metabolic issues (e.g., ACE inhibitors, calcium channel blockers and β-blockers) began to replace it (Officers et al., 2002; Grossman and Messerli, 2006). Meanwhile, chlorthalidone was also effective in treating hypertension, as seen in the large ALLHAT trial (Elliott, 1996), which compared different blood pressure medications. Despite some criticisms of this trial (McInnes, 2003; Hebert et al., 2007), its findings heavily influenced future treatment guidelines, promoting the use of thiazide diuretics (Chobanian et al., 2003; Ernst and Moser, 2009). However, these guidelines largely focused on hydrochlorothiazide, not chlorthalidone or other thiazide-like drugs like indapamide (Stafford et al., 2010; Messerli and Bangalore, 2011). At this point, it is important to recognize that the term thiazide has often been used loosely to refer to hydrochlorothiazide, chlorthalidone and indapamide, despite their pharmacokinetic and pharmacodynamic differences between them (Kurtz, 2010). Over time, each of these diuretics has inherited the benefits and drawbacks of the most commonly prescribed and studied one, i.e., hydrochlorothiazide.

Indeed, early studies did suggest that (hydrochloro)thiazide diuretics might be linked to negative effects on glucose metabolism (Johnston and Cornish, 1959; Goldner et al., 1960; Zatuchni and Kordasz, 1961; Runyan, 1962; Author Anonymous, 1963; Anderson, 1966; Author Anonymous, 1971b; Hollenberg and Mickiewicz, 1989; Pollare et al., 1989; Lithell et al., 1990; Plavinik et al., 1992). However, not all research confirmed these findings, with some studies failing to show any such connections (Cornish et al., 1961; Runyan, 1962; Jackson and Nellen, 1966; Andersen and Persson, 1968; Chaudhury et al., 1968; Healy et al., 1970; Berglund et al., 1986; Grimm et al., 1996; Lakshman et al., 1999). The differences in results may be due to variations in study design, dosages, or the populations studied. At any rate, the metabolic effects of hydrochlorothiazide were considered mild or irrelevant from the standpoint of managing hypertension. More recently, however, research has focused on the use of hydrochlorothiazide, alone or in combination with other drugs, for treating hypertension in specific groups (Omboni et al., 2009), such as the elderly and those with obesity, MetS or T2D (Klauser et al., 1991; Middeke et al., 1997; Reisin et al., 1997; Maitland-van der Zee et al., 2005; Siegel et al., 2008; Cooper-DeHoff et al., 2010; Manrique et al., 2010; Gong et al., 2014; Brown et al., 2016; Huang et al., 2016; Georgianos and Agarwal, 2019). Interestingly, no substantial effects of hydrochlorothiazide or chlorthalidone on plasma insulin were reported in several of these and other trials (Klauser et al., 1991; Plavinik et al., 1992; Price et al., 2013). However, there has been less emphasis on thiazide-like diuretics, despite evidence that both chlorthalidone and indapamide may offer better metabolic outcomes compared to hydrochlorothiazide (Kostis et al., 1997; Black et al., 2008; Karnes et al., 2014; Singh et al., 2018). With apparently few exceptions (Jian-Liang et al., 2004), these drugs can reduce blood pressure with less impact on blood glucose and cholesterol levels (Liang et al., 2017). Despite this, many studies, especially those involving hydrochlorothiazide, have led to the widespread belief that diuretics invariably affect glucose metabolism, regardless of their class or specific characteristics.

The story of loop diuretics began in the early 1960s, when ethacrynic acid was found to increase urine production in both animals and humans (Beyer et al., 1962; Bernstein et al., 1965; Cannon et al., 1965). Ethacrynic acid became the first non-sulfonamide loop diuretic used in clinical settings, leading to the development of more powerful loop diuretics. In the mid-to-late 1960s, furosemide was synthetized and quickly gained popularity due to its strong diuretic effects, fast action and effectiveness, especially in treating heart failure, hypertension, edema and kidney failure (Ingram, 1964; Godwin and Gunton, 1965; Laragh et al., 1966; Stason et al., 1966; Davidov et al., 1967; Earley, 1967; Walker, 1967; Joynt and Morrin, 1968; Kirkendall and Stein, 1968; Cannon and Kilcoyne, 1969; Shanoff, 1969). By the 1970s, furosemide became one of the most commonly prescribed diuretics, with fewer undesired effects compared to earlier diuretics, including hydrochlorothiazide (Feit, 1971; Wertheimer et al., 1971; Cannon, 1972; Valmin and Hansen, 1975; Mahabir and Bacchus, 1976; Finnerty et al., 1977; Araoye et al., 1978; Coodley et al., 1979; Dettelbach and Bennett, 1979). Other loop diuretics with better bioavailability and longer-lasting effects, such as bumetanide, torsemide, azosemide and piretanide, were introduced around this time as well (Asbury et al., 1972; Murdoch and Auld, 1975; Hettiarachchi et al., 1977; Jayakumar and Puschett, 1977; Benet, 1979; Brater et al., 1979; Konecke, 1981; Whelton, 1981; Stroobandt et al., 1982; Halstenson and Matzke, 1983; McNabb et al., 1984; Ward and Heel, 1984; Clissold and Brogden, 1985; Car et al., 1988). Many of these diuretics are still in use today (Blose et al., 1995; Bagshaw et al., 2007; Carone et al., 2016; Mentz et al., 2016; Rahhal et al., 2019; Singh et al., 2023). Even after 40 years, loop diuretics remain a key treatment for conditions involving excess fluid retention, as supported by ongoing clinical trials (Blake, 1990; Kissling and Pickworth, 2014; Ozieranski et al., 2019; Eid et al., 2021; Verbrugge and Menon, 2022; Greene et al., 2023; Mentz et al., 2023; Cuthbert and Clark, 2024; Kapelios et al., 2024; Krim et al., 2024).

However, most clinical trials on loop diuretics over the past 50 years have primarily and understandably focused on how they affect edema and electrolyte balance, rather than their potential impact on glucose metabolism. As a result, there is limited evidence linking loop diuretics to metabolic issues, especially compared to hydrochlorothiazide. However, early on, loop diuretics seemed to inherit the perceived metabolic effects of hydrochlorothiazide (Toivonen and Mustala, 1966). This concern may have originated from a 1959 study that first raised the possibility of diuretics affecting glucose metabolism (Freis and Finnerty, 1959) based on the effects of hydrochlorothiazide. Although there are few direct studies connecting loop diuretics (such as furosemide) to metabolic problems (Lavender and McGill, 1974; Tasker and Mitchell-Heggs, 1976; Khaleeli and Wyman, 1978), isolated cases of glucose intolerance or diabetes in patients using furosemide have been reported. Nevertheless, one study in 1966 found that furosemide had little effect on glucose tolerance over 3 months in both healthy people and those with hypertension (Jackson and Nellen, 1966). Another study suggested that ethacrynic acid also had minimal effects on glucose levels in mildly hypertensive patients (Andersen and Persson, 1968). Yet, later reports documented some cases of glucose intolerance associated with furosemide (Coni et al., 1974; Cowley and Elkeles, 1978; Kobayakawa et al., 2003). On the other hand, short-term studies showed no significant impact on blood sugar levels from either furosemide or bumetanide in both healthy individuals and patients with T2D (Asbury et al., 1972; Kaldor et al., 1975). Similarly, studies in 1980 indicated that neither diuretic had a significant effect on insulin or glucagon secretion (Giugliano et al., 1980b; Luyckx et al., 1980), though furosemide did slightly alter insulin and glucagon responses without affecting glycemia (Giugliano et al., 1980a). In 1981, a study found that bumetanide even improved glucose tolerance, but furosemide did not (Robinson et al., 1981). Further research indicated that, unlike hydrochlorothiazide, bumetanide had no significant effect on insulin or other hormone levels in dog pancreas models (Hermansen et al., 1985). The introduction of piretanide in the 1980s also did not consistently affect glucose tolerance or insulin levels, though both piretanide and furosemide were linked to changes in cholesterol levels in hypertensive patients (Valimaki et al., 1983; Weidmann et al., 1983; Campbell et al., 1998). However, later studies did not confirm these findings consistently (Harno et al., 1985; Chaudhuri and Catania, 1988; Weidmann et al., 1993; Lind et al., 1995; van der Heijden et al., 1998). In fact, piretanide (Harno et al., 1985) and likely bumetanide (Harno et al., 1985) increased insulin secretion in humans.

Therefore, overall, it appears that the “diabetogenic” risks commonly associated with diuretics are more strongly linked to hydrochlorothiazide (Padwal and Laupacis, 2004; Stump et al., 2006) than other classes of diuretic or anti-hypertensive medications. Although “meta-analysis (97 comparisons across 95 trials) demonstrated a statistically significant but clinically unimportant increase in FPG [fasting plasma glucose]” (Hall et al., 2020), the impact of any diuretic on glucose homeostasis seems to depend on several factors, including the type of diuretic used and the specific metabolic context on which these diuretics are being studied (Grossman et al., 2011).

The process by which nutrients, particularly glucose, trigger insulin secretion from islet β-cells is complex and involves many signals (Di Fulvio et al., 2014). However, medical textbooks often oversimplify this process. Typically, the consensus mechanism is described as follows (see Figure 2): When glucose enters β-cells, it undergoes glycolysis, which raises intracellular ATP levels. This increase in ATP closes ATP-sensitive K⁺ channels (K_ATP channels), causing depolarization of the cell membrane. As a result, voltage-gated Ca²⁺ channels open, allowing Ca²⁺ to flow into the cell. The influx of Ca²⁺ triggers the release of insulin from the β-cells into the bloodstream. While this mechanism is important, it is incomplete (Henquin et al., 2009; Merrins and Kibbey, 2024). Indeed, Cl⁻ channels and Cl⁻ transporters also help regulate β-cell membrane potential and excitability, both crucial for insulin release (Best et al., 2010). In fact, recent studies have clearly defined the roles of some of these Cl⁻ channels in islet physiology (Crutzen et al., 2016; Kang et al., 2018; Stuhlmann et al., 2018; Di Fulvio et al., 2020) and importantly, some Cl⁻ transporters help maintain the intracellular Cl⁻ concentration ([Cl⁻]_i) above its predicted thermodynamic equilibrium, facilitating the movement of Cl⁻out of the cell and through Cl⁻ channels in an electrogenic manner. Notably, some of these Cl⁻ transporters in β-cells can be directly targeted by thiazide and loop diuretics (Di Fulvio and Aguilar-Bryan, 2019). In fact, hydrochlorothiazide (Hoskins and Jackson, 1978; Sandstrom et al., 1993; Kucharczyk et al., 2023), trichlormethiazide (Seltzer and Allen, 1969), hydroflumethiazide (Hermansen et al., 1985), bumetanide (Hermansen et al., 1985; Sandstrom, 1990), furosemide (Aynsley-Green and Alberti, 1973; Hermansen et al., 1986; Sandstrom and Sehlin, 1988c; Eberhardson et al., 2000) and indapamide (Hermansen et al., 1986) can all influence insulin secretory responses in vitro and in vivo in animal models. In addition, hydrochlorothiazide, bumetanide and furosemide were also consistently linked to altered blood glucose and impaired glucose tolerance in a variety of animal models (Foy, 1967; Weller and Borondy, 1967; Foy and Furman, 1969; Foy and Furman, 1971; Foy and Furman, 1972; Hoskins and Jackson, 1978;Papaccio and Esposito, 1987; Sandstrom, 1988; Sandstrom and Sehlin, 1988d; Sandstrom and Sehlin, 1988b; Ray et al., 1993; Sandstrom et al., 1993). Therefore, these data support the hypothesis that the metabolic effects associated with the use of thiazide, thiazide-like, loop-diuretics are related, at least in part, to direct or indirect effects on islet β-cell secretory function.

The kidneys produce and release glucose primarily through gluconeogenesis (Weber, 1961; Schoolwerth et al., 1988) (Figure 3A). In fact, the kidneys contribute ∼50% of the total glucose released into the systemic circulation under fasting conditions (Gerich et al., 2001). Moreover, increased renal glucose production is a possible contributor to the development of hyperglycemia in patients with insulin resistance and MetS (Legouis et al., 2022). Indeed, insulin regulates renal gluconeogenesis by influencing enzyme production or activity associated with the availability of gluconeogenic precursors (Cano, 2001), an influence anticipated to be diminished or impaired in individuals with MetS or obesity-related insulin resistance (Rebelos et al., 2024). Yet, it remains uncertain whether any individual component of MetS, either alone or in combination, affects the gluconeogenic capacity of the kidneys. Much less certain is the potential effects that diuretics may have on renal glucose production.

Nevertheless, studies performed ∼30 years ago have shown that furosemide and ethacrynic acid can inhibit mitochondrial electron transport in renal tissues (Manuel and Weiner, 1976; Orita et al., 1983) and that diuretics, in general, appear to have adverse effects on renal (and muscular) glycolysis and gluconeogenesis, at least in rodents (Jones and Landon, 1967; Yoshida et al., 1970; Klahr et al., 1971; Fulgraff et al., 1972; Cohen and Little, 1976; Vinay et al., 1987; Dimitriadis et al., 1988; Dimitriadis et al., 1993; Amores et al., 1994). More recently, a potential functional link between NKCC2 and renal glucose fate has been suggested. Indeed, fructose-bisphosphate aldolase B, an enzyme involved in both gluconeogenesis and glycolysis, and primarily located in the kidneys, liver and intestines, may bind to, sequester, and reduce the functional expression of NKCC2 (Benziane et al., 2007) in a manner dependent of fructose 1,6-bisphosphate (F1,6BP), the enzyme’s substrate. Moreover, fructose, once activated to fructose-1-phosphate, can also serve as a substrate for aldolase B and has been shown to increase NKCC2 functional expression in the kidney (Ares et al., 2019). Although it is unknown whether NKCC2 can modulate the enzymatic activity of aldolase B in tubular cells, or if loop diuretics in general directly influence this interaction, these findings suggest a complex regulatory relationship with potential clinical implications. On one hand, there appears to be a negative regulatory link between renal gluconeogenesis and NKCC2 function. On the other hand, dietary fructose consumption, a potential contributor to MetS (Reungjui et al., 2007), is linked to NKCC2 function.

Although the role of NKCC2 in renal handling of glucose remains poorly defined, NKCC2A^KO mice developed several aspects of MetS including increased basal glycemia, glucose intolerance and insulin resistance (Kelly et al., 2019), but not hypertension, at least when mice were young (Oppermann et al., 2007). In addition, these mice showed enhanced glucose responses to alanine (Kelly et al., 2019), a substrate almost exclusively converted into glucose in the liver (Stumvoll et al., 1998; Meyer et al., 2002a; Mithieux et al., 2004; Mutel et al., 2011; Alsahli and Gerich, 2017; Sasaki et al., 2017). While these findings suggest increased hepatic de novo gluconeogenesis, the gluconeogenic response of NKCC2A^KO mice to exogenous pyruvate, which is converted into glucose in the liver, kidneys and small intestines (Stumvoll et al., 1998; Meyer et al., 2002a; Mithieux et al., 2004) remained normal (Kelly et al., 2019). Therefore, these observations suggest that NKCC2A^KO mice might have compromised renal gluconeogenesis. Moreover, aged NKCC2A^KO male mice developed overweight and consumed excessive food and water indicating that, unsurprisingly, the kidneys and other organs may contribute to the impaired glucose homeostasis observed in NKCC2A^KO mice (Kelly et al., 2019). In that regard, NKCC2 has been detected in other organs at much lower levels than those found in the kidneys, including small intestines (Xue et al., 2009) and hypothalamic regions of the brain (Konopacka et al., 2015). Even though the specific roles that NKCC2 in these organs may have in glucose homeostasis remain unexplored, the potential relevance of extrarenal NKCC2 is underlined by the following: i) the gluconeogenic capacity (Watford, 2005) of the small intestine supplies circulating glucose (Penhoat et al., 2014) and prevents obesity-related hepatic steatosis (Vily-Petit et al., 2020), ii) the hypothalamus plays a central role in endocrine integration of fuel homeostasis, control of water/energy intake and feeding behavior (Schwartz et al., 2000; Coll et al., 2007; Begg and Woods, 2013), and iii) as it has been known for a long time, diet and food intake affect renal gluconeogenesis and water balance (Author Anonymous, 1971a). Therefore, the metabolic phenotype of NKCC2A^KO mice likely stems from complex, age-dependent and long-term functional interactions between the brain, pancreatic islets, kidneys and intestines as well as other tissues where NKCC2 may be expressed, even in minimal quantities relative to the kidneys.

The kidneys utilize ∼10% of the total glucose used by the body in a daily basis, filtering 180 g of glucose per day, which is then almost entirely brought back into circulation (Ross et al., 1986; Alsahli and Gerich, 2017). Glucose is actively reabsorbed in the proximal convoluted tubule via the Na⁺-glucose transporter 2 (SGLT2), which couples the transport of the sugar with that of Na⁺ following its electrochemical gradient created by the Na⁺/K⁺ ATPase on the basolateral membrane of the tubular cells (see Figure 3B). Once inside the tubular cell, glucose is transported across the basolateral membrane into the peritubular capillaries by GLUT2 to reach back the bloodstream (Kanai et al., 1994). Importantly, SGLT2 is targeted by a class of highly efficacious drugs known as gliflozins, which reduce renal glucose reabsorption, thereby aiding in the management of glycemia and improving cardiovascular and metabolic health (Teo et al., 2021; Matthews, 2024). Notably, there has long been awareness that at least two loop diuretics, i.e., furosemide and ethacrynic acid can moderately decrease glucose reabsorption in the proximal tubule (Bowman et al., 1973; Arruda et al., 1975; Boonjarern et al., 1977; Wen et al., 1978). However, the potential of loop diuretics (or thiazide and thiazide-like diuretics) to promote glycosuria through this or any mechanism remains uncertain. It is worth noting that SGLT2 inhibitors not only enhance glycemic control but also reduce hypertension and mitigate MetS in animal models co-administered with furosemide or hydrochlorothiazide (Rahman et al., 2016), as well as in clinical settings involving patients with chronic heart failure (Grodin and Tang, 2020; Ibrahim et al., 2020).

Hepatic gluconeogenesis is a highly regulated process that serves as a backup for synthesizing glucose and glycogen from non-sugar sources (Zhang et al., 2018). Like the liver, muscle cells store glucose as glycogen. However, muscle glycogen is used locally for energy rather than being released into the circulation. During muscle activity, for instance, glycogen is broken down into glucose-6-phosphate for ATP production through glycolysis. This process can occur either aerobically or anaerobically, the latter leading to lactate production and release. Muscle-derived lactic acid is converted into alanine, transported to the liver, converted back to lactic acid and then used in de novo gluconeogenesis to synthesize glucose (see Figure 1). Glucagon effectively stimulates gluconeogenesis from amino acids and other non-carbohydrate substrates in the liver, but not in muscle, while insulin has the opposite effect, i.e., it inhibits hepatic glucose production and release (Puigserver et al., 2003; Adeva-Andany et al., 2019). Importantly, hepatic gluconeogenesis produces glucose-6-phosphate, which together with that produced from glycogen degradation (glycogenolysis) must be hydrolyzed by glucose-6-phosphatase in the endoplasmic reticulum to be released as glucose into the circulation (Cahill et al., 1959). Therefore, tissue glucose-6-phosphatase plays a major role in the maintenance of glycemia, particularly under fasting conditions.

Very little is understood about the metabolic effects that thiazides, thiazide-like and loop diuretics may have in hepatic glucose production and/or degradation. Nonetheless, early evidence did suggest that mechanisms sensitive to loop diuretics, possibly involving NKCC1 and/or KCCs, may contribute to the phosphorylation of numerous protein substrates in the liver (Lang et al., 1998). Among these proteins, is the serum- and glucocorticoid-dependent kinase (Waldegger et al., 1997), which is now recognized for its role in promoting hepatic insulin resistance (Zhou et al., 2021). Although this kinase was shown to regulate plasma membrane trafficking of NKCC2 in vitro (Fillon et al., 2001), the specific role of loop diuretics in developing hepatic insulin resistance remains unclear. It has been suggested that loop diuretics might contribute to insulin resistance in the liver (Schliess et al., 2001) and as such contribute to increased hepatic gluconeogenesis, while thiazides may exacerbate insulin resistance in general (Ramsay et al., 1992; Eriksson et al., 2008). Despite these findings, our current knowledge about the overall impact of diuretics on hepatic gluconeogenesis related to insulin resistance remains very limited.

Also poorly understood is the potential relationship that may exist between hepatocyte swelling in response to amino acids, the obligatory KCC-dependent K⁺/Cl⁻ extrusion, the resulting reduction in [Cl⁻]_i and glycogen synthesis via activation of the Cl⁻-dependent enzyme glycogen synthase phosphatase (Meijer et al., 1992). As Cl⁻ions can directly inhibit this enzyme (Meijer et al., 1992) as well as glucose-6-phosphatase (Pederson et al., 1998) one would expect that changes in [Cl⁻]_i may inversely correlate with glycogen biosynthesis or glucose production. However, like β-cells, the likely redundancy of diuretic-sensitive mechanisms involved in the regulation of hepatocyte [Cl⁻]_i makes it challenging to study the role of loop diuretics on hepatic glucose metabolism. In addition, the potential of hydrochlorothiazide to indirectly impair the function of pyruvate carboxylase by inhibiting carbonic anhydrase Vb (Kucharczyk et al., 2023) may have widespread physiological implications; this enzyme is widely distributed and plays an essential role in de novo gluconeogenesis and lipogenesis (Jitrapakdee and Wallace, 1999).

In comparison, virtually nothing is known about the metabolic effects that diuretics may directly have on muscle glucose homeostasis.