Good Guys, Bad Guys, Guesses, and Near Misses in Nephrology

• FGF23
• phosphate
• SGLT2 inhibitors
• CKD
• clinical trials
• nephrology
• diuretics
• Research Personnel
• SLC5A2 protein
• human
• Sodium-Glucose Transporter 2
• Renal Insufficiency, Chronic
• Kidney Tubules
• Proximal
• Hypoglycemic Agents
• Biotechnology
• Glucose

From seemingly out of nowhere, inhibitors of sodium-glucose cotransporter 2 (SGLT2) stand poised to revolutionize the care of patients with CKD. No clinical practitioner, scientific investigator, pharmaceutical developer, or biotechnology investor could have honestly predicted that simply blocking glucose reabsorption in the proximal tubule in patients with diabetes would translate into long-term improvements in kidney function and startling reductions in cardiovascular events and mortality (1,2). No one could have expected that a drug class with relatively weak diuretic and antihyperglycemic effects could have such outsized benefits on clinical outcomes. If SGLT2 inhibitors prove to enhance clinical outcomes in nondiabetic CKD, it would cement the concept that these agents fundamentally alter critical aspects of kidney pathophysiology that were previously underappreciated (3). Although exact mechanisms are presently unknown, large-scale clinical outcomes trials clearly inform us that SGLT2 inhibitors are “good guys.”

In contrast, alterations in calcium and phosphate homeostasis are considered among the most notorious of “bad guys” in CKD. As kidney function declines, serum concentrations of fibroblast growth factor 23 (FGF23), parathyroid hormone (PTH), and phosphate rise, whereas serum 1,25-dihydroxyvitamin D and calcium fall (4). On their own and in aggregate, these changes have been implicated as factors that contribute to bone disease, kidney disease progression, anemia, pruritus, arterial calcification, left ventricular hypertrophy, heart failure, and death (4). On the basis of an abundance of association data, guidelines were written and updated. Their overarching message: clinicians should strive to push the biochemical parameters of mineral metabolism toward their normal ranges to mitigate adverse effects. The implicit corollary: factors that push the numbers in the wrong direction are to be avoided.

In this edition of the Clinical Journal of the American Society of Nephrology, de Jong et al. (5) present a secondary analysis of a randomized, double-blinded, placebo-controlled trial of the SGLT2 inhibitor dapagliflozin, in 31 patients with type 2 diabetes complicated by CKD stages 2–4 and albuminuria. Dapagliflozin or placebo was initially administered for 6 weeks. This was followed by a 6-week washout period, after which all patients crossed over to the other treatment. The authors collected blood and urine at the onset and end of each 6-week treatment period.

Compared with placebo, 6 weeks of dapagliflozin treatment increased serum phosphate by 9%, FGF23 by 19%, and PTH by 16%, and decreased 1,25-dihydroxyvitamin D by 12%, whereas tubular reabsorption of phosphate and serum calcium did not significantly differ. Dapagliflozin induced a 4 ml/min per 1.73 m² decrease in eGFR, but this reduction did not correlate with any of the concomitant changes in mineral metabolism, suggesting GFR-independent effects.

These results in patients with mild-to-moderate diabetic CKD are similar to those from a previous study of healthy volunteers in whom canagliflozin increased serum phosphate by 16%, FGF23 by 20%, and PTH by 25%, and decreased 1,25-dihydroxyvitamin D by 10%, compared with controls within 24–72 hours of initiating therapy (6). Unlike the dapagliflozin study, the increase in serum phosphate in response to canagliflozin coincided with a significant increase in tubular reabsorption of phosphate. This minor difference likely relates to the earlier and more frequent data collection in the canagliflozin study. Taken together, the pair of studies suggest a class effect of SGLT2 inhibitors on mineral metabolism.

What accounts for these effects? In a complex endocrine system characterized by multiple interacting negative feedback loops, a systematic consideration of all possibilities is needed to decipher the singular mechanistic root cause of the constellation of increased serum phosphate, FGF23, and PTH, with reduced 1,25-dihydroxyvitamin D and normal serum calcium. A primary increase in FGF23 is ruled out by lack of hypophosphatemia. A primary increase in PTH is ruled out by lack of an increase in serum calcium, decreased rather than increased 1,25-dihydroxyvitamin D, and increased rather than decreased serum phosphate. A primary decrease in 1,25-dihydroxyvitamin D is ruled out by elevated rather than depressed FGF23. A primary reduction in klotho expression or function (which could not be measured) that results in increased serum phosphate because of renal resistance to FGF23 is ruled out by low rather than high 1,25-dihydroxyvitamin D and increased rather than decreased tubular reabsorption of phosphate, at least in the canagliflozin study. That leaves a primary increase in serum phosphate as the root cause to explain all of the other downstream findings.

As proposed by the authors of the dapagliflozin and canagliflozin studies, their data suggest that the primary perturbation in mineral metabolism effectuated by SGLT2 inhibitors is a consequence of a role reversal driven by the proximal tubule’s appetite for sodium: increased phosphate-dependent sodium reabsorption due to decreased glucose-dependent sodium reabsorption. The resulting increase in phosphate reabsorption would be expected to raise FGF23 through a mechanism that is similar to how dietary phosphate loading stimulates FGF23 (although the mechanisms of phosphate-sensing that elicit the FGF23 response are unknown). High FGF23 accounts for suppressed 1,25-dihydroxyvitamin D, which increases PTH, both directly, via reduced feedback inhibition, and indirectly, by threatening serum calcium, which is reflected in the decreased 24-hour urinary calcium excretion in the canagliflozin study that suggests less dietary calcium absorption. Lack of change in serum calcium in either study is not contradictory; it reflects intact PTH function to tightly regulate serum calcium.

If compensatory increases in proximal tubular sodium reabsorption via sodium-phosphate cotransporters underlie the alterations in mineral metabolism induced by SGLT2 inhibitors, the effects might be expected to be exaggerated in patients with diabetes in whom an increase in luminal SGLT2 expression in response to chronic hyperglycemia substantially increases the SGLT2 contribution to total sodium reabsorption (7,8). However, contrary to this hypothesis, the results of the dapagliflozin-in-diabetes and canagliflozin-in-health studies were strikingly similar. Perhaps reduced GFR among participants in the dapagliflozin study decreased the filtered load of glucose, which offset the effect of hyperglycemia to increase it. Additional studies should investigate whether SGLT2 inhibitors induce larger excursions in phosphate homeostasis among patients with diabetes and normal GFR, among those with poorly controlled glycemia, and among those with hyperfiltration due to early diabetic nephropathy. If the phosphate-elevating effects are more severe in these settings, it might be worthwhile for clinicians to first control glycemia with alternate antihyperglycemic agents before initiating SGLT2 inhibitors in certain subgroups of patients. Indeed, in a study of patients with diabetes already on metformin, the addition of dapagliflozin therapy did not change serum phosphate, PTH, or bone mineral density after 50 weeks of treatment (9).

Additional studies are also needed to define the long-term clinical consequences of SGLT2 inhibitors on mineral metabolism. Among their adverse effects are decreased bone density and increased incidence of fractures (10). Whether changes in calcium-phosphate homeostasis underlie these effects and whether they represent compound-specific or class effects require further investigation. In large studies of SGLT2 inhibitors in patients with diabetes, the initial decrease in eGFR eventually stabilizes and leads to more preserved eGFR over the long-term when compared with controls (1,2). The long-term improvement in eGFR trajectory with extended SGLT2 inhibitor therapy may ultimately translate into favorable effects on calcium-phosphate homeostasis.

What broader lessons can be learned from this important study? Deficiency of definitive outcomes trials creates a void into which questionable connect-the-dots guesswork invades. Familiar biochemical measures monitored in clinical practice become surrogate canvases onto which we project assumed effects of treatments that have not been subject to detailed investigation in outcomes trials. We hope to avoid bad guys—reduced GFR, hyperkalemia, anemia, hyperphosphatemia, and hyperparathyroidism, to name a few—and can easily fall into the trap of assuming that therapies that nudge our trusted surrogates in the wrong directions will have harmful effects on long-term outcomes and those that shift the surrogates in the opposite direction will be beneficial. In reality, treatments have multiple effects, some good, others not. Some we can predict on the basis of known physiology and others we cannot. SGLT2 inhibitors are a case in point. The ultimate arbiter of the net accounting of these pluses and minuses on a treatment’s ledger must be clinical outcomes trials.

In the case of SGLT2 inhibitors, what if their seemingly adverse effects on mineral metabolism reported by de Jong et al. (5) emerged before the results of the clinical outcomes trials? Would we have abandoned SGLT2 inhibitors on the basis of a phosphate–FGF23–PTH–1,25-dihydroxyvitamin D safety signal? Consider the epic scale of the public health blunder this would have entailed. This cautionary near-miss should inform us to no longer rely on or accept oversimplified, physiology-driven, good guy/bad guy guesswork, regardless of how well intended and well informed it may be. Instead, we must recall that angiotensin-converting enzyme inhibitors and calcineurin inhibitors lower GFR and raise serum potassium but deliver life-changing benefits to millions. At the other end of the spectrum, erythropoietin-stimulating agents raise hemoglobin and bardoxolone improves GFR, yet each can induce significant harm. As for SGLT2 inhibitors, they save lives despite causing glycosuria and raising serum phosphate, FGF23, and PTH, as well as lowering 1,25-dihydroxyvitamin D. They remind us that we must be prepared with an open mind for the next revolutionary treatment around the corner, and must demand its testing by clinical outcomes trial.

• Copyright © 2019 by the American Society of Nephrology