Comparative effectiveness of sodium-glucose cotransporter-2 inhibitors for recurrent nephrolithiasis among patients with pre-existing nephrolithiasis or gout: target trial emulation studies

Objective To emulate target trials comparing recurrence of nephrolithiasis among patients with pre-existing nephrolithiasis (overall and stratified by concomitant gout) initiating sodium-glucose cotransporter-2 (SGLT-2) inhibitors versus an active comparator.

Design Target trial emulation studies.

Setting Canadian population database, January 2014 to June 2022.

Participants 20 146 patients with nephrolithiasis and type 2 diabetes, including those with concomitant gout at baseline, a high risk group.

Interventions Initiation of an SGLT-2 inhibitor or glucagon-like peptide-1 (GLP-1) receptor agonist, with a dipeptidyl peptidase-4 (DPP-4) inhibitor as alternative comparator.

Main outcome measures The primary outcome was recurrent nephrolithiasis events ascertained from diagnoses during emergency department visits, hospital admissions, or outpatient visits. Secondary outcomes included nephrolithiasis resulting in hospital admission or emergency department visits and flare-up of gout, as well as a positive control outcome (genital infection) and negative control outcomes (osteoarthritis encounter and appendicitis). Poisson and Cox proportional hazards regression models were used (primary analyses), as well as overlap weighting.

Results After inverse probability of treatment weighting, 1924 recurrent nephrolithiasis events occurred among the 14 456 weighted patients who used an SGLT-2 inhibitor (105.3 per 1000 person years), compared with 853 events among the 5877 weighted patients who used a GLP-1 receptor agonist (156.4 per 1000 person years). The adjusted rate ratio was 0.67 (95% confidence interval (CI) 0.57 to 0.79) and rate difference was −51 (95% CI −63 to −40) per 1000 person years, with a number needed to treat (NNT) of 20. Among those with recently active nephrolithiasis, the absolute rate difference was 219 per 1000 person years (NNT of 5). Protective associations persisted for nephrolithiasis events that required emergency department visits, hospital admissions, or procedures, and when an SGLT-2 inhibitor was compared with a DPP-4 inhibitor (rate ratio 0.73 (0.68 to 0.78), rate difference −38 (−46 to −29) per 1000 person years (NNT of 26)). Protective associations also persisted among patients with nephrolithiasis and concomitant gout, with a rate ratio of 0.67 (0.57 to 0.79) and rate difference of –53 (95% CI –78 to –27) per 1000 person years versus a GLP-1 receptor agonist (NNT of 19), and 0.63 (0.55 to 0.72) and–62 (–81 to –42) per 1000 person years, respectively, versus a DPP-4 inhibitor (NNT of 16). Furthermore, SGLT-2 inhibitor use was associated with a lower rate of gout flare-ups (rate ratio 0.72, 0.54 to 0.95, rate difference –16, –31 to –1 per 1000 person years) compared with GLP-1 receptor agonists (0.65, 0.52 to 0.82, and –21, –33 to –9 per 1000 person years) compared with DPP-4 inhibitors. SGLT-2 inhibitor initiators showed higher risk of genital infection (eg, hazard ratio 2.21, 95% CI 1.68 to 2.90, and rate difference 13 per 1000 person years), but no altered risk of osteoarthritis encounter (0.87, 0.68 to 1.1, and –2 per 1000 person years) or appendicitis (1.07, 0.69 to 1.67, and 1 per 1000 person years). Results were similar when propensity score overlap weighting was applied.

Conclusions The benefits associated with SGLT-2 inhibitor for patients with nephrolithiasis in these target trial emulations suggest they may be a useful addition to current treatments to simultaneously manage nephrolithiasis recurrence and comorbidities, including gout.

Target trial

We followed the framework proposed by Hernan and Robins36 to emulate two arm randomized clinical trials (the target trials) that would randomize patients with nephrolithiasis and type 2 diabetes to initiate an SGLT-2 inhibitor versus GLP-1 receptor agonist (as well as SGLT-2 inhibitor versus DPP-4 inhibitor as an alternative comparator), in an unblinded fashion. We first specified the framework for the target trial emulation that would address our research questions (see supplementary tables S1-S4), including eligibility criteria, treatment strategies and assignment, start and end of follow-up, outcomes, causal contrasts (eg, intention-to-treat effect or per protocol effect), and analysis plan. We then specified how we would use observational data to emulate components of the protocol and conduct the respective analyses.3637

Source population

The source population for the target trials was the entire population of the province of British Columbia (BC) in Canada. We used Population Data BC, population based linked administrative databases that provide deidentified individual level data for nearly all of British Columbia’s five million residents on provincially funded outpatient medical visits and hospital discharges (since 1990), emergency department encounters (since 2011), and dispensed prescriptions, regardless of age or funding source (since 1996). Vital statistics were available for all individuals. These data have been used in previous studies of diabetes3839 and gout.274041424344 The current study followed the recommendations of the STROBE (strengthening the reporting of observational studies in epidemiology) initiative.45 We used six datasets: BC Medical Services Plan, Consolidation Database, Hospital Separations, PharmaNet, National Ambulatory Care Reporting System, and Vital Statistics Deaths.

Study population and design

For our primary emulated target trial, we performed a new user cohort study with inverse probability weighting to compare the risk of recurrent nephrolithiasis among people with type 2 diabetes and pre-existing nephrolithiasis who initiated an SGLT-2 inhibitor or GLP-1 receptor agonist (an active comparator). In the trials, we included adults aged ≥18 years with type 2 diabetes whose first ever dispensing for one of the comparison drugs was between 1 January 2014 and 30 June 2022, after at least one year of continuous enrolment with the provincial medical services plan (see supplementary figure S1). Diagnoses were based on the presence of at least one ICD-9 or ICD-10 (international classification of diseases, ninth revision (outpatient encounters) and 10th revision (inpatient and emergency department encounters), respectively)2734464748 code (see supplementary table S5). Our primary target trial emulation (see supplementary table S1) for recurrent nephrolithiasis included those with a nephrolithiasis diagnosis at baseline (any time before the drug index date) based on ICD-9 and ICD-10 codes, whereas our three secondary target trial emulations included those with pre-existing nephrolithiasis and pre-existing gout (ie, diagnoses of both conditions recorded any time before the index date; see supplementary table S2), those with pre-existing nephrolithiasis but no history of gout (no recorded diagnosis of gout before the index date; see supplementary table S3), and those without a diagnosis of nephrolithiasis at baseline (no recorded diagnosis of nephrolithiasis before the index date; see supplementary table S4). In all analyses, we also excluded individuals who were dispensed an SGLT-2 inhibitor combination pill or an SGLT-2 inhibitor and GLP-1 receptor agonist at the same time (see supplementary figure S1).

Outcome assessment

The primary outcome was recurrent nephrolithiasis events during follow-up, defined by an emergency department visit or hospital admission with a primary discharge diagnosis of nephrolithiasis via ICD-10 codes, or having an ICD-9 code for nephrolithiasis recorded during an outpatient encounter (see supplementary table S5); the latter has been found to have positive predictive value of 91.8% and 95.9% for an active nephrolithiasis diagnosis and any nephrolithiasis diagnosis, respectively.46 Recurrent nephrolithiasis was defined as having an active nephrolithiasis event among those with a history of nephrolithiasis. Secondary outcomes included the first recurrent nephrolithiasis event during follow-up and recurrent nephrolithiasis counts resulting in emergency department visits or hospital admission (as the primary discharge diagnosis) and those that required interventions including procedural or surgical removal of stones (see supplementary table S5). In the emulated trial of patients with concurrent gout (see supplementary table S2), recurrent flare-ups were an additional outcome, defined by emergency department visit or hospital admission with a primary discharge diagnosis of gout, or an ICD-9 code for gout recorded during an outpatient encounter together with at least one of the following treatments dispensed within seven days: intra-articular or oral corticosteroids, colchicine, or non-steroidal anti-inflammatory drugs.27 A first gout related visit to an emergency department or hospital admission49 and the combination of gout related visit and drug dispensing or procedure5051 have been found to accurately ascertain gout flare-ups, with a positive predictive value of 95% for that combination in identifying patients with at least one flare-up during a 15 month study period,50 closely resembling our ascertainment of flare-up counts over a similar follow-up period.

Covariate assessment

Covariates, selected a priori based on previous literature1952535455 and subject and methodologic expertise, included sociodemographic factors (age, sex, area level socioeconomic status, and residence in one of British Columbia’s five geographic health regions), calendar year of first dispensing, time since first recorded diagnosis of diabetes, chronic comorbidities (eg, cardiovascular disease, hypertension, chronic kidney disease), diabetes complications (neuropathy, nephropathy, and retinopathy), risk factors for nephrolithiasis (nephrolithiasis duration, frequency of nephrolithiasis events in the past 12 months, gout diagnosis, obesity, thiazide diuretic use, and use of urate lowering treatment), relevant drugs, and healthcare utilization (table 1). We also included gout duration, baseline flare-up rate, and use of drugs for flare-ups in the secondary target trial emulation among patients with nephrolithiasis and concomitant gout. Medical conditions were assessed from 1990 or enrolment, whichever came later, whereas healthcare utilization and drug use were assessed during the year before the target trial index date.

Cohort follow-up

In our primary analysis, follow-up started the day of first dispensing (target trial index date) and continued until end of the study period (30 June 2022), deregistration from the provincial medical plan, occurrence of nephrolithiasis (except recurrent nephrolithiasis counts), discontinuation of the index drug (>60 days after the expiration date of the last dispensed supply2734535455) (see supplementary tables S1-S4), switching to or addition of the comparator drug, or death.

Statistical analysis

To account for non-random allocation of patients to the treatment groups, we used stabilized inverse probability of treatment weighting for the average treatment effect in the whole population.5657 Weights were calculated based on all covariates described above in a multivariable logistic regression model to predict each patient’s probability (ie, propensity score) of filling a prescription for an SGLT-2 inhibitor or GLP-1 receptor agonist before the start of follow-up. Inverse probability of treatment weights based on propensity scores of the cohort result in a weighted cohort where treatment assignment is independent of measured confounders.58 Stabilized weights are recommended to account for the otherwise large weights assigned to individuals with a low probability of receiving the treatment of interest, and resultant large variance for the inverse probability of treatment weights estimator.5659 We assessed the balance of the distribution of covariates before and after weighting using standardized differences between treatment groups (<0.1 denoting negligible difference).58 In a sensitivity analysis we applied overlap weighting of the propensity score instead of inverse probability of treatment weighting to balance baseline characteristics.6061 The overlap weighting estimate can be interpreted as the average treatment effect in the overlap population, representing patients with a realistic likelihood of receiving either treatment.57 This approach results in exact balance of the mean of every measured covariate.61

We report crude and weighted numbers of events, person time, incidence rates per 1000 person years, and rate differences per 1000 person years for each outcome by treatment group. To facilitate interpretation, we also calculate the number needed to treat (NNT). For each comparison, weighted Poisson or Cox proportional hazards models were used to estimate rate ratios or hazard ratios and corresponding 95% confidence intervals (CIs),62 while accounting for competing risk of death.63 An E value was calculated to evaluate the robustness of our primary outcome to unmeasured confounders.64 We obtained rate differences and corresponding CIs using the methods of Rothman and colleagues.65 Missing categories were used for the two variables with missing data: region (0.13%) and area level socioeconomic status (1.0%); no other data were missing. All analyses were performed using SAS statistical software, version 9.4 (SAS Institute), and statistical significance was defined as two tailed P value of ≤0.05.

Subgroup and sensitivity analyses

We conducted prespecified analyses in subgroups based on age, sex, thiazide use, and nephrolithiasis over the previous year (as a proxy for recently active nephrolithiasis). Within each subgroup, propensity scores were recalculated with patients reweighted. We conducted several sensitivity analyses to ensure the robustness of our findings, including analyses: with follow-up truncated after one year; with patients censored one year after the target trial index date unless they had already been censored based on the previously mentioned criteria except drug discontinuation (carrying forward use of the index drug up to one year, akin to an intention-to-treat analysis48); restricted to individuals treated with metformin monotherapy at baseline (ie, using no other glucose lowering drugs, to improve comparability of treatment groups); and comparing initiation of an SGLT-2 inhibitor with a DPP-4 inhibitor instead of a GLP-1 receptor agonist.66 Furthermore, to evaluate reproducibility with established relations and the presence of spurious associations, we assessed the risk of a positive control outcome, genital infection (see supplementary table S5), which is positively associated with SGLT-2 inhibitors relative to GLP-1 receptor agonists, per randomized trials,67 and for risk of any osteoarthritis encounter and appendicitis (see supplementary table S5), two negative control outcomes for which we expected null associations.68

Patient and public involvement

No patients were involved in setting the research question or the outcome measures, nor were they involved in developing plans for design or implementation of the study. No patients were asked to advise on the interpretation or writing up of results.

Nephrolithiasis recurrence among patients with pre-existing nephrolithiasis

Before inverse probability weighting, 1892 recurrent nephrolithiasis events occurred among 14 400 unweighted patients who initiated an SGLT-2 inhibitor (99.7 per 1000 person years), compared with 847 events among 5746 unweighted patients who initiated a GLP-1 receptor agonist (177.9 per 1000 person years). The unweighted rate ratio associated with SGLT-2 inhibitor initiation was 0.57 (95% CI 0.52 to 0.62), and the unadjusted rate difference was −78 (95% CI −91 to −65) per 1000 person years. After inverse probability weighting, 1924 recurrent nephrolithiasis events occurred among the 14 456 weighted patients who initiated an SGLT-2 inhibitor (105.3 per 1000 person years), compared with 853 events among the 5877 weighted patients who initiated a GLP-1 receptor agonist (156.4 per 1000 person years) (table 2). This corresponded to an adjusted rate ratio of 0.67 (95% CI 0.57 to 0.79) and rate difference of −51 (−63 to −40) per 1000 person years, with a corresponding NNT of 20, comparing initiation of SGLT-2 inhibitors with initiation of GLP-1 receptor agonists (table 2). Findings were similar when the SGLT-2 inhibitor group was restricted to those initiating empagliflozin (n=9716), with an adjusted rate ratio of 0.66 (0.60 to 0.72). When we restricted our endpoint to recurrent nephrolithiasis events diagnosed during an emergency department visit or hospital admission, the rate ratio of 0.61 (0.51 to 0.73) and rate difference of −12 (−17 to −7) per 1000 person years; the associations for events defined by requiring a procedure were similar (table 2).

In the sensitivity analysis that applied overlap weighting in place of inverse probability weighting (see supplementary table S9), the adjusted rate ratio was 0.68 (0.60 to 0.78) and the rate difference was −52 (−71 to −34) per 1000 person years (table 2). When use of the index drug was carried forward for up to one year (akin to intention-to-treat analysis), the adjusted rate ratio for recurrent nephrolithiasis was 0.77 (0.71 to 0.84) and the rate difference was −37 (−50 to −24) per 1000 person years (table 2). Restricting to individuals using metformin but no other glucose lowering drugs at baseline, the rate ratio was 0.63 (0.57 to 0.70) and the rate difference was −59 (−75 to −44) per 1000 person years (table 2). The inverse association between initiation of an SGLT-2 inhibitor and recurrent nephrolithiasis also persisted when DPP-4 inhibitors were used as comparator (n=11 911 SGLT-2 inhibitor initiators and n=8843 DPP-4 inhibitor initiators, after inverse probability weighting (table 1 and supplementary table S10 and figure S2); the rate ratio associated with SGLT-2 inhibitor initiation was 0.73 (0.68 to 0.78) and the rate difference was −38 (−46 to −29) per 1000 person years (table 2). Results persisted regardless of sex, age, or baseline thiazide diuretic use (table 3). The rate ratio also persisted regardless of the recency of nephrolithiasis at baseline, although the absolute rate difference was higher in patients with recently active nephrolithiasis (−219 (−274 to −164) v −42 (−51 to −33) per 1000 person years) (table 3). The corresponding NNTs were 5 and 24.

As with recurrent nephrolithiasis counts, cumulative incidence of the first recurrent nephrolithiasis event was lower among patients initiating an SGLT-2 inhibitor than among those initiating a GLP-1 receptor agonist (fig 1), with incidence rates of 54.2 and 90.3 per 1000 person years, respectively, resulting in a hazard ratio of 0.68 (95% CI 0.58 to 0.79) and rate difference of −36 (95% CI −45 to −27) per 1000 person years. The risk for a first recurrent nephrolithiasis event was also lower for events resulting in an emergency department visit or hospital admission (see supplementary table S11).

Fig 1

Cumulative incidence of first recurrent nephrolithiasis overall among patients initiating an SGLT-2 inhibitor versus GLP-1 receptor agonist. Cumulative incidence among GLP-1 receptor agonist initiators was adjusted using stabilized inverse probability of treatment weighting. GLP-1=glucagon-like peptide-1; SGLT-2=sodium-glucose cotransporter-2

Recurrence of nephrolithiasis and gout among patients with concurrent gout

Overall, 4409 patients with gout at baseline initiated an SGLT-2 inhibitor (n=3159) or GLP-1 receptor agonist (n=1250), before inverse probability weighting. Similar to the primary target trial emulation, even before weighting, risk factors for nephrolithiasis, such as frequency of nephrolithiasis events in the past year, nephrolithiasis duration, gout flare-up rate, gout duration, thiazide use, and use of urate lowering treatment were similar between treatment comparison groups, except for a higher prevalence of obesity among GLP-1 receptor agonist users (see supplementary table S12). After inverse probability weighting, all baseline covariates (including obesity) were well balanced, with 3159 initiators of an SGLT-2 inhibitor and 1272 initiators of a GLP-1 receptor agonist; the mean age was 65 years and 73% were men (see supplementary table S12). About 24% of patients in each treatment group were using thiazide diuretics and 34% were using urate lowering treatment. In total, 479 recurrent nephrolithiasis events occurred among the 3159 weighted patients who initiated an SGLT-2 inhibitor (122.4 per 1000 person years), compared with 218 events among the 1272 weighted patients who initiated a GLP-1 receptor agonist (174.1 per 1000 person years), with a rate ratio of 0.67 (95% CI 0.57 to 0.79) and rate difference of −53 (95% CI −78 to −27) per 1000 person years (table 4). When using DPP-4 inhibitors as the comparator group, 2668 weighted initiators of an SGLT-2 inhibitor and 2028 weighted initiators of a DPP-4 inhibitor with concurrent gout at baseline were included (76% men, mean age 66 years). Overall, 418 recurrent nephrolithiasis events occurred among SGLT-2 inhibitor initiators (113.5 per 1000 person years) and 439 among DPP-4 inhibitor initiators (175.0 per 1000 person years), with a rate ratio of 0.63 (0.55 to 0.72) and rate difference of −62 (−81 to −42).

In terms of gout outcome, recurrent flare-up counts were lower among those initiating an SGLT-2 inhibitor. Overall, 161 recurrent gout flare-up events occurred among the 3159 weighted patients with gout initiating an SGLT-2 inhibitor, and 71 flare-up events among the 1272 patients initiating a GLP-1 receptor agonist, corresponding to a rate ratio of 0.72 (0.54 to 0.95) and absolute risk reduction of 16 fewer flare-up events per 1000 person years (table 4). Initiation of an SGLT-2 inhibitor was also associated with a lower rate of recurrent gout flare-up events compared with initiation of a DPP-4 inhibitor, with a rate ratio of 0.65 (0.52 to 0.82) and absolute risk reduction of 21 fewer recurrent gout flare-up events per 1000 person years (table 4).

Rate ratios in our target trial of patients with nephrolithiasis without concurrent gout (see supplementary table S3) were similar to those of patients with concurrent gout, but the rate difference tended to be smaller (see supplementary table S13), with a more pronounced difference when comparing initiation of an SGLT-2 inhibitor with initiation of a DPP-4 inhibitor, with 62 versus 31 fewer recurrent nephrolithiasis events per 1000 person years among initiators of an SGLT-2 inhibitor with and without gout (see supplementary table S13).

Incident nephrolithiasis

We also identified 153 445 initiators of an SGLT-2 inhibitor or GLP-1 receptor agonist without a history of nephrolithiasis at baseline (since 1990 or enrolment); 112 771 initiating an SGLT-2 inhibitor and 40 674 initiating a GLP-1 receptor agonist (see supplementary table S14). After inverse probability weighting, 1072 new nephrolithiasis events occurred among 112 728 weighted initiators of an SGLT-2 inhibitor (7.4 per 1000 person years) and 499 new events among 41 623 weighted initiators of a GLP-1 receptor agonist (12.8 per 1000 person years). This corresponded to an adjusted rate ratio of 0.58 (0.52 to 0.64) and rate difference of −5 (−7 to −4) per 1000 person years (see supplementary table S15). Results were consistent when applying overlap weighting of the propensity score, with a rate ratio of 0.56 (0.49 to 0.68) and rate difference of −6 (−8 to −4) per 1000 person years.

Potential impact of unmeasured confounding

The E value for our primary target trial outcome comparing rates of recurrent nephrolithiasis events between initiators of an SGLT-2 inhibitor versus a GLP-1 receptor agonist was 2.35 overall (and 2.66 for nephrolithiasis events requiring emergency department visits or hospital admission); the corresponding E value among patients with gout was also 2.35. This finding indicates that a hypothetical unmeasured confounder would need to have a rate ratio of 2.35 (or 2.66) for both initiation of an SGLT-2 inhibitor and risk of nephrolithiasis to nullify the findings.

Control outcomes

In the positive control outcome analyses, initiation of an SGLT-2 inhibitor versus GLP-1 receptor agonist was associated with a higher risk of genital infection (incidence rates 24.1 and 10.7 per 1000 person years, respectively), with a hazard ratio of 2.21 (95% CI 1.68 to 2.90) and rate difference of 13 (95% CI 10 to 17) per 1000 person years (table 5). For our negative control outcome of osteoarthritis encounter, comparing an SGLT-2 inhibitor with a GLP-1 receptor agonist, the hazard ratio was 0.87 (0.68 to 1.14) and the rate difference was −2 (−6 to 2) per 1000 person years, respectively. Similarly, for the negative control outcome of appendicitis, the corresponding hazard ratio and rate difference were 1.07 (0.69 to 1.67) and 1 (−2 to 3) per 1000 person years (table 5). These findings for all control outcomes persisted when comparing with a DPP-4 inhibitor (table 5) and when using overlap weighting instead of inverse probability of treatment weighting to emulate randomization (see supplementary table S16).

Biological mechanisms

The mechanism underlying the lower risk of recurrent nephrolithiasis with SGLT-2 inhibitor use could partly be attributed to increased urinary output from osmotic diuresis.293031 Randomized controlled trials have shown increasing fluid intake lowers the risk for recurrent nephrolithiasis72 by increasing urinary volume and reducing the concentration of stone forming solutes. Indeed, SGLT-2 inhibitors can increase urinary volume up to twofold, although this effect may be most pronounced at the start of treatment.73 Furthermore, SGLT-2 inhibitors also lower serum urate levels 2347 and have been associated with lower risk of incident and recurrent gout.27477475 Glycosuria is thought to be a central mechanism for the urate lowering effect, competing with soluble urate for transporter mediated reabsorption in the proximal tubule76 and enhancing the excretion of uric acid. This uricosuria could raise the potential to transiently increase the risk of uric acid stones, as do prototypic uricosuric urate lowering drugs such as probenecid or benzbromarone7778; however, our results suggest that the net effect for nephrolithiasis recurrence is beneficial. Indeed, urine pH levels, rather than hyperuricosuria itself, are thought to be the main determinant of uric acid crystallization and precipitation (and potentially calcium oxalate stone formation) since a more alkaline environment will enhance solubility of uric acid crystals and prevent stone formation.3379 To that end, empagliflozin can raise urine pH by inhibiting the renal sodium-hydrogen exchanger and increasing urinary bicarbonate excretion,8081 whereas dapagliflozin can increase urinary citrate excretion, which can also increase urine pH and inhibit crystallization of calcium salts.3382 At the same time, canagliflozin has been shown to decrease urine pH.83 These apparently conflicting effects among SGLT-2 inhibitors could be explained by differences in mineral metabolism between agents, but this warrants further investigation, as does the association between SGLT-2 inhibitors and different compositions of stones. More information about the potential mechanisms of action may come from the SWEETSTONE trial,84 an exploratory, phase 2 randomized trial (n=50) assessing the treatment potential of empagliflozin to prevent the recurrence of nephrolithiasis, with a primary outcome of urinary supersaturations.

Type 2 diabetes is associated with a higher risk of nephrolithiasis, which is thought to be due in part to impaired kidney ammoniagenesis, leading to low urinary pH, a primary factor in the formation of uric acid stones. Acidic urine increases the concentration of the insoluble undissociated uric acid, facilitating precipitation at levels of urine uric acid within or even below the normal range.8586 For example, in a study involving 2464 patients with nephrolithiasis, type 2 diabetes was found to be the strongest predictor of uric acid stones, with an odds ratio of 6.9.87 Similarly, gout was associated with a twofold higher risk of kidney stones in a large male health professional prospective cohort study,14 which was also thought to be through insulin resistance (a close correlate for gout) interfering with renal ammoniagenesis leading to acidic urine.8586 The prevalence of pure and mixed uric acid stones has been found be up to 43-88% in patients with gout,8889 although interestingly, a previous randomized trial found allopurinol reduced recurrent events of calcium oxalate stones, suggesting potential benefit of urate reduction for preventing these stones.90 To that end, the current study addressed overall risk of kidney stones irrespective of composition.

Clinical implications

Among those with the additional risk factor of concurrent gout, SGLT-2 inhibitor use was associated with simultaneous risk reduction in gout recurrence, with a trend of a larger absolute risk reduction in nephrolithiasis recurrence, given their high background rate (particularly compared with DPP-4 inhibitors). These data suggest that SGLT-2 inhibitors might be useful agents in this multimorbidity clinical context with gout, together with their established cardiovascular-kidney benefits. Furthermore, our study provides important outcome data on emergency department visits and hospital admissions for newly diagnosed nephrolithiasis as well as for patients requiring intervention; all costly and burdensome sources of healthcare utilization.91

Comparison with other studies

The findings for recurrent nephrolithiasis from our emulated trial agree with the few studies that focused on risk of newly diagnosed nephrolithiasis.333435 Without prespecified ascertainment, clinical trials of SGLT-2 inhibitors have substantially underreported nephrolithiasis as an adverse event. As such, a previous meta-analysis of such trials with a limited number of nephrolithiasis events (n=106) reported no significant association32; however, a recent meta-analysis of empagliflozin trials with a larger number of reported events (n=183) found a 36% lower risk of urolithiasis compared with placebo.33 Nevertheless, the latter meta-analysis had only nine patients with pre-existing nephrolithiasis, of which one experienced recurrence of nephrolithiasis. Two observational studies with primary outcomes of newly diagnosed nephrolithiasis also found similar rate ratio reductions (26% to 49% lower risk compared with GLP-1 receptor agonist or DPP-4 inhibitor).3435 The current emulated trials focused on the primary outcome of nephrolithiasis recurrence among patients with pre-existing nephrolithiasis, an endpoint thus directly relevant to clinical care.

Strengths and limitations of this study

Our study design, target trial emulations with an active comparator new user design to assess risk of nephrolithiasis with SGLT-2 inhibitor use, reduced bias and supported causal inference by balancing baseline characteristics and specifying the start of follow-up and assigned treatment group.37 We used an entire population based database, which makes our findings likely generalizable, with data on dispensed drugs (as opposed to issued prescriptions) reducing misclassification of drug use. We also conducted several sensitivity and subgroup analyses, and results were consistent with our primary findings, supporting the robustness of the findings. The results for the control outcomes lend specificity to our findings; notably, genital infection (due to SGLT-2 inhibitor induced glycosuria) is considered mild to moderate.21 As we could not capture nephrolithiasis events that did not receive medical attention, our data likely underestimated the overall event risk experienced by our cohort. However, our ascertainment approach for the captured nephrolithiasis events has had a positive predictive value of ≥90%,46 and our approach for capturing recurrent gout flare-ups has had a positive predictive value of 95%50 (with accuracy nearing 100% for gout flare-ups requiring an emergency department visit or hospital admission49). The findings in patients with nephrolithiasis events requiring intervention, including procedural or surgical treatment of stones, as well as those requiring an emergency department visit or hospital admission, should be even more accurate; any residual non-differential misclassification of outcomes would bias our findings towards the null. SGLT-2 inhibitors have recently been approved in patients without diabetes, to treat chronic kidney disease or heart failure, but we were only able to meaningfully analyze people with diabetes so it remains to be determined how the benefit would extend to those without diabetes.

As in any observational study, residual unmeasured confounding is possible. Although we incorporated several proxy measures of nephrolithiasis and diabetes severity and comorbidities into the propensity scores, including baseline nephrolithiasis recurrence rate, baseline gout flare-up rate, drug use, diabetes complications, cardiovascular risk factors, history of heart failure and chronic kidney disease, and healthcare utilization, we did not have access to data on body mass index or laboratory measures such as serum urate or glycated hemoglobin (HbA_1c). However, it is unlikely these or other unmeasured factors would affect the choice between SGLT-2 inhibitor and GLP-1 receptor agonist (or DPP-4 inhibitor) with respect to nephrolithiasis risk, as none were approved or clinically considered for nephrolithiasis (or gout) care during the study period. In addition, several key risk factors for recurrent nephrolithiasis, including duration of nephrolithiasis, frequency of recurrent nephrolithiasis, frequency of nephrolithiasis procedures, urologist and nephrologist visits, gout diagnosis, thiazide use, and use of urate lowering therapy, were similar between treatment groups even before weighting (table 1). Furthermore, our E value64 indicated an unmeasured covariate would need to be associated with both risk of nephrolithiasis and use of an SGLT-2 inhibitor by a risk ratio of up to 2.35 to nullify our findings.

Conclusion

These target trial emulation studies of patients with nephrolithiasis and type 2 diabetes showed that initiation of an SGLT-2 inhibitor was associated with a lower recurrence of nephrolithiasis compared with initiation of a GLP-1 receptor agonist or DPP-4 inhibitor. A similar rate ratio reduction was observed among those with concurrent gout, with simultaneous reduction in gout recurrence (ie, dual added benefits beyond established diabetes treatment). Together with the multiple cardiometabolic-kidney benefits associated with SGLT-2 inhibitors, for patients with an existing indication (eg, type 2 diabetes, heart failure, chronic kidney disease), this class of drugs may be a useful addition to the current treatments for nephrolithiasis to simultaneously tackle the high burden of kidney stones and comorbidities, including gout.