Dapagliflozin in chronic kidney disease: cost-effectiveness beyond the DAPA-CKD trial

ABSTRACT Background The Dapagliflozin and Prevention of Adverse Outcomes in CKD (DAPA-CKD) trial enrolled patients with estimated glomerular filtration rate 25–75 mL/min/1.73 m2 and urine albumin-to-creatinine ratio >200 mg/g. The Dapagliflozin Effect on CardiovascuLAR Events-Thrombolysis in Myocardial Infarction 58 (DECLARE-TIMI 58) trial enrolled patients with type 2 diabetes, a higher range of kidney function and no albuminuria criterion. The study objective was to estimate the cost-effectiveness of dapagliflozin in a broad chronic kidney disease population based on these two trials in the UK, Spain, Italy and Japan. Methods We adapted a published Markov model based on the DAPA-CKD trial but to a broader population, irrespective of urine albumin-to-creatinine ratio, using patient-level data from the DAPA-CKD and DECLARE-TIMI 58 trials. We sourced cost and utility inputs from literature and the DAPA-CKD trial. The analysis considered healthcare system perspectives over a lifetime horizon. Results Treatment with dapagliflozin was predicted to attenuate disease progression and extend projected life expectancy by 0.64 years (12.5 versus 11.9 years, undiscounted) in the UK, with similar estimates in other settings. Clinical benefits translated to mean quality-adjusted life year (QALY; discounted) gains between 0.45 and 0.68 years across countries. Incremental cost-effectiveness ratios in the UK, Spain, Italy and Japan ($10 676/QALY, $14 479/QALY, $7771/QALY and $13 723/QALY, respectively) were cost-effective at country-specific willingness-to-pay thresholds. Subgroup analyses suggest dapagliflozin is cost-effective irrespective of urinary albumin-to-creatine ratio and type 2 diabetes status. Conclusion Treatment with dapagliflozin may be cost-effective for patients across a wider spectrum of estimated glomerular filtration rates and albuminuria than previously demonstrated, with or without type 2 diabetes, in the UK, Spanish, Italian and Japanese healthcare systems.


INTRODUCTION
The prevalence of chronic kidney disease ( CKD) is rising and is estimated to be 8%-16% worldwide [1 ], affecting an estimated 840 million people [2 , 3 ].CKD is usually a progressive condition, typically described by the level of estimated glomerular filtration rate ( eGFR) and the presence and degree of albuminuria.Patients with reduced eGFR and/or albuminuria typically incur higher rates of hospitalization, outpatient visits, ambulance use and emergency room visits relative to patients with milder disease [4 ].Accordingly, CKD has a detrimental effect on healthrelated quality of life ( HRQoL) and yields a substantial burden to healthcare systems and society [5 ].
In addition to lifestyle changes, standard-of-care pharmacologic treatment options for CKD include angiotensin-converting enzyme inhibitors or angiotensin receptor blockers which aim to delay disease progression and prevent or ameliorate associated complications, yet many patients continue to progress toward advanced CKD.
Sodium-glucose co-transporter-2 ( SGLT2) inhibitors are an established therapeutic option in patients with T2D.Trials designed to ensure the cardiovascular safety of SGLT2 inhibitors also demonstrated substantial benefits vis-à-vis cardiovascular outcomes distinct from their glucose lowering action.These and other studies showed protective effects on kidney function, including the potential to reduce the rate of eGFR decline and the risk of end-stage kidney disease ( ESKD) in patients with type 2 diabetes ( T2D) [6 -12 ].
The efficacy and safety of dapagliflozin in addition to standard therapy in patients with CKD and albuminuria [eGFR 25-75 mL/min/1.73m 2 ; urinary albumin-to-creatine ratio ( UACR) ≥200 mg/g] was investigated in the Dapagliflozin and Prevention of Adverse Outcomes in CKD ( DAPA-CKD) trial [13 ].The trial was ended prematurely due to overwhelming efficacy in patients with or without T2D.Subsequently, a cost-effectiveness analysis based on the DAPA-CKD trial demonstrated cost-effectiveness in multiple national healthcare systems [14 ].While patients with CKD with lower levels of albuminuria or without albuminuria are generally at lower risk of adverse clinical outcomes, they may also gain benefit from treatment with an SGLT2 inhibitor [15 , 16 ].However, the potential value of treatment in these patients has not been studied.
The DAPA-CKD-like population typically have a higher per patient burden than the broad CKD population described in this analysis [17 ].However, patients with lower UACR are more prevalent and therefore exert a higher burden at a population level on national healthcare systems [18 ].Therefore, the largest benefit to patients and national healthcare systems could be attained by addressing the residual risk in patients outside of the DAPA-CKD population.
By adapting a previously published model driven by patientlevel data from the DAPA-CKD trial, the objective of the present study is to assess the cost-effectiveness of dapagliflozin within the broader CKD population.Based on data from both the DAPA-CKD and Dapagliflozin Effect on CardiovascuLAR Events-Thrombolysis in Myocardial Infarction 58 ( DECLARE-TIMI 58) trials, the model assesses dapagliflozin from national healthcare perspectives in the UK, Spain, Italy and Japan.

Source data
A broad data set was derived by combining individual patientlevel data from two placebo-controlled randomized controlled trials of dapagliflozin plus standard therapy.The DAPA-CKD trial included patients with eGFR 25-75 mL/min/1.73m 2 and UACR 200-5000 mg/g, with or without T2D.The DECLARE-TIMI 58 study included patients with T2D and estimated creatinine clearance ≥60 mL/min/1.73m 2 and no inclusion criteria based on UACR.Study designs, patient characteristics and outcomes have been published [8 , 13 , 19 -23 ].

The broad CKD population
We pooled individual patient data on patients with CKD from the two trials, including only the DECLARE-TIMI 58 subset of patients with eGFR < 60 mL/min/1.73m 2 calculated by the CKD Epidemiology Collaboration equation [24 , 25 ] and/or with albuminuria at baseline ( > 30 mg/g) , referred to hereafter as the DECLARE CKD population.Baseline patient characteristics of the pooled CKD population are provided in Table 1 .
Patients with lower levels of albuminuria ( UACR < 200 mg/g) were taken primarily from the DECLARE CKD data set with T2D patients only.We therefore estimated event rates and the dapagliflozin treatment effects for participants without T2D with a UACR < 200 mg/g using a Poisson distribution because neither trial enrolled patients with low UACR but without T2D [26 ].To do this, we applied adjustment factors for UACR and T2D status to the DECLARE CKD data to estimate rates for patients without T2D with lower levels of albuminuria, relative to the treatment  effect observed in patients with or without T2D in the DAPA-CKD trial ( Supplementary data, Tables S1 and S2) .The pr ev alence of T2D in the broad CKD population was assumed equivalent to the DAPA-CKD trial.

Economic model
This analysis adapts a published Markov model for the DAPA-CKD trial population [27 ], which modelled disease progression by discrete eGFR-defined health states and health states for patients on kidney replacement therapy ( KRT) .
The adapted model estimated outcomes separately according to T2D status since the extrapolation estimates effects for patients without T2D and lower levels of albuminuria.The model outcomes are a weighted average according to a defined proportion of lower versus higher UACR ( Fig. 1 ) .The model assumes that 75% of the broad CKD population consists of patients with lower UACR ( < 200 mg/g) and 25% with higher UACR ( ≥200 mg/g) , as observed in the DECLARE CKD population, which had no albuminuria inclusion criterion.
The model assumes a constant rate of discontinuation, which was applied to all patients receiving treatment with dapagliflozin in each modelled cycle.Upon discontinuing treatment with dapagliflozin, patients were assumed to be treated with standard therapy only and subject to the risk of outcomes associated with that treatment arm.

Disease progression
For patients with higher UACR, treatment dependent transition matrices were derived from the DAPA-CKD trial for the first 4 months of follow-up and from Month 4 thereafter, so as to replicate observed patterns of eGFR trajectory ( see Supplementary data, Tables S3 and S4) [13 ].For patients with lower UACR, transition matrices were derived by similarly  [34 ] using DECLARE CKD data for CKD G1-G3 and DAPA-CKD for CKD G4 and G5, consistent with risk stratification according to guidelines ( see Supplementary data, Tables S5 and S6) [28 ].The model derives transitions for post-KRT outcomes from a systematic literature review of CKD modelling [29 ], as these events were rare in the DECLARE-TIMI 58 and DAPA-CKD trials.
To avoid double counting of mortality, we derived transition probabilities excluding the transition to death.We calculated the role of altered kidney function on the risk of death by parametric survival modelling, described below.

Mortality and event incidence
We extrapolated the incidence of all-cause mortality using adjusted parametric survival equations, in accordance with established guidelines [30 ].Health state defined by eGFR were timeupdated to estimate the dynamic mortality risk associated with progression.Of the distributions evaluated, we considered the Weibull distribution to yield the most consistent fits by albuminuria stratification to the pooled trial data ( Supplementary data, Fig. S1) .
We used adjusted generalized estimating equations to capture the incidence of hospitalization for heart failure ( HHF) , including first and recurrent events.As with mortality, timeupdated eGFR captured the effect of disease evolution on event risk.We conducted all analyses from an intention-to-treat perspective and validated against observed data ( Supplementary data, Fig. S2) .
Patients could experience treatment-related adverse events of interest at a constant rate, incurring management costs and utility decrements in the incident cycle only.Considered adverse events were volume depletion, major hypoglycaemic events, fractures, diabetic ketoacidosis and amputation.

Resource use and costs
The analysis considered only direct costs from healthcare payer perspectives.We discounted costs at an annual rate of 3.5% in the UK, 3% in Spain and Italy and 2% in Japan, according to local guidelines [30 -33 ].Cost inputs specific to each country are presented in 2022 US dollars and native currencies in Supplementary data, Tables S7 and S8, respectively.

Health-related quality of life
As EuroQol-5 Dimension 5 Levels ( EQ-5D-5L) data were not recorded in the DECLARE-TIMI 58 trial, utility estimates were solely derived from patient-level EQ-5D-5L data from the DAPA-CKD trial ( Table 2 ) .Notably, utility values for each setting were derived using country-specific life tables and tariffs; further details are available in published material [14 ].The model assumes that patients with lower UACR are adequately represented by HRQoL outcomes in patients with higher albuminuria; in effect, that patient utility is driven only by eGFR-defined stage and the incidence of complications, in line with other published CKD models [29 ].We also discounted benefits annually at a rate of 3.5% in the UK, 3% in Spain and Italy, and 2% in Japan [30 -33 ].

Subgroup analyses
Key subgroups were established for the analysis to assess model outcomes stratified by T2D status ( with and without) and level of albuminuria ( lower: < 200 mg/g; higher: ≥200 mg/g) to yield four subgroups for evaluation.

Scenario analyses
The DAPA-CKD and DECLARE-TIMI 58 trials had median followup periods of 2.4 years and 4.2 years, respectively.Given that our model predicts outcomes over a lifetime horizon, we considered varying time horizons to characterize the evolution of the costeffectiveness of dapagliflozin.

Sensitivity analyses
We conducted one-way sensitivity analyses to demonstrate the influence of varying key parameter values on modelled outcomes.We conducted probabilistic sensitivity analysis to assess the influence of uncertainty across all model parameters Note that results are presented for the UK setting, there will be some variation in outcomes owing to the application of country-specific life tables.
on health economic outcomes.For each analysis, 1000 replicates were generated in which the parameters sampled included patient baseline demographic characteristics, direct costs, utilities, adverse event risks and treatment efficacy.Probabilistic sensitivity analyses were also conducted separately for the assessed subgroups for each of the four countries.Specific to the extrapolative analysis in the population with low UACR without T2D, the parameters estimating HHF and mortality effects were also sampled.Results were evaluated on the cost-effectiveness plane and plotted with 95% credibility ellipses.

Base case analysis
In the broad CKD population, dapagliflozin is predicted to slow the rate of CKD progression and lower the incidence of HHF.Patients treated with dapagliflozin were likely to spend 1.08 more years in the earlier stages of CKD ( stages G1 to G4) versus those treated with standard therapy alone ( dapagliflozin 14.9 years; standard therapy 13.8 years; Table 3 ) .Patients treated with dapagliflozin had a lower rate of HHF ( 119 versus 140 per 1000 patient-years) .
These clinical benefits lead to extended life expectancy versus patients treated with standard therapy alone ( Table 4 ) , with mean predicted increase in life expectancy by 0.60 years in those treated with dapagliflozin ( 12.5 years) versus those treated with standard therapy alone ( 11.9 years) in the UK setting.Other settings demonstrated similar improvements in patient life expectancy ( Spain 0.64; Italy 0.66 years; Japan 0.84) .
Treatment with dapagliflozin in the broad CKD population led to mean lifetime QALY gains of 0.47, 0.54, 0.57 and 0.68 in the UK, Spain, Italy and Japan, respectively ( Table 4 ) .Differences between setting are borne from country specific utility tariffs, life tables and discounting rates.
Treatment with dapagliflozin led to increased overall costs versus placebo in the UK ( $5011) , Spain ( $7807) , Italy ( $4406) and Japan ( $9271) .The main contributors of additional cost in those treated with dapagliflozin were additional drug acquisition costs and disease management costs resulting from extended life expectancy.These costs were partially offset by reductions in costs associated with KRT through delayed time to dialysis or transplant and a reduced rate of HHF over the lifetime model horizon.Dapagliflozin was considered cost-effective with ICERs of $10 676/QALY, $14 479/QALY, $7771/QALY and $13 723/QALY in the UK, Spain, Italy and Japan, well below WTP thresholds ( $28 777/QALY, $39 063/QALY, $35 261/QALY, $53 671/QALY, respectively) and estimated net monetary benefit of $8496, $13 255, $15 587 and $26 989 respectively.Health economic Figure 2: Subgroup analyses ( top to bottom) in the UK, Spain, Italy and Japan dependent on UACR severity and T2D status.'Low' and 'elevated' refer to albuminuria status ( UACR < 200 mg/g and UACR ≥200 mg/g respectively) while 'Base' refers to the base case analysis of 75% low and 25% elevated albuminuria.Dashed lines designate the WTP threshold of each country.
outcomes are presented in their native currency in Supplementary data, Table S9.

Subgroup analysis
Across all subgroups analysed, dapagliflozin was considered cost effective at the corresponding WTP thresholds attributed to each country ( Fig. 2 ) .In all settings, the ICERs for patients with lower UACR and without T2D ( UK $25 750; Spain $16 429; Italy $15 124; Japan $25 973) were highest, while ICERs for the subgroup with higher UACR and without T2D were lowest ( UK $5080; Spain $10 103; Italy $4681; Japan $11 254) .In all settings, a similar spread of incremental QALYs was observed; in terms of costs, T2D subgroups were associated with higher incremental costs than non-T2D subgroups and this differential was larger for the lower UACR than for the higher UACR groups.

Scenario analysis
Treatment with dapagliflozin was estimated to lead to fewer patients progressing to ESKD and becoming hospitalized due to heart failure over a lifetime horizon.The largest changes in ICERs were observed during the early period, up to approximately 15 years, where increasing time horizons are associated with large decreases in the ICERs ( Fig. 3 ) .Beyond this point, the ICERs stabilize and show little change up to a lifetime time horizon.Similarly, the estimated net monetary benefit was negative over the initial years of treatment, before rising until 20 years from baseline, and the rate of increase levelled off across most settings by 25 years.
Over the patient lifetime, treatment and disease management costs accrue, and these costs are offset in the dapagliflozin arm by reductions in HHF and KRT events ( Fig. 4 ) .While average, per-patient HHF cost reductions remain constant at longer time horizons, those related to avoided KRT exhibit a local minimum at around 15 years, followed by a slight increase before levelling off, likely due to a combined effect of survival and changes in disease progression as more patients transition to later stages of CKD.

Sensitivity analysis
Deterministic sensitivity analyses revealed that, aside from discounting, subpopulation selection and time horizon had the greatest influence on outcomes.Analyses remained cost effective at the defined thresholds for each country for all parameter variations tested ( Supplementary data, Table S10) .
Probabilistic sensitivity analysis showed that the model was robust to joint uncertainty of all parameters.In the UK, 99.8% of simulations were found to be cost effective at a threshold of $28 777/QALY ( £20 000/QALY) , 99.9% were cost effective at a threshold of $39 063 ( €30 000/QALY) in Spain, whilst 100% were cost effective at thresholds of $35 261 ( €25 000/QALY) in Italy and $53 671/QALY ( ¥5 000 000/QALY) in Japan ( Fig. 5 ) .Probabilistic sensitivity analyses across subgroups showed that costeffectiveness outcomes were robust across simulations of the elevated UACR populations and low UACR population with T2D, while simulations in the population with low UACR without T2D showed less certainty in the cost-effectiveness outcomes for this population ( see Supplementary data, Fig. S3) .

DISCUSSION
The present study sought to bridge a data gap to explore costeffectiveness of dapagliflozin in patients with CKD including lower risk patients with lower levels of albuminuria with or without T2D.To model this broader population, individual patient data were pooled from randomized large outcome clinical trials of dapagliflozin plus standard therapy versus standard therapy alone that could inform overall survival and cardiorenal event risks.After adaptation and incorporation of these risk equations into a previously published cost-effectiveness model, results of the present study expand the evidence base of health economic outcomes in a population with CKD that is broader than the inclusion criteria of the DAPA-CKD trial.Analyses revealed ICERs below attributed WTP thresholds and positive net monetary benefit in all settings ( UK, Spain, Italy and Japan) , and outcomes were generally supportive of cost-effectiveness in patients with lower UACR ( < 200 mg/g) and without T2D.Outcomes were typically robust across deterministic and probabilistic sensitivity analyses.
Building on the previous cost-effectiveness analysis in the DAPA-CKD population [14 ], this analysis suggests that the clinical and economic value of dapagliflozin in comparison with usual care alone could be generalizable ( albeit somewhat less pronounced) to patients with CKD who are not represented in the DECLARE-TIMI 58 or DAPA-CKD trials, including patients with albuminuria stage A2 without T2D.In line with the previous analysis, patients treated with dapagliflozin spent more time in CKD stages 1-4 ( broad CKD population, 1.1 years; DAPA-CKD, 1.7 years) versus patients treated with standard therapy alone [14 ].Moreover, dapagliflozin slowed the rate of HHF ( broad CKD population, 21 events avoided; DAPA-CKD, 19 events avoided) and increased life expectancy ( broad CKD population, 0.7 years; DAPA-CKD, 1.7 years) versus standard therapy alone.In comparison with the previous analysis in the higher UACR population only, potential benefits were lower as reflected by smaller incremental QALY and slightly smaller incremental costs in the context of the broader CKD population.The effect is expected since, while the rates of mortality, ESKD and cardiovascular events were lower relative to patients with more advanced eGFR and albuminuria stage of CKD, the presence of CKD will increase the risk of cardiorenal events and so the treatment effect of dapagliflozin was consistent [39 ].Nevertheless, the analysis suggests that results would be considered cost effective at all country-specific thresholds for the broad CKD population, and even when restricted to the population with lower UACR.The substantial clinical benefits observed in the medium-to long-term suggest that, given management of CKD in patients within this broader population will typically lie with primary care physicians, earlier enrolment on reno-protective therapies such as dapagliflozin could be important in reducing the population level impact of CKD progression.
This study, as with any modelling analysis, is subject to several limitations.The most significant source of uncertainty relates to the requirement to extrapolate outcomes in patients without T2D and with lower UACR.Additionally, given the relative health of these patients, there are limited data available against which to externally validate these projections.
This analysis takes advantage of two trial data sets to generate a pooled population including patient-level data from 10 273 patients, including patients at lower risk of progression.Therefore, these patients would be expected to experience fewer events such as HHF or progression to advanced CKD, potentially confounding extrapolations for these individuals.The limitation is partly mitigated through the longer observation time for these patients ( median trial follow-up was 4.2 years) and controlling for multiple variables including albuminuria and CKD state to exploit fully all available data to inform estimates.
The transition probabilities for post-KRT outcomes were estimated from a systematic review of published literature on modelling in CKD [29 ], given the small number of observed events in either trial.
Finally, the prevalence of albuminuria may vary between countries, and therefore country-specific prevalence data would best reflect outcomes specific to each setting.National prevalence data stratified by eGFR and albuminuria are limited and importantly, the DAPA-CKD trial threshold definition of higher albuminuria was 200 mg/g, a level that differs from KDIGO categories of albuminuria ( A2, 30-300 mg/g; A3, > 300 mg/g) [18 , 24 ].The relative proportions were instead approximated from the DECLARE CKD trial data in which patients were not selected according to albuminuria status.
In conclusion, this analysis addresses a data gap that currently exists in patients with CKD.The country-specific results of the model suggest that dapagliflozin could be a cost-effective treatment option for the broader CKD population and irrespective of eGFR, UACR and T2D status in the UK, Spain, Italy and Japan.

Figure 1 :
Figure 1: Model schematic for estimation of outcomes in the broad CKD population.HF: heart failure; UACR: urine albumin to creatinine ratio.

Figure 3 :
Figure 3: ( a) ICER ( US dollars per QALY) and ( b) net monetary benefit evolution over time in the UK, Spain, Italy and Japan.

Figure 5 :
Figure 5: Cost-effectiveness acceptability curves for the UK, Spain, Italy and Japan.The probability of dapagliflozin in addition to standard therapy being cost-effective at the given WTP threshold versus standard therapy only.The specified WTP threshold for each country is denoted by the corresponding-coloured dotted line parallel to the y-axis.