Introduction
The population of western countries is progressively aging. As the prevalence of diabetes increases with age, a large proportion of people with diabetes falls into elderly categories, often > 70 years old [
1]. Treatment of diabetes in such elderly group has specific needs related to targets, complications, and adverse events [
2]. The aging diabetic population exhibits particularly high rates of heart failure (HF) and kidney disease, both acute and chronic [
3]. Prevention of these conditions is key to improve quality of life in aged people with diabetes. Trials with sodium glucose cotransporter-2 inhibitors (SGLT2i) have demonstrated extensive benefits on multiple hard endpoints in diversified populations [
4]. Among individuals with type 2 diabetes (T2D), SGLT2i reduced the rates of hospitalization for HF, the progression of chronic kidney disease (CKD), and the development of acute kidney injury (5; 6). Among people with a history of HF or CKD and with or without T2D, SGLT2i improved overall outcomes [
7‐
9]. Based on these effects, SGLT2i appear particularly suited for the treatment of elderly people with diabetes. However, diabetic patients aged 70 years or older represented a minority of those enrolled in phase III trials and studies examining whether SGLT2i maintain their effects in aging individuals with acceptable safety and tolerability profiles are scant. According to a re-analysis of the DECLARE trial, dapagliflozin maintained its glucose-lowering efficacy and was equally effective in preventing cardiovascular death or hospitalization for heart failure in patients aged < 65 years, 65–75 years, or > 75 years [
10]. Nonetheless, clinicians may be concerned with some rare side effects of SGLT2i, like volume depletion, that may be particularly dangerous in the elderly. In Italy, there has been caution in adopting SGLT2i treatment in the individuals with T2D age > 70 years, mainly in favour of dipeptidyl-peptidase-4 inhibitors (DPP-4i) [
3]. In the interval trial, people with T2D aged 70 years or older who were randomized to vildagliptin versus placebo exhibited a threefold higher probability of reaching their individualized HbA1c target, without new safety signal [
11]. Yet, it is questionable that DPP-4i are preferable over SGLT2i in elderly patients, since DPP-4i exert no protection against cardio-renal complications [
12,
13].
In older patients, glycaemic targets need to be adjusted to the degree of frailty and life expectancy, such that the use of individualized targets is recommended. In 2015, based on a consensus among key worldwide opinion-leading diabetologists, Cahn et al. proposed an algorithm for calculating the individualized glycaemic goals in patients with T2D [
14]. There is a general paucity of studies specifically focusing on the elderly population, even in the real-world evidence (RWE) setting, and the adoption of individualized glycaemic targets is still uncommon.
In this study, we aimed to compare the probability of attaining individualized HbA1c target among elderly patients with T2D who initiated the SGLT2i dapagliflozin or a DPP-4i under specialist care in Italy.
Methods
Study design
DARWIN-FUP (dapagliflozin real-world evidence follow-up) was a retrospective multicentre study conducted at 56 diabetes specialist outpatient clinics in Italy. The study collected data on patients who received for the first time a prescription of dapagliflozin or a DPP-4 inhibitor from 2015 to 2017. In the study period, SGLT2i and DPP4i could be prescribed only by diabetes specialists. The primary objective was comparing the effectiveness of dapagliflozin versus DPP-4 inhibitors on a composite endpoint of HbA1c, body weight, and blood pressure reduction. Details on the study design, along with results of such primary analysis, have been published elsewhere [
15]. The protocol was approved by the Ethical Committees of all participating centres. In agreement with the National regulations on observational studies, the need for informed consent was waived.
Cohort identification
In this secondary analysis of the DARWIN-FUP study, we included only data on male and female patients with a diagnosis of type 2 diabetes since at least 1 year, aged 70–80 years, who received first prescription of dapagliflozin or DPP4i on top of metformin with or without insulin and had at least one follow-up visit available for the evaluation of effectiveness. The lower limit of 70 years was chosen to match a more modern definition of “elderly” as opposed to the traditional threshold of 65 years, accounting for population aging, as suggested by the United Nations [
16]. The upper limit of 80 years was defined by the DARWIN-FUP protocol [
15]. General exclusion criteria applied in the entire dataset, such as other forms of diabetes, chronic kidney disease stage III or higher (which was a contraindication to prescription of SGLT2i), and prior use of dapagliflozin or DPP4i. We selected only patients with HbA1c levels above the personalized target, defined accordingly to the short form previously described [
14], which is based on life expectancy, disease duration, hypoglycaemia risk from treatment, comorbidities, and complications. Life expectancy was based on age- and sex-adjusted national survival curves [
17], whereas the degree of comorbidity was inferred from the number of concomitant medications other than diabetes drugs.
Data collection
Data were extracted automatically from the same electronic chart system at all centres. The baseline date was set as the date patients received the first prescription of dapagliflozin, or DPP4i, whereas follow-up was collected at the last routine visit at the same clinic, at least 3 months but less than 12 months after baseline. At baseline, we collected information on demographics, anthropometrics, risk factors, laboratory values, complications, and therapies (for details, see [
15]). At the follow-up date, we collected endpoint data to evaluate effectiveness, i.e., HbA1c, body weights, systolic blood pressure, and whether the patients continued receiving dapagliflozin or stopped the drug.
Endpoints
The primary endpoint of this analysis was the proportion of patients achieving the individualized HbA1c target. Secondary outcomes were the change in HbA1c, body weight, and systolic blood pressure.
Statistical analysis
Continuous data are presented as mean and standard deviation (SD), whereas categorical variables are shown as percentages. Non-normal variables upon the Kolmogorov–Smirnov test were log-transformed before analysis with parametric tests. The comparison of baseline characteristics between the two groups was performed using Student’s t test for continuous variables and the Chi-square test for categorical variables. The within-group changes in endpoint variables was assessed using the paired Student’s t test with two tails.
We used three different approaches to control the confounding by indication (channelling bias), as depicted in Figure S1. In the primary analysis, we used the inverse probability of treatment weighting (IPTW), estimating a propensity score (PS) for the probability of being treated with dapagliflozin. Propensity scores (PS) were calculated from the following baseline covariates: age, sex, duration of diabetes, baseline body weight, systolic and diastolic blood pressure, HbA1c, HbA1c target, fasting plasma glucose, total and HDL cholesterol, triglycerides, eGFR, micro- or macro-albuminuria, diabetic retinopathy, diabetic macular oedema, microangiopathy, macroangiopathy, carotid atherosclerosis, history of stroke/TIA, coronary revascularization, ischemic heart disease (IHD), coronary heart disease (CHD), history of heart failure, left-ventricular hypertrophy (LVH), use of other GLM (metformin and insulin), and other medications (angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, calcium channel blockers, anti-platelet therapies, beta-blockers, diuretics, and statins). To reduce bias arising from immortal time and time lag, we also included in PS models the number of GLM classes used by the patients before starting DPP-4i or dapagliflozin and the calendar year of index date. The residual of imbalances of the IPTW analyses was evaluated comparing the weighted SMD between dapagliflozin and DPP4i group (i.e., SMD ≥ 0.1 and p ≤ 0.05). Direct comparison of the outcome was allowed when there was no residual imbalance. Thus, the proportion of patients meeting the primary and secondary endpoints were compared with log-binomial regression or linear regression model without any further adjustment, or with further adjustment in case of residual imbalances.
We performed sensitivity analyses in the entire dataset by means of multivariable adjusted (MVA) linear or log-binomial regression models (or, whenever the latter failed to converge, using Poisson regression model with robust error variances). These MVA analyses were adjusted for all clinical characteristics used to compute PS, as listed above. Additional sensitivity analysis was performed on the primary outcome after 1:1 propensity score matching (PSM) based on the same PS used in the IPTW analysis.
For IPTW, PSM, and MVA, full datasets of baseline variables were needed to compute PS or to be entered in the regression models. Therefore, missing data were handled with multiple imputation (MI). MI was performed as previously described [
18], with a fully conditional specification (FCS) algorithm [
19] and obtaining ten imputed datasets including only covariates with less than 50% of missing values. Outcome variables were not imputed. Outcome analyses with IPTW and MVA were performed on each imputed dataset and pooled estimated treatment difference (ETD) are presented [
20]. Relative risk (RR) with 95% confidence interval (CI) was calculated for binomial outcomes.
The primary analyses were conducted following an intention to treat (ITT) approach (i.e., including all patients regardless of whether they continued to be prescribed such treatment at follow-up). Additional sensitivity analyses were conducted in the “as-treated” (AT) dataset, including only patients for whom the prescription of DPP-4i or dapagliflozin was confirmed at the follow-up visit. Information on reasons for stopping or on drug refills rates was not available to evaluate adherence.
All analyses were stratified by sex, disease duration (< > 15 years), body mass index (< > 30 kg/m2), baseline HbA1c (< > 8.5%), eGFR (quartiles), concomitant insulin treatment, history of major adverse cardiovascular events (MACE), and formal interaction analyses were performed to evaluate whether ETD was influenced by these possible moderators.
A two-tailed p value < 0.05 was considered statistically significant. Statistical analyses were performed using SAS version 9.4 (TS1M4), and graphs were produced with GraphPad Prism ver. 8.
Discussion
Among people with T2D who were aged 70–80 years and who attended Italian diabetes specialist clinics under routine care in 2015–2017, initiation of dapagliflozin was associated with a similar reduction in HbA1c as compared to initiation of DPP-4i, but with a smaller proportion attaining the individualized HbA1c target. This difference was observed mainly among patients with lower eGFR, for whom SGLT2i is known to exert a blunted glycaemic effect [
21], due to reduced glucose excretion. In the TriMaster randomized crossover trial, among patients with T2D (mean age 62) and an eGFR between 60 and 90 ml/min/1.73 m
2, sitagliptin reduced HbA1c more than canagliflozin [
22]. In our study, patients with longer diabetes duration also reached HbA1c target less frequently with dapagliflozin than with DPP-4i.
The extra-glycaemic benefits appear to be preserved in older dapagliflozin-treated patients, who experienced greater improvements in blood pressure and reduction in body weight than initiators of DPP-4i. These are important endpoints, as blood pressure control remains an unmet need in elderly individuals with diabetes [
23]. Though weight loss may not always be beneficial in the elderly, therapy with SGLT2i is expected to result in loss of adipose and ectopic fat [
24,
25], possibly enhancing the overall metabolic improvement. Results on the overall efficacy of dapagliflozin versus placebo are re-assuring on the possibility that SGLT2i maintain their cardiovascular protective effects in elderly patients under routine care, as demonstrated in the stratified analysis of the DECLARE trial [
10]. This is supported by results of other trials showing superiority of dapagliflozin versus placebo with regards to heart failure and CKD outcomes including a proportion of elderly patients with and without T2D [
7,
9,
26]. For example, in the DELIVER trial, mean age was 72 years and 2806 patients had T2D [
26].
It should be noted that the average characteristics of patients included in this study are those of an aged population with advanced and poorly controlled T2D. This scenario differs from that of elderly onset T2D, for whom the development of chronic complications may be less of a concern. Indeed, aged patients with > 10 year diabetes duration, an HbA1c of 8.3%, and highly prevalent complications should be considered particularly at risk of developing adverse diabetes-related outcomes, including heart failure and CKD. Use of SGLT2i in this population is particularly appropriate for the potential to improve disease-related outcomes beyond glycaemic control. Therefore, decisions on the best treatment strategy for this population of patients should not be limited to the evaluation of glycaemic targets. Here, we show that dapagliflozin maintained its effectiveness on extra-glycaemic endpoints in the aged population. Although we do not have data on hard endpoints, it is arguable that simultaneous improvements in glucose, weight, and blood pressure control could translate into improved cardiovascular outcomes. On the other side, no substantial improvement in hard endpoints is expected during treatment with DPP-4i. Therefore, despite a greater proportion of patients attained glycaemic targets with DPP-4i, extra-glycaemic effects were negligible and DPP-4i provide no protection against cardio-renal disease. Therefore, while DPP-4i maintained efficacy in the elderly and are generally well tolerated, we argue that they may be more suited for the treatment of elderly onset diabetes when prevention of heart failure and kidney disease is less of a concern. For elderly patients with high risk for cardio-renal disease and in need of intensifying the glucose-lowering regimen, SGLT2i remain the best option as suggested by treatment algorithms. The combination of SGLT2i and DPP4i is rationale and may be particularly effective and safe [
27], also in the elderly.
We did not have information on adverse events, which can be particularly relevant in elderly patients. We observed a persistence of 71.1% on dapagliflozin, which was lower than that observed for DPP-4i (78.8%), but in line with the general population of the DARWIN-FUP study [
15], suggesting no specific safety issue leading to treatment discontinuation in this elderly population. Previously, using a similar database, we detected no specific clinical feature leading to discontinuation in patients receiving dapagliflozin, as opposed to those receiving a range of different glucose-lowering medications [
28].
We wish to underline that our findings are divergent from those of head-to-head comparative trials. Though with some differences, dapagliflozin and empagliflozin appeared to be non-inferior to saxagliptin and linagliptin, respectively [
29,
30], but patients in both trials were much younger than in our study and baseline HbA1c was quite diversified (7.9% and 8.9%) with no use of individualized targets. The reasons why a smaller proportion of patients attained HbA1c targets with dapagliflozin versus DPP-4i in our study probably reside in its specific design and setting. First, there are limitations inherent to the observational design. Though we carefully addressed confounding by indication with gold standard methodologies for comparative effectiveness research, the risk of residual confounding is high, especially because the two populations of patients were very different at baseline. Despite good matching on baseline HbA1c, the higher variability of baseline HbA1c in the dapagliflozin group may be one of the reasons why the proportion of patients reaching HbA1c targets was lower than in the DPP-4i group, even without differences in the change of HbA1c on a continuous scale. Furthermore, systematic factors driving baseline differences between the two groups may have influenced the outcome. During the study period, different reimbursement restrictions, with regards to possible combinations and upper limit of baseline HbA1c, applied to the two classes of drugs, creating a true channelling bias. Therefore, generalizability of our findings needs to be carefully scrutinized.
The glucose-lowering effect of SGLT2i is proportional to eGFR [
21] and is supposed to be independent from beta cell function. Thus, it remains unclear why longer diabetes duration was associated with lower proportions of patients attaining HbA1c targets among patients initiated on dapagliflozin versus DPP-4i. While we cannot exclude this finding is due to residual confounding, an inverse association between diabetes duration and glycaemic effectiveness of SGLT2i was noted in prior real-world studies [
31,
32], and deserves future investigation. It should also be mentioned that, during the study period, SGLT2i and DPP4i could be prescribed only by diabetes specialists making results not immediately transferrable to primary care. Finally, we acknowledge that the duration of observation was short (7.5 months on average) and a longer follow-up may provide different data with regards to persistence of the glucose-lowering effect of the two treatments.
Nonetheless, our data contribute to building evidence on the effectiveness of SGLT2i in elderly patients with T2D. We detected that a smaller proportion of patients attained individualized targets with dapagliflozin than with DPP-4i, but extra-glycaemic effects of dapagliflozin were preserved in this elderly population and were stronger than those exerted by DPP-4i. In view of the expected benefits of SGLT2i on hard outcomes in this specific population, we advocate for randomized controlled trials dedicated to elderly people with T2D.
Acknowledgements
We would like to thank the technical support of Alessia Russo, Italian Diabetes Society. Composition of the DARWIN-FUP study network: Federica Ginestra, Gloria Formoso, Agostino Consoli (UOC Endocrinologia e Malattie del Metabolismo Ospedale di Pescara). Francesco Andreozzi, Giorgio Sesti (Medicina Interna, Dipartimento di Scienze Mediche e Chirurgiche Università Magna Graecia di Catanzaro Catanzaro). Salvatore Turco (Servizio di Diabetologia - Centro di Medicina Preventiva, Quarto, Napoli). Luigi Lucibelli (Ambulatorio di Diabetologia, ASL3 Napoli 3 Sud, distretto 56). Adriano Gatti (U.O. Malattie del Metabolismo, Presidio Ospedaliero San Gennaro, Napoli). Raffaella Aldigeri, Alessandra Dei Cas (Endocrinologia Azienda Ospedaliera Universitaria di Parma). Giuseppe Felace (SOC Medicina Interna Spilimbergo, Pordenone). Patrizia Li Volsi (S.S.D. di Endocrinologia e Malattie del metabolismo, ASFO Pordenone). GianPio Sorice, Andrea Giaccari (Endocrinologia e Diabetologia, Policlinico Universitario Agostino Gemelli, Università Cattolica del Sacro Cuore, Roma). Carmen Mignogna, Raffaella Buzzetti (U.O. Dipartimentale di Diabetologia, Dipartimento di Medicina Sperimentale, Azienda Ospedaliero-Universitaria Policlinico Umberto I, Sapienza Università di Roma, Roma). Tiziana Filardi, Susanna Morano (U.O.S. Diabetes Complications, Clinica Medica V, Azienda Ospedaliera-Universitaria Policlinico Umberto I, Dipartimento di Medicina Sperimentale, Università di Roma “La Sapienza”). Ilaria Barchetta, Maria Gisella Cavallo (UOS Diabetologia; Azienda Ospedaliera Policlinico Umberto, Università di Roma “La Sapienza”, Roma). Ilaria Malandrucco, Simona Frontoni (UOC Endocrinologia, Diabetologia e Malattie Metaboliche Ospedale Fatebenefratelli Isola Tiberina Roma). Silvia Carletti, Paola D’Angelo (UOC Diabetologia - Ospedale Sandro Pertini). Gaetano Leto, Frida Leonetti (UOC Diabetologia Universitaria dell'Ospedale SM Goretti, Latina). Paola Silvia Morpurgo, Paolo Fiorina (S.S.D. Malattie Endocrine – Diabetologia, ASST Fatebenefratelli Sacco, P.O. Fatebenefratelli e Oftalmico, Milano. Università degli Studi di Milano). Eva Palmieri, Emanuela Orsi (UO Endocrinologia e Malattie Metaboliche, Fondazione IRCCS Ca' Granda-Ospedale Maggiore Policlinico di Milano). Enzo Mantovani (SSD Diabetologia e Malattie Metaboliche, Presidio Ospedaliero di Mantova). Ivano Franzetti (SSD Endocrinologia-Diabetologia, ASST Valle Olona). Fabrizio Querci (Diabetologia, ASST Bergamo Est). Antonio Bossi (Diabetologia, ASST Bergamo Ovest). Federica Turchi (UOC Diabetologia e Malattie Metaboliche INRCA Ancona). Silvana Manfrini (U.O.S.D Diabetologia e Malattie Metaboliche, Senigallia). Danila Guida, Giuseppe Placentino (SOSD Endocrinologia e Diabetologia, ASL VCO). Guglielmo Beccuti, Fabio Broglio (S.C.D.U. Endocrinologia, Diabetologia e Malattie del Metabolismo A.O.U. Città della Salute e della Scienza di Torino - Università degli Studi di Torino). Franco Cavalot (SSD Malattie Metaboliche e Diabetologia, A.O.U. San Luigi Gonzaga, Orbassano, Torino). Alessandro Nuzzo, Gianluca Aimaretti (Endocrinologia, Azienda Ospedaliera di Novara). Olga Lamacchia (Endocrinology Unit, Department of Medical and Surgical Sciences, University of Foggia). Angelo Cignarelli, Luigi Laviola, Francesco Giorgino (Section of Internal Medicine, Endocrinology, Andrology and Metabolic Diseases; Division of Endocrinology, University Hospital Policlinico Consorziale, Bari). Eleonora Devangelio (Ambulatorio e Day Service di Diabetologia, DSS2 ASL Taranto). Giuliana Cazzetta (ASL Lecce - DSS Gagliano del Capo – Tricase, Lecce). Roberta Chianetta, Roberto Citarrella (Policlinico Universitario di Palermo). Andrea Tumminia, Lucia Frittitta (Centro di Diabetologia, Azienda Ospedaliera Garibaldi, Catania). Massimiliano Anzaldi, Massimo Buscema (U.O.C. di Endocrinologia e Malattie del Ricambio e della Nutrizione, A.O.E. Cannizzaro - Catania). Salvatore Piro, Antonino Di Pino, Francesco Purrello (UOC Medicina Interna, Azienda Ospedaliera Garibaldi, Catania). Antonino Di Benedetto, Giuseppina Russo (Azienda Ospedaliera Universitaria di Messina). Roberto Anichini (U.O. Diabetologia USL Centro, Area Pistoiese). Anna Solini (Department of Surgical, Medical, Molecular and Critical Area Pathology University of Pisa). Monia Garofolo, Stefano Del Prato (UO Malattie Metaboliche e Diabetologia; Azienda Ospedaliera Universitaria Pisana). Bruno Fattor (Servizio di Diabetologia, Ospedale Centrale di Bolzano, Azienda Sanitaria Alto Adige). Gian Paolo Fadini, Angelo Avogaro (UOC Malattie del Metabolismo, Azienda Ospedale Università di Padova). Annunziata Lapolla, Giovanni Sartore (Servizio di Diabetologia ULSS6, Padova). Michele D’Ambrosio, Virgilio Da Tos (UOSD Diabetologia, ULSS6 Monselice, Padova). Vera Frison, Natalino Simioni (Medicina Interna e Diabetologia, Ospedale di Cittadella, Padova). Massimo Cigolini, Enzo Bonora (UOC Endocrinologia, Diabetologia e Malattie Del Metabolismo, Azienda Ospedaliera Universitaria Integrata di Verona). Elisabetta Brun, Marco Strazzabosco, (UOC Malattie Endocrine, del Ricambio e della Nutrizione, AULSS 8 Berica). Maurizio Poli (Servizio di Diabetologia, Ospedale San Bonifacio, AULSS9 Verona). Mauro Rigato, Agostino Paccagnella (UOC Malattie Endocrine, del Ricambio e della Nutrizione Azienda AULSS 2 della Marca Trevigiana). Carmela Vinci (Servizio di Diabetologia, ULSS4 Veneto).
Declarations
Conflict of interest
MLM received lecture fees or advisory board fees from Eli Lilly, MSD, Novo Nordisk, SLAPharma, and Servier. IR Advisory Board: AstraZeneca, Eli Lilly and Company, Novo Nordisk Consultant: AstraZeneca, Curolife, BOL Pharma Speaker’s Bureau: AstraZeneca, Eli Lilly and Company, Novo Nordisk. AC has consulting relationships with Astra Zeneca and Novo Nordisk; has participated in advisory panels for Abbot, Astra Zeneca, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Mundipharma, Novo-Nordisk, Sanofi-Aventis, Takeda; has received speaker fees from Abbot, Astra Zeneca, Boehringer Ingelheim, Bruno Farmaceutici, Eli Lilly, Merck Sharp & Dohme, Mundipharma, Novo-Nordisk, Sanofi-Aventis, Takeda; has received research supporting grants from Astra Zeneca, Eli Lilly, Novo Nordisk. MR received lecture and advisory board fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Mundipharma, and Novo Nordisk. AL received grant support, lecture or advisory board fees from Novo Nordisk, Sanofi, Abbott, and Eli Lilly. FB was a consultant for Boehringer Ingelheim–Eli Lilly and Company and received lecture fees from Boehringer Ingelheim–Eli Lilly and Company and AstraZeneca. EB received honoraria from Eli Lilly and Company, Novo Nordisk, and Sanofi; travel support from Eli Lilly and Company and Servier Laboratories; and has participated on advisory boards for Abbott, AstraZeneca, Bayer, Becton Dickinson, Boehringer Ingelheim, Daiichi-Sanyo, Eli Lilly and Company, MSD, Novo Nordisk, and Sanofi. AA received research grants, lecture, or advisory board fees from Merck Sharp & Dome, AstraZeneca, Novartis, Boeringher-Ingelheim, Sanofi, Mediolanum, Janssen, Novo Nordisk, Lilly, Servier, and Takeda. GPF received lecture or consultancy fees or grant support from Abbott, AstraZeneca, Boehringer, Lilly, Merck-Sharp-Dome, Mundipharma, Novartis, Novo Nordisk, Sanofi, and Servier.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.