Long-term Trends in Antidiabetes Drug Usage in the U.S.: Real-world Evidence in Patients Newly Diagnosed With Type 2 Diabetes

A broad choice of “old” and “new” antidiabetes drugs (ADDs) is available, which differ not only in their mechanisms of action but also in their glycemic and extraglycemic effects (1). While treatment guidelines for type 2 diabetes are regularly updated based on new evidence, real-world prescription trends may also be driven by other factors, such as medication costs, side effect profile, and provider and patient preferences. With the development of new classes of antidiabetes therapies since 2005, including incretin-based drugs and sodium–glucose cotransporter 2 inhibitors (SGLT2i), the paradigm of therapy options for patients with highly heterogeneous glycemic and cardiovascular risk factors has changed significantly. However, the way in which this has occurred in real-world practice, especially in the trade-off between older and new classes of ADDs as initial and intensification therapy options, has not been studied thoroughly.

The newer ADDs have been shown to be associated with significantly lower risk of hypoglycemia compared with the sulfonylureas (SU) and insulin (INS) (2). The weight neutrality or benefits of weight reductions have also been well established for new therapies including the incretins (3,4). Given the glycemic and extraglycemic benefits of these agents, one would expect a fall in the use of SU or INS as intensification therapies. However, studies evaluating the possible benefits of using newer ADDs in terms of delaying the need for INS are scarce (5,6). In this context, understanding the changing patterns of therapy initiation and intensifications with second- and third-line therapies, in conjunction with the heterogeneous patients’ characteristics, is a fundamental background requirement.

Cohen et al. (7) explored ADD-prescribing patterns in the U.S. from 1997 to 2000 and reported decreasing use of SU and increasing trends in metformin (MET) and thiazolidinedione (TZD) prescription over time. Utilizing a claims database, Desai et al. (8) reported an increasing proportional share of MET and decreasing prescriptions for TZD between 2006 and 2008. One of the reasons for the reduction in the use of TZD was the safety concerns (9–12). While Berkowitz et al. (13) evaluated treatment initiations with MET, SU, and dipeptidyl peptidase-4 inhibitors (DPP-4i) between 2009 to 2013 in the U.S., utilization patterns of other ADDs and the changing utilization trend over time were not explored (14). Lipska et al. (15) have evaluated the temporal trend in the use of ADDs from 2006 to 2013 using claims data from the U.S. A small number of studies have explored the clinical characteristics of patients according to the type of ADD prescribed, but only over relatively short time periods, and did not evaluate treatment intensification with second- or third-line ADDs over a long period of time (16,17).

To the best of our knowledge, the progressive changes in the proportional distributions across all new and old ADDs, and the patterns and determinants of therapy intensification with second- and third-line ADDs, have not been explored comprehensively in any study. With recognition of the growing disease burden and increasing volumes of dispensed medications (18–21), the primary aim of this study was to provide a comprehensive up-to-date exploration of the treatment pattern changes for type 2 diabetes in the U.S. using the nationally representative Centricity Electronic Medical Records (CEMR) from primary and secondary ambulatory care systems. Specifically, the aims were to 1) explore temporal changes in prescribing patterns from 2005 to 2016 with respect to the drug initiation order, 2) explore therapy intensification with second and third ADDs, 3) explore patient characteristics including risk factors and comorbidities according to ADD therapy prescribed, 4) explore the temporal patterns in the rates of intensification with SU and INS, and 5) evaluate whether use of incretin-based therapies as second-line therapy delays the need for intensification with third-line ADDs and with INS any time during follow-up.

CEMR represents a variety of ambulatory and primary care medical practices, including solo practitioners, community clinics, academic medical centers, and large integrated delivery networks in the U.S. More than 34 million individuals’ longitudinal electronic medical records (EMRs) were available from 1995 to April 2016. More than 35,000 physicians and other providers from all U.S. states contribute to the CEMR, of whom ∼75% are primary care providers. The database is generally representative of the U.S. population, with a diabetes prevalence of 7.1% (identified by diagnostic codes) that is similar to the national diabetes prevalence of 6.7% (diagnosed diabetes in 2014) (14). The CEMR has been used extensively for academic research worldwide (3,22,23).

This database contains comprehensive patient-level information on demographic, anthropometric, clinical, and laboratory variables including age, sex, ethnicity, and longitudinal measures of BMI, blood pressure, glycated hemoglobin (HbA_1c), and lipids. All disease events along with dates are coded with ICD-9, ICD-10, or SNOMED CT codes. Medication data include brand names and doses for individual medications prescribed, along with start/stop dates and specific fields to track treatment alterations. This data set also contains patient-reported medications, including prescriptions received outside the EMR network and over-the-counter medications.

Eleven antidiabetes therapeutic classes were considered in this study: MET, SU, TZD, α-glucosidase inhibitors (AGI), amylin, dopamine receptor agonists (DOPRA), meglitinides (MEG), DPP-4i, glucagon-like peptide 1 receptor agonists (GLP-1RA), SGLT2i, and INS. For each patient, these ADDs were arranged chronologically according to the initiation dates. Same-day initiations (including combination therapies) were prioritized in the order as listed above, with highest order priority assigned to MET and lowest to INS. Additions or switches were defined by comparing stop dates and start dates of corresponding therapies. Details on the medication data structure, associated data-mining challenges, and description of an algorithm applied to extract and aggregate patient-level medication data from CEMR have recently been published (24).

For convenience, AGI, amylin, DOPRA, and MEG were combined into the “other” category. Saxenda (a version of liraglutide) was excluded from the GLP-1RA list, as it was approved in 2014 for weight lowering and not as an ADD (25). Although Welchol (colesevelam) was approved for the treatment of type 2 diabetes, it was mainly prescribed to reduce cholesterol levels; therefore, we did not include colesevelam in our analyses (18).

Patients with diabetes were identified on the basis of diagnostic codes; those with a diagnosis of type 1 diabetes or only gestational diabetes mellitus were identified and excluded. For identified patients with type 2 diabetes, the following inclusion criteria were applied: 1) age at diagnosis ≥18 and <80 years, 2) diagnosis date strictly after first registered activity in the database, 3) diagnosis date on or after 1 January 2005, and 4) initiation of any ADD.

Demographic variables included sex, age, and ethnicity. HbA_1c values on the date of diagnosis and first-, second-, and third-line ADD therapy initiation were obtained as the closest observations within a 3-month window. Body weight, BMI, systolic/diastolic blood pressure (SBP/DBP), lipids (LDL, HDL, and triglycerides), and heart rate were calculated as the average of available measurements within a 3-month window of the diagnosis or ADD initiation date. Obesity was defined as BMI ≥30 kg/m².

The presence of comorbidities prior to the first and second drug initiation was explored. Cardiovascular disease (CVD) was defined as ischemic heart disease (including myocardial infarction), peripheral vascular disease, heart failure, or stroke. Cancer was defined as any malignancy except malignant neoplasm of skin. Charlson Comorbidity Index was defined and calculated following the algorithm described by Quan et al. (26).

The characteristics of patients were summarized by ADD classes—at first prescription and at second ADD initiation when added to MET. Separate analyses were conducted to explore the pattern of addition or switch to third ADD by major classes of second-line ADDs. Study variables were summarized as number (%), mean (SD), or median (first quartile [Q1], third quartile [Q3]) as appropriate. In patients who had a second-line ADD added after MET, and had at least 1 year of follow-up post–second-line initiation, the “restricted mean survival time” estimation approach was used to compare the mean time to the third ADD/INS initiation among major second-line ADD groups. This method computes survival time as time to third ADD/INS if initiated, and otherwise as time to the end of follow-up (date of patient’s last available record within the database). Standard life table methods were used to estimate rates per 100 person-years (95% CI) of second-line ADD, INS, and SU initiations in patients with a minimum of 1 year follow-up post–MET initiation.

From the 2,624,954 patients identified with type 2 diabetes, 2,590,853 were aged between ≥18 and <80 years at the date of diagnosis with mean/median 3.9/2.7 years of follow-up. Of these patients, 1,305,686 (50%) were newly diagnosed after 1 January 2005, while 1,023,340 patients initiated any ADD (study cohort) (Supplementary Fig. 1) during the available mean/median 3.4/2.8 years of follow-up time. In the study cohort, 46% were male, mean (SD) age was 58 years (13), and 68% were white Caucasians, 12% blacks, and 2% Asians (Table 1).

Figure 1A presents the proportional distribution of the first-line ADD by year of initiation. The proportional share of MET as the first choice increased consistently from 60% in 2005 to 77% in 2016 (first quarter of the year). SU’s share declined from 20 to 8%, while INS’s share ranged from 8 to 10%. Starting at 11% in 2005, TZD’s proportional share dropped progressively to 0.7% in 2016. Other drugs were chosen as first-line in 3% of cases or less.

In the study cohort of 1,023,340 patients, the distribution of prescription patterns for individual ADDs at any time from January 2005 to April 2016 and as the first ADD are presented in Table 1. The demographic and clinical characteristics of the patients, along with the prevalence of comorbidities at the time of first ADD initiation, are also presented in Table 1.

In the study cohort, 79% received MET any time during the recorded follow-up, and 72% received MET as the first ADD. The mean time to initiation of MET as the first ADD and the available follow-up time since initiation were 3.7 months and 3.3 years, respectively. The proportions of patients with HbA_1c level ≥7.5% (58 mmol/mol) and 8% (64 mmol/mol) at MET initiation were 48% and 37%, respectively. Those who initiated with GLP-1RA, DPP-4i, TZD, and SU had a similar mean HbA_1c of 8.0, 7.7, 7.8, and 8.0%, respectively (64, 61, 62, and 64 mmol/mol). INS was initiated at an average HbA_1c of 8.9% (74 mmol/mol), with 71% and 59% having HbA_1c ≥7.5% (58 mmol/mol) and 8% (64 mmol/mol), respectively.

Patients who initiated treatment with MET were younger (mean age 57 years, with 19% ≥70 years) than those who initiated with SU (mean age 64 years, 43% ≥70 years), with INS (mean 60 years, 29% ≥70 years), with TZD (mean 62 years, 32% ≥70 years), or with DPP-4i (mean 64 years, 39% ≥70 years). Those who had GLP-1RA and SGLT2i as the first ADD were younger, more likely to be white Caucasian, female, and obese, as compared with those who initiated with MET, DPP-4i, INS, TZD, or SU.

The proportions of patients with CVD, chronic kidney disease (CKD), cancer, or depression at first ADD initiation were 19%, 4%, 4%, and 11%, respectively. Those patients initiating therapy with INS had a significantly higher prevalence of CVD (27%, P < 0.01), CKD (11%, P < 0.01), and higher Charlson Comorbidity Index with mean (SD) of 1.84 (1.31), compared with those initiating with MET, DPP-4i, GLP-1RA, or TZD (Table 1 for comparative estimates).

Among patients with at least 1 year of follow-up (n = 813,826), the proportions of patients discontinuing the first-line ADD within 1 year by individual ADDs are presented in Table 1. While only 8% of patients discontinued MET within a year, 20%, 17%, and 25% of patients discontinued GLP-1RA, DPP-4i, and SGLT2i within a year, respectively.

Among 740,478 patients who initiated therapy with MET, 357,482 (48%, subcohort 1) (Supplementary Fig. 1) initiated a second ADD, with an annual mean rate of 10.7 initiations per 100 person-years (minimum 10.2, maximum 14.0) during a mean 3.3 years of available follow-up, at an average HbA_1c level of 8.4% (68 mmol/mol), with 60% and 48% having HbA_1c ≥7.5% (58 mmol/mol) and 8.0% (64 mmol/mol), respectively. The proportional share of second-line ADD (post-MET) over time is presented in Fig. 1B. The demographic and clinical characteristics of the patients along with the time to second ADD, and the prevalence of comorbidities at the time of second drug initiation, are presented in Table 2.

Although the proportional share of SU as a second-line therapy gradually decreased from 60 to 46% over time (Fig. 1B), it remained the most popular choice (53%) of therapy intensification post–MET initiation across the whole time period. SU was initiated as second-line ADD at an average HbA_1c level of 8.4% (68 mmol/mol), with 62% and 49% having HbA_1c ≥7.5% (58 mmol/mol) and 8.0% (64 mmol/mol), respectively (Table 2). Among patients with a second ADD and a minimum 1 year of follow-up, only 10% discontinued SU within 1 year compared with significantly higher discontinuation proportions in other second-line non-INS ADDs.

The proportional share of DPP-4i as a therapy intensification option post–MET initiation sharply increased from 0.4% in 2006 (approved in October 2006) to 20% in 2016 (Fig. 1B). DPP-4i were initiated at an average HbA_1c of 8.2% (66 mmol/mol), with 57% and 43% having HbA_1c ≥7.5% (58 mmol/mol) and 8.0% (64 mmol/mol), respectively. While 18% discontinued DPP-4i within a year of initiation, the proportions of patients discontinuing second-line GLP-1RA, TZD, or SGLT2i within a year were higher.

The proportional share of patients receiving GLP-1RA as a second ADD increased from 3% in 2006 to 7% in 2016. Initiation of GLP-1RA occurred at relatively lower HbA_1c levels of 7.8% (62 mmol/mol) and at the highest BMI levels among second-line ADD groups. Twenty-one percent of patients discontinued GLP-1RA therapy within a year of commencing it as a second ADD. After approval of the first SGLT2i in 2013, the proportional share of those receiving it as a second ADD reached 7% in 2016. One-quarter of patients discontinued SGLT2i therapy within a year of adding it as second-line ADD. The proportional share of patients receiving TZD as a second-line therapy dropped from 30 to 4% (Fig. 1B), with 21% of patients discontinuing therapy within 1 year.

The proportional share of patients receiving INS as a second ADD post–MET initiation has consistently increased from 7% in 2005 to 17% in 2016 (Fig. 1B). The intensification with INS occurred at a 9.3% (78 mmol/mol) average HbA_1c level, with 73% and 64% having HbA_1c ≥7.5% (58 mmol/mol) and 8.0% (64 mmol/mol), respectively. Only 7% patients discontinued INS within 1 year of initiation.

Among patients in subcohort 1, 78% had at least 1 year of follow-up from the second ADD initiation (subcohort 2; n = 278,188). Of these patients, 144,106 (52%) initiated a third ADD, with an annual mean rate of 12.6 initiations per 100 person-years (minimum 11.4, maximum 14.9) during a mean follow-up of 4 years post–second-line initiation. Table 3 presents treatment intensification patterns by the major second-line ADDs. Most of the patients (84% [n = 121,559]) added a third drug on top of the second ADD, while 16% (n = 22,547) ceased the second ADD and switched to a third ADD. Addition of the third drug occurred at higher HbA_1c levels (8.5% [69 mmol/mol]) compared with switching (8.2% [66 mmol/mol]).

Among patients with SU as the second ADD, 49% (n = 73,776) added and 6% (n = 8,204) switched to a third drug during a mean follow-up of 4.1 years. The most popular third ADD addition was DPP-4i (34% of those who added a third ADD), followed by INS (28%) and TZD (26%). Among those who switched, almost one-half (49%) switched to INS, while 30% and 8% switched to DPP-4i and GLP-1RA, respectively.

SU, DPP-4i, and GLP-1RA were added to INS in 32%, 26%, and 22% of patients (from those who added a third ADD), respectively. Only 3% of patients ceased INS therapy to switch to another ADD during a mean 3.6 years of follow-up. In the second-line DPP-4i group (n = 46,822), 40% added and 11% switched to a third drug during a mean 3.4 years of follow-up. The most popular third ADD addition was SU (40% of those who added a third ADD), followed by INS (29%) and GLP-1RA (9%). Of those who switched from DPP-4i, one-half of the patients moved to SU, followed by INS and GLP-1RA (17%).

Among those who had a GLP-1RA as the second ADD, 52% added INS (of those who added a third ADD) and 18% switched to INS (of those who switched to a third ADD) during a mean 3.9 years of follow-up; 11% added and 34% switched to DPP-4i. In the TZD group, 43% added and 22% switched to a third ADD during 5.4 years of follow-up. Among those who switched, 45% chose SU while 35% moved to DPP-4i.

Among patients with first-line MET and a minimum 1 year of follow-up, the annual rates per 100 person-years of INS/SU initiation (irrespective of order of therapy intensification) are presented in Fig. 1C. The rates did not significantly decline from 2005 to 2014.

The Kaplan-Meier analyses, based on restricted mean years to adding or moving to a third ADD, in major second-line ADD groups are presented in Table 3. The mean time to intensification with a third ADD was marginally longer in incretin groups (DPP-4i 4.1 years [95% CI 4.1, 4.2] and GLP-1RA 4.2 years [4.1, 4.3]) compared with that in patients with SU as the second-line ADD (3.9 years [3.8, 3.9]; P = 0.04). The restricted mean times to intensification with INS any time during follow-up were 6.3, 7.1, and 6.6 years in the SU, DPP-4i, and GLP-1RA groups, respectively (all comparative P < 0.05).

This longitudinal exploratory study of a large cohort of patients with type 2 diabetes observed between 2005 and 2016 from primary and ambulatory care systems in the U.S. provides 1) a detailed account of glycemic states, clinical characteristics, and comorbidities at first-line and second-line therapy initiation by different drug classes, as well as new insights into 2) the changes in the choice of first- and second-line ADDs over the last 10 years, 3) patterns of therapy intensification with third-line ADDs and with INS, separately for major second-line ADDs, 4) changes in the annual rates of therapy intensification with SU and INS over time, and 5) possible benefits of using newer novel antidiabetes therapies in terms of delaying the need for third-line therapy intensification, including the need for initiating INS.

With 3.4 years of mean follow-up in more than one million patients with a diagnosis of type 2 diabetes from 2005, this study provides robust and detailed information on the changing clinical practices for the management of type 2 diabetes in a real-world setting. We are not aware of any study that simultaneously evaluated the changing prescribing patterns of old and new ADDs as first-line therapy and as intensification options at various levels of glycemia and comorbidities.

The proportional share of MET as the first-line therapy choice has increased from 60 to 77%, while that for SU has decreased from 20 to 8%, over the last decade. However, SU continue to be the most popular second-line therapy intensification option, although with a declining share (from 60 to 46% over the last decade). The discontinuation rate of SU was found to be the lowest among non-INS second-line ADDs. Among those who intensified with a third-line therapy, the ratio of addition to switching to third ADD was highest in the SU group (9.0), followed by DPP-4i (3.8) and GLP-1RA (3.2), during >4 years of mean follow-up post–second-line ADD initiation.

We observed that second-line SU users initiate a third ADD marginally sooner compared with incretin users. A study based on EMR data from the U.K. reported the opposite results, with an average of 1.6 and 2.4 years to third ADD initiation in the 3,080 and 15,508 patients treated with MET + DPP-4i and MET + SU, respectively (5).

The proportions of patients who added INS were similar between patients who had a DPP-4i and an SU as the second ADD. However, among those who switched to a third ADD, only 17% of patients in the DPP-4i group switched to INS, compared with almost 50% in the SU group. We also observed that the mean time to INS initiation was significantly shorter for second-line SU users (6.3 years) than in the DPP-4i group (7.1 years). This finding is similar to a study (2015) based on 3,864 matched pairs of patients treated with DPP-4i or SU when added to MET, where Inzucchi et al. (6) reported that those treated with DPP-4i were significantly less likely to initiate INS compared with those treated with SU.

We observed an increasing proportional share of INS as a second-line therapy option over the last 10 years, despite the availability of novel therapies that were found to have similar or better glycemic efficacy in clinical trials. Also, the annual rates of intensification with INS remained similar over the last decade. In a similar study, Lipska et al. (15) observed that the overall rate of severe hypoglycemia did not reduce from 2006 to 2013. This may reflect pressure to achieve glycemic targets rapidly and an increasing recognition that for people with very poor glycemic control, INS may be the only drug likely to achieve targets.

Compared with rates for older ADDs, high discontinuation rates of new therapeutic classes, particularly of DPP-4i, are surprising. The higher cost of these newer drugs may be relevant and may also contribute to the fairly low rates of initiation of these drugs overall. More studies, utilizing additional data sources, are needed to specifically test hypotheses for the differences in initiation, adherence, and persistence between drug classes.

The HbA_1c level at pharmacological therapy initiation was found to be 8.2% (66 mmol/mol), with 50% having HbA_1c ≥7.5% (58 mmol/mol). The HbA_1c levels at first-line ADD initiation were similar across all ADDs, except in those who initiated with INS, whose average HbA_1c was 8.9% (74 mmol/mol). Although the mean time to second ADD post–MET initiation was only 8 months, it occurred at a high HbA_1c level of 8.4% (68 mmol/mol), with 60% and 48% of patients having HbA_1c ≥7.5% (58 mmol/mol) and 8% (64 mmol/mol), respectively. Among those with a minimum of 1 year of follow-up post–second ADD, ∼52% intensified with a third ADD at an average HbA_1c level of 8.5% (69 mmol/mol). These findings reflect the continued glycemic risk burden in patients with type 2 diabetes (27–30). While the persistent therapeutic inertia (28) and the long-term consequences of therapeutic inertia (27,29) for glycemic control in primary care systems have been evaluated, exploration of the glycemic state at therapy initiation and intensification by ADD classes is scarce. Our study provides a detailed account of the glycemic state in people with type 2 diabetes at therapy initiation and intensifications during a reasonable follow-up period in primary and ambulatory care settings.

The main strength of this study is the availability of data from the patients’ medication lists that included prescribed medications within the EMR network and also medication information that could be prescribed outside of the EMR, as well as data on glycemic control and comorbidities. The CEMR database tracks longitudinal treatment adjustments and contains comprehensive clinical information, which is usually not available in claims databases.

The limitations of this study include the nonavailability of complete and reliable data on 1) medication adherence and side effects, 2) diet and exercise, 3) socioeconomic status, and 4) insurance type. We did not include dosage changes or brands’ distribution in our analyses. The findings of this study should be interpreted with caution: EMR data are in general biased toward unhealthy populations and commercially insured individuals, white Caucasians are overrepresented in the CEMR, and the results are subject to limited follow-up.

Less popular ADDs such as MEG, AGI, DOPRA, and amylin were included in our study for multiple reasons: first, to assess utilization data of such medications, as these drugs are usually omitted, and second, to ensure market shares of other drugs are not overestimated. We observed that only 39,549 patients with type 2 diabetes were using SGLT2i during the available follow-up period, which is not surprising, given that they were first approved in 2013.

To conclude, while we have observed significant increase in the use of MET as the first-line therapy over the last 10 years, the second- and third-line therapy intensification choices are highly heterogeneous. While increasing popularity of “new” drugs, especially DPP-4i and SGLT2i, was observed as the second and third drugs choices, SU remain the most popular therapy intensification choice and have a lower discontinuation rate compared with other non-INS ADDs. The proportional share of INS as a second-line therapy choice has also increased significantly over the last decade. Incretin-based therapies were found to delay the need for therapy intensification only marginally compared with other ADDs. Contrary to the guidelines for proactive glycemic management, pharmacological therapy initiation and the intensifications occurred at very high levels of HbA_1c, with 48% of patients having HbA_1c ≥8.0% (58 mmol/mol) at second-line therapy initiation.

Acknowledgments. O.M. acknowledges her cosupervisors Ross Young and Louise Hafner of Queensland University of Technology.

Funding. J.J.A. and S.K.P. acknowledge project grant support provided by the Royal Brisbane and Women’s Hospital Foundation. O.M. acknowledges a PhD scholarship from Queensland University of Technology. J.S. is supported by a National Health and Medical Research Council Research Fellowship. Melbourne EpiCentre received support from the National Health and Medical Research Council and the Australian Government’s National Collaborative Research Infrastructure Strategy initiative through Therapeutic Innovation Australia.

Duality of Interest. J.S. has received speaker honoraria, consultancy fees, and/or travel sponsorship from AstraZeneca, Boehringer Ingelheim, Lilly, Sanofi, Mylan, Novo Nordisk, Merck Sharp & Dohme, and Novartis. J.J.A. has received speaker honoraria, consultancy fees, and/or travel sponsorship from AstraZeneca, Boehringer Ingelheim, Lilly, and Novartis. S.K.P. has acted as a consultant and/or speaker for Novartis, GI Dynamics, Roche, AstraZeneca, Guangzhou Zhongyi Pharmaceutical, and Amylin Pharmaceuticals. S.K.P. has received grants in support of investigator and investigator-initiated clinical studies from Merck, Novo Nordisk, AstraZeneca, Hospira, Amylin Pharmaceuticals, Sanofi, and Pfizer. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. O.M. and S.K.P. were responsible for the primary design of the study. J.S. and J.J.A. contributed significantly in the study design. O.M. conducted the data extraction. O.M. and S.K.P. jointly conducted the statistical analyses. F.S. contributed in the interpretation of the results. The first draft of the manuscript was developed by O.M. and S.K.P., and all authors contributed to the finalization of the manuscript. S.K.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.