Copyright. Jenny Chun Ye Wong

Transcription

1 Copyright by Jenny Chun Ye Wong 2016

2 The Thesis Committee for Jenny Chun Ye Wong Certifies that this is the approved version of the following thesis: Evaluation of Glycemic Control and Medication Utilization Patterns in Patients with Type 2 Diabetes Mellitus on Sodium Glucose Cotransporter 2 Inhibitors Compared to Dipeptidyl Peptidase IV Inhibitors APPROVED BY SUPERVISING COMMITTEE: Supervisor: Co-Supervisor: Karen L. Rascati Co-Supervisor: James P. Wilson Paul J. Godley

3 Evaluation of Glycemic Control and Medication Utilization Patterns in Patients with Type 2 Diabetes Mellitus on Sodium Glucose Cotransporter 2 Inhibitors Compared to Dipeptidyl Peptidase IV Inhibitors by By Jenny Chun Ye Wong, B.S., PharmD Thesis Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Master of Science in Pharmaceutical Sciences The University of Texas at Austin May 2016

4 Dedication I dedicate this thesis to, My Husband My Parents My Brother For their unconditional love, patience and support, For supporting me in my pursuit of academic achievement

5 Acknowledgements I would like to thank Dr. Paul Godley, Dr. Jim Wilson, Dr. Karen Rascati and Dr. Delaney Ivy for their guidance and support throughout my fellowship and for serving on my thesis committee. I would also like to thank Dr. Laurel Copeland and Dr. I-Chia Liao for their efforts and assistance in providing laboratory data for analysis and statistical expertise. v

6 Abstract Evaluation of Glycemic Control and Medication Utilization Patterns in Patients with Type 2 Diabetes Mellitus on Sodium Glucose Cotransporter 2 Inhibitors Compared to Dipeptidyl Peptidase IV Inhibitors Jenny Chun Ye Wong, MSPS The University of Texas at Austin, 2016 Supervisors: James P. Wilson and Karen L. Rascati Objectives: To compare patient demographics, glycemic control and medication utilization patterns in type-2 diabetes patients initiating Sodium Glucose Cotransporter 2 Inhibitors (SGLT2-I) or Dipeptidyl Peptidase IV Inhibitors (DPP4-I). Methods: Retrospective analysis of medical and pharmacy claims from the Scott and White Health Plan. Patient > 18 and < 62 years old with a diagnosis of type 2 DM ([ICD-9-CM] of 250.X0 or 250.X2) with a six-month continuous enrollment criterion during the study period were included. Patients on either a SGLT2-I or DPP4-I and a glucagon like peptide-1 agents were excluded. Demographic, clinical, and economic characteristics were assessed at baseline. Adherence was measured using proportion of days covered greater than 80% and persistence was defined as number of days until treatment discontinuation with a gap in therapy of 45 days. Mann-Whitney U or chi- vi

7 square tests were used to compare cohorts. Multivariable linear regression models assessed baseline factors associated with change in A1C. Results: Analysis included a total of 99 SGLT2-I and 201 DPP4-I patients. The average age (50.9 vs 52.2, p=0.1559), and Charlson comorbidity index (2.25 vs 2.11, p=0.1359) were comparable. The SGLT2-I cohort had a lower proportion of patients receiving sulfonylureas (45% vs. 65%, p=0.0011), a higher proportion receiving insulin (62% vs. 31%, p <0.001) and a higher baseline A1C (8.9 vs. 8.3, p=0.0324). The SGLT2-I cohort was found to be less persistent than the DPP4-I cohort with a mean time to treatment discontinuation of days compared to days (p = ). Though the change in A1C was not statistically significant, clinically the SGLT2-I had a larger decrease in A1C of -0.7 compared to -0.5 in the DPP4-I cohort. Conclusion: In our sample, there was no statistical difference in the improvement of glycemic control (A1C) between a SGLT2-I or a DPP4-I. Medication utilization patterns also did not differ between either cohorts, except that patients in the SGLT2-I cohort were found to be less persistent than the DPP4-I cohort. Limitations of the study include a small sample size and a limited study period of one year. vii

8 Table of Contents List of Tables... ix List of Figures...x CHAPTER 1: INTRODUCTION...1 Epidemiology of Diabetes Mellitus...1 Pathophysiology of DM...1 DM Treatment Guidelines...5 Approaches to Glycemic Treatment...6 Changes in glycemic treatment guidelines from 2012 to present DPP4 inhibitors vs. SGLT2 inhibitors...15 Study Rationale...16 CHAPTER 2: METHODOLOGY...20 Study Objectives and Hypotheses...20 Study Design and Data Source...20 Outcome Measures...22 Analysis Plan...23 Feasibility Analysis...24 CHAPTER 3: RESULTS...25 CHAPTER 4: DISCUSSION...34 Study Limitations...36 Conclusions...37 REFERENCES...38 viii

9 List of Tables Table 1. Criteria for diagnosis of DM... 2 Table 2. Risk factors for Type 2 DM... 4 Table 3. Glycemic goals of therapy by ADA and AACE... 5 Table 4. ADA and AACE guidelines on therapeutic Lifestyle Changes... 6 Table 5. Available insulin preparations... 7 Table 6. Agents for the treatment of type 2 DM (excluding Insulin, see Table 5)... 8 Table 7. Effects of Diabetes Drug Action Table 8. FDA approval dates of DPP4-I and SGLT2-I Table 9. ICD-9-CM Codes Used for Exclusion Criteria Table 10. Study Hypotheses Table 11. Variables analyzed in the study Table 12. Baseline Demographics, Comorbidities, and Laboratory Values Table 13. Change in patients still using insulin six months during, before or after the initiation of index drug Table 14. Summary of hypotheses tested ix

10 List of Figures Figure 1. ADA/EASD 2016 Anti-Hyperglycemic Therapy in Type 2 DM Figure 2. AACE Glycemic Control Algorithm Figure 3. AACE Glycemic Control Algorithm Figure 4. Flow-chart of patient identification Figure 5. Linear Regression Results (N = 1684) Figure 6. Linear Regression Results specific to SGLT2-I and DPP4-I Cohorts x

11 CHAPTER 1: INTRODUCTION Epidemiology of Diabetes Mellitus Diabetes Mellitus (DM) is a metabolic disorder characterized by persistent hyperglycemia caused by the body s resistance to use of or lack of secretion of insulin, or both. It is also associated with abnormalities in the metabolism of carbohydrate, fat and proteins. The Centers for Disease Control and Prevention (CDC) estimates that 9.3% (29.1 million) of the United States (US) population in 2012 have DM, with an additional 1.4 million US adults newly diagnosed every year. Eighty-six million US adults (aged 20 or older) have prediabetes, defined as hyperglycemia above normal but below diabetic diagnostic levels (hemoglobin A1C 5.7-<6.4) with only 11% that have been informed of the diagnosis. 1,2 An economic report by the American Diabetes Association (ADA) in 2012 estimates the total economic burden of diagnosed DM at $245 billion, including $176 billion in direct medical costs and $69 billion in indirect costs (disability, work loss, premature mortality, absenteeism). Anti-diabetic agents and supplies contributed to 12.2% of total costs. Adults with diagnosed DM are also likely to incur 2.3 times more healthcare costs than those without DM. 2,3 Pathophysiology of Diabetes Mellitus The majority of DM cases are classified into two types: 1) Type 1 DM, defined by an absolute deficiency in the production of insulin, and 2) Type 2 DM, defined by insulin resistance or inadequate insulin secretion. Approximately 5% of the diagnosed population has Type 1 DM and 90-95% has Type 2 DM. Gestational diabetes (development of diabetes during pregnancy; GDM) and other uncommon types of DM make up the rest of the diagnosed diabetic population (1-5%). 4-6,9-12 1

12 Diagnosis of Diabetes Mellitus DM is typically diagnosed based on glycemic cut points. Both the ADA and the American Association of Clinical Endocrinologist (AACE) have published criteria for the diagnosis of DM (shown in Table 1). Table 1. Criteria for diagnosis of DM 6,9,12 ADA 1. A1C 6.5 percent. The test should be performed in a laboratory using a method that is NGSP certified and standardized to the DCCT assay.* 2. FPG 126 mg/dl (7.0 mmol/l). Fasting is defined as no caloric intake for at least eight hours.* 3. Two-hour plasma glucose 200 mg/dl (11.1 mmol/l) during an OGTT. The test should be performed as described by the World Health Organization, using a glucose load containing the equivalent of 75-gram anhydrous glucose dissolved in water.* 4. In a patient with classic symptoms of hyperglycemia or hyperglycemic crisis, a random plasma glucose 200 mg/dl (11.1 AACE 1. FPG concentration (after 8 or more hours of no caloric intake) 126 mg/dl 2. Plasma glucose concentration 200 mg/dl 2 hours after ingesting a 75-g oral glucose load in the morning after an overnight fast of at least 8 hours 3. Symptoms of hyperglycemia (e.g., polyuria, polydipsia, polyphagia) and a random (casual, nonfasting) plasma glucose concentration 200 mg/dl 4. A1C level 6.5% (secondary)** mmol/l). A1C: glycated hemoglobin; DCCT: diabetes control and complications trial; FPG: fasting plasma glucose; NGSP: national glycohemoglobin standardization program; OGTT: oral glucose tolerance test. * In the absence of unequivocal hyperglycemia, criteria 1 to 3 should be confirmed by repeat testing.; ** Glucose criteria are preferred for the diagnosis of DM. In all cases, the diagnosis should be confirmed on a separate day by repeating glucose or A1C testing. When A1C is used for diagnosis, follow-up glucose testing should be done when possible to help manage DM Recommendations are either the plasma glucose or A1C measure be repeated on different days to confirm diagnosis. However, a glucose level 200 mg/dl while in the presence of DM symptoms does not need to be confirmed. An International Expert Committee has also recommended that an A1C level 6.5% as a criterion for diagnosis of DM. 12 Type 1 Diabetes Mellitus Type 1 DM develops due to the lack of insulin caused by an autoimmune destruction of beta cells in the pancreas. It often manifests in children and adolescents but can occur at any age. Classic symptoms include polyuria, polyphagia, weight loss and fatigue, followed by hyperglycemia. About 20-40% will present with diabetic ketoacidosis (DKA). Screening is not recommended due to the acute onset of symptoms. 4-6,9-12 2

13 Type 2 Diabetes Mellitus Patients diagnosed with type 2 DM do not normally present with symptoms, though lethargy, polyuria, nocturia and polydipsia can often be found. They are usually characterized by defects in insulin secretion, insulin resistance involving muscle, liver and the adipocyte, excess glucagon secretion, and deficiency in the glucagon-like peptide 1 (GLP-1). Glucagon and insulin are closely linked in that when one increases, the other decreases to keep plasma glucose level within normal limits. The result of hyperinsulinemia suppresses the production of glucose in the liver, enable more glucose uptake in the peripheral tissues and suppress glucagon release. 4,5,6,11,12 Impaired insulin secretion thus results in excess glucose in the blood. Weight gain often leads to insulin resistance as well, hence most patients with type 2 DM are found to be overweight or obese. The CDC estimated that in 2012, 85% of patients with type 2 DM were overweight or obese. Both the ADA and the AACE recommend screening for any individuals in the presence of risk factors for DM (shown in Table 2). 1,2 3

14 6, 9,12 Table 2. Risk factors for Type 2 DM ADA Overweight or obese Physical inactivity First degree relative with diabetes or high-risk ethnicity/race Women who delivered a baby > 9 lbs (>4kg) History of GDM Hypertension High triglycerides Low HDL-C Polycystic ovary syndrome (PCOS) Previous diagnosis of prediabetes Acanthosis nigricans History of cardiovascular disease AACE Age 45 years without other risk factors CVD or family history of Type 2 DM Overweight or obese* Sedentary lifestyle Member of an at-risk racial or ethnic group: Asian, African American, Hispanic, Native American (Alaska Natives and American Indians), or Pacific Islander HDL-C <35 mg/dl (0.90 mmol/l) and/or a triglyceride level >250 mg/dl (2.82 mmol/l) IGT, IFG, and/or metabolic syndrome PCOS, acanthosis nigricans, NAFLD Hypertension (BP >140/90 mm Hg or on therapy for hypertension) History of gestational diabetes or delivery of a baby weighing more than 4 kg (9 lb) Antipsychotic therapy for schizophrenia and/or severe bipolar disease Chronic glucocorticoid exposure Sleep disorders in the presence of glucose intolerance (A1C >5.7%, IGT, or IFG on previous testing),including OSA, chronic sleep deprivation, and night-shift occupation BP = blood pressure; CVD = cardiovascular disease; HDL-C = high-density lipoprotein cholesterol; IFG = impaired fasting glucose; IGT = impaired glucose tolerance; NAFLD = nonalcoholic fatty liver disease; OSA = obstructive sleep apnea; PCOS = polycystic ovary syndrome.*testing should be considered in all adults who are obese (BMI 30 kg/m2), and those who are overweight (BMI 25 to <30 kg/m2) and have additional risk factors. At-risk BMI may be lower in some ethnic groups, in whom parameters such as waist circumference and other factors may be used 4

15 Diabetes Mellitus Treatment Guidelines Hemoglobin A1C goals Reducing the risk of microvascular and macrovascular complications are the primary goals in managing DM. Measurements of hemoglobin A1C remain the gold standard for longterm glycemic control. Depending on the patient history and health status, the A1C may vary. Less stringent A1C goals may be appropriate for patients with a history of hypoglycemia, severe microvascular or macrovascular complications, the elderly or younger children. 4-6,9-12 Published guidelines from both the ADA and the AACE have varied A1C goals (see Table 3), however both organizations agree that having a controlled A1C of less than 7% can lead to reduced complications if implemented soon after a DM diagnosis. These guidelines are supported by studies have shown that glycemic control is vital to reducing such risk in both Type 1 and Type 2 DM patients. One study showed that patients with type 1 and type 2 DM patients and a mean A1C of 10% or more would incur twice the adjusted rate of diabetes-related hospitalization costs compared to those with an A1C of less than 7% ($6,759 vs $2,792). 7 Shetty et al reported that patients with type 2 DM who maintained an A1C of less 7% incurred lower diabetes-related costs compared to those above 7% ($1,171 vs. $1,540, p<0.001). 8 Table 3. Glycemic goals of therapy by ADA and AACE 6,9,12 ADA AACE Hemoglobin A1C < 7% < 6.5% Preprandial plasma glucose mg/dl < 110 mg/dl Postprandial plasma glucose < 180 mg/dl < 140 mg/dl 5

16 Approaches to Glycemic Treatment Nonpharmacologic therapy Therapeutic lifestyle changes have been shown to reduce the risk of or delay the onset of Type 2 DM. Randomized controlled trials have shown that after 3 years, programs involving lifestyle change and a hypoglycemic medication program may reduce the risk of DM by up to 58%. Follow-up studies such as the U.S. Diabetes Prevention Program Outcomes Study have shown sustained risk reduction of 34% at 10 years. 6,10 Therapeutic lifestyle changes typically involve healthy eating habits, regular physical activity, limited alcohol consumption and stress reduction. Evidence has shown that obesity management can help delay the progression of pre-diabetes to type 2 DM, thus weight reduction is recommended for overweight and obese patients with pre-diabetes. Improved glycemic control coupled with sustained weight loss can also help reduce the need for pharmacologic therapy. Guidelines on therapeutic lifestyle changes from both the ADA and AACE can be found on Table 4. Table 4. ADA and AACE guidelines on therapeutic Lifestyle Changes 4-6,9-12 ADA AACE Weight loss (for overweight Reduce by 7% Reduce by 5% to 10% and obese patients) Physical activity 150min/week minimum High intensity ( 16 sessions in 6 months) and focus on diet, physical activity, and behavioral strategies to achieve a kcal/day energy deficit Diet Achieve kcal/day energy deficit or provide approximately: 1,200 1,500 kcal/day for women and 1,500 1,800 kcal/day for men 150 min/week of moderate-intensity exercise (eg, brisk walking) plus flexibility and strength training Eat regular meals and snacks; avoid fasting to lose weight Consume plant-based diet (high in fiber, low in calories/glycemic index, and high in phytochemicals/antioxidants) 6

17 Pharmacologic therapy Type 1 Diabetes Mellitus Insulin remains the gold standard of treatment for Type 1 DM patients. Insulin, an anabolic and anticatabolic hormone, plays a major role in the metabolism of protein, carbohydrate and fat. Type 1 DM patients should be treated with multiple injections of daily and basal insulin or with continuous subcutaneous insulin (CS2). Table 5 shows a comprehensive list of available insulin preparations, time to onset of their effects, timing of peak levels and duration of effects. Table 5. Available insulin preparations 4,5 Type Agent Onset (h) Peak (h) Duration (h) Brand Names Rapid acting Aspart min Novolog Lispro min Humalog Glulisine min Apidra Short acting Regular hr Humulin R, Novolin R Intermediate NPH 2 4 hr Humulin N, Novolin N Long Acting Determir 2 hr n/a Levemir Glargine 4-5 hr n/a Lantus Degludec 1 hr n/a 42 Tresiba Type 2 Diabetes Mellitus In addition to insulin, 11 additional classes of agents are approved to treat type 2 DM, nine of which are oral agents. Use of any approved agents will depend on the patients medical 4-6, 9-12 needs and treatment goals. The CDC has shown that 14% of patients diagnosed with type 2 DM take insulin only,14.7% take both insulin and oral medications, 56.9% take oral medication only, and an additional 14.4% do not take either. 1 A study by the ADA on the economic costs of diabetes in 2012 has shown that prescription medications account for 12% of an estimated $176 billon of direct medical expenditures. 3 An overview of agents approved to treat type 2 DM is shown in Table 6 and Table 7. 7

18 Table 6. Agents for the treatment of type 2 DM (excluding Insulin, see Table 5) 4-6,9-12 Drug Class Compound Name Mechanism of action Efficacy Advantages Disadvantages Biguanides Metformin Hepatic glucose production A1C 1.5-2% FPG 60-80mg/dL Sulfonylureas Meglitinides Thiazolidinediones Alpha Glucosidase inhibitors Alpha Glucosidase inhibitors DPP-4 Inhibitors Glyburide Glipizide Glimepiride Repaglinide Nateglinide Pioglitazone Rosiglitazone Acarbose Miglitol Acarbose Miglitol Sitagliptin Vildagliptin Saxagliptin Linagliptin Alogliptin Insulin secretion A1C 1.5-2% FPG 60-70mg/dL Insulin secretion A1C 1.7% FPG 60-80mg/dL Insulin sensitivity A1C ~1-1.5% FPG 60-70mg/dL Slows intestinal carbohydrate digestion/absorption Slows intestinal carbohydrate digestion/absorption Insulin secretion Glucagon secretion (both glucose dependent) Bile acid sequestrants Colesevelam? Incretin levels? Hepatic glucose production Dopamine-2 agonists Bromocriptine Modulates hypothalamic regulation of metabolism insulin sensitivity A1C 0.3-1% FPG ~10% reduction A1C 0.3-1% FPG ~10% reduction A1C 0.7-1% A1C 0.4% when added to stable metformin, sulfonylurea, or insulin FPG 5-10mg/dL A1C 0.3%-0.6% Extensive experience No hypoglycemia CVD (UKPDS) Extensive experience Microvascular risk (UKPDS) Post prandial glucose Dosing flexibility No hypoglycemia Durability HDL-C Triglycerides (pioglitazone) No hypoglycemia Post prandial glucose excursions Nonsystemic No hypoglycemia Post prandial glucose excursions Nonsystemic No hypoglycemia Well tolerated No hypoglycemia LDL-C No hypoglycemia? CVD events Diarrhea, abdominal cramping Vit B12 deficiency Lactic acidosis (rare) Hypoglycemia weight Hypoglycemia weight Frequent dosing schedule weight Edema/heart failure Bone fractures LDL-C (rosiglitazone) Modest A1C efficacy Flatulence, diarrhea Frequent dosing schedule Modest A1C efficacy Flatulence, diarrhea Frequent dosing schedule Angioedema/ Urticarial and other immune mediated dermatological effects? Acute pancreatitis? heart failure hospitalizations Generally modest A1C efficacy Constipation Triglycerides May absorption of other medications Generally modest A1C efficacy Dizziness/syncope Nausea Fatigue Rhinitis 8

19 Table 7. Agents for the treatment of type 2 DM (excluding Insulin, see Table 5) 4-6,9-12 Drug Class Compound Name Mechanism of action Efficacy Advantages Disadvantages SGLT2 inhibitors Canagliflozin Dapagliflozin Empagliflozin A1C 0.5-1% GLP-1 receptor agonists Amylin mimetics Exenatide Liraglutide Albiglutide Lixisenatide Dulaglutide Pramlintide (typically used in combination with insulin) Blocks glucose reabsorption by the kidney, increasing glucosuria Insulin secretion Glucagon secretion (both glucose dependent) Slows gastric emptying Satiety Glucagon secretion Slows gastric emptying Satiety A1C % A1C 0.6% No hypoglycemia weight blood pressure Effective at all stages of type 2 DM No hypoglycemia weight post prandial glucose excursions some cardiovascular risk factors weight post prandial glucose excursions Genitourinary infections Polyuria Volume depletion/ hypotension/ dizziness LDL-C Creatinine Diabetic ketoacidosis Urinary tract infections Nausea, vomiting, diarrhea heart rate Injectable Generally modest A1C efficacy Nausea, vomiting Hypoglycemia unless insulin dose is simultaneously reduced Injectable Frequent dosing schedule A1C = glycosylated hemoglobin A1C; FPG = fasting plasma glucose; DPP4-i=Dipeptidyl Peptidase-4 inhibitors; SGLT2-I =Sodium dependent Glucose Cotransporter-2 inhibitors; GLP-1 = Glucagon-like Peptide-1 agonists 9

20 FPG lowering Table 8. Effects of Diabetes Drug Action 11,12 GLP-1 SGLT2- Met DPP4-I TZD AG-I Coles RA I Mod Mild to mod* BCR- QR Mod Mild Mod Neutral Mild Neutral SU/Glinide Insulin Pram SU: mod Glinide: mild Mod to marked (basal insulin or premixed) Mild PPG lowering Mild Mod to marked Mild Mod Mild Mod Mild Mild Mod Mod to marked (short/ rapidacting insulin or premixed) Mod to marked NAFLD benefit Mild Mild Neutral Neutral Mod Neutral Neutral Neutral Neutral Neutral Neutral Hypoglycemia Neutral Neutral Neutral Neutral Neutral Neutral Neutral Neutral SU: mod to severe Glinide: mild to mod Mod to severe* Neutral Weight Renal impairment/ GU GI adverse effects Slight loss Contraindicated in stage 3B, 4, 5 CKD Loss Loss Neutral Gain Neutral Neutral Neutral Gain Gain Loss Exenatide contraindicated CrCl <30 mg/ml Not effective w egfr < 45 / GU infection risk Dose adjustment (except linagliptin) Neutral Neutral Neutral Neutral Increased hypoglycemia risk Increased risks of hypoglycemia and fluid retention Mod Mod* Neutral Neutral* Neutral Mod Mild Mod Neutral Neutral Mod Possible CHF Neutral Neutral benefit Neutral Mod Neutral Neutral Neutral Neutral Neutral Neutral Possible Possible CVD Neutral Neutral Neutral Neutral Neutral Safe? Neutral Neutral benefit benefit Mod Bone Neutral Neutral Neutral Neutral Neutral Neutral Neutral Neutral Neutral Neutral bone loss AGI = α-glucosidase inhibitors; BCR-QR = bromocriptine quick release; CHF = congestive heart failure; CKD = chronic kidney disease; Coles = colesevelam; CrCl = creatinine clearance; CV = cardiovascular; DPP4I = dipeptidyl peptidase 4 inhibitors; FPG = fasting plasma glucose; GI = gastrointestinal; GLP1RA = glucagon-like peptide 1 receptor agonists; GU = genitourinary; Met = metformin; NAFLD = nonalcoholic fatty liver disease; PI = prescribing information; PPG = postprandial glucose; SGLT2I = sodiumglucose cotransporter 2 inhibitors; SU = sulfonylureas; TZD = thiazolidinediones. Boldface type highlights a benefit or potential benefit; italic type highlights adverse effects.*especially with short/ rapid-acting or premixed. With the addition of new drug classes and recommendations to achieve glycemic control, choosing a pharmacotherapy regimen for patients in type 2 DM can depend on the patients glycemic state (level of A1C), patient preferences, and other factors. Type 2 DM patients with Neutral 10

21 an A1C below or approaching 7% may be treated with therapeutic lifestyle changes alone or may be treated with an agent that does not cause hypoglycemia such as metformin. Patients with A1C between 7% and 8.5% may be treated with one single agent or a combination of agents. Patients with a A1C > 9% will likely require more than 2 agents and may begin to involve the use of insulin. 4-6,9-12 Due to the differing characteristics of available agents, a patient-centered approach including patient preferences, cost and potential side effects needs to be taken when selecting an agent. The ADA and the European Association for the Study of Diabetes (EASD) have published a position statement together with a treatment algorithm for patients with type 2 DM (see Figure 1). Figure 1. ADA/EASD 2016 Anti-Hyperglycemic Therapy in Type 2 DM 10 The AACE has also published a treatment algorithm for patients with Type 2 DM. The AACE guidelines differ from the ADA/EASD guidelines in that the AACE provides a preferential order 11

22 of recommended agents that do not increase the risk of hypoglycemia and weight gain (see Figure 2). Figure 2. AACE Glycemic Control Algorithm While both the ADA/EASD and the AACE guidelines may differ in recommendations of certain anti-diabetic agents, they do agree that in addition to therapeutic lifestyle modifications, metformin remains the agent of first choice to begin treatment for patients with type 2 DM. Metformin has a long history of evidence for efficacy and safety and is considerably less expensive compared to other agents and has shown a decrease in total mortality. 1,2,26 Changes in glycemic treatment guidelines from 2012 to present 2016 The last class of anti-diabetic agents to be approved by the Food and Drug Administration (FDA) were the dipeptidyl peptidase 4 inhibitors (DDP4-I), which are oral agents. The DPP4-I compounds, also known as gliptins, include sitagliptin, saxagliptin, linagliptin, and alogliptin. Most recently in August 2015, the FDA added a new Warning and 12

23 Precaution about the risk of joint pain that can be severe and disabling; this warning is required on the labels of all medicines in the DDP4-I class. 14 The FDA had also reviewed heart failure risk with diabetes drug saxagliptin (marketed as Onglyza and Kombiglyze XR) in February 2014 and concluded that more patients receiving saxagliptin- or alogliptin-related agents were hospitalized for heart failure compared to placebo. In the saxagliptin study, 3.5% of patients were hospitalized for heart failure compared to 2.8% of patients who received a placebo. In the alogliptin study, 3.9% of alogliptin-treated patients were hospitalized for heart failure compared to 3.3% in the placebo group. Results of both studies led the FDA to add a new Warnings and Precaution statement for saxagliptin or alogliptin about the potential increased risk of heart failure, but it was not added to other molecules in the same class. 25 As such, the ADA/EASD guidelines suggest that this class be used cautiously in type 2 DM patients with heart failure. 10 A major change from the ADA/EASD guidelines since its publication of a treatment algorithm in 2012 is the inclusion of sodium glucose co-transporter-2 inhibitors (SGLT2-I), an oral class of anti-diabetic agents approved in 2013 by the FDA. The SGLT2-I compounds, also known as gliflozins, include canagliflozin, empagliflozin, and dapagliflozin. Table 8 shows the FDA approval timeline of drugs in both DPP4-I and SGLT2-I classes. 15 SGLT2-I works in the kidney by blocking the re-absorption of glucose. This allows more glucose to pass through the urine and thus lowers serum glucose and allows weight loss. The mechanism of action for SGLT2-I also enables it to be used at any stage of type 2 DM as SGLT2-I are independent of insulin. 4-6 Because more glucose is passed through the urine, potential side effects include genital mycotic infections and a slight increase in urinary tract infections and increased urine calcium excretion. In September 2015, the FDA strengthened the warning on increased risk of bone 13

24 fractures and added new information about decreased bone mineral density to the product labeling of canagliflozin. Furthermore, in December 2015, the FDA revised the labels of SGLT2-I for diabetes to include warnings about excess acid in the blood and risk of serious urinary tract infections. 13 The FDA had identified 19 cases of life-threatening blood infections (urosepsis) and kidney infections (pyelonephritis) that started as urinary tract infections with the SGLT2-I reported from March 2013 through October All 19 patients were hospitalized, however few required admission to an intensive care unit or dialysis in order to treat kidney failure. Table 9. FDA approval dates of DPP4-I and SGLT2-I 15 DPP4 Inhibitor FDA approval Date SGLT2 Inhibitor FDA Approval Date Sitagliptin (Januvia ) October 2006 Canagliflozin (Invokana ) March 2013 Sitagliptin and Metformin March 2007 Dapagliflozin (Farxiga ) January 2014 (Janumet ) Saxagliptin (Onglyza ) July 2009 Canagliflozin and Metformin August 2014 (Invokamet ) Saxagliptin and Metformin November 2010 Empagliflozin (Jardiance ) August 2014 (Kombiglyze XR ) Linagliptin (Tradjenta ) May 2011 Dapagliflozin and Metformin October 2014 extended release (Xigduo XR ) Linagliptin and Metformin January 2012 Empagliflozin and Metformin August 2015 (Jentadueto ) (Synjardy ) Sitagliptin and Metformin February 2012 extended release (Janumet XR ) Alogliptin (Nesina) January 2013 Alogliptin and Metformin January 2013 (Kazano) Alogliptin and Pioglitazone (Oseni) January 2013 Combination DDP4-I and SGLT2-I Empagliflozin and Linagliptin (Glyxambi) January 2015 When the first SGLT2-I, canagliflozin (Invokana ) was approved in 2013, the AACE guidelines of 2013 (Figure 2) ranked SGLT2-I as a fifth line treatment along with use with caution based on early clinical trial data. Three years later, with more clinical evidence and more SGLT2-I approved, the AACE guidelines of 2015 ranked SGLT2-I as a third line treatment given that the class had few adverse events and/or possible benefits. Despite the FDA labeling 14

25 changes in 2015 for SGLT2-I, the 2016 AACE algorithm on glycemic control has not changed from its 2015 recommendations (see Figure 3). 16 Figure 3. AACE Glycemic Control Algorithm DPP4 inhibitors vs. SGLT2 inhibitors Head-to-head clinical comparisons of SGLT2-I to DPP4-I have also been shown to favor the SGLT2-I. Schernthaner et al. compared canagliflozin to sitagliptin (DPP4-I) in patients who have not achieved glycemic control while on metformin and sulfonylurea. 18 Results showed that at 52 weeks, canagliflozin 300mg/day showed non-inferiority and subsequent superiority to sitagliptin 100mg/day in A1C-lowering (-1.03% vs. 0.66%, 95% CI -0.5, -0.25). Canagliflozin also showed greater reductions in fasting plasma glucose (-1.3mmol/L or mg/dl, p<0.001) and body weight (-2.4kg, p <0.001). 17 Rosenstock et al. (2012) showed that canagliflozin 15

26 (100mg up to 300mg once daily from baseline) reduced A1C by -0.70% to -0.95% after 12 weeks of treatment compared to sitagliptin 100mg, which reduced A1C by -0.74%. Furthermore, clinical trial evidence has shown that a combination of a SGLT2-I and DPP4-I was better at reducing A1C compared to single agents of either class. Rosenthal et al. (2015) found that saxagliptin and dapagliflozin produced better A1C reductions compared to single agent saxagliptin and dapagliflozin (all three cohorts were also treated with metformin). 19 Results at 24 weeks showed differences in adjusted mean changes in A1C between saxa+dapa+met and saxa+met (p<0.001) and saxa+dapa+met and dapa+met, (p = 0.016). While a statistical analysis was not performed to compared saxa+met to dapa+met, the reduction in A1C among those two cohorts was in favor of the SGLT2-I cohort. Defronzo et al. (2015) found that the combination of empagliflozin and linagliptin was superior in reduction of A1C compared to single-agent empagliflozin and linagliptin (along with metformin). 20 At 24 weeks, reductions in A1C were -0.58% for emp+lina compared to empagliflozin alone and -0.50% compared to linagliptin alone (both p<0.001). Study Rationale While clinical trials have shown favorable results towards SGLT2-I compared to DPP4-I, there is still limited evidence on the effectiveness of such agents in a real-world setting. A review of the literature have yielded only two studies, both focused on canagliflozin, despite the fact that the agent was approved in 2013 and five additional new SGLT2-I agents (though only 3 actual molecules) have been approved by the FDA since then. Buysman et al. was the first to study the use of canagliflozin in a real-world setting. It was a descriptive study aimed at presenting patient characteristics, treatment patterns and outcomes of patients using canagliflozin within the first six months of its FDA approval. 21 The study found that patients on SGLT2-I had 16

27 a mean A1C that decreased from 8.54 to 7.76 (p<0.001) and a decreased usage of other antidiabetic agents following initiation of canagliflozin. While the study was able to determine change in A1C, the study period was limited to a three-month follow-up period after the first canagliflozin claim. This limited timeframe is not enough to determine long-term effects of canagliflozin, as A1C is a measure of a person s average level of blood glucose over the past three months. In essence, this study period would have enabled only one evaluation of A1C laboratory data with the study period and would have failed to capture the full response to canagliflozin. Garbner et al. was the second study that compared baseline demographic, clinical and economic characteristics of patients with type 2 DM on canagliflozin compared to DPP4-I. 22 The study showed that canagliflozin initiators were more likely to have higher A1C levels and more likely to incur higher pharmacy utilization and costs compared to DPP4-I initiators. Limitations of the study included the amount of available laboratory data, the study period and differences in baseline patient characteristics. Patients with A1C laboratory results accounted for only 30% of the overall patient population, and the study period of January 2011 to September 2013 would only allow, at most, a 3-6 month follow up for patients on canagliflozin - thus limiting the ability to identify trends in treatment patterns or a more comprehensive assessment of outcome post-initiation. Finally, the study did not control for differences in patient characteristics, as early initiators of canagliflozin may differ from patients who waited until the guidelines included the SGLT2-Is. While both studies assessed descriptive changes of short-term outcomes and clinical profiles of patients that initiated SGLT2-I and the DPP4-I, neither study looked into patient adherence or persistence. Given that both classes of drug are oral agents and are taken once 17

28 daily, however, it is possible (but cannot be assumed ) that adherence and persistence do not differ. Aside from clinical efficacy and safety profiles of anti-diabetic drugs, patient adherence continues to be a factor in determining patient glycemic control. Studies have shown that nonadherent patients have both statistically and clinically worsened outcomes compared to adherence patients. Pladevall et al showed that an 10% increase in nonadherence can lead to a 0.14% increase in A1C, while Schectman et al showed that a 10% increase in adherence can lead to a A1C decrease of 0.16% (p< ). 23,24 Rozenfeld et al found an inverse relationship between oral antidiabetic medication and A1C, demonstrating that each 10% increase in adherence led to a 0.1% decrease in A1C (p = 0.004). 27 Al-Qazaz et al found significant correlations between A1C and adherence (p< 0.05), showing that higher scores on the Morisky Medication Adherence Scale were found in patients with lower A1C. 28 Medication regimen can also play a factor in adherence as Odegard et al demonstrated that taking more than 2 doses of diabetes medication daily were significantly associated with higher A1C (p =0.2) and Donnan et al showed that there was a significant linear trend of poorer adherence with each increase in the daily number of tablets taken (p=0.001) It has been more than 3 years since the approval of the first SGLT2-I compound in 2013, with five additional approved SGLT2-I agents on the market. The ADA/EASD and the AACE guidelines have since incorporated that drug class into their treatment algorithms for glycemic control. With over nine oral drug classes available and the DPP4-I and SGLT2-I drug classes gaining more favor as treatment options for patients with type 2 DM, more studies are needed to determine the effectiveness of these medications in the real world. Patient adherence and persistence should also be evaluated as those factors may also contribute to effectiveness. This 18

29 study will contribute significantly to the literature by conducting a comparison of A1C reduction, and quantifying differences in adherence, persistence and costs of treatment for patients initiating SGLT2-I compared to DPP4-I. The study is feasible and timely, now that the SGLT2-I has had enough time to establish itself as a treatment option for patients with type 2 DM. 19

30 CHAPTER 2: METHODOLOGY Study Objectives and Hypotheses This study evaluated the level of A1C reduction between SGLT2-Is compared to DPP4-Is and documented changes in medication utilization patterns among those two classes of oral hypoglycemic agents in patients with type 2 DM within the Scott & White Health Plan. The specific objectives of this study are listed below; specific hypotheses are listed in Table 10: 1. To determine if patient characteristics (age, gender, race, or comorbidities) differ between patients taking SGLT2-Is versus DPP4-Is. 2. To determine if A1C reductions differ between patients taking SGLT2-Is versus DPP4-Is. 3. To determine if change in body mass index at initiation and after 6 months from initiation differs between patients taking SGLT2-Is versus DPP4-Is. 4. To determine if patient adherence differs between patients taking SGLT2-Is versus DPP4-Is. 5. To determine if patient persistence differs between patients taking SGLT2-Is versus DPP4-Is. 6. To determine if the use of insulin at initiation and after 6 months from initiation differs between patients that were on insulin while initiating SGLT2-Is versus DPP4-Is. Study Design and Data Source This retrospective cohort study took place from October 1, 2014 to September 30, Patients selected for this study were required to have one or more pharmacy claims for index drug of SGLT2-I or DPP4-I. Patients were required to have continuous enrollment in the health plan for at least 6 months during the study period. Patient-level data were extracted from pharmacy claims, medical claims, electronic medical records, and health plan eligibility status for both SGLT2-I and DPP4-I patients during the study period. Eligibility data were collected from the health plan s internal enrollment data source. The SWHP has up to 300,000 covered 20

31 lives, with over 60% as commercial lives. As part of an integrated health system, the SWHP has access to medical laboratory data through electronic medical records. This study was approved by Baylor Scott & White Health and the University of Texas at Austin institutional review boards (IRBs) following expedited review. Inclusion/Exclusion Criteria Patients were required to be between the age of 18 and 62 years old with a confirmed diagnosis of type 2 DM (International classification of diseases, ninth revision, clinical modification [ICD- 9-CM] of 250.X0 or 250.X2) and continuous enrollment within the SWHP for at least six months for the duration of the study. Patients on either a SGLT2 or a DPP4-I and a GLP-1 were excluded. Patients with a medical claim for any of the ICD-9-CM listed in Table 9 were also excluded. 22 Table 10. ICD-9-CM Codes Used for Exclusion Criteria Condition ICD-9-CM Code Type 1 diabetes 250.x1, 250.x3 Gestational diabetes 648.8x Hyperglycemia not otherwise specified 790.6x Neonatal diabetes mellitus 775.1x Nonclinical diabetes Diabetes with hyperosmolar coma 250.2x Table 11. Study Hypotheses H 0 1: There is no statistical difference in patient characteristics between patients taking a SGLT2-I versus a DPP4-I. H 0 2: There is no statistical difference in A1C between patients taking a SGLT2-I versus a DPP4-I. H 0 3: The is no statistical difference in BMI differences in patients taking a SGLT2-I versus a DPP4-I. H 0 4: There is no statistical difference in the patient adherence between patients taking a SGLT2-I versus a DPP4-I. H 0 5: There is no statistical difference in patient persistence between patients taking a SGLT2-I versus a DPP4-I. H 0 6: There is no statistical difference in the use of insulin between patients taking a SGLT2-I versus a DPP4-I. 21

32 Outcome Measures Table 12. Variables analyzed in the study Category Type Measure Source Independent Demographics Age in years? Independent Demographics Gender (M, F) EHR Independent Demographics Race/ethnicity (white, black, other race, EHR Hispanic) (Inclusion Clinical Diabetes Type 2 Claims criterion) Independent Clinical Charlson Comorbidity Index Claims Dependent Clinical Change in Body mass index EHR Dependent Clinical Change in A1c EHR Dependent Clinical Use of insulin (yes/no) Claims Dependent Clinical Use of DPP4i (yes/no) Claims Dependent Clinical Use of SGLT2i (yes/no) Claims Dependent Clinical Adherence: PDC 80% (yes/no) Claims Dependent Clinical Persistence in days Claims Dependent Healthcare Costs of DM drugs per 6-month period Claims utilization EHR = electronic health records 1. Baseline patient characteristics (age, Charlson co-morbidity index, race, ethnicity, drug therapeutic class) were evaluated from historical data in EHR and pharmacy drug claims. 2. Change in A1C was calculated from historical data laboratory in EHR 3. Change in body mass index was calculated from historical data laboratory in EHR 4. Adherence a. Adherence defined as a proportion of days covered (PDC): Number of days covered by prescription fills during the denominator period Number of days between the first fill of the medication during the measurement period and the end of the measurement period), Limited to those patients with at least one prescription If the PDC >= 80%, then adherent 5. Persistence a. Persistence: Number of days until discontinuation 22

33 Time from the initial prescription fill until the patient has a gap in therapy (time gap of > 45 days) 6. Change in insulin use was evaluated by selecting the subset of patients that were on insulin while on a SGLT2-I and a DPP4-I and determining whether the patient was still using insulin 6 months after initiation of the novel hypoglycemic agent (YES/NO) Analysis Plan Descriptive statistics (mean and standard deviation or median and interquartile range) will characterize demographic data before and after matching. o Continuous variables: mean, standard deviation, median, interquartile range, minimum, maximum o Categorical variables: N (frequency), % (percent frequency) Tests for normality were conducted to determine appropriate statistical tests o Chi-square analyses was conducted to analyze differences in baseline co-morbidities between the two treatment groups o Mann-Whitney U tests were conducted for nonparametric continuous variable comparisons between groups (eg. Baseline A1C) Linear regression modeling was conducted to assess the adjusted effect of SGLT2-I versus DPP4-I on the interval-level outcomes, A1C, BMI, and costs, controlling for baseline values and clinical and demographic covariates. Baseline differences were assessed and handled as above. Distributions of the dependent measures will be assessed for normality; appropriate transformation, such as log transform, was applied to cost measures as needed. A value of p<0.05 was used to determine statistical significance. 23

34 Statistical analyses were computed using SAS 9.4 TS1M2 (SAS Institute Inc, Cary, North Carolina). A value of p<0.05 was used to determine statistical significance. Feasibility Analysis Sample size calculations were performed using G*Power software: o Using a conventional estimation for small effect size (w = 0.55), a total sample size of 118 would be required to determine a difference in A1C between treatment groups, assuming an alpha of 0.05 and power of Post hoc sample size calculations were performed using G*Power software to compute the power achieved o Using an estimation for small effect size (w = 0.134), a total sample size of 300 and assuming an alpha of 0.05, the calculated achieved power was low at

35 CHAPTER 3: RESULTS Patient Selection A total of 300 patients met the inclusion criteria for this study. Among these, 99 patients met criteria for inclusion in the SGLT2-I cohort, and 201 met criteria for inclusion in the DPP4-I cohort. (Figure 4). Figure 4. Flow-chart of patient identification 25

36 Demographic Characteristics Baseline demographic information for the 300 patients are shown in table 12. The mean age (SD) was 51 (7.7) years for SGLT2-I and 52 (7.1) years for DPP4 initiators. Gender distributions for both cohorts were almost equal. Among other antidiabetic therapies, a chisquare analysis showed that a greater proportion of patients in the DPP4-I cohort were prescribed sulfonylureas compared to the SGLT2-I cohort. (x 2 = ; df=1; p = ). A Wilcoxon- Mann-Whitney test showed that the baseline A1C in the DPP4-I cohort was significantly lower than the SGLT2-I cohort (mean [SD], 8.9 [2.0] vs. 8.4 [1.8], p = ). Chi-square analysis between insulin use (yes/no) at baseline and medication cohort showed that patients in the SGLT2-I cohort were prescribed significantly more insulin compared to the DPP4-I cohort. (x 2 = ; df=1; p < 0.001). In addition, chi-square analysis also revealed that among those prescribed insulin in both cohorts, the SGLT2-I group was more likely to have been prescribed insulin prior to the initiation of SGLT2-I (X 2 = ; df=1; p < 0.001). 26

37 Table 13. Baseline Demographics, Comorbidities, and Laboratory Values SGLT2-1 (n = 99) DPP4-I (n = 201) p-value Age at index (years) mean (SD) 51 (7.7) 52 (7.11) Charlson Comorbidity Index mean (SD) 2.25 (1.5) 2.11 (1.88) Gender Male, n (%) Female, n (%) Race (missing 545 [32.4% from sample]) White, n (%) Black, n (%) Other, n (%) Ethnicity (missing 546 [32.4% from sample]) Non-Hispanic, n (%) hispanic, n (%) Medication Class AGI, n (%) Biguanides, n (%) Sulfonylurea, n (%) Meglinitides, n (%) TZD, n (%) Combination productions, n (%) 51 (51.5) 48 (48.5) 40 (67.8) 7 (11.9) 12 (20.3) 48 (81.4) 11 (18.6) 2 (2.02) 76 (76.8) 45 (45.5) 6 (6.06) 2 (2.02) 4 (4.04) 94 (46.8) 107 (53.2) (64.4) 20 (15.1) 27 (20.5) (80.5) 26 (19.5) (2.49) 149 (74.1) 131 (65.2) 0 (0%) 23 (11.4) 7 (3.5) Mean follow-up in days (SD) (114.5) (118.9) Baseline A1C mean (SD) 8.9 (2.0) 8.4 (1.8) Baseline BMI (kg/m 2 ) mean (SD) 34.5 (7.1) 34.4 (7.2) Insulin use Categories, N (%) 1) Neither period 37 (37.4) 138 (68.6) < ) Regardless of before or after index drug 62 (62.6) 63 (31.4) < A)Before initiation of index drug 46 (46.5) 23 (11.4) < B) After initiation of index drug 16 (16.16) 40 (20.0) <0.001 AGI= alpha glucosidase inhibitors; TZD: Thiazolidinedione 27

38 Change in A1C The change in A1C was calculated based on difference between the initial A1C at baseline and the final A1C at the end of the study period. The difference (mean [sd]) in A1C for the SGLT2-I cohort was -0.7 (1.6) compared to -0.5 (1.7) for the DPP4-I cohort. A Wilcoxon- Mann-Whitney test showed that the difference in A1C was not statistically significant. (Z = , p = ). Among the subset of patients that were not on insulin at baseline (N = 16 SGLT2-I and N = 40 DPP4-I), a Wilcoxon-Mann-Whitney test also showed no statistical significance (Z = , p = ) Change in Body Mass Index (BMI) The change in BMI was calculated based on the difference between the initial BMI at baseline and the final BMI at the end of the study period. The difference (mean [sd]) in BMI for the SGLT2-I cohort was (2.55) compared to (2.21) for the DPP4-I cohort. A Wilcoxon-Mann-Whitney test showed that the difference in A1C was not statistically significant. (Z = , p = ). Adherence Adherence was defined by a PDC 80%. PDC was calculated using the number of days covered by prescription fills during the denominator period divided by the number of days between the first fill of the medication during the measurement period and the end of the measurement period). Among the SGLT2-I cohort, 67 (67%) patients were adherent compared to 147 (73%) in the DPP4-I cohort. A chi square analysis showed that the difference in adherence was not statistically significant (X 2 = ; df=1; p = ) Persistence Persistence was determined by evaluating the number the days until treatment discontinuation (based on a gap in therapy of greater than 45 days). The mean time to treatment 28

39 discontinuation was days (sd= days) for the SGLT2-I cohort compared to days (sd = days) for the DPP4-I cohort. A Wilcoxon-Mann-Whitney test showed that the days to treatment discontinuation was statistically significant. (Z = , p = ). Insulin use Insulin use was determined by selecting the subset of patients that were on SGLT2-I or DPP4-I and concurrently with insulin during the course of the study period. Among the subset of patients on insulin, 63% (n=62) of patients from the SGLT2-I cohort was on insulin compared to 31% (n=63) patients from the DPP4-I cohort. A chi square analysis showed that the difference in insulin use was statistically significant (X 2 = ; df=1; p < 0.001). However, while the change in number of patients still on insulin decreased, the results were not statistically different between both cohorts (X 2 = ; df=1; p = ). Chi square analysis also showed that there was a significant difference between both cohorts regarding the order of insulin use. (X 2 = ; df=1; p < 0.001). Results showed that 46.5% (n=46) of patients from the SGLT2-I cohort were on insulin prior to starting a SGLT2-I compared to 11.4% (n=23) of patients from the DPP4-I cohort who were on insulin prior to starting a DPP4-I. However, while the change in number of patients still on insulin decreased after the index drug was initiated, the results were not statistically different between both cohorts (Fisher s exact test; p = ). A summary of results related to insulin is displayed in Table