Web Appendix: 4-Year Cost Trajectories in Real-World Patients Matched to Themetabolic

Web Appendix: 4-Year Cost Trajectories in Real-World Patients Matched to theMetabolic Profiles of Trial Subjects Before/After Treatment with Phentermine-Topiramate

Authors: James Li, MD; Nancy L Reaven, MA2; Susan E Funk; MBA2; Karen McGaughey, PhD3; Martin Neovius, PhD4

Affiliations:

MilesMemorialHospital, Damariscotta, Maine, USA

2 Strategic Health Resources, La Canada, CA USA

3CaliforniaPolytechnicStateUniversity, San Luis Obispo, CAUSA

4Karolinska Institute, Stockholm, Sweden

Corresponding Author:

Nancy L Reaven

Considerations in the Matching Process

The population selection and matching process is depicted in Figure S1.

Starting with 1,487,977 patients in the original database, we excluded 654,187 patients for whom hospitalization information was unavailable and 700,671 patients who did not meet our minimum requirementof having ≥3 years of follow-up data from the index date. (We defined the index date as the date of first valid BMI reading preceded by at least 6 months of individual patient data).This left 133,119 patients in the eligible database pool.

Since no CONQUER subject exceeded 71 years in age, 19,030 patients were excluded due to age ≥75 years.Additional exclusions included1,364 patients who had unreliable BMI data;1,099 patients who had <3 years of patient data following their index date; 2,529 patients for having outlier BMI, blood pressure, and triglyceride values making them impossible to match toany CONQUER subject (BMI >51.4, DBP>119, SBP>180, triglycerides>1135); and 368 patients for having diagnosis codes normally associated with care of newborns and infants, suggesting the potential for blended parent-child data or inaccurate presentation of age.

The CONQUER study included the following subjective exclusionary criterion: “clinically significant abnormalities elicited by history, physical examination or tests that the investigator would consider as a contraindication to administration of the study drug, affect compliance, interfere with study evaluations, or confound interpretation of study results.” Consistent with this statement, we reviewed all diagnosis codes occurring during the pre-index period in the remaining pool of eligible patients and excluded 10 additional patients for having one or more of the following diagnoses: severe depressive psychosis (296.23), thyrotoxicity (376.21), cardiac arrest (427.5), coma (780.01), persistent vegetative state (780.03), malignant ascites (789.51), or palliative care (V66.7).

Finally, 38,865 patients were excluded due to havingone or more CONQUER-specific exclusion criteria. Among these excluded patients, 38% met more than one criterion. These criteria, in order of frequency, included: use of anti-diabetic drugs other than metformin; history of bipolar disorder or psychosis; history of malignancy (other than surgically excised basal or squamous cell skin carcinoma or cervical cancer); type I diabetes; current significant depressive symptoms; more than one episode of major depression; clinically significant thyroid dysfunction; unstable angina; known or suspected clinically significant valvular heart disease; history of glaucoma or increased intraocular pressure; history of nephrolithiasis; congestive heart failure; clinically significant renal, hepatic or psychiatric disease; pregnancy or breastfeeding on or after index date; use of tricyclic antidepressants or MAO-inhibitors at index date; coronary revascularization within 6 months of index date; history of myocardial infarction; malignant hypertension; coronary revascularization; history of stroke; cholelithiasis; history of bariatric surgery; history of life-threatening arrhythmia; obesity of known genetic or endocrine origin; steroid hormone therapy; history of suicidal behavior or ideation with intent to act; history of eating disorder, drug, or alcohol abuse within the preceding year; and use of an investigational medication or device within 1 month.Following all exclusions, the remaining pool of patients eligible for matching numbered 69,854.

The 817 CONQUER ITT subjects matched with replacement are represented by 3,559 unique database patients, of whom1,522 matched only subjects in the pre-trial ITT study arm, 1,645 matched only subjects in the post-trial ITT study arm, and 392 matched one or more subjects both study arms. Of the 3,559 unique database patients used, 2,841 (80%) were matched once, 483 (14%) twice, 174 (5%) 3 times, 46 (1%) 4 times. 13 (0.4%) 5 times, and 2 patients (0.1%) 6 times. For patients with a larger number of matches, investigators examined inpatient utilization to test for potential skewing, identifying no outliers among these patients.

Medications

Medications used in categorizing patients for the matching process were defined according to drug lists developed using ePocrates.com.

Anti-diabetic medicationsconsisted of Amaryl, Byetta, Diabeta, exenatide, glimepiride, glipizide, Glucotrol, Glucotrol-XL, glyburide, insulin, Micronase, nateglinide, pioglitazone, Prandin, repaglinide, rosiglitazone, sitagliptin, Starlix, and rosiglitazone.

Antihypertensive medications consisted of ACE inhibitors,alpha antagonists,angiotensin II receptor blockers,beta blockers,calcium channel blockers,diuretics, vasodilators, aliskiren, eplerenone, fenoldopam, phenoxybenzamine, phentolamine,and reserpine.

Anti-lipid medicationsconsisted of fibrates, sequestrants, statins, ezetimibe, niacin, and omega-3acid.

A prescription was considered in effect from the date written until the end of supply obtainable through authorized refills, assuming a 30-day supply per fill or refill. Only prescriptions in effect as of the index date were used in evaluating patients for matching purposes.

Inpatient Cost Methodology Details

Cost for each hospital inpatient stay was estimated using actual length of stay in conjunction with allowed cost per day for private insurance admissions in the same Medicare Severity Diagnosis Related Group (MS-DRG) and mean physician cost per inpatient day, obtained for 2011 private insurance plans from OptumInsight, Inc. of Minneapolis. An MS-DRG (Medicare Severity DRG) was assigned to each inpatient admission using all available diagnosis.Using a certified professional coder, we developed a list of procedures that were appropriate for consideration as the principal procedure of an inpatient stay. Admissions with no procedures on this list were deemed medical admissions. Codes were cross-walked as needed between CPT and ICD-9 formats using EncoderPro Expert (also from OptumInsight).

Primary diagnoses were not designated in the source data. All diagnosis codes during the inpatient stay were categorized usingmultiple parameters: the number of occurrences; whether the diagnosis was flagged as a discharge diagnosis; the earliest day of the hospitalization on which the diagnosis appeared; whether the diagnosis was acute or chronic (using the chronic condition indicator for ICD-9 diagnosis codes); whether the code designated a sign/symptom, event (E*) or history (V*) condition; proximity to the discharge of the last occurrence of the diagnosis code during the hospitalization; the specificity of the code; the designation as a major complication or comorbidity by the Centers for Medicare and Medicaid Services; and the relationship to the principal procedure. The diagnosis code with the best fit using these parameters was considered the primary diagnosis for MS-DRG assignment. MS-DRGs were assigned without regard to present-on-admission status using a batch grouper (M+H Consulting, LLC, Branford, CT).

Page 1