William J. Huster

Effects of raloxifene on bone mineral density, serum cholesterol concentrations, and uterine endometrium in postmenopausal women.

BACKGROUNDnLong-term estrogen therapy can reduce the risk of osteoporotic fracture and cardiovascular disease in postmenopausal women. At present, however, these beneficial effects are not separable from undesirable stimulation of breast and endometrial tissues.nnnMETHODSnWe studied the effect of raloxifene, a nonsteroidal benzothiophene, on bone mineral density, serum lipid concentrations, and endometrial thickness in 601 postmenopausal women. The women were randomly assigned to receive 30, 60, or 150 mg of raloxifene or placebo daily for 24 months.nnnRESULTSnThe women receiving each dose of raloxifene had significant increases from base-line values in bone mineral density of the lumbar spine, hip, and total body, whereas those receiving placebo had decreases in bone mineral density. For example, at 24 months, the mean (+/-SE) difference in the change in bone mineral density between the women receiving 60 mg of raloxifene per day and those receiving placebo was 2.4+/-0.4 percent for the lumbar spine, 2.4+/-0.4 percent for the total hip, and 2.0+/-0.4 percent for the total body (P<0.001 for all comparisons). Serum concentrations of total cholesterol and low-density lipoprotein cholesterol decreased in all the raloxifene groups, whereas serum concentrations of high-density lipoprotein cholesterol and triglycerides did not change. Endometrial thickness was similar in the raloxifene and placebo groups at all times during the study. The proportion of women receiving raloxifene who reported hot flashes or vaginal bleeding was not different from that of the women receiving placebo.nnnCONCLUSIONSnDaily therapy with raloxifene increases bone mineral density, lowers serum concentrations of total and low-density lipoprotein cholesterol, and does not stimulate the endometrium.

Adverse events reported by postmenopausal women in controlled trials with raloxifene

OBJECTIVEnTo assess the incidence of adverse events in postmenopausal women treated with raloxifene compared with placebo, hormone replacement therapy (HRT), or unopposed estrogen.nnnMETHODSnCommon treatment groups were pooled across eight randomized, parallel clinical trials (6-30 months duration) of raloxifene to create the following three databases: placebo-controlled, HRT-controlled, and estrogen-controlled databases. Incidence and severity of all treatment-emergent adverse events, defined as events that first occurred or worsened during treatment, were compared among groups in each of the databases.nnnRESULTSnDiscontinuation rates overall, and those related to adverse events, were not significantly different between treatment groups in any database. There was no significant difference in incidence of vaginal bleeding or breast discomfort between women treated with raloxifene (60 mg/d) or placebo. Both of these events were reported more frequently in women receiving HRT or estrogen. Vaginal bleeding was responsible for significantly more discontinuations from the HRT groups compared with the raloxifene group. Hot flashes was the only event common to all three databases that was significantly increased in the raloxifene group, but this event did not increase the discontinuation rates. The incidence of leg cramps was greater in raloxifene-treated women compared with placebo-treated women in the placebo-controlled database, but did not cause any discontinuations of therapy. Raloxifene had no effect on the incidence of vaginal symptoms or central nervous system events.nnnCONCLUSIONnRaloxifene had an adverse event profile distinct from HRT and unopposed estrogen and was well tolerated by postmenopausal women.

Health-Related Quality-of-Life Results From Multinational Clinical Trials of Insulin Lispro: Assessing benefits of a new diabetes therapy

OBJECTIVE To compare health-related quality of life (HRQOL) in patients with diabetes receiving insulin lispro with patients receiving regular human insulin (Humulin R). RESEARCH DESIGN AND METHODS We performed two randomized comparative studies over a 6-month period (3 months per treatment). Primary analyses used crossover baseline to 3-month changes in HRQOL scores. Ninety-three principal investigators in Canada, France, Germany, and the U.S. participated in these studies. One HRQOL crossover study included 468 patients with type I diabetes; the other HRQOL crossover study included 474 patients with type II diabetes. In both studies, patients were taking insulin at least 2 months before enrollment. Primary outcomes included two generic HRQOL domains, energy/fatigue and health distress, and two diabetes-specific domains, treatment satisfaction and treatment flexibility. Thirty secondary outcomes included both generic and diabetes-specific measures. Secondary outcome domains were controlled for multiplicity in the analyses. RESULTS Primary analyses showed that treatment satisfaction scores (P < 0.001) and treatment flexibility scores (P = 0.001) were higher for insulin lispro in type I diabetic patients. No other significant treatment differences were detected using the data from these 6-month crossover studies. CONCLUSIONS Treatment satisfaction and treatment flexibility were significantly improved in patients with type I diabetes using insulin lispro. Other HRQOL findings were comparable for insulin lispro and regular human insulin. Insulin lispro appears to have a measurable impact on lifestyle benefits in patients with type I diabetes, as demonstrated by increased treatment satisfaction and treatment flexibility.

Development and validation of the Diabetes Quality of Life Clinical Trial Questionnaire.

OBJECTIVEnThe objective of this study was to develop a valid and reliable health-related quality of life (HRQOL) questionnaire for use in multinational clinical trials of patients with type I and type II diabetes.nnnMETHODSnThrough patient focus groups and expert clinician panels in the United States (US) and France, relevant HRQOL domains for patients with type I and type II diabetes were identified. A draft questionnaire was developed by including validated, widely used generic and diabetes-specific domains and by developing original questions as required. A pilot study (n = 123) was conducted to evaluate the psychometric properties of the draft questionnaire with revisions being subsequently made. Data collected from two multinational clinical trials of patients with type I and type II diabetes were used to further validate and enhance the questionnaire (DQLCTQ).nnnRESULTSnA total of 942 patients were recruited in the clinical trials from Canada, France, Germany, and the United States. The mean age was 33.8 years for patients with type I diabetes (n = 468) and 58.2 years for patients with type II diabetes (n = 474). The mean HbAlc level at baseline was 8.6. The revised version of the questionnaire (DQLCTQ-R) contains a total of 57 questions comprising 8 generic and disease-specific domains, as follows: Physical Function; Energy/Fatigue; Health Distress; Mental Health; Satisfaction; Treatment Satisfaction; Treatment Flexibility; and Frequency of Symptoms. Intraclass correlation coefficients range from 0.74 to 0.90 and Cronbachs alphas range from 0.77 to 0.90. With very few exceptions, all eight domains were able to discriminate between type I and type II diabetes, tight and poor metabolic control, male and female, and good and poor self perceived control of diabetes. Four domains (Treatment Satisfaction, Health/Distress, Mental Health, and Satisfaction) were responsive to clinical change in metabolic control.nnnCONCLUSIONnThe DQLCTQ-R is a reliable, valid, and comprehensive HRQOL instrument. It is suitable in multinational clinical trials to evaluate new or alternative treatments for patients with type I and type II diabetes.

Endometrial response to raloxifene compared with placebo, cyclical hormone replacement therapy, and unopposed estrogen in postmenopausal women.

OBJECTIVEnTo determine the endometrial effects of raloxifene 60 mg/day in postmenopausal women as assessed by vaginal bleeding and endometrial thickness.nnnDESIGNnData from 1157 postmenopausal women were analyzed from a database consisting of four independent, double-blind, randomized, placebo-controlled trials (range = 6-30 months duration), a 24-month open-label randomized, cyclical hormone replacement therapy (HRT)-controlled trial, and a 6-month double-blind, randomized, unopposed estrogen-controlled trial. Vaginal bleeding rate was derived from self-reported adverse events collected at least every 6 months. Endometrial thickness was measured by ultrasonography at regular intervals.nnnRESULTSnRaloxifene 60 mg/day was not significantly different from placebo with regard to the incidence of vaginal bleeding, the baseline-to-endpoint change in endometrial thickness, or the proportion of women experiencing an increase in endometrial thickness above baseline after either 12 or 24 months of therapy. Unexpected bleeding was reported significantly more frequently in the unopposed estrogen groups compared with the raloxifene group (raloxifene 60 mg/day, 0% versus estrogen, 50%; p = 0.002). A significantly greater baseline-to-endpoint increase in endometrial thickness was observed in both the HRT and estrogen groups compared with their respective raloxifene comparison group (raloxifene 60 mg/day, 0.01 +/- 2.0 mm versus HRT, 1.8 +/- 3.2; p < 0.001; raloxifene 60 mg/day, 1.1 +/- 1.7 mm versus estrogen, 7.8 +/- 3.8; p < 0.001). No cases of endometrial hyperplasia or cancer were diagnosed in the placebo or raloxifene 60 mg/day groups. Endometrial hyperplasia was diagnosed in one case in the HRT group and in two cases in the estrogen group.nnnCONCLUSIONnRaloxifene 60 mg/day for up to 30 months is not associated with vaginal bleeding or increased endometrial thickness in postmenopausal women.

Antiestrogenic Properties of Raloxifene

This 21-day, open-label study evaluated the effects of raloxifene and tamoxifen on estrogen-induced changes in serum levels of anterior pituitary hormones (prolactin, luteinizing hormone, and follicle-stimulating hormone), sex steroids (testosterone, estradiol), and binding globulins [thyroid binding globulin (T3 resin uptake), transcortin, sex steroid binding globulin]. Seventeen healthy male volunteers completed the study after being randomized to one of three treatments: raloxifene, tamoxifen, or placebo. Six subjects received raloxifene (200 mg daily) for 10 days, 6 subjects received tamoxifen [20 mg twice a day (b.i.d.)] for 10 days, and 5 subjects received placebo for 10 days. All subjects received ethinyl estradiol (20 micrograms b.i.d.) for 7 days starting 3 days after initiation of study drug or placebo treatment. Results of the primary analysis of this study indicate that for six of the seven analyzable parameters of estrogen action (excluding luteinizing hormone) raloxifene blunted the estrogen response; this effect was significant only for T3 resin uptake. Tamoxifen administration significantly blunted or reversed the estrogen effect in all six of these parameters. Raloxifene, an effective antiestrogen in animal models, is also antiestrogenic in humans.

A Bayesian approach to the statistical analysis of device preference studies.

Drug delivery devices are required to have excellent technical specifications to deliver drugs accurately, and in addition, the devices should provide a satisfactory experience to patients because this can have a direct effect on drug compliance. To compare patients experience with two devices, cross-over studies with patient-reported outcomes (PRO) as response variables are often used. Because of the strength of cross-over designs, each subject can directly compare the two devices by using the PRO variables, and variables indicating preference (preferring A, preferring B, or no preference) can be easily derived. Traditionally, methods based on frequentist statistics can be used to analyze such preference data, but there are some limitations for the frequentist methods. Recently, Bayesian methods are considered an acceptable method by the US Food and Drug Administration to design and analyze device studies. In this paper, we propose a Bayesian statistical method to analyze the data from preference trials. We demonstrate that the new Bayesian estimator enjoys some optimal properties versus the frequentist estimator.

A Comparison of Different Statistical Methods Analyzing Hypoglycemia Data Using Bootstrap Simulations

Hypoglycemia has long been recognized as a major barrier to achieving normoglycemia with intensive diabetic therapies. It is a common safety concern for the diabetes patients. Therefore, it is important to apply appropriate statistical methods when analyzing hypoglycemia data. Here, we carried out bootstrap simulations to investigate the performance of the four commonly used statistical models (Poisson, negative binomial, analysis of covariance [ANCOVA], and rank ANCOVA) based on the data from a diabetes clinical trial. Zero-inflated Poisson (ZIP) model and zero-inflated negative binomial (ZINB) model were also evaluated. Simulation results showed that Poisson model inflated type I error, while negative binomial model was overly conservative. However, after adjusting for dispersion, both Poisson and negative binomial models yielded slightly inflated type I errors, which were close to the nominal level and reasonable power. Reasonable control of type I error was associated with ANCOVA model. Rank ANCOVA model was associated with the greatest power and with reasonable control of type I error. Inflated type I error was observed with ZIP and ZINB models.

Crossover design and its application in late-phase diabetes studies.

Crossover design has been widely used in late‐phase clinical studies, as well as in pharmacokinetic and pharmacodynamic, bioequivalence, and medical device studies; however, its interpretability and applicability continue to be debated. Herein we provide discussions around a crossover designs scientific benefit, applicability, and how it can be implemented in late‐phase diabetes studies by properly handling key issues: carryover effect, washout period, and baseline selection. Specifically, detailed considerations are provided about the validity and situations of having appropriate length of study duration to deal with carryover effects so that a washout period may not be needed. A simulation study and data mining results on 12 crossover late‐phase insulin clinical trials are presented to examine the discussion points and proposals.

Statistical and Operational Issues Arising in an Interim Analysis When the Study Will Continue

Guidelines for conducting interim analyses in clinical trials sponsored by the pharmaceutical industry have been recently published (1). Usually, the clinical trial will terminate or the design will change when the interim analysis shows outstanding efficacy results. There are situations, however, where the interim analysis shows outstanding efficacy results and yet the study continues, for example, when regulatory requirements in the United States and Europe differ concerning study duration. A case study is presented which describes the statistical and operational issues encountered while performing a two-year interim analysis of a three-year registration study when the study was to continue to the three-year timepoint with the same design regardless of the outcome of the interim analysis. The statistician plays a central role in developing and implementing the strategy to effectively resolve these issues.