Walter W. Hauck

A new statistical procedure for testing equivalence in two-group comparative bioavailability trials

Northwestern University Cancer Center and Department of Community Health and Preventive Medicine, Northwestern University Medical School, Chicago, 111. 2American Critical Care, McGaw Park, Ill., and Department of Epidemiology and Biometry, University of Illinois, School of Public Health, Chicago, I11. 3Address correspondence and reprint requests to Dr. Walter W. Hauck, Northwestern University Cancer Center, Biometry Section, Ward 3-332, 303 E. Chicago Ave., Chicago, IL 60611.

Bioequivalence: Performance of Several Measures of Extent of Absorption

The highest point of the plasma concentration-time profile, Cmax , is currently used by regulatory agencies to assess the rate of drug absorption after single dose administration of oral products. It is, however, quite insensitive, and a number of new measures of rate have been proposed. Using simulations, several approaches toward measuring rate were tested. A set of model scenarios for drugs with typical mean characteristics and statistical distributions was investigated. Using different kinetic models of disposition, the time course of the concentration in plasma was simulated. Intraindividual and interindividual variability and assay error were modeled using Monte Carlo techniques. The accuracy, precision, and ease of use of the various measures of rate were evaluated by simulating crossover design clinical trials and then determining the probability of declaring bioequivalence as a function of differences in rates of absorption between test and reference formulations. All of the rate measures tested showed a degree of insensitivity to changes in rate and no universally superior measure was found. Indeed, the main conclusion is that the choice of a measure should be based on simulations of the particular situation in a bioequivalence trials.

Smokeless tobacco, cardiovascular risk factors, and nicotine and cotinine levels in professional baseball players

BACKGROUND The use of smokeless tobacco (ST), which has increased in popularity over the past 2 decades, results in considerable systemic exposure to nicotine. Nicotine might contribute to atherosclerosis by an effect on cardiovascular risk factors. METHODS The effects of ST use on cardiovascular risk factors and cotinine and nicotine levels were studied in 1061 professional baseball players during spring training in 1988 and 1989. RESULTS Of the study participants 477 (45%) were users. ST use was more common among Whites (55%) than among Blacks (29%) or Hispanics (21%), and users reported heavier consumption of alcohol (p less than .001) and had higher mean serum caffeine levels (p less than .001) than nonusers. ST users did not differ from nonusers in adjusted levels of systolic and diastolic blood pressure, pulse, and total or HDL-cholesterol. Among ST users, participants using snuff had higher mean serum cotinine levels than those who used chewing tobacco (p less than .001). There was no association between serum cotinine levels and adjusted levels of any cardiovascular risk factor studied. However, higher diastolic blood pressures were associated with higher mean serum nicotine levels (p = .02). CONCLUSIONS Smokeless tobacco use has at most a modest effect on cardiovascular risk factors in young physically fit men.

Factors associated with obtaining nipple aspirate fluid: Analysis of 1428 women and literature review

SummaryStudies of cytologic and biochemical constituents of nipple aspirates of breast fluid have contributed to understanding the natural history of benign and malignant breast disease. We conducted multivariate analyses using 1428 women from a recent case-control study of breast disease to determine which factors were independently associated with the ability to obtain breast fluid from nonlactating women. We then compared results from these analyses to the results from five previous studies that also used the aspiration technique of Sartorius. Four factors were consistently associated across studies with increased ability to obtain breast fluid: 1) age up to 35 to 50 years; 2) earlier age at menarche; 3) non-Asian compared to Asian ethnicity; and 4) history of lactation. Exogenous estrogen use, endogenous estrogen concentrations, phase of menstrual cycle, family history of breast cancer, type of menopause, and less than full-term pregnancy consistently did not influence ability to obtain fluid.New findings from this study shed light on some apparently contradictory findings from the previous studies. In particular, this study showed that the effects of age on ability to obtain fluid appeared to be independent of the effects of menopause. Furthermore, discrepancies in previous findings on the effects of parity on ability to obtain fluid may be explained by our finding that the increased ability to obtain fluid from parous compared to nulliparous women applied only to parous women who had breastfed.

The effects of propranolol on cognitive function and quality of life: a randomized trial among patients with diastolic hypertension ∗

PURPOSE We sought to determine whether propranolol has adverse effects on cognitive function, depressive symptoms, and sexual function in patients treated for diastolic hypertension. SUBJECTS AND METHODS We performed a placebo-controlled trial among 312 men and women, 22 to 59 years of age, who had untreated diastolic hypertension (90 to 104 mm Hg). Patients were randomly assigned to treatment with propranolol (80 to 400 mg/day) or matching placebo tablets. Thirteen tests of cognitive function were assessed at baseline, 3 months, and 12 months. Five tests measured reaction time to, or accuracy in, interpreting visual stimuli; one test measured the ability to acquire, reproduce, and change a set of arbitrary stimulus-response sets; and seven tests measured memory or learning verbal information. Depressive symptoms and sexual function were assessed by questionnaires at baseline and 12 months. RESULTS There were no significant differences by treatment assignment for 11 of the 13 tests of cognitive function at either 3 or 12 months of follow-up. Compared with placebo, participants treated with propranolol had slightly fewer correct responses at 3 months (33 +/- 3 [mean +/- SD] versus 34 +/- 2, P = 0.02) and slightly more errors of commission at 3 months (4 +/-5 versus 3 +/- 3, P = 0.04) and at 12 months (4 +/- 4 versus 3 +/- 3, P = 0.05). At 12 months, depressive symptoms and sexual function and desire did not differ by treatment assignment. CONCLUSIONS Treatment of hypertension with propranolol had limited adverse effects on tests of cognitive function that were of questionable clinical relevance, and there were no documented adverse effects on depressive symptoms or sexual function. Selection of beta-blockers for treatment of hypertension should be based on other factors.

Immediate side effects of stereotactic radiotherapy and radiosurgery.

PURPOSE Despite increased utilization of fractionated stereotactic radiation therapy (SRT) or stereotactic radiosurgery (SRS), the incidence and nature of immediate side effects (ISE) associated with these treatment techniques are not well defined. We report immediate side effects from a series of 78 patients. MATERIALS AND METHODS Intracranial lesions in 78 adult patients were treated with SRT or SRS, using a dedicated linear accelerator. Those lesions included 13 gliomas, 2 ependymomas, 19 metastatic tumors, 15 meningiomas, 12 acoustic neuromas, 4 pituitary adenomas, 1 optic neuroma, 1 chondrosarcoma, and 11 arteriovenous malformations (AVM). SRT was used in 51 and SRS in 27 patients. Mean target volume was 9.0 cc. Eleven patients received prior external-beam radiation therapy within 2 months before SRT/SRS. Any side effects occurring during and up to 2 weeks after the course of radiation were defined as ISE and were graded as mild, moderate, or severe. The incidence of ISE and the significance of their association with several treatment and pretreatment variables were analyzed. RESULTS Overall, 28 (35%) of 78 patients experienced one or more ISE. Most of the ISE (87%) were mild, and consisted of nausea (in 5), dizziness/vertigo (in 5), seizures (in 6), and new persistent headaches (in 17). Two episodes of worsening neurological deficit and 2 of orbital pain were graded as moderate. Two patients experienced severe ISE, requiring hospitalization (1 seizure and 1 worsening neurological deficit). ISE in 6 cases prompted computerized tomography of the brain, which revealed increased perilesional edema in 3 cases. The incidence of ISE by diagnosis was as follows: 46% (6 of 13) for gliomas, 50% (6 of 12) for acoustic neuromas, 36% (4 of 11) for AVM, 33% (5 of 15) for meningiomas, and 21% (4 of 19) for metastases. A higher incidence of dizziness/vertigo (4 of 12 = 33%) was seen among acoustic neuroma patients than among other patients (p< 0.01). There was no significant association of dizziness/vertigo with either a higher average and maximum brainstem dose (p = 0.74 and 0.09, respectively) or with 2-Gy equivalents of the average and maximum brainstem doses (p = 0.28 and 0.09, respectively). Higher RT dose to the margin and higher maximum RT dose were associated with a higher incidence of ISE (p = 0.05 and 0.01, respectively). However, when RT dose to the margin was converted to a 2-Gy dose-equivalent, it lost its significance as predictor of ISE (p = 0.51). Recent conventional external-beam radiation therapy, target volume, number of isocenters, collimator size, dose inhomogeneity, prescription isodose, pretreatment edema, dose of prior radiation, fraction size (2.0-7.0 Gy with SRT and 13.0-21.0 Gy with SRS), fractionation schedule, and dose to brainstem were not significantly associated with ISE. ISE occurred in 26% (8) of 31 patients taking corticosteroids prior to SRT/SRS and in 20 (42%) of 47 patients not taking them (p = 0.15). CONCLUSION ISE occur in one third of patients treated with SRT and SRS and are usually mild or moderate and self-limited. Dizziness/vertigo are common and unique for patients with acoustic neuromas and are not associated with higher brainstem doses. We are unable to detect a relationship between ISE and higher margin or maximum RT doses. No specific conclusion can be drawn with regard to the effect of corticosteroids, used prior to SRS/SRT, on the occurrence of ISE.

A consequence of omitted covariates when estimating odds ratios

In the epidemiologic literature, one finds three criteria for confounding, which we will call the classical (marginal), operational (change-in-estimate) and conditional criteria. We define mavericks to be covariates that satisfy the operational criterion, but not the classical criterion. We present what is known about the problems of mavericks for estimating odds ratios and clarify the interpretation of odds ratios. Key results are: (1) omitting mavericks biases odds ratios towards 1; (2) omitting mavericks cannot artificially introduce an effect in contrast to omitting classical confounders; (3) the operational criterion for confounding corresponds to the conditional criterion when estimating odds ratios, but for relative risks, there are no mavericks (i.e. the classical and operational criterion correspond); and (4) the interpretation of odds ratios obtained from standard methods is that of comparing proportions, not of individual risk.

A Comparison of Large-Sample Confidence Interval Methods for the Difference of Two Binomial Probabilities

Abstract A simulation study was done to compare seven confidence interval methods, based on the normal approximation, for the difference of two binomial probabilities. Cases considered included minimum expected cell sizes ranging from 2 to 15 and smallest group sizes (NMIN) ranging from 6 to 100. Our recommendation is to use a continuity correction of 1/(2 NMIN) combined with the use of (N − 1) rather than N in the estimate of the standard error. For all of the cases considered with minimum expected cell size of at least 3, this method gave coverage probabilities close to or greater than the nominal 90% and 95%. The Yates method is also acceptable, but it is slightly more conservative. At the other extreme, the usual method (with no continuity correction) does not provide adequate coverage even at the larger sample sizes. For the 99% intervals, our recommended method and the Yates correction performed equally well and are reasonable for minimum expected cell sizes of at least 5. None of the methods perfor...

Multiple sexual partners and their psychosocial correlates: The population-based AIDS in Multiethnic Neighborhoods (AMEN) Study.

In this study we examined multiple partners in a household probability sample of heterosexuals. Thirty-seven percent reported 2 or more partners in the past year, and 6% reported 5 or more partners. Significant interactions among gender, ethnicity, and relationship status revealed complex relations. African American men without a primary partner were the most likely to have multiple partners; ethnic minority women with primary partners were the least likely. Psychosocial factors associated with multiple partners were examined with the AIDS Risk Reduction Model (ARRM). Situational factors influenced how ARRM variables related to multiple partners. Perceived risk, commitment to monogamy, and communication were all important correlates. Prevention efforts must focus on increasing awareness of the risks of heterosexual transmission of HIV for people in dating relationships and on strengthening sexual communication skills with new partners.

Proportional hazards (cox) regression

PROPORTIONAL HAZARDS ( C o x ) REGRESSION is a powerful analytic tool for testing whe the r several factors (e.g., cigarette smoking, hyper tens ion) are independent ly related to the rate (over t ime) of a specific event (e.g., heart attack yes /no) . It can also be used to control for baseline differences be t ween groups in nonrandomized studies and randomized clinical trials (RCTs). The availabili ty of desktop compute r s and userfr iendly software has resul ted in a marked increase in the use of propor t iona l hazards regression by clinical researchers. However , most detai led reviews of the t echnique 1-5 cannot be unders tood by non-statisticians. In this article we begin wi th a rev iew of s impler types of survival analyses, highl ight ing the concepts of rate of ou t come and censored observations. Building on these two concepts , we descr ibe the statistical propert ies , under lying assumptions, interpretat ion, and applicat ion of propor t iona l hazards regression. Also we describe t ime-dependen t covariates, the use of proport ional hazards regression versus logistic regression, and other technical aspects of propor t iona l hazards regression. Finally, we illustrate the appl icat ions of this techn ique by reviewing 80 articles f rom the New England Journal o f Medicine and the Annals of Internal Medicine that used propor t iona l hazards regression dur ing 1984, 1987, and 1990. Our goal is to enable non-statisticians to in terpret these models and to provide guidelines for clinical researchers per forming this type of analysis.