Eric W. Lee

Cox-Type Regression Analysis for Large Numbers of Small Groups of Correlated Failure Time Observations

The Cox regression model has been used extensively to analyze survival data. For data that consist of large numbers of small groups of correlated failure time observations, we show that the standard maximum partial likelihood estimate of the regression coefficient in the Cox model is still consistent and asymptotically normal. However, the corresponding standard variance-covariance estimate may no longer be valid due to the dependence among members in the groups. In this article, a correct variance-covariance estimate that takes account of the intra-group correlation is proposed. Power comparisons are performed to show the advantage of the new proposal. Examples are provided for illustration.

Heterosexual transmission of HIV in a cohort of couples in New York City

Objective:Since heterosexual transmission of HIV in the United States is occurring at an increasing rate, especially among black and Hispanic couples and those in which one member has a history of intravenous drug use, we sought to study the heterosexual transmission of HIV in couples. Design:Multiple logistic regression analysis of risks for HIV infection in female partners. Methods:We enrolled 158 non-intravenous drug user (IVDU) steady heterosexual partners of HIV-infected individuals (indexes) in this study. Of these, 93% were women, 54% were Hispanic whites, 23% were black and 65.6% were partners of IVDU. Results:In a multiple logistic regression analysis of risks for HIV infection in female partners, the strongest predictors of transmission were AIDS or AIDS-related complex (ARC) in the index [adjusted odds ratios (OR), 16.81; P< 0.001 and 12.53; P = 0.003, respectively], a history of anal intercourse (adjusted OR, 10.81; P< 0.001) and bleeding as a result of intercourse (adjusted OR, 4.90; P< 0.05). Female-to-male transmission was detected in seven out of 11 couples at risk. Ethnicity, number of episodes of vaginal intercourse, number of other sexual partners and history of sexually transmitted infections were not significantly associated with transmission to women. Conclusion:Our study demonstrates that health of the index, anal intercourse and bleeding as a result of intercourse are the major determinants of sexual transmission of HIV to women in couples.

Predicting ten‐year survival of patients with primary cutaneous melanoma

Recently, the Pigmented Lesion Group at the University of Pennsylvania described a 4‐variable model for predicting 10‐year survival for patients with primary cutaneous melanoma. The variables are tumor thickness, anatomic site of the lesion, age, and gender. The objective of the current study was to test the validity of this model, employing the large data base of the New York University Melanoma Cooperative Group.

Impaired early growth of infants perinatally infected with human immunodeficiency virus: Correlation with viral load

OBJECTIVEnTo evaluate the effect of viral load on the early growth of infants infected with human immunodeficiency virus (HIV).nnnMETHODSnPlasma concentrations of p24-antigen and HIV ribonucleic acid were measured retrospectively and correlated with growth parameters for the first 18 months of life in a cohort of 47 term infants born to HIV-infected mothers prospectively enrolled in a study of perinatal HIV transmission. Comparisons of the mean weight and length of the 18 HIV-infected and 29 uninfected infants for each interval and across intervals were made. Viral load was correlated with standard deviation scores. Infants were stratified by high and low viral load during the first 6 months of life.nnnRESULTSnAt birth, no difference in weight and length was observed between HIV-infected and uninfected infants. Between birth and 6 months of age, the infected infants grew less rapidly than the uninfected infants, a finding temporally associated with an exponential increase in HIV viremia. The linear growth of infected infants remained consistently less than that of the uninfected infants after 6 months of life, although the differences were no longer statistically significant and tended to decrease with age in parallel with declines in viral load. The median plasma concentration of HIV ribonucleic acid was significantly higher at 3, 6, 12, and 18 months in infected infants in whom growth failure developed. Infants who had a high viral load in the first 6 months of life were significantly more likely to have severe growth failure. Though the mean SD for weight of the infected infants was always less than that of the uninfected infants, the differences were small and not significant.nnnCONCLUSIONSnOur results confirm the observation that stunting is an early frequent finding in perinatal HIV infection. The deleterious effect of HIV on linear growth appears to be correlated with the level of postnatal HIV viremia, although the exact mechanism of this association remains to be elucidated.

Linear regression analysis for highly stratified failure time data

Abstract In this article we consider cases where data consist of many small and independent groups of correlated failure time observations. For each failure time, which may be censored, some important covariates are also recorded. Our goal is to examine the covariate effects on the individual failure time observations. We assume that the logarithm of each failure time is linearly related to its covariates. We then take a population-averaged model approach to obtain inference procedures for the regression parameters without specifying the joint distribution of the observations within the group. The new proposals do not need complicated and unstable nonparametric estimates for the hazard function of the error term. Their properties are extensively examined for practical sample sizes. Comparisons among various procedures are also performed. All the methods studied in this article are illustrated with examples.

Hepatic Arterial Embolization for the Treatment of Metastatic Neuroendocrine Tumors

Neuroendocrine tumors (NETs) have a high predilection for metastasizing to the liver and can cause severe debilitating symptoms adversely affecting quality of life. Although surgery remains the treatment of choice, many liver metastases are inoperable at presentation. Hepatic arterial embolization procedures take advantage of the arterial supply of NET metastases. The goals of these therapies are twofold: to increase overall survival by stabilizing tumor growth, and to reduce the morbidity in symptomatic patients. Patients treated with hepatic arterial embolization demonstrate longer progression-free survival and have 5-year survival rates of nearly 30%. The safety of repeat embolizations has also been proven in the setting of recurrent symptoms or progression of the disease. Despite not being curative, hepatic arterial embolization should be used in the management of NETs with liver metastases. Long-term survival is not uncommon, making aggressive palliation of symptoms an important component of treatment.

THE ANALYSIS OF CORRELATED PANEL DATA USING A CONTINUOUS-TIME MARKOV MODEL

We consider the analysis of correlated panel data in which two or more correlated multistate processes are periodically observed on each individual and the exact transition times between states are unknown. We describe a procedure that models each process marginally under a time-homogeneous Markov model allowing for covariates. The resulting estimators are shown to be asymptotically jointly normal with a covariance matrix that can be consistently estimated. Simultaneous inference procedures are also proposed. Methods are illustrated using data from an AIDS clinical trial to compare the toxic effects of two treatments on two hematologic variables, hemoglobin and absolute neutrophil count.

TWO SAMPLE COMPARISON FOR LARGE GROUPS OF CORRELATED BINARY RESPONSES

Although there is much literature on testing the treatment effect in experiments with correlated binary outcomes, the existing methodologies only work well when the cluster size is small. In this paper, I propose a simple method to test the treatment effect in experiments with large groups of correlated binary outcomes. I use the weighted estimating equations approach to estimate the treatment effect. The proposal is appropriate under any correlation structure. Power comparisons show the advantage of the new procedure.