Joseph W. Hogan

The Prevention and Treatment of Missing Data in Clinical Trials

Missing data in clinical trials can have a major effect on the validity of the inferences that can be drawn from the trial. This article reviews methods for preventing missing data and, failing that, dealing with data that are missing.

Day 3 serum inhibin-B is predictive of assisted reproductive technologies outcome

OBJECTIVE To determine if women with day 3 serum inhibin-B concentrations < 45 pg/mL (conversion factor to SI unit, 1.00) demonstrate a poorer response to ovulation induction and assisted reproductive technologies outcome relative to women with inhibin-B values > or = 45 pg/mL. DESIGN Analysis of inhibin-B, FSH, and E2 concentrations in day 3 serum samples. SETTING Academic clinical practice. PATIENT(S) One hundred fifty-six women who underwent 178 assisted reproductive technology (ART) cycles with luteal phase GnRH agonist suppression plus hMG and urofollitropin stimulation. MAIN OUTCOME MEASURE(S) Serum E2 on day of hCG, number of oocytes retrieved per patient, fertilization rate, cleavage rate, clinical pregnancy rate (PR) per initiated cycle, cancellation rate per initiated cycle, and spontaneous abortion rate. RESULT(S) Women with day 3 serum inhibin-B < 45 pg/mL demonstrated 70% of the E2 response, had 66.6% of the number of oocytes retrieved per patient, with 28% of the clinical PR per initiated cycle, and three times the cancellation rate per initiated cycle than women with day 3 inhibin-B > or = 45 pg/mL. After controlling for age, day 3 serum FSH, day 3 serum E2, patient cycle number, and method of ART, day 3 serum inhibin-B > or = 45 pg/mL was noted to be prognostic of the number of oocytes retrieved and clinical PR. The adjusted odds ratio of clinical pregnancy for those with day 3 serum inhibin-B > or = 45 pg/mL versus those with inhibin-B < 45 pg/mL was 6.8 (95% confidence interval 1.8 to 25.6). CONCLUSION(S) Women with low day 3 serum inhibin-B concentrations demonstrate a poorer response to ovulation induction and are less likely to conceive a clinical pregnancy through ART relative to women with high day 3 inhibin-B.

MIXTURE MODELS FOR THE JOINT DISTRIBUTION OF REPEATED MEASURES AND EVENT TIMES

Many long-term clinical trials collect both a vector of repeated measurements and an event time on each subject; often, the two outcomes are dependent. One example is the use of surrogate markers to predict disease onset or survival. Another is longitudinal trials which have outcome-related dropout. We describe a mixture model for the joint distribution which accommodates incomplete repeated measures and right-censored event times, and provide methods for full maximum likelihood estimation. The methods are illustrated through analysis of data from a clinical trial for a new schizophrenia therapy; in the trial, dropout time is closely related to outcome, and the dropout process differs between treatments. The parameter estimates from the model are used to make a treatment comparison after adjusting for the effects of dropout. An added benefit of the analysis is that it permits using the repeated measures to increase efficiency of estimates of the event time distribution.

MODEL‐BASED APPROACHES TO ANALYSING INCOMPLETE LONGITUDINAL AND FAILURE TIME DATA

Since Wu and Carroll (Biometrics 44, 175-188) proposed a model for longitudinal progression in the presence of informative dropout, several researchers have developed and studied models for situations where both a vector of repeated outcomes and an event time is available for each subject. These models have been developed for either longitudinal studies with dropout or for survival studies in which a random, time-varying covariate is measured repeatedly across time. When inference about the longitudinal variable is of interest, event times are treated as covariates and are often incomplete due to censoring. If survival or event time is the primary endpoint, repeated outcomes observed prior to the event are viewed as covariates; this covariate process is often incomplete, measured with error, or observed at unscheduled times during the study. We review several models which are used to handle incomplete response and covariate data in both survival and longitudinal studies.

Association between bacterial vaginosis and expression of human immunodeficiency virus type 1 RNA in the female genital tract.

We assessed the effect of lower genital tract infections on human immunodeficiency virus type 1 (HIV-1) RNA shedding in the female genital tract. Bacterial vaginosis was significantly associated with HIV-1 RNA expression in the female genital tract of HIV-infected women.

A problem-solving approach to stress reduction among younger women with breast carcinoma: a randomized controlled trial.

Previous research indicates that younger women (i.e., ≤ 50) with breast carcinoma experience greater emotional distress than older women (i.e., > 50) and that coping style is significantly related to the psychosocial adjustment of women with this disease. The purpose of this study was to evaluate through a randomized controlled trial the effectiveness of a problem‐solving training intervention designed to empower women with breast carcinoma to cope with a range of difficulties when diagnosed in mid‐life.

Misclassification of First-Line Antiretroviral Treatment Failure Based on Immunological Monitoring of HIV Infection in Resource-Limited Settings

BACKGROUND The monitoring of patients with human immunodeficiency virus (HIV) infection who are treated with antiretroviral medications in resource-limited settings is typically performed by use of clinical and immunological criteria. The early identification of first-line antiretroviral treatment failure is critical to prevent morbidity, mortality, and drug resistance. Misclassification of failure may result in premature switching to second-line therapy. METHODS Adult patients in western Kenya had their viral loads (VLs) determined if they had adhered to first-line therapy for >6 months and were suspected of experiencing immunological failure (ie, their CD4 cell count decreased by 25% in 6 months). Misclassification of treatment failure was defined as a 25% decrease in CD4 cell count with a VL of <400 copies/mL. Logistic and tree regressions examined relationships between VL and 4 variables: CD4 T cell count (hereafter CD4 cell count), percentage of T cells expressing CD4 (hereafter CD4 cell percentage), percentage decrease in the CD4 T cell count (hereafter CD4 cell count percent decrease), and percentage decrease in the percentage of T cells expressing CD4 (hereafter CD4% percent decrease). RESULTS There were 149 patients who were treated for 23 months; they were identified as having a 25% decrease in CD4 cell count (from 375 to 216 cells/microL) and a CD4% percent decrease (from 19% to 15%); of these 149 patients, 86 (58%) were misclassified as having experienced treatment failure. Of 42 patients who had a 50% decrease in CD4 cell count, 18 (43%) were misclassified. In multivariate logistic regression, misclassification odds were associated with a higher CD4 cell count, a shorter duration of therapy, and a smaller CD4% percent decrease. By combining these variables, we may be able to improve our ability to predict treatment failure. CONCLUSIONS Immunological monitoring as a sole indicator of virological failure would lead to a premature switch to valuable second-line regimens for 58% of patients who experience a 25% decrease in CD4 cell count and for 43% patients who experience a 50% decrease in CD4 cell count, and therefore this type of monitoring should be reevaluated. Selective virological monitoring and the addition of indicators like trends CD4% percent decrease and duration of therapy may systematically improve the identification of treatment failure. VL testing is now mandatory for patients suspected of experiencing first-line treatment failure within the Academic Model Providing Access to Healthcare (AMPATH) in western Kenya, and should be considered in all resource-limited settings.

Prevalence of Lower Genital Tract Infections Among Human Immunodeficiency Virus (HIV)—Seropositive and High-Risk HIV-Seronegative Women

This study was undertaken to assess whether the prevalence of lower genital tract infections among human immunodeficiency virus (HIV)-seropositive women was higher than among high-risk HIV-seronegative women at their baseline visit for the HIV Epidemiology Research Study. Results were available for 851 HIV-seropositive and 434 HIV-seronegative women. Human papilloma virus (HPV) infection was more prevalent among HIV-seropositive women (64% vs. 28%). Bacterial vaginosis was common (35% vs. 33%), followed by trichomoniasis (12% vs. 10%), syphilis (8% vs. 6%), Chlamydia trachomatis infection (4% vs. 5%), candidal vaginitis (3% vs. 2%), and Neisseria gonorrhoeae infection (0.8% vs. 0.3%). Alcohol use (odds ratio [OR], 1.8; 95% confidence interval [CI], 1. 3-2.4) and smoking (OR, 1.8; 95% CI, 1.3-2.5) were associated with bacterial vaginosis. Bacterial vaginosis (OR, 2.3; 95% CI, 1.5-3.4), trichomoniasis (OR, 2.3; 95% CI, 1.1-4.7), and syphilis (OR, 3.1; 95% CI, 1.3-7.4) were found to be more prevalent among black women. Our study showed no statistically significant difference in the prevalence of lower genital tract infections except for HPV between HIV-infected and demographically and behaviorally similar HIV-uninfected high-risk women.

The efficacy of moderate-intensity exercise as an aid for smoking cessation in women: a randomized controlled trial.

Evidence suggests that vigorous-intensity exercise interventions may be effective for smoking cessation among women; however, few studies have examined the efficacy of a moderate-intensity exercise program. The present study examined the efficacy of moderate-intensity exercise for smoking cessation among female smokers. Healthy, sedentary female smokers (N = 217) were randomly assigned to an 8-week cognitive-behavioral smoking cessation program plus moderate-intensity exercise (CBT+EX) or to the same cessation program plus equal contact (CBT). A subsample received nicotine replacement therapy. Results indicated that the CBT+EX and CBT groups were equally likely to attain smoking cessation at the end of treatment, as measured by cotinine-verified 7-day point-prevalence abstinence (20.2% for CBT+EX vs. 18.5% for CBT). The CBT+EX group was more likely to report smoking cessation, as measured by 7-day point prevalence at the 3-month follow-up (11.9% vs. 4.6%, p<.05), compared with the CBT group. No group differences were found at 12 months by either 7-day point prevalence (7.3% for CBT+EX vs. 8.3% for CBT) or continuous abstinence (0.9% for CBT+EX vs. 0.9% for CBT). Additionally, among participants in the CBT+EX group, those with higher adherence to the exercise prescription were significantly more likely to achieve smoking cessation at the end of treatment than were participants reporting lower adherence to exercise. Our findings indicate that the empirical support for moderate-intensity exercise as an adjunctive treatment to CBT for smoking cessation may be limited. Perhaps future studies could compare moderate- vs. vigorous-intensity physical activity to test their relative efficacy.

Instrumental variables and inverse probability weighting for causal inference from longitudinal observational studies

Inferring causal effects from longitudinal repeated measures data has high relevance to a number of areas of research, including economics, social sciences and epidemiology. In observational studies in particular, the treatment receipt mechanism is typically not under the control of the investigator; it can depend on various factors, including the outcome of interest. This results in differential selection into treatment levels, and can lead to selection bias when standard routines such as least squares regression are used to estimate causal effects. Interestingly, both the characterization of and methodology for handling selection bias can differ substantially by disciplinary tradition. In social sciences and economics, instrumental variables (IV) is the standard method for estimating linear and nonlinear models in which the error term may be correlated with an observed covariate. When such correlation is not ruled out, the covariate is called endogenous and least squares estimates of the covariate effect are typically biased. The availability of an instrumental variable can be used to reduce or eliminate the bias. In public health and clinical medicine (e.g., epidemiology and biostatistics), selection bias is typically viewed in terms of confounders, and the prevailing methods are geared toward making proper adjustments via explicit use of observed confounders (e.g., stratification, standardization). A class of methods known as inverse probability weighting (IPW) estimators, which relies on modeling selection in terms of confounders, is gaining in popularity for making such adjustments. Our objective is to review and compare IPW and IV for estimating causal treatment effects from longitudinal data, where the treatment may vary with time. We accomplish this by defining the causal estimands in terms of a linear stochastic model of potential outcomes (counterfactuals). Our comparison includes a review of terminology typically used in discussions of causal inference (e.g., confounding, endogeneity); a review of assumptions required to identify causal effects and their implications for estimation and interpretation; description of estimation via inverse weighting and instrumental variables; and a comparative analysis of data from a longitudinal cohort study of HIV-infected women. In our discussion of assumptions and estimation routines, we try to emphasize sufficient conditions needed to implement relatively standard analyses that can essentially be formulated as regression models. In that sense this review is geared toward the quantitative practitioner. The objective of the data analysis is to estimate the causal (therapeutic) effect of receiving combination antiviral therapy on longitudinal CD4 cell counts, where receipt of therapy varies with time and depends on CD4 count and other covariates. Assumptions are reviewed in context, and resulting inferences are compared. The analysis illustrates the importance of considering the existence of unmeasured confounding and of checking for ‘weak instruments.’ It also suggests that IV methodology may have a role in longitudinal cohort studies where potential instrumental variables are available.