Estimating mono- and bi-phasic regression parameters using a mixture piecewise linear Bayesian hierarchical model.

The dynamics of tumor burden, secreted proteins or other biomarkers over time, is often used to evaluate the effectiveness of therapy and to predict outcomes for patients. Many methods have been proposed to investigate longitudinal trends to better characterize patients and to understand disease progression. However, most approaches assume a homogeneous patient population and a uniform response trajectory over time and across patients. Here, we present a mixture piecewise linear Bayesian hierarchical model, which takes into account both population heterogeneity and nonlinear relationships between biomarkers and time. Simulation results show that our method was able to classify subjects according to their patterns of treatment response with greater than 80% accuracy in the three scenarios tested. We then applied our model to a large randomized controlled phase III clinical trial of multiple myeloma patients. Analysis results suggest that the longitudinal tumor burden trajectories in multiple myeloma patients are heterogeneous and nonlinear, even among patients assigned to the same treatment cohort. In addition, between cohorts, there are distinct differences in terms of the regression parameters and the distributions among categories in the mixture. Those results imply that longitudinal data from clinical trials may harbor unobserved subgroups and nonlinear relationships; accounting for both may be important for analyzing longitudinal data.

Can computer-based feedback improve emergency department patient uptake of rapid HIV screening?

OBJECTIVE: We determine whether (1) an audiocomputer-delivered tailored feedback intervention increases emergency department (ED) patient uptake of opt-in, nontargeted rapid HIV screening; and (2) uptake is greater among patients who report more HIV risk and among those whose self-perceived HIV risk increases from baseline after completion of an HIV risk assessment. METHODS: ED patients aged 18 to 64 years were randomly assigned to receive either an assessment about reported and self-perceived HIV risk or an identical assessment plus feedback about their risk for having or acquiring an HIV infection, tailored according to their reported risk. All participants were offered a fingerstick rapid HIV test. Two-sample tests of binomial proportions were used to compare screening uptake by study arm. Multivariable logistic regression was used to assess the relationship of reported HIV risk and an increase in self-perceived HIV risk with uptake of HIV screening. RESULTS: Of the 566 participants, the median age was 29 years, 62.2% were women, and 66.9% previously had been tested for HIV. Uptake of HIV screening was similar in the intervention and no intervention arms (54.1% versus 55.5% [Δ=-0.01%; 95% confidence interval {CI} -0.09% to 0.07%]). An increase in self-perceived HIV risk predicted greater uptake of HIV screening for women (odds ratio 2.15; 95% CI 1.08 to 4.28) but not men (odds ratio 1.61; 95% CI 0.60 to 4.30). Uptake of HIV screening was not related to reported HIV risk. CONCLUSION: Uptake of rapid HIV screening in the ED was not improved by this feedback intervention. Other methods need to be investigated to improve uptake of HIV screening by ED patients.

Statistical interpretation of the RV144 HIV vaccine efficacy trial in Thailand: a case study for statistical issues in efficacy trials.

Recently, the RV144 randomized, double-blind, efficacy trial in Thailand reported that a prime-boost human immunodeficiency virus (HIV) vaccine regimen conferred ∼30% protection against HIV acquisition. However, different analyses seemed to give conflicting results, and a heated debate ensued as scientists and the broader public struggled with their interpretation. The lack of accounting for statistical principles helped flame the debate, and we leverage these principles to provide a more scientific interpretation. We first address interpretation of frequentist results, including interpretation of P values, synthesis of results from multiple analyses (ie, intention-to-treat versus per-protocol/fully immunized), and accounting for external efficacy trials. Second, we address how Bayesian statistics, which provide clearly interpretable statements about probabilities that the vaccine efficacy takes certain values, provide more information for weighing the evidence about efficacy than do frequentist statistics alone. Third, we evaluate RV144 for completeness of end point ascertainment and integrity of blinding, necessary tasks for establishing robustly interpretable results.

The relationship of reported HIV risk and history of HIV testing among emergency department patients.

OBJECTIVE: Among a random sample of emergency department (ED) patients, we sought to determine the extent to which reported risk for human immunodeficiency virus (HIV) is related to ever having been tested for HIV. METHODS: A random sample of patients (aged 18-64 years) from an adult, urban, northeastern United States, academic ED were surveyed about their history of ever having been tested for HIV and their reported HIV risk behaviors. A reported HIV risk score was calculated from the survey responses and divided into 4 levels, based on quartiles of the risk scores. Pearson's X(2) testing was used to compare HIV testing history and level of reported HIV risk. Logistic regression models were created to investigate the association between level of reported HIV risk and the outcome of ever having been tested for HIV. RESULTS: Of the 557 participants, 62.1% were female. A larger proportion of females than males (71.4% vs 60.6%; P < 0.01) reported they had been tested for HIV. Among the 211 males, 11.4% reported no HIV risk, and among the 346 females, 10.7% reported no HIV risk. The proportion of those who had been tested for HIV was greater among those reporting any risk compared with those reporting no risk for females (75.4% vs 37.8%; P < 0.001), but not for males (59.9% vs 66.7%; P < 0.52). However, certain high-risk behaviors, such as a history of injection-drug use, were associated with prior HIV testing for both genders. In the logistic regression analyses, there was no relationship between increasing level of reported HIV risk and a history of ever having been tested for HIV for males. For females, a history of ever having been tested was related to increasing level of reported risk, but not in a linear fashion. CONCLUSIONS: The relationship between reported HIV risk and history of testing among these ED patients was complex and differed by gender. Among these patients, having greater risk did not necessarily mean a higher likelihood of ever having been tested for HIV.

Demographic variations in HIV testing history among emergency department patients: implications for HIV screening in US emergency departments.

OBJECTIVE: To determine the proportion of emergency department (ED) patients who have been tested for human immunodeficiency virus (HIV) infection and assess if patient history of HIV testing varies according to patient demographic characteristics. DESIGN: From July 2005-July 2006, a random sample of 18-55-year-old English-speaking patients being treated for sub-critical injury or illness at a northeastern US ED were interviewed on their history of HIV testing. Logistic regression models were created to compare patients by their history of being tested for HIV according to their demography. Odds ratios (ORs) with 95% confidence intervals (CIs) were estimated. RESULTS: Of 2107 patients surveyed who were not known to be HIV-infected, the median age was 32 years; 54% were male, 71% were white, and 45% were single/never married; 49% had private health-care insurance and 45% had never been tested for HIV. Of the 946 never previously tested for HIV, 56.1% did not consider themselves at risk for HIV. In multivariable logistic regression analyses, those less likely to have been HIV tested were male (OR: 1.32 [1.37-2.73]), white (OR: 1.93 [1.37-2.73]), married (OR: 1.53 [1.12-2.08]), and had private health-care insurance (OR: 2.10 [1.69-2.61]). There was a U-shaped relationship between age and history of being tested for HIV; younger and older patients were less likely to have been tested. History of HIV testing and years of formal education were not related. CONCLUSION: Almost half of ED patients surveyed had never been tested for HIV. Certain demographic groups are being missed though HIV diagnostic testing and screening programmes in other settings. These groups could potentially be reached through universal screening.

Emergency department patient perceptions and preferences on opt-in rapid HIV screening program components.

The aim of this investigation was to assess emergency department (ED) patients' perceptions and preferences about an opt-in, universal, rapid HIV screening program and identify patient groups who expressed stronger beliefs about components of the testing program. From July 2005 to July 2006, ED patients in the opt-in, universal, rapid HIV screening program were interviewed in person. Multivariable regression models were used to compare participants on their beliefs about the program components. Of the 561 participants, 62.0% had previously been tested for HIV. The majority of participants (58.8%) believed the rapid and standard/conventional HIV tests to be equally accurate, 27.7% believed the rapid test to be less or much less accurate, and 8.7% believed the rapid test to be more or much more accurate. Almost two-thirds (65.1%) favored having a rapid instead of a standard/conventional HIV test, 94.6% wanted the test results within one hour, and 61.3% would be likely or very likely to undergo testing in the ED if it prolonged their ED visit. Almost all (92.5%) believed that their medical care was "not at all" delayed because of being tested, 94.1% believed that testing did "not at all" divert attention from the reason for their ED visit, and 80.9% thought that testing in the ED was "not at all" stressful. In multivariable logistic regression models, males and those with more than 12 years of formal education showed greater concerns about the rapid HIV test's accuracy. Hispanic/Latinos, participants with governmental insurance, and those previously HIV tested were more apt to be screened for HIV even if testing delayed their ED departure. Overall, participants were highly accepting of the components of this opt-in rapid HIV screening program. However, concerns regarding the accuracy of the rapid HIV test might limit test acceptance and should be addressed during pre-test information procedures.

Video as an effective method to deliver pretest information for rapid human immunodeficiency testing.

OBJECTIVES: Video-based delivery of human immunodeficiency virus (HIV) pretest information might assist in streamlining HIV screening and testing efforts in the emergency department (ED). The objectives of this study were to determine if the video "Do you know about rapid HIV testing?" is an acceptable alternative to an in-person information session on rapid HIV pretest information, in regard to comprehension of rapid HIV pretest fundamentals, and to identify patients who might have difficulties in comprehending pretest information. METHODS: This was a noninferiority trial of 574 participants in an ED opt-in rapid HIV screening program who were randomly assigned to receive identical pretest information from either an animated and live-action 9.5-minute video or an in-person information session. Pretest information comprehension was assessed using a questionnaire. The video would be accepted as not inferior to the in-person information session if the 95% confidence interval (CI) of the difference (Delta) in mean scores on the questionnaire between the two information groups was less than a 10% decrease in the in-person information session arm's mean score. Linear regression models were constructed to identify patients with lower mean scores based upon study arm assignment, demographic characteristics, and history of prior HIV testing. RESULTS: The questionnaire mean scores were 20.1 (95% CI = 19.7 to 20.5) for the video arm and 20.8 (95% CI = 20.4 to 21.2) for the in-person information session arm. The difference in mean scores compared to the mean score for the in-person information session met the noninferiority criterion for this investigation (Delta = 0.68; 95% CI = 0.18 to 1.26). In a multivariable linear regression model, Blacks/African Americans, Hispanics, and those with Medicare and Medicaid insurance exhibited slightly lower mean scores, regardless of the pretest information delivery format. There was a strong relationship between fewer years of formal education and lower mean scores on the questionnaire. Age, gender, type of insurance, partner/marital status, and history of prior HIV testing were not predictive of scores on the questionnaire. CONCLUSIONS: In terms of patient comprehension of rapid HIV pretest information fundamentals, the video was an acceptable substitute to pretest information delivered by an HIV test counselor. Both the video and the in-person information session were less effective in providing pretest information for patients with fewer years of formal education.

Effectiveness of increasing emergency department patients' self-perceived risk for being human immunodeficiency virus (HIV) infected through audio computer self-interview-based feedback about reported HIV risk behaviors.

OBJECTIVES: Prior research has demonstrated that emergency department (ED) patient acceptance of human immunodeficiency virus (HIV) screening is partially dependent on patients' self-perceived risk of infection. The primary objective of this study was to determine the effectiveness of audio computer-assisted self-interview (ACASI)-based feedback. The intervention aimed to increase patient's self-perceived risk of being HIV infected by providing immediate feedback on their risk behaviors. METHODS: This 1-year, randomized, controlled trial at a U.S. ED enrolled a random sample of 18- to 64-year-old subcritically ill or injured adult patients who were not known to be HIV infected. All participants completed an anonymous, ACASI-based questionnaire about their HIV risk behaviors related to injection drug use and sex, as well as their self-perceived risk for being HIV infected. Participants were randomly assigned to one of two study groups: an intervention group in which participants received immediate ACASI-based feedback in response to each of their reported risk behaviors or a no-intervention group without feedback. Participants were asked to indicate their level of HIV risk on a five-point scale before and after they answered the questions. Change in level of self-perceived HIV risk was calculated and compared by study group using Pearson's chi-square test. An HIV risk behavior score that summarized reported HIV risk behavior was devised. Because HIV risk behaviors differ by sex, scores were calculated separately for each sex. Linear regression models that adjusted for study group and same subject covariance were employed to determine if higher HIV risk behavior scores were associated with an increase in self-perceived HIV risk. RESULTS: Of the 566 trial participants, the median age was 29 years (interquartile range [IQR] = 22-43 years), 62.2% were females, and 66.9% had been tested previously for HIV. After answering the reported HIV risk behavior questions, 12.6% of participants had an increase, 79.9% had no change, and 7.5% had a decrease in self-perceived HIV risk. Of the 46.6% of participants who initially indicated that they were not at risk for HIV, 11.4% had an increase in self-perceived HIV risk after answering the reported HIV risk behavior questions. Change in self-perceived HIV risk did not differ by study group (p = 0.77). There were no differences in reported HIV risk scores between the intervention and no-intervention groups for females (p = 0.78) or males (p = 0.86). In the linear regression models, a greater increase in self-perceived HIV risk was associated with higher reported HIV risk behavior scores for females (beta = 0.59, 95% confidence interval [CI] = 0.15, 1.04) but not for males (beta = 1.00, 95% CI = -0.13 to 2.14). CONCLUSIONS: Some ED patients can be moved, although modestly, to recognize their risk for being HIV infected by asking about their HIV risk behaviors. However, ACASI-based feedback messages about HIV risk behaviors do not increase subjects' self-perceived HIV risk. Female ED patients appear to increase their self-perceived HIV risk more than males when queried about their HIV risk behaviors.

Accommodating uncertainty in a tree set for function estimation.

Multiple branching trees have been used to model the acquisition of HIV drug resistance mutations, and several different algorithms have been developed to construct the tree set that best describes the data. These algorithms have mainly focused on the structure of the tree set. The focal point of this paper is estimation of functions of the tree set parameters that incorporate uncertainty in the tree set. The functions of interest are the state probabilities, the co-occurrence of mutations and the order of acquisition. Such functions are of interest because they help characterize the genetic pathways that lead to multi-drug resistance. We propose a bootstrap technique to account for the additional variability in estimates due to uncertainty in the tree set. The methods are applied to genetic sequences of patients from a database compiled by the Forum for Collaborative HIV Research in an effort to characterize genetic pathways to resistance to drugs from the nucleoside reverse transcriptase inhibitor (NRTI) class. The main results were that patients with a 211K mutation in the RT region of the viral genome were more likely to have a 215Y mutation and less likely to have a 70R mutation compared to patients without a 211K mutation.

Emergency department patient acceptance of opt-in, universal, rapid HIV screening.

OBJECTIVES: We assessed emergency department (ED) patient acceptance of opt-in, rapid human immunodeficiency virus (HIV) screening and identified demographic characteristics and HIV testing-history factors associated with acceptance of screening. METHODS: A random sample of 18- to 55-year-old ED patients was offered rapid HIV screening. Patient acceptance or decline of screening and the reasons for acceptance or decline were analyzed with multivariable regression models. Odds ratios (ORs) with 95% confidence intervals (CIs) were estimated for the logistic regression models. RESULTS: Of the 2,099 participants, 39.3% accepted HIV screening. In a multinomial regression model, participants who were never married/not partnered, did not have private health insurance, and had 12 or fewer years of education were more likely to be screened due to concern about a possible HIV exposure. In a multivariable logistic regression model, the odds of accepting screening were greater among those who were younger than 40-years-old (OR=1.61, 95% CI 1.32, 2.00), nonwhite (OR=1.28, 95% CI 1.04, 1.58), not married (OR=1.82, 95% CI 1.44, 2.28), lacking private health insurance (OR=1.40, 95% CI 1.13, 1.74), and who had 12 or fewer years of education (OR=1.43, 95% CI 1.16, 1.75). Despite use of a standardized protocol, patient acceptance of screening varied by which research assistant asked them to be screened. Patients not previously tested for HIV who were white, married, and 45 years or older and who had private health insurance were more likely to decline HIV screening. CONCLUSIONS: In an opt-in, universal, ED HIV screening program, patient acceptance of screening varied by demography, which indicates that the impact of such screening programs will not be universal. Future research will need to determine methods of increasing uptake of ED HIV screening that transcend patient demographic characteristics, HIV testing history, and motivation for testing.

Two-sample tests of area-under-the-curve in the presence of missing data.

The commonly used two-sample tests of equal area-under-the-curve (AUC), where AUC is based on the linear trapezoidal rule, may have poor properties when observations are missing, even if they are missing completely at random (MCAR). We propose two tests: one that has good properties when data are MCAR and another that has good properties when the data are missing at random (MAR), provided that the pattern of missingness is monotonic. In addition, we discuss other non-parametric tests of hypotheses that are similar, but not identical, to the hypothesis of equal AUCs, but that often have better statistical properties than do AUC tests and may be more scientifically appropriate for many settings.

Comparison of patient comprehension of rapid HIV pre-test fundamentals by information delivery format in an emergency department setting.

BACKGROUND: Two trials were conducted to compare emergency department patient comprehension of rapid HIV pre-test information using different methods to deliver this information. METHODS: Patients were enrolled for these two trials at a US emergency department between February 2005 and January 2006. In Trial One, patients were randomized to a no pre-test information or an in-person discussion arm. In Trial Two, a separate group of patients were randomized to an in-person discussion arm or a Tablet PC-based video arm. The video, "Do you know about rapid HIV testing?", and the in-person discussion contained identical Centers for Disease Control and Prevention-suggested pre-test information components as well as information on rapid HIV testing with OraQuick. Participants were compared by information arm on their comprehension of the pre-test information by their score on a 26-item questionnaire using the Wilcoxon rank-sum test. RESULTS: In Trial One, 38 patients completed the no-information arm and 31 completed the in-person discussion arm. Of these 69 patients, 63.8% had twelve years or fewer of formal education and 66.7% had previously been tested for HIV. The mean score on the questionnaire for the in-person discussion arm was higher than for the no information arm (18.7 vs. 13.3, p < or = 0.0001). In Trial Two, 59 patients completed the in-person discussion and 55 completed the video arms. Of these 114 patients, 50.9% had twelve years or fewer of formal education and 68.4% had previously been tested for HIV. The mean score on the questionnaire for the video arm was similar to the in-person discussion arm (20.0 vs. 19.2; p < or = 0.33). CONCLUSION: The video "Do you know about rapid HIV testing?" appears to be an acceptable substitute for an in-person pre-test discussion on rapid HIV testing with OraQuick. In terms of adequately informing ED patients about rapid HIV testing, either form of pre-test information is preferable than for patients to receive no pre-test information.

A resampling-based approach to multiple testing with uncertainty in phase.

Characterizing the genetic correlates to complex diseases requires consideration of a large number of potentially informative biological markers. In addition, attention to alignment of alleles within or across chromosomal pairs, commonly referred to as phase, may be essential for uncovering true biological associations. In the context of population based association studies, phase is generally unobservable. Preservation of type-1 error in a setting with multiple testing presents a further analytical challenge. This manuscript combines a likelihood-based approach to handling missing-ness in phase with a resampling method to adjust for multiple testing. Through simulations we demonstrate preservation of the family-wise error rate and reasonable power for detecting associations. The method is applied to a cohort of 626 HIV-1 infected individuals receiving highly active anti-retroviral therapies, to ascertain potential genetic contributions to abnormalities in lipid profiles. The haplotypic effects of 2 genes, hepatic lipase (HL) and endothelial lipase (EL), on high-density lipoprotein cholesterol (HDL-C) are tested.

What constitutes efficacy for a human immunodeficiency virus vaccine that ameliorates viremia: issues involving surrogate end points in phase 3 trials.

Initial human immunodeficiency virus (HIV) vaccines are unlikely to prevent acquisition of HIV in all recipients. Moreover, several HIV vaccines are under evaluation that are designed to reduce viremia after acquisition of infection. Such vaccines could provide important benefits to delay HIV progression and to reduce transmission. The decision to license a vaccine on the basis of observed effects on virus load and other postinfection surrogate end points in an efficacy trial is complicated by uncertainty about whether the vaccine effects will persist and reliably predict clinical effects, and by the challenge in interpreting the data posed by treatment of some seroconverters with antiretroviral drugs. Here, we evaluate how analyses of certain surrogate end points can be used for inferring clinically significant vaccine effects and propose end points that could be evaluated in efficacy trials to support licensure. The assessment suggests that a vaccine demonstrating moderately durable effects to delay therapy and to ameliorate viremia merits consideration for licensure.

Baseline predictors of CD4 T-lymphocyte recovery with combination antiretroviral therapy.

CD4 T-cell recovery with potent antiretroviral therapy varies considerably among HIV-1-infected patients. Data from two studies that enrolled subjects at different stages of disease progression were retrospectively combined. This analysis assessed the association between patient-specific factors and three measures of CD4 T-cell recovery after the initiation of triple-drug therapy: overall changes in CD4 cell counts; changes in CD4 cell counts during the first 8 weeks (phase I); and changes in CD4 cell counts during weeks 8-24 (phase II). Higher initial HIV-1 RNA values corresponded to greater increases in overall and phase I changes in mean CD4 cell counts, particularly among subjects with less advanced disease. In the overall and phase II cases, those subjects with suppressed HIV RNA levels had consistently higher increases in mean CD4 cell counts across both baseline HIV-1 RNA levels and CD4 cell counts than did the respective unsuppressed group. Based on a multivariate model, increases in mean phase I CD4 cell counts corresponded to higher log baseline HIV-1 RNA levels (p =.0001) and log changes in HIV-1 RNA levels at week 4 (p =.03). Patients with earlier stages of disease (p =.0001) and females (p =.01) had higher increases in phase I changes. Phase II CD4 cell counts did not depend solely on baseline HIV-1 RNA levels and CD4 cell counts but on their interaction (p =.0001) as well as on achieving virologic suppression (p =.0009).

Modelling HIV viral rebound using non-linear mixed effects models.

Individuals infected with the human immunodeficiency virus type 1 (HIV-1) who initiate antiretroviral therapy typically experience a marked decline in concentrations of HIV-1 RNA in plasma. Often, however, viral rebound occurs within the first year of treatment and this rebound may be associated with resistance to antiretroviral therapy. For this reason, it is important to study the patterns of virological response of HIV-1 RNA to treatment. In particular, there is interest in the relationship between the lowest level of plasma HIV-1 RNA attained after initiation of therapy (nadir value) and the time until rebound. To investigate this question, we implement a simple and flexible non-linear mixed effects model for the trajectory of the HIV-1 RNA until rebound. This model is also consistent with biological insights into the effects of treatment. We also show how the problem of censoring of HIV-1 RNA values at the lower limit of assay quantification can be addressed using a multiple imputation scheme. The algorithm is simple to implement and is based on accessible software. Our application makes use of data from clinical trial 315 conducted by the AIDS Clinical Trials Group (ACTG 315). We find a strong relationship between HIV-1 RNA nadir and time to rebound, with potentially important consequences for the management of HIV-infected individuals.