Showing papers by "Matthew P. Fox published in 2009"

Applying Quantitative Bias Analysis to Epidemiologic Data

Abstract: This text provides the first-ever compilation of bias analysis methods for use with epidemiologic data. It guides the reader through the planning stages of bias analysis, including the design of validation studies and the collection of validity data from other sources. Three chapters present methods for corrections to address selection bias, uncontrolled confounding, and classification errors. Subsequent chapters extend these methods to multidimensional bias analysis, probabilistic bias analysis, and multiple bias analysis. The text concludes with a chapter on presentation and interpretation of bias analysis results. Although techniques for bias analysis have been available for decades, these methods are considered difficult to implement. This text not only gathers the methods into one cohesive and organized presentation, it also explains the methods in a consistent fashion and provides customizable spreadsheets to implement the solutions. By downloading the spreadsheets (available at links provided in the text), readers can follow the examples in the text and then modify the spreadsheet to complete their own bias analyses. Readers without experience using quantitative bias analysis will be able to design, implement, and understand bias analyses that address the major threats to the validity of epidemiologic research. More experienced analysts will value the compilation of bias analysis methods and links to software tools that facilitate their projects.

Global strategies to reduce the price of antiretroviral medicines: evidence from transactional databases.

Abstract: OBJECTIVE: To estimate the impact of global strategies, such as pooled procurement arrangements, third-party price negotiation and differential pricing, on reducing the price of antiretrovirals (ARVs), which currently hinders universal access to HIV/AIDS treatment. METHODS: We estimated the impact of global strategies to reduce ARV prices using data on 7253 procurement transactions (July 2002-October 2007) from databases hosted by WHO and the Global Fund to Fight AIDS, Tuberculosis and Malaria. FINDINGS: For 19 of 24 ARV dosage forms, we detected no association between price and volume purchased. For the other five ARVs, high-volume purchases were 4-21% less expensive than medium- or low-volume purchases. Nine of 13 generic ARVs were priced 6-36% lower when purchased under the Clinton Foundation HIV/AIDS Initiative (CHAI). Fifteen of 18 branded ARVs were priced 23-498% higher for differentially priced purchases compared with non-CHAI generic purchases. However, two branded, differentially priced ARVs were priced 63% and 73% lower, respectively, than generic non-CHAI equivalents. CONCLUSION: Large purchase volumes did not necessarily result in lower ARV prices. Although current plans for pooled procurement will further increase purchase volumes, savings are uncertain and should be balanced against programmatic costs. Third-party negotiation by CHAI resulted in lower generic ARV prices. Generics were less expensive than differentially priced branded ARVs, except where little generic competition exists. Alternative strategies for reducing ARV prices, such as streamlining financial management systems, improving demand forecasting and removing barriers to generics, should be explored.

Global Strategies to Reduce the Price of Antiretroviral Medicines: Evidence from Transactional databases/Strategies Mondiales Pour Reduire le Prix Des Medicaments Antiretroviraux : Elements Provenant Des Bases De Donnees De transaction/Estrategias Mundiales De Reduccion del Precio De Los Antirretrovirales: Evidencia Extraida De Bases De Datos Transaccionales

Abstract: Introduction New goals on providing universal access to HIV/AIDS services by 2010 were announced in 2007 by WHO, the Joint United Nations Programme on HIV/AIDS (UNAIDS) and the United Nations Children's Fund (UNICEF) (1) The need for life-long HIV/AIDS treatment and the high cost of antiretroviral (ARV) agents present challenges to achieving and sustaining universal access targets During the past decade, various large-scale strategies have been used to reduce ARV prices in low- and middle-income countries This paper focuses on three price-reduction strategies: procurement arrangements designed to increase purchase volumes, third-party price negotiation for generic ARVs and differential pricing for branded ARVs The first strategy, procurement arrangements to increase purchase volumes, often involves pooled procurement schemes that group multiple purchasers into a single purchasing unit in the hope that economies of scale will lead to lower prices A pooled procurement mechanism is currently being developed at the Global Fund to Fight AIDS, Tuberculosis and Malaria (Global Fund) (2,3) The second large-scale strategy involves third-party consultation and price negotiation with generic ARV suppliers, a practice introduced by the Clinton Foundation HIV/AIDS Initiative (CHAI) in 2003 (4) In practice, CHAI attempts to make ARVs more affordable by negotiating price ceilings that reflect suppliers' costs plus reasonable and sustainable profit margins (4) Moreover, CHAI furthers this strategy by providing direct technical assistance to some suppliers to help lower their production costs (4) The resulting ceiling prices are made available to all members of the CHAI procurement consortium (4) Countries that wish to become part of the consortium sign a memorandum of understanding with CHAI and manufacturers are required to offer ARVs to these countries at prices equal to or less than CHAI-negotiated ceiling prices (4) The third strategy involves differential pricing, sometimes referred to as price discrimination or tiered pricing In 2000, the Accelerating Access Initiative, a collaborative endeavour of multiple international agencies and pharmaceutical manufacturers, first launched such a strategy for ARVs (5) Whereas CHAI price negotiation deals exclusively with generic ARVs, differential pricing pertains to branded ARVs and was introduced at a time when generic ARVs were not yet available Under differential-pricing schemes, each manufacturer selects certain branded ARVs to be sold to low- and middle-income countries at prices lower than those charged in high-income countries (5) Each manufacturer determines which countries are eligible to purchase ARVs under their differential-pricing scheme, with eligibility typically being based on the country's income level and prevalence of HIV infection Data on transactions involving the procurement of ARVs with donor funds are made public by the Global Fund and WHO (6,7) The Global Fund and WHO databases can be used to monitor and examine the global ARV marketplace Although some analyses of these databases have been carried out, (8-11) none has examined the global impact of the various ARV price-reduction strategies mentioned above We used the Global Fund and WHO databases to test the following hypotheses on three different ARV price-reduction strategies: prices for high-volume ARV purchases are less than for low-volume purchases; prices for generic ARVs purchased within the CHAI consortium are less than for generic ARVs purchased outside the consortium; and prices for branded ARVs purchased under differential-pricing schemes are equal to or less than those for generic ARVs Methods Data sources We used data on ARV procurement transactions from the Global Fund Price Reporting Mechanism and the WHO Global Price Reporting Mechanism (GPRM) for the period between July 2002 and October 2007 (6,7) The Global Fund posts details of ARV procurements reported by their international aid recipients on the web-based Price Reporting Mechanism …

High frequency of multidrug-resistant gram-negative rods in 2 neonatal intensive care units in the Philippines.

Abstract: Background.Although hospital-acquired infections appear to be a growing threat to the survival of newborns in the developing world, the epidemiology of this problem remains poorly characterized.Methods.During a 10-month period, we conducted prospective longitudinal surveillance for colonization and bloodstream infection caused by gram-negative rods among all infants hospitalized in the 2 largest neonatal intensive care units in Manila, the Philippines. We determined antibiotic susceptibilities and calculated adjusted odds ratios for risk factors for bacteremia by means of multivariate logistic regression.Results.Of 1,831 neonates enrolled during a 10-month period, 1,017 (55.5%) became newly colonized and 358 (19.6%) became bacteremic with a drug-resistant gram-negative rod, most commonly Klebsiella species, Enterobacter species, Acinetobacter species, and Pseudomonas aeruginosa. Of the invasive isolates, 20% were resistant to imipenem, 41% to trimethoprim-sulfamethoxazole, 52% to amikacin, 63% to ampicillin-sulbactam, 67% to ceftazidime, and 80% to tobramycin. The factors significantly associated with an increased risk of bacteremia were mechanical ventilation and prematurity. Additionally, colonization with a drug-resistant gram-negative rod was an independent risk factor for bacteremia (odds ratio, 1.4 [95% confidence interval, 1.0-1.9]).Conclusions.Colonization with a drug-resistant gram-negative rod was an independent risk factor for sepsis. If our data are typical, the unusually high intensity of colonization pressure and disease caused by multidrug-resistant gram-negative rods at these 2 neonatal intensive care units indicates an emerging healthcare crisis in the developing world. Improved infection control methods are therefore critically needed in developing countries.

Do the socioeconomic impacts of antiretroviral therapy vary by gender? A longitudinal study of Kenyan agricultural worker employment outcomes

Abstract: As access to antiretroviral therapy (ART) has grown in Africa, attention has turned to evaluating the socio-economic impacts of ART. One key issue is the extent to which improvements in health resulting from ART allows individuals to return to work and earn income. Improvements in health from ART may also be associated with reduced impaired presenteeism, which is the loss of productivity when an ill or disabled individual attends work but accomplishes less at his or her usual tasks or shifts to other, possibly less valuable, tasks. Longitudinal data for this analysis come from company payroll records for 97 HIV-infected tea estate workers (the index group, 56 women, 41 men) and a comparison group of all workers assigned to the same work teams (n = 2485, 1691 men, 794 women) for a 37-month period covering two years before and one year after initiating ART. We used nearest neighbour matching methods to estimate the impacts of HIV/AIDS and ART on three monthly employment outcomes for tea estate workers in Kenya – days plucking tea, days assigned to non-plucking assignments, and kilograms harvested when plucking. The female index group worked 30% fewer days plucking tea monthly than the matched female comparison group during the final 9 months pre-ART. They also worked 87% more days on non-plucking assignments. While the monthly gap between the two groups narrowed after beginning ART, the female index group worked 30% fewer days plucking tea and about 100% more days on non-plucking tasks than the comparison group after one year on ART. The male index group was able to maintain a similar pattern of work as their comparison group except during the initial five months on therapy. Significant impaired presenteeism continued to exist among the female index group after one year on ART. Future research needs to explore further the socio-economic implications of HIV-infected female workers on ART being less productive than the general female workforce over sustained periods of time.

Role of breastfeeding cessation in mediating the relationship between maternal HIV disease stage and increased child mortality among HIV-exposed uninfected children

Abstract: Background Maternal CD4 count predicts child mortality in HIV-uninfected children born to HIV-infected women Methods To explore the mediating role of breastfeeding cessation in this relationship, we compared marginal structural models of maternal CD4 count on child death with and without adjustment for breastfeeding Results In crude analyses, children of mothers with CD4 500 Earlier breastfeeding cessation was also associated with low CD4 (HR 18; CI 12–27) After adjusting for breastfeeding and low birth weight using a marginal structural model, the low CD4 count-child mortality association through 18 months was reduced 17% The change was overestimated using a traditional Cox proportional hazards model (35% reduction in HR from 34 to 25) Conclusions Our analysis suggests that only a small part of the effect of low vs high CD4 count on child mortality through 18 months is mediated through breastfeeding cessation Our results must be taken into account when deciding whether or not to recommend breastfeeding for infants of HIV-infected mothers

Probabilistic Bias Analysis

Abstract: To this point we have considered situations in which the bias parameters for a bias analysis are known with certainty or modeled as if they are known with certainty (i.e., simple bias analysis, see Chaps. 4– 6). We have also considered models of the impact of combinations of bias parameters on an observed estimate of association (i.e., multidimensional bias analysis, see Chap. 7). Simple bias analysis is an improvement over conventional analyses, which implicitly assume that all the bias parameters are fixed at values that confer no bias. However, the usefulness of simple bias analysis is limited by its assumption that the bias parameters are known without error, a situation that is rarely, if ever, a reality. Multidimensional bias analysis improves on simple bias analysis by examining the impact of more than one set of bias parameters, but even this approach only examines the bias conferred by a limited set of bias parameters. For any analysis, many other possible combinations of bias parameters are plausible, and a multidimensional analysis will not describe the impact of these possibilities. More important, multidimensional analysis gives no sense of which corrected estimate of association is the most likely under the assumed bias model, which can make interpretation of the results challenging.

Periconceptional maternal vitamin supplementation and childhood leukaemia : an uncertainty analysis. Commentary

Abstract: Background: Recent studies in childhood cancer suggest that maternal vitamin supplementation may reduce the risk of leukaemia, neuroblastoma and certain types of childhood brain tumours. For example, a previous study found a significantly reduced risk of acute lymphoblastic leukaemia (ALL) but not acute myeloid leukaemia (AML) in children with Down syndrome whose mothers reported any vitamin supplement use prior to knowledge of pregnancy (ALL OR adjusted for confounders 0.51, 95% confidence limits (CL): 0.30, 0.89; AML OR adjusted for confounders 0.92, 95% CL 0.48, 1.76). Recall of exposures, including maternal vitamin supplement use, however, may be difficult and subject to error. Epidemiologists are encouraged to quantitatively adjust for systematic error in study results, but often do not. Methods: The impact that misclassification of maternal vitamin supplement use may have had on the observed ORs in this study was quantified. Uncertainty analysis was used to calculate ORs adjusted for inaccurate reporting of vitamin supplement use under assumed probability distributions for exposure misclassification parameters. Results: Given our assumptions, adjustment for exposure misclassification yielded ORs that were predominantly more protective for ALL than the crude OR. Conclusions: Uncertainty analysis can give important insights into the magnitude and direction of error in study results due to exposure misclassification.

A Guide to Implementing Quantitative Bias Analysis

Abstract: Estimates of association from nonrandomized epidemiologic studies are susceptible to two types of error: random error and systematic error. Random error, or sampling error, is often called chance, and decreases toward zero as the sample size increases and the data are more efficiently distributed in the categories of the adjustment variables. The amount of random error in an estimate of association is measured by its precision. Systematic error, often called bias, does not decrease toward zero as the sample size increases or with more efficient distributions in the categories of the analytic variables. The amount of systematic error in an estimate of association is measured by its validity.

Unmeasured and Unknown Confounders

Abstract: Confounding occurs when the effect of the exposure of interest mixes with the effects of other variables that are causes of the exposure or that share common causal ancestors with the exposure (Kleinbaum et al.,1982). Understanding and adjusting for confounding in epidemiologic research is central to addressing whether an observed association is indeed causal. It is imperative to control confounding because it can make an association appear greater or smaller than it truly is, and can even reverse the apparent direction of an association. Confounding can also make a null effect (i.e., no causal relation between the exposure and the disease) appear either causal or preventive. For a covariate variable to induce confounding, there must be a relation between both the exposure and the covariate in the source population and between the covariate and the disease (among those unexposed). In addition, the covariate must not be affected by the exposure. In nonrandomized epidemiologic studies, confounding can be controlled by design or in the analysis, but only for known and measured confounders.

Multidimensional Bias Analysis

Abstract: The preceding three chapters have described the techniques for conducting simple bias analysis to assess errors caused by selection bias, residual confounding, or misclassification. However, simple bias analysis implies that the researcher has one and only one estimate to assign to each of the values for the error model’s bias parameters. In many situations, that is not the case. There are many bias parameters for which validation data do not exist, so the values assigned to the bias parameter are educated guesses. In this situation, the analyst is better served by making more than one educated guess for each value and then combining values in different sets. In other situations, multiple different measures of the bias parameter may exist, and there may be no basis for the analyst to select just one as the best estimate of the truth from among those available. For example, when both internal and external validation studies have been conducted, or there were multiple external estimates each in populations slightly different to the one under study, the analyst has no basis to select one value for the bias parameter over another. Frequently, internal estimates are more useful than external estimates because they derive from the same source population as yielded the study’s estimate of association. If there is the possibility of selection bias into the internal validation study, however, then it is possible that the subjects included in the validation study do not provide a good estimate of the bias parameter in the remainder of the study population. In this situation, the analyst may want to use values informed by all of the available validation studies as independent estimates.

Presentation and Inference

Abstract: Throughout the text we have illustrated methods to present the results of bias analysis and explained the inferences that might derive from those results. This chapter will briefly describe the overarching considerations we recommend for presentation and inference, and the reader should refer to specific examples throughout the text for the detailed implementations of these principles.

Data Sources for Bias Analysis

Abstract: All bias analyses modify a conventional estimate of association to account for bias introduced by systematic error. These quantitative modifications revise the conventional estimate of association (e.g., a risk difference or a rate ratio) with equations that adjust it for the estimated impact of the systematic error. These equations have parameters, called bias parameters, that ultimately determine the direction and magnitude of the adjustment. For example: The proportions of all eligible subjects who participate in a study, simultaneously stratified into subgroups of persons with and without the disease outcome and within categories of the exposure variable of interest, are bias parameters. These parameters determine the direction and magnitude of selection bias. The sensitivity and specificity of exposure classification, within subgroups of persons with and without the disease outcome of interest, are bias parameters that affect the direction and magnitude of bias introduced by exposure misclassification. The strength of association between an unmeasured confounder and the exposure of interest and between the unmeasured confounder and the disease outcome of interest are bias parameters that affect the direction and magnitude of bias introduced by an unmeasured confounder.

Multiple Bias Modeling

Abstract: Many nonrandomized epidemiologic studies are susceptible to more than one threat to validity (i.e., multiple biases). Bias analysis applied to these studies requires a strategy to address each important threat. The methods described in earlier chapters can be applied serially or in parallel to quantify bias and uncertainty from these multiple biases.