Objective
To define confounding bias in difference‐in‐difference studies and compare regression‐ and matching‐based estimators designed to correct bias due to observed confounders.
Data sources
We ...simulated data from linear models that incorporated different confounding relationships: time‐invariant covariates with a time‐varying effect on the outcome, time‐varying covariates with a constant effect on the outcome, and time‐varying covariates with a time‐varying effect on the outcome. We considered a simple setting that is common in the applied literature: treatment is introduced at a single time point and there is no unobserved treatment effect heterogeneity.
Study design
We compared the bias and root mean squared error of treatment effect estimates from six model specifications, including simple linear regression models and matching techniques.
Data collection
Simulation code is provided for replication.
Principal findings
Confounders in difference‐in‐differences are covariates that change differently over time in the treated and comparison group or have a time‐varying effect on the outcome. When such a confounding variable is measured, appropriately adjusting for this confounder (ie, including the confounder in a regression model that is consistent with the causal model) can provide unbiased estimates with optimal SE. However, when a time‐varying confounder is affected by treatment, recovering an unbiased causal effect using difference‐in‐differences is difficult.
Conclusions
Confounding in difference‐in‐differences is more complicated than in cross‐sectional settings, from which techniques and intuition to address observed confounding cannot be imported wholesale. Instead, analysts should begin by postulating a causal model that relates covariates, both time‐varying and those with time‐varying effects on the outcome, to treatment. This causal model will then guide the specification of an appropriate analytical model (eg, using regression or matching) that can produce unbiased treatment effect estimates. We emphasize the importance of thoughtful incorporation of covariates to address confounding bias in difference‐in‐difference studies.
Effects of Testosterone Treatment in Older Men Snyder, Peter J; Bhasin, Shalender; Cunningham, Glenn R ...
The New England journal of medicine,
02/2016, Letnik:
374, Številka:
7
Journal Article
Recenzirano
Odprti dostop
In this study, men 65 years of age or older with low serum testosterone and symptoms of hypoandrogenism received testosterone or placebo for a year. Testosterone had a moderate benefit in sexual ...function and some benefit in mood but no benefit in vitality or walking distance.
Testosterone concentrations in men decrease with increasing age.
1
,
2
Many symptoms and conditions similar to those that are caused by low testosterone levels in men with pituitary or testicular disease become more common with increasing age. Such symptoms include decreases in mobility, sexual function, and energy. These parallels suggest that the lower testosterone levels in older men may contribute to these conditions.
Previous trials of testosterone treatment in men 65 years of age or older, however, have yielded equivocal results. Although testosterone treatment consistently increased muscle mass and decreased fat mass,
3
,
4
effects on physical performance,
3
,
5
,
6
sexual function, . . .
We develop a method to estimate subject-level trajectory functions from longitudinal data. The approach can be used for patient phenotyping, feature extraction, or, as in our motivating example, ...outcome identification, which refers to the process of identifying disease status through patient laboratory tests rather than through diagnosis codes or prescription information. We model the joint distribution of a continuous longitudinal outcome and baseline covariates using an enriched Dirichlet process prior. This joint model decomposes into (local) semiparametric linear mixed models for the outcome given the covariates and simple (local) marginals for the covariates. The nonparametric enriched Dirichlet process prior is placed on the regression and spline coefficients, the error variance, and the parameters governing the predictor space. This leads to clustering of patients based on their outcomes and covariates. We predict the outcome at unobserved time points for subjects with data at other time points as well as for new subjects with only baseline covariates. We find improved prediction over mixed models with Dirichlet process priors when there are a large number of covariates. Our method is demonstrated with electronic health records consisting of initiators of second-generation antipsychotic medications, which are known to increase the risk of diabetes. We use our model to predict laboratory values indicative of diabetes for each individual and assess incidence of suspected diabetes from the predicted dataset.
IMPORTANCE: As men age, they experience decreased serum testosterone concentrations, decreased bone mineral density (BMD), and increased risk of fracture. OBJECTIVE: To determine whether testosterone ...treatment of older men with low testosterone increases volumetric BMD (vBMD) and estimated bone strength. DESIGN, SETTING, AND PARTICIPANTS: Placebo-controlled, double-blind trial with treatment allocation by minimization at 9 US academic medical centers of men 65 years or older with 2 testosterone concentrations averaging less than 275 ng/L participating in the Testosterone Trials from December 2011 to June 2014. The analysis was a modified intent-to-treat comparison of treatment groups by multivariable linear regression adjusted for balancing factors as required by minimization. INTERVENTIONS: Testosterone gel, adjusted to maintain the testosterone level within the normal range for young men, or placebo gel for 1 year. MAIN OUTCOMES AND MEASURES: Spine and hip vBMD was determined by quantitative computed tomography at baseline and 12 months. Bone strength was estimated by finite element analysis of quantitative computed tomography data. Areal BMD was assessed by dual energy x-ray absorptiometry at baseline and 12 months. RESULTS: There were 211 participants (mean SD age, 72.3 5.9 years; 86% white; mean SD body mass index, 31.2 3.4). Testosterone treatment was associated with significantly greater increases than placebo in mean spine trabecular vBMD (7.5%; 95% CI, 4.8% to 10.3% vs 0.8%; 95% CI, −1.9% to 3.4%; treatment effect, 6.8%; 95% CI, 4.8%-8.7%; P < .001), spine peripheral vBMD, hip trabecular and peripheral vBMD, and mean estimated strength of spine trabecular bone (10.8%; 95% CI, 7.4% to 14.3% vs 2.4%; 95% CI, −1.0% to 5.7%; treatment effect, 8.5%; 95% CI, 6.0%-10.9%; P < .001), spine peripheral bone, and hip trabecular and peripheral bone. The estimated strength increases were greater in trabecular than peripheral bone and greater in the spine than hip. Testosterone treatment increased spine areal BMD but less than vBMD. CONCLUSIONS AND RELEVANCE: Testosterone treatment for 1 year of older men with low testosterone significantly increased vBMD and estimated bone strength, more in trabecular than peripheral bone and more in the spine than hip. A larger, longer trial could determine whether this treatment also reduces fracture risk. TRIAL REGISTRATION: clinicaltrials.gov Identifier: NCT00799617
IMPORTANCE: In one-third of older men with anemia, no recognized cause can be found. OBJECTIVE: To determine if testosterone treatment of men 65 years or older with unequivocally low testosterone ...levels and unexplained anemia would increase their hemoglobin concentration. DESIGN, SETTING, AND PARTICIPANTS: A double-blinded, placebo-controlled trial with treatment allocation by minimization using 788 men 65 years or older who have average testosterone levels of less than 275 ng/dL. Of 788 participants, 126 were anemic (hemoglobin ≤12.7 g/dL), 62 of whom had no known cause. The trial was conducted in 12 academic medical centers in the United States from June 2010 to June 2014. INTERVENTIONS: Testosterone gel, the dose adjusted to maintain the testosterone levels normal for young men, or placebo gel for 12 months. MAIN OUTCOMES AND MEASURES: The percent of men with unexplained anemia whose hemoglobin levels increased by 1.0 g/dL or more in response to testosterone compared with placebo. The statistical analysis was intent-to-treat by a logistic mixed effects model adjusted for balancing factors. RESULTS: The men had a mean age of 74.8 years and body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared) of 30.7; 84.9% were white. Testosterone treatment resulted in a greater percentage of men with unexplained anemia whose month 12 hemoglobin levels had increased by 1.0 g/dL or more over baseline (54%) than did placebo (15%) (adjusted OR, 31.5; 95% CI, 3.7-277.8; P = .002) and a greater percentage of men who at month 12 were no longer anemic (58.3%) compared with placebo (22.2%) (adjusted OR, 17.0; 95% CI, 2.8-104.0; P = .002). Testosterone treatment also resulted in a greater percentage of men with anemia of known cause whose month 12 hemoglobin levels had increased by 1.0 g/dL or more (52%) than did placebo (19%) (adjusted OR, 8.2; 95% CI, 2.1-31.9; P = .003). Testosterone treatment resulted in a hemoglobin concentration of more than 17.5 g/dL in 6 men who had not been anemic at baseline. CONCLUSIONS AND RELEVANCE: Among older men with low testosterone levels, testosterone treatment significantly increased the hemoglobin levels of those with unexplained anemia as well as those with anemia from known causes. These increases may be of clinical value, as suggested by the magnitude of the changes and the correction of anemia in most men, but the overall health benefits remain to be established. Measurement of testosterone levels might be considered in men 65 years or older who have unexplained anemia and symptoms of low testosterone levels. TRIAL REGISTRATION: clinicaltrials.gov Identifier: NCT00799617
BACKGROUND: Follicular lymphoma (FL) is an indolent, incurable lymphoma with a heterogeneous clinical course characterized by multiple relapses. Recent data indicate a potential synergism between ...tazemetostat and lenalidomide and an additive effect with rituximab, offering a clinical rationale for this as a combination strategy (Batlevi, Blood 2022). SYMPHONY-1 Phase 1b, Stage 1 (EZH-302; NCT04224493) is an open-label, single-arm, safety run-in trial to evaluate the safety and tolerability of tazemetostat combined with lenalidomide plus rituximab (R2). This study indirectly compares the efficacy results of the SYMPHONY-1 Phase 1b single arm trial with an external control arm (ECA) derived from the Flatiron Health Research Database (FHRD), consisting of real-world FL patients treated with R2 after receiving ≥1 prior systemic therapy. METHODS: This study utilized individual patient data from SYMPHONY-1 based on a data cutoff date of July 10, 2023, and the FHRD, a large database of electronic health records from oncology practices in the United States, including data through December 31, 2022. The treatment arm consisted of patients enrolled in the safety run-in portion of SYMPHONY-1 treated with a combination of tazemetostat plus R2, followed from enrollment to the data cutoff date. The ECA was constructed by applying eligibility criteria adapted from SYMPHONY-1 to real-world patients from the FHRD treated with R2 in second line (2L); patients were followed from the start date of the earliest eligible line of therapy (LOT) to the end of the patient's record. The primary endpoint was objective response rate (ORR), defined as the proportion of patients who achieved a complete or partial response. In SYMPHONY-1, ORR was assessed by an independent review committee according to the 2014 Lugano Classification. In the ECA, real-world ORR (rwORR) was determined by the treating clinician's interpretation of change in disease burden. The analysis set included patients from the treatment arm with ≥1 evaluable response assessment and patients from the ECA with ≥1 radiographic assessment documented in the FHRD. The statistical analysis used propensity score weighting (PSW) to adjust the distribution of seven prespecified baseline patient characteristics of the ECA to align with the treatment arm. These characteristics were age, Eastern Cooperative Oncology Group Performance Status, disease grade at diagnosis, progression of disease within 24 months (POD24) from diagnosis, lactate dehydrogenase (LDH) > upper limit of normal (ULN), size of the largest node, and Follicular Lymphoma International Prognostic Index (FLIPI) risk group. An impute-then-exclude multiple imputation strategy was used to handle missing data for variables required in the eligibility criteria and PSW. The primary endpoint was indirectly compared by estimating the difference between the ORR of the treatment arm and rwORR of the balanced ECA on each imputed dataset, and the associated standard errors were approximated using bootstrapping. Rubin's rules were applied to combine the results into a single estimate with a 95% confidence interval and corresponding two-sided p-value. RESULTS: The primary analysis compared 42 patients in the treatment arm from SYMPHONY-1 and a median of 36 patients (range: 34 to 38) in the ECA among 815 total patients in the FHRD across ten imputed datasets. The unweighted analysis found an ORR of 95.2% in the treatment arm and 75.7% in the ECA, resulting in a difference of 19.6% (95% confidence interval CI: 3.6-35.5) (Table 2). After PSW, the distribution of baseline characteristics for the ECA was comparable with the treatment arm (all standardized mean differences < 0.16), and the median effective sample size was reduced to 12.7 (range: 4.9 to 19.1) (Table 1). ORR was 95.2% in the treatment arm and 77.9% in the ECA, resulting in a difference of 17.3% (95% CI: -17.5-52.1; two-sided P = 0.16) (Table 2). CONCLUSIONS: This ECA study found that patients treated with tazemetostat combined with R2 were estimated to have a higher ORR than those treated with R2, though inference is limited due to small sample sizes. These findings help contextualize the SYMPHONY-1 Phase 1b single-arm trial results and combination potential of tazemetostat plus R2. There is a randomized Phase 3 confirmatory trial underway to confirm the clinical benefit of tazemetostat plus R2 (SYMPHONY-1 Stage 2, NCT04224493).
We propose a general Bayesian nonparametric (BNP) approach to causal inference in the point treatment setting. The joint distribution of the observed data (outcome, treatment, and confounders) is ...modeled using an enriched Dirichlet process. The combination of the observed data model and causal assumptions allows us to identify any type of causal effect—differences, ratios, or quantile effects, either marginally or for subpopulations of interest. The proposed BNP model is wellsuited for causal inference problems, as it does not require parametric assumptions about the distribution of confounders and naturally leads to a computationally efficient Gibbs sampling algorithm. By flexibly modeling the joint distribution, we are also able to impute (via data augmentation) values for missing covariates within the algorithm under an assumption of ignorable missingness, obviating the need to create separate imputed data sets. This approach for imputing the missing covariates has the additional advantage of guaranteeing congeniality between the imputation model and the analysis model, and because we use a BNP approach, parametric models are avoided for imputation. The performance of the method is assessed using simulation studies. The method is applied to data from a cohort study of human immunodeficiency virus/hepatitis C virus co-infected patients.
Context:
The Testosterone Trials are a coordinated set of seven trials to determine the efficacy of T in symptomatic men ≥65 years old with unequivocally low T levels. Initial results of the Sexual ...Function Trial showed that T improved sexual activity, sexual desire, and erectile function.
Objective:
To assess the responsiveness of specific sexual activities to T treatment; to relate hormone changes to changes in sexual function; and to determine predictive baseline characteristics and T threshold for sexual outcomes.
Design:
A placebo-controlled trial.
Setting:
Twelve academic medical centers in the United States.
Participants:
A total of 470 men ≥65 years of age with low libido, average T <275 ng/dL, and a partner willing to have sexual intercourse at least twice a month.
Methods:
Men were assigned to take T gel or placebo for 1 year. Sexual function was assessed by three questionnaires every 3 months: the Psychosexual Daily Questionnaire, the Derogatis Interview for Sexual Function, and the International Index of Erectile Function.
Results:
Compared with placebo, T administration significantly improved 10 of 12 measures of sexual activity. Incremental increases in total and free T and estradiol levels were associated with improvements in sexual activity and desire, but not erectile function. No threshold T level was observed for any outcome, and none of the 27 baseline characteristics predicted responsiveness to T.
Conclusions:
In older men with low libido and low T levels, improvements in sexual desire and activity in response to T treatment were related to the magnitude of increases in T and estradiol levels, but there was no clear evidence of a threshold effect.
Testosterone Treatment and Sexual Function in Older Men with Low Testosterone Levels. Improvements in sexual desire and activity were related to the magnitude of increases in testosterone.
Bayesian Additive Regression Trees (BART) is a flexible machine learning algorithm capable of capturing nonlinearities between an outcome and covariates and interactions among covariates. We extend ...BART to a semiparametric regression framework in which the conditional expectation of an outcome is a function of treatment, its effect modifiers, and confounders. The confounders are allowed to have unspecified functional form, while treatment and effect modifiers that are directly related to the research question are given a linear form. The result is a Bayesian semiparametric linear regression model where the posterior distribution of the parameters of the linear part can be interpreted as in parametric Bayesian regression. This is useful in situations where a subset of the variables are of substantive interest and the others are nuisance variables that we would like to control for. An example of this occurs in causal modeling with the structural mean model (SMM). Under certain causal assumptions, our method can be used as a Bayesian SMM. Our methods are demonstrated with simulation studies and an application to dataset involving adults with HIV/Hepatitis C coinfection who newly initiate antiretroviral therapy. The methods are available in an R package called semibart.
Pneumococcal nasopharyngeal colonization is a pre-requisite for pneumococcal disease; the risk for pneumococcal disease is high in children born to women living with human immunodeficiency virus ...(HIV). We investigated pneumococcal colonization, serotype distribution and antibiotic susceptibility of Streptococcus pneumoniae isolates carried by perinatal HIV-infected and HIV-exposed-uninfected (HEU) children.Serial nasopharyngeal swabs were collected from 331 HIV-infected and 491 HEU children, at up to 6 scheduled timepoints, between median ages of 25 to 181 weeks. Pneumococcus was identified by culture; serotyping and antibiotic susceptibility testing were done by conventional methods. No pneumococcal vaccine was given.HIV-infected children were less likely to be colonized with 7-valent pneumococcal conjugate vaccine 7 serotypes than HEU at a median of 25 weeks of age (23% vs 36%; P < .001); however, no differences in colonization between the 2 groups were observed at subsequent study-visits. Over the 36-months study-period pneumococcal colonization increased in both HIV-infected (from 45% to 77%) and HEU (from 57% to 61%) children. Over the study-period, pneumococcal isolates non-susceptible to cotrimoxazole decreased from 92% to 57% and had a similar trend to penicillin (from 65% to 42%) in HIV-infected children. Similarly, pneumococcal nonsusceptible to cotrimoxazole decreased from 93% to 57% and to penicillin from 69% to 37% in HEU children.Vaccine serotype colonization was common in this population and similar rates were observed in HIV-infected and HEU children. The prevalence of pneumococcal isolates non-susceptible to cotrimoxazole and penicillin decreased with age.
PDF
