Propensity Score Matching in Randomized Clinical Trials

Summary Cluster randomization trials with relatively few clusters have been widely used in recent years for evaluation of health‐care strategies. On average, randomized treatment assignment achieves balance in both known and unknown confounding factors between treatment groups, however, in practice investigators can only introduce a small amount of stratification and cannot balance on all the important variables simultaneously. The limitation arises especially when there are many confounding variables in small studies. Such is the case in the INSTINCT trial designed to investigate the effectiveness of an education program in enhancing the tPA use in stroke patients. In this article, we introduce a new randomization design, the balance match weighted (BMW) design, which applies the optimal matching with constraints technique to a prospective randomized design and aims to minimize the mean squared error (MSE) of the treatment effect estimator. A simulation study shows that, under various confounding scenarios, the BMW design can yield substantial reductions in the MSE for the treatment effect estimator compared to a completely randomized or matched‐pair design. The BMW design is also compared with a model‐based approach adjusting for the estimated propensity score and Robins‐Mark‐Newey E‐estimation procedure in terms of efficiency and robustness of the treatment effect estimator. These investigations suggest that the BMW design is more robust and usually, although not always, more efficient than either of the approaches. The design is also seen to be robust against heterogeneous error. We illustrate these methods in proposing a design for the INSTINCT trial.     

Performance of the marginal structural models under various scenarios of incomplete marker's values: A simulation study

Marginal structural models (MSMs) have been proposed for estimating a treatment's effect, in the presence of time‐dependent confounding. We aimed to evaluate the performance of the Cox MSM in the presence of missing data and to explore methods to adjust for missingness. We simulated data with a continuous time‐dependent confounder and a binary treatment. We explored two classes of missing data: (i) missed visits, which resemble clinical cohort studies; (ii) missing confounder's values, which correspond to interval cohort studies. Missing data were generated under various mechanisms. In the first class, the source of the bias was the extreme treatment weights. Truncation or normalization improved estimation. Therefore, particular attention must be paid to the distribution of weights, and truncation or normalization should be applied if extreme weights are noticed. In the second case, bias was due to the misspecification of the treatment model. Last observation carried forward (LOCF), multiple imputation (MI), and inverse probability of missingness weighting (IPMW) were used to correct for the missingness. We found that alternatives, especially the IPMW method, perform better than the classic LOCF method. Nevertheless, in situations with high marker's variance and rarely recorded measurements none of the examined method adequately corrected the bias.     

Estimating Antihypertensive Effects of Combination Therapy in an Observational Study Using Marginal Structural Models

The evaluation of the antihypertensive effect of multiple antihypertensive drugs using data from an observational study requires adjustment for time‐dependent confounders. Marginal structural models (MSMs) have been proposed to address this type of confounding through inverse probability weighting. Generally, the probabilities are estimated using logistic regression models that assume linearity between the logistic link and the predictors, but the linearity might be inaccurate. In this article, we proposed MSMs to assess the blood pressure‐lowering effects of combination therapy with olmesartan medoxomil (OLM) plus calcium channel blockers (CCB) (OLM+CCB) in an observational study of OLM, and extended estimation methods of the probabilities for the MSMs using generalized additive models (GAMs). The estimation using GAMs was suggested to improve the balance of the distributions of confounder values between the therapy groups in the pseudo‐population. We obtained estimated changes in systolic blood pressure (SBP) and diastolic blood pressure (DBP) for OLM+CCB combination therapy after 12 wk compared with OLM monotherapy of ?4.3 mmHg (95% confidence interval (CI): ?7.7 and ?0.9 mmHg) and ?2.9 mmHg (95% CI: ?5.1 and ?0.7 mmHg), respectively. The estimated target BP (SBP<140 mmHg and DBP<90 mmHg) achievement rates for OLM+CCB combination therapy and OLM monotherapy were 62.0 and 46.7%, respectively. The results of the MSMs were closer to those in the randomized controlled trial, such as the combination of OLM and amlodipine besylate in controlling high blood pressure study, than those of conventional methods. The proposed MSMs provided useful information to evaluate the effects of combination therapy of antihypertensive drugs in the context of an observational study.     

Marginal structural models for estimating principal stratum direct effects under the monotonicity assumption

In epidemiological and clinical research, investigators are frequently interested in estimating the direct effect of a treatment on an outcome that is not relayed by intermediate variables. In 2009, VanderWeele presented marginal structural models (MSMs) for estimating direct effects based on interventions on the mediator. This paper focuses on direct effects based on principal stratification, i.e. principal stratum direct effects (PSDEs), which are causal effects within latent subgroups of subjects where the mediator is constant, regardless of the exposure status. We propose MSMs for estimating PSDEs. We demonstrate that the PSDE can be estimated readily using MSMs under the monotonicity assumption.     

The effects of small sample size and sample bias on threshold selection and accuracy assessment of species distribution models

Species distribution models are used for a range of ecological and evolutionary questions, but often are constructed from few and/or biased species occurrence records. Recent work has shown that the presence‐only model Maxent performs well with small sample sizes. While the apparent accuracy of such models with small samples has been studied, less emphasis has been placed on the effect of small or biased species records on the secondary modeling steps, specifically accuracy assessment and threshold selection, particularly with profile (presence‐only) modeling techniques. When testing the effects of small sample sizes on distribution models, accuracy assessment has generally been conducted with complete species occurrence data, rather than similarly limited (e.g. few or biased) test data. Likewise, selection of a probability threshold – a selection of probability that classifies a model into discrete areas of presences and absences – has also generally been conducted with complete data. In this study we subsampled distribution data for an endangered rodent across multiple years to assess the effects of different sample sizes and types of bias on threshold selection, and examine the differences between apparent and actual accuracy of the models. Although some previously recommended threshold selection techniques showed little difference in threshold selection, the most commonly used methods performed poorly. Apparent model accuracy calculated from limited data was much higher than true model accuracy, but the true model accuracy was lower than it could have been with a more optimal threshold. That is, models with thresholds and accuracy calculated from biased and limited data had inflated reported accuracy, but were less accurate than they could have been if better data on species distribution were available and an optimal threshold were used.     

Upper limb musculoskeletal stress markers among middle Holocene foragers of Siberia's Cis‐Baikal region

This evaluation of musculoskeletal stress markers (MSMs) in the Cis‐Baikal focuses on upper limb activity reconstruction among the region's middle Holocene foragers, particularly as it pertains to adaptation and cultural change. The five cemetery populations investigated represent two discrete groups separated by an 800–1,000 year hiatus: the Early Neolithic (8000–7000/6800 cal. BP) Kitoi culture and the Late Neolithic/Bronze Age (6000/5800–4000 cal. BP) Isakovo‐Serovo‐Glaskovo (ISG) cultural complex. Twenty‐four upper limb MSMs are investigated not only to gain a better understanding of activity throughout the middle Holocene, but also to independently assess the relative distinctiveness of Kitoi and ISG adaptive regimes. Results reveal higher heterogeneity in overall activity levels among Early Neolithic populations—with Kitoi males exhibiting more pronounced upper limb MSMs than both contemporary females and ISG males—but relative constancy during the Late Neolithic/Bronze Age, regardless of sex or possible status. On the other hand, activity patterns seem to have varied more during the latter period, with the supinator being ranked high among the ISG, but not the Kitoi, and forearm flexors and extensors being ranked generally low only among ISG females. Upper limb rank patterning does not distinguish Early Neolithic males, suggesting that their higher MSM scores reflect differences in the degree (intensity and/or duration), rather than the type, of activity employed. Finally, for both Kitoi and ISG peoples, activity patterns—especially the consistently high‐ranked costoclavicular ligament and deltoid and pectoralis major muscles—appear to be consistent with watercraft use. Am J Phys Anthropol, 2009. © 2008 Wiley‐Liss, Inc.     

DOES NICHE DIVERGENCE ACCOMPANY ALLOPATRIC DIVERGENCE IN APHELOCOMA JAYS AS PREDICTED UNDER ECOLOGICAL SPECIATION?: INSIGHTS FROM TESTS WITH NICHE MODELS

The role of ecology in the origin of species has been the subject of long‐standing interest to evolutionary biologists. New sources of spatially explicit ecological data allow for large‐scale tests of whether speciation is associated with niche divergence or whether closely related species tend to be similar ecologically (niche conservatism). Because of the confounding effects of spatial autocorrelation of environmental variables, we generate null expectations for niche divergence for both an ecological‐niche modeling and a multivariate approach to address the question: do allopatrically distributed taxa occupy similar niches? In a classic system for the study of niche evolution—the Aphelocoma jays—we show that there is little evidence for niche divergence among Mexican Jay (A. ultramarina) lineages in the process of speciation, contrary to previous results. In contrast, Aphelocoma species that exist in partial sympatry in some regions show evidence for niche divergence. Our approach is widely applicable to the many cases of allopatric lineages in the beginning stages of speciation. These results do not support an ecological speciation model for Mexican Jay lineages because, in most cases, the allopatric environments they occupy are not significantly more divergent than expected under a null model.     

Utilizing systems biology to unravel stomatal function and the hierarchies underpinning its control

Stomata control the concomitant exchange of CO₂ and transpiration in land plants. While a constant supply of CO₂ is need to maintain the rate of photosynthesis, the accompanying water losses must be tightly regulated to prevent dehydration and undesired metabolic changes. The factors affecting stomatal movement are directly coupled with the cellular networks of guard cells. Although the guard cell has been used as a model for characterization of signaling pathways, several important questions about its functioning remain elusive. Current modeling approaches describe the stomatal conductance in terms of relatively few easy‐to‐measure variables being unsuitable for in silico design of genetic manipulation strategies. Here, we argue that a system biology approach, combining modeling and high‐throughput experiments, may be used to elucidate the mechanisms underlying stomata control and to determine targets for modulation of stomatal responses to environment. In support of our opinion, we review studies demonstrating how high‐throughput approaches have provided a systems‐view of guard cells. Finally, we emphasize the opportunities and challenges of genome‐scale modeling and large‐scale data integration for in silico manipulation of guard cell functions to improve crop yields, particularly under stress conditions which are of pertinence both to climate change and water use efficiency.     

Comparing mortality in renal patients on hemodialysis versus peritoneal dialysis using a marginal structural model

When comparing the causal effect of peritoneal dialysis (PD) and hemodialysis (HD) treatment on lowering mortality in renal patients, using observational data, it is necessary to adjust for different forms of confounding and informative censoring. Both the type of dialysis treatment that is started with and mortality are affected by baseline covariates. Longitudinal and baseline variables can affect both the probability of switching from one type of dialysis to the other, and mortality. Longitudinal and baseline variables can also affect the probability of receiving a kidney transplant, possibly causing informative censoring. Adjusting for longitudinal variables by including them as covariates in a regression model potentially causes bias, for instance by losing a possible indirect effect of dialysis on mortality via these longitudinal variables. Instead, we fitted a marginal structural model (MSM) to estimate the causal effect of dialysis type, adjusted for confounding and informative censoring. We used the MSM to compare the hazard of death as well as cumulative survival between the potential treatment trajectories "always PD" and "always HD" over time, conditional on age and diabetes mellitus status. We used inverse probability weighting (IPW) to fit the MSM.     

Efficient search,mapping, and optimization of multi‐protein genetic systems in diverse bacteria

Developing predictive models of multi‐protein genetic systems to understand and optimize their behavior remains a combinatorial challenge, particularly when measurement throughput is limited. We developed a computational approach to build predictive models and identify optimal sequences and expression levels, while circumventing combinatorial explosion. Maximally informative genetic system variants were first designed by the RBS Library Calculator, an algorithm to design sequences for efficiently searching a multi‐protein expression space across a > 10,000‐fold range with tailored search parameters and well‐predicted translation rates. We validated the algorithm's predictions by characterizing 646 genetic system variants, encoded in plasmids and genomes, expressed in six gram‐positive and gram‐negative bacterial hosts. We then combined the search algorithm with system‐level kinetic modeling, requiring the construction and characterization of 73 variants to build a sequence‐expression‐activity map (SEAMAP) for a biosynthesis pathway. Using model predictions, we designed and characterized 47 additional pathway variants to navigate its activity space, find optimal expression regions with desired activity response curves, and relieve rate‐limiting steps in metabolism. Creating sequence‐expression‐activity maps accelerates the optimization of many protein systems and allows previous measurements to quantitatively inform future designs.     

A joint model for longitudinal continuous and time‐to‐event outcomes with direct marginal interpretation

Joint modeling of various longitudinal sequences has received quite a bit of attention in recent times. This paper proposes a so‐called marginalized joint model for longitudinal continuous and repeated time‐to‐event outcomes on the one hand and a marginalized joint model for bivariate repeated time‐to‐event outcomes on the other. The model has several appealing features. It flexibly allows for association among measurements of the same outcome at different occasions as well as among measurements on different outcomes recorded at the same time. The model also accommodates overdispersion. The time‐to‐event outcomes are allowed to be censored. While the model builds upon the generalized linear mixed model framework, it is such that model parameters enjoy a direct marginal interpretation. All of these features have been considered before, but here we bring them together in a unified, flexible framework. The model framework's properties are scrutinized using a simulation study. The models are applied to data from a chronic heart failure study and to a so‐called comet assay, encountered in preclinical research. Almost surprisingly, the models can be fitted relatively easily using standard statistical software.     

Impact of landscape predictors on climate change modelling of species distributions: a case study with Eucalyptus fastigata in southern New South Wales,Australia

Aim We consider three questions. (1) How different are the predicted distribution maps when climate‐only and climate‐plus‐terrain models are developed from high‐resolution data? (2) What are the implications of differences between the models when predicting future distributions under climate change scenarios, particularly for climate‐only models at coarse resolution? (3) Does the use of high‐resolution data and climate‐plus‐terrain models predict an increase in the number of local refugia? Location South‐eastern New South Wales, Australia. Methods We developed two species distribution models for Eucalyptus fastigata under current climate conditions using generalized additive modelling. One used only climate variables as predictors (mean annual temperature, mean annual rainfall, mean summer rainfall); the other used both climate and landscape (June daily radiation, topographic position, lithology, nutrients) variables as predictors. Predictions of the distribution under current climate and climate change were then made for both models at a pixel resolution of 100 m. Results The model using climate and landscape variables as predictors explained a significantly greater proportion of the deviance than the climate‐only model. Inclusion of landscape variables resulted in the prediction of much larger areas of existing optimal habitat. An overlay of predicted future climate on the current climate space indicated that extrapolation of the statistical models was not occurring and models were therefore more robust. Under climate change, landscape‐defined refugia persisted in areas where the climate‐only model predicted major declines. In areas where expansion was predicted, the increase in optimal habitat was always greater with landscape predictors. Recognition of extensive optimal habitat conditions and potential refugia was dependent on the use of high‐resolution landscape data. Main conclusions Using only climate variables as predictors for assessing species responses to climate change ignores the accepted conceptual model of plant species distribution. Explicit statements justifying the selection of predictors based on ecological principles are needed. Models using only climate variables overestimate range reduction under climate change and fail to predict potential refugia. Fine‐scale‐resolution data are required to capture important climate/landscape interactions. Extrapolation of statistical models to regions in climate space outside the region where they were fitted is risky.     

Inverse‐probability‐of‐treatment weighted estimation of causal parameters in the presence of error‐contaminated and time‐dependent confounders

Inverse‐probability‐of‐treatment weighted (IPTW) estimation has been widely used to consistently estimate the causal parameters in marginal structural models, with time‐dependent confounding effects adjusted for. Just like other causal inference methods, the validity of IPTW estimation typically requires the crucial condition that all variables are precisely measured. However, this condition, is often violated in practice due to various reasons. It has been well documented that ignoring measurement error often leads to biased inference results. In this paper, we consider the IPTW estimation of the causal parameters in marginal structural models in the presence of error‐contaminated and time‐dependent confounders. We explore several methods to correct for the effects of measurement error on the estimation of causal parameters. Numerical studies are reported to assess the finite sample performance of the proposed methods.     

Standardizing fishery-dependent catch and effort data in complex fisheries with technology change

Standardization of commercial catch and effort data is important in fisheries where standardized abundance indices based on fishery-dependent data are a fundamental input to stock assessments. The goal of the standardization is then to minimize bias due to the confounding of apparent abundance patterns with fishing power. There is a high risk of confounding between fishing power and abundance in fisheries where the fleet has altered their fishing technology over the years. Also, the spatial aspects and the fishing history can be so heterogeneous that any standardization really involves an extrapolation, for example to a hypothetical standard vessel. When the standardization involves an extrapolation, then the appropriate modeling strategy is to build a so-called estimation model, rather than a predictive model. Strategies to build such an estimation model from fishery-dependent data include: pay careful attention to subject matter, and collect information about potential confounding effects to include in the model (putting a high value on the acquisition of data on covariates); model variable catchability at a highly disaggregated scale; aim for realistic coefficients when fitting the model and pay relatively less attention to achieving precision or maximizing explained variance; adopt modern statistical methods to combine data from different sources; and if data are deficient, then apply precautionary allowances. These strategies offer some protection against bias due to confounding, in the absence of formal criteria for identifying the best model.     

Hydrological and environmental variables outperform spatial factors in structuring species,trait composition,and beta diversity of pelagic algae

There has been increasing interest in algae‐based bioassessment, particularly, trait‐based approaches are increasingly suggested. However, the main drivers, especially the contribution of hydrological variables, of species composition, trait composition, and beta diversity of algae communities are less studied. To link species and trait composition to multiple factors (i.e., hydrological variables, local environmental variables, and spatial factors) that potentially control species occurrence/abundance and to determine their relative roles in shaping species composition, trait composition, and beta diversities of pelagic algae communities, samples were collected from a German lowland catchment, where a well‐proven ecohydrological modeling enabled to predict long‐term discharges at each sampling site. Both trait and species composition showed significant correlations with hydrological, environmental, and spatial variables, and variation partitioning revealed that the hydrological and local environmental variables outperformed spatial variables. A higher variation of trait composition (57.0%) than species composition (37.5%) could be explained by abiotic factors. Mantel tests showed that both species and trait‐based beta diversities were mostly related to hydrological and environmental heterogeneity with hydrological contributing more than environmental variables, while purely spatial impact was less important. Our findings revealed the relative importance of hydrological variables in shaping pelagic algae community and their spatial patterns of beta diversities, emphasizing the need to include hydrological variables in long‐term biomonitoring campaigns and biodiversity conservation or restoration. A key implication for biodiversity conservation was that maintaining the instream flow regime and keeping various habitats among rivers are of vital importance. However, further investigations at multispatial and temporal scales are greatly needed.     

Afforestation driving long‐term surface water browning

Increase in surface water color (browning), caused by rising dissolved organic carbon (DOC) and iron concentrations, has been widely reported and studied in the last couple of decades. This phenomenon has implications to aquatic ecosystem function and biogeochemical carbon cycling. While recovery from acidification and changes in climate‐related variables, such as precipitation and length of growing season, are recognized as drivers behind browning, land‐use change has received less attention. In this study, we include all of the above factors and aim to discern their individual and combined contribution to water color variation in an unprecedentedly long (1940–2016) and highly resolved dataset (~20 times per month), from a river in southern Sweden. Water color showed high seasonal variability and a marked long‐term increase, particularly in the latter half of the dataset (~1980). Short‐term and seasonal variations were best explained by precipitation, with temperature playing a secondary role. All explanatory variables (precipitation, temperature, S deposition, and land‐use change) contributed significantly and together predicted 75% of the long‐term variation in water color. Long‐term change was best explained by a pronounced increase in Norway spruce (Picea abies Karst) volume—a measure of land‐use change and a proxy for buildup of organic soil layers—and by change in atmospheric S deposition. When modeling water color with a combination of explanatory variables, Norway spruce showed the highest contribution to explaining long‐term variability. This study highlights the importance of considering land‐use change as a factor behind browning and combining multiple factors when making predictions in water color and DOC.     

Estimating causal treatment effects from longitudinal HIV natural history studies using marginal structural models

Several recently completed and ongoing studies of the natural history of HIV infection have generated a wealth of information about its clinical progression and how this progression is altered by therepeutic interventions and environmental factors. Natural history studies typically follow prospective cohort designs, and enroll large numbers of participants for long-term prospective follow-up (up to several years). Using data from the HIV Epidemiology Research Study (HERS), a six-year natural history study that enrolled 871 HIV-infected women starting in 1993, we investigate the therapeutic effect of highly active antiretroviral therapy regimens (HAART) on CD4 cell count using the marginal structural modeling framework and associated estimation procedures based on inverse-probability weighting (developed by Robins and colleagues). To evaluate treatment effects from a natural history study, specialized methods are needed because treatments are not randomly prescribed and, in particular, the treatment-response relationship can be confounded by variables that are time-varying. Our analysis uses CD4 data on all follow-up visits over a two-year period, and includes sensitivity analyses to investigate potential biases attributable to unmeasured confounding. Strategies for selecting ranges of a sensitivity parameter are given, as are intervals for treatment effect that reflect uncertainty attributable both to sampling and to lack of knowledge about the nature and existence of unmeasured confounding. To our knowledge, this is the first use in "real data" of Robins's sensitivity analysis for unmeasured confounding (Robins, 1999a, Synthese 121, 151-179). The findings from our analysis are consistent with recent treatment guidelines set by the U.S. Panel of the International AIDS Society (Carpenter et al., 2000, Journal of the American Medical Association 280, 381-391).     

The Semiparametric Case‐Only Estimator

Summary We propose a semiparametric case‐only estimator of multiplicative gene–environment or gene–gene interactions, under the assumption of conditional independence of the two factors given a vector of potential confounding variables. Our estimator yields valid inferences on the interaction function if either but not necessarily both of two unknown baseline functions of the confounders is correctly modeled. Furthermore, when both models are correct, our estimator has the smallest possible asymptotic variance for estimating the interaction parameter in a semiparametric model that assumes that at least one but not necessarily both baseline models are correct.     

Thermodynamic Constraints in Kinetic Modeling: Thermodynamic‐Kinetic Modeling in Comparison to Other Approaches

Kinetic models of reaction networks may easily violate the laws of thermodynamics and the principle of detailed balance. In large network models, the constraints that are imposed by these laws are particularly difficult to address. This hinders modeling of biochemical reaction networks. Thermodynamic‐kinetic modeling is a method that provides a thermodynamically sound and formally appealing way for deriving dynamic model equations of reaction systems. State variables of this approach are thermokinetic potentials that describe the ability of compounds to drive a reaction. A compound has a parameter called capacity, which is the ratio of its concentration and thermokinetic potential. A reaction is described by its resistance which is the ratio of the thermokinetic driving force and flux. In these aspects, the formalism is similar to the modeling formalism for electrical networks and an analogous graphical representation is possible. The thermodynamic‐kinetic modeling formalism is equivalent to the traditional kinetic modeling formalism with the exception that it is not possible to build thermodynamically infeasible models. Here, the thermodynamic‐kinetic modeling formalism is reviewed, compared to other approaches, and some of its advantages are worked out. In contrast to other approaches, thermodynamic‐kinetic modeling does not rely on an explicit enumeration of stoichiometric cycles. It is capable of describing rate laws far from equilibrium. Further, the parameterization by capacities and resistances is particularly intuitive and powerful.     

Weighted regression analysis to correct for informative monitoring times and confounders in longitudinal studies

We address estimation of the marginal effect of a time‐varying binary treatment on a continuous longitudinal outcome in the context of observational studies using electronic health records, when the relationship of interest is confounded, mediated, and further distorted by an informative visit process. We allow the longitudinal outcome to be recorded only sporadically and assume that its monitoring timing is informed by patients' characteristics. We propose two novel estimators based on linear models for the mean outcome that incorporate an adjustment for confounding and informative monitoring process through generalized inverse probability of treatment weights and a proportional intensity model, respectively. We allow for a flexible modeling of the intercept function as a function of time. Our estimators have closed‐form solutions, and their asymptotic distributions can be derived. Extensive simulation studies show that both estimators outperform standard methods such as the ordinary least squares estimator or estimators that only account for informative monitoring or confounders. We illustrate our methods using data from the Add Health study, assessing the effect of depressive mood on weight in adolescents.