PubMed Health⌕ Search

Biomedical subjects

Robert J Glynn

Publications and source records attributed to Robert J Glynn.

At least 19 recordsLinked to original sources
PubMed2007

Interval estimation for rank correlation coefficients based on the probit transformation with extension to measurement error correction of correlated ranked data.

The Spearman (rho(s)) and Kendall (tau) rank correlation coefficient are routinely used as measures of association between non-normally distributed random variables. However, confidence limits for rho(s) are only available under the assumption of bivariate normality and for tau under the assumption of asymptotic normality of tau. In this paper, we introduce another approach for obtaining confidence limits for rho(s) or tau based on the arcsin transformation of sample probit score correlations. This approach is shown to be applicable for an arbitrary bivariate distribution. The arcsin-based estimators for rho(s) and tau (denoted by rho(s,a), tau(a)) are shown to have asymptotic relative efficiency (ARE) of 9/pi2 compared with the usual estimators rho(s) and tau when rho(s) and tau are, respectively, 0. In some nutritional applications, the Spearman rank correlation between nutrient intake as assessed by a reference instrument versus nutrient intake as assessed by a surrogate instrument is used as a measure of validity of the surrogate instrument. However, if only a single replicate (or a few replicates) are available for the reference instrument, then the estimated Spearman rank correlation will be downwardly biased due to measurement error. In this paper, we use the probit transformation as a tool for specifying an ANOVA-type model for replicate ranked data resulting in a point and interval estimate of a measurement error corrected rank correlation. This extends previous work by Rosner and Willett for obtaining point and interval estimates of measurement error corrected Pearson correlations.

Confidence Intervals↗
PubMed2006

Statins and the risk of lung, breast, and colorectal cancer in the elderly.

BACKGROUND: Although most randomized trials and meta-analyses suggest a slight or no increase in the risk of cancer in statin users, results from observational studies have been conflicting, and some have even suggested a large protective effect of statins on certain cancers. Long-term statin users tend to be healthier, less frail, and more adherent to therapy than nonusers, however. This could explain such apparent "protective" effects. METHODS AND RESULTS: We conducted the present cohort study by linking data from a large state drug benefit program with cancer registry data and Medicare healthcare utilization data. We identified all initiators of statins; initiators of glaucoma medications, another preventive drug, served as a comparison group. Outcomes included all registry-identified cases of colorectal, lung, and breast cancer. Multivariable Cox proportional models were used to adjust for confounding. Patient characteristics were similar in both groups, but statin initiators (n=24,439) were slightly younger and used some services more frequently than glaucoma drug initiators (n=7284). The mean follow-up was 2.9 years, with the longest follow-up being 8.4 years. Incidence rates of colorectal, lung, and breast cancers in both groups were very similar to rates in the general population. Adjusted hazard ratios were 0.96 (95% CI, 0.70 to 1.31) for colorectal cancer, 1.11 (95% CI, 0.77 to 1.60) for lung cancer, and 0.99 (95% CI, 0.74 to 1.33) for breast cancer. CONCLUSIONS: These data from a large population of typical older patients who began using statins indicate that it is unlikely that statins confer a clinically important decrease or increase in the risk of colorectal, lung, or breast cancer over the durations studied.

Aged↗
PubMed2006

Impact of traditional and novel risk factors on the relationship between socioeconomic status and incident cardiovascular events.

BACKGROUND: Persons of lower socioeconomic status have greater cardiovascular risk than those of higher socioeconomic status. However, the mechanism through which socioeconomic status affects cardiovascular disease (CVD) is uncertain. Virtually no data are available that examine the prospective association between novel inflammatory and hemostatic CVD risk indicators, socioeconomic status, and incident CVD events. METHODS AND RESULTS: We assessed the relationship between 2 indicators of socioeconomic status (education and income), traditional and novel CVD risk factors (high sensitivity C-reactive protein, soluble intercellular adhesion molecule-1, fibrinogen, and homocysteine), and incident CVD events among 22,688 apparently healthy female health professionals participating in the Women's Health Study. These women were followed up for 10 years for the development of myocardial infarction, ischemic stroke, coronary revascularization, and cardiovascular death. More educated women were less likely to be smokers; had a lower prevalence of hypertension, diabetes, and obesity; and were more likely to participate in vigorous physical activity than less educated women. At baseline, median total cholesterol, low-density lipoprotein, triglyceride, C-reactive protein, intercellular adhesion molecule-1, fibrinogen, and homocysteine levels for women in 5 categories of education (< 2 years of nursing education, 2 to < 4 years of nursing education, a bachelor's degree, a master's degree, and a doctoral degree) and 6 categories of income [< or = 19,999 dollars, 20,000 dollars to 29,999 dollars, 30,000 dollars to 39,999 dollars, 40,000 dollars to 49,999 dollars, 50,000 dollars to 99,999 dollars, and > or = 100,000 dollars) decreased progressively with increasing education or income levels (all P<0.001), whereas an opposite pattern was observed for high-density lipoprotein (P<0.001). Overall, in age-adjusted Cox proportional hazards models, the relative risk of incident CVD events decreased with increasing education (1.0, 0.7, 0.5, 0.4, and 0.5; P for trend <0.001) and income (1.0, 1.0, 0.9, 0.7, 0.6, and 0.4; P for trend <0.001) categories. In multivariate models that assessed the impact of traditional and novel CVD risk factors on the relationship between education/income and CVD events, the relative hazard of incident CVD associated with a 1-category-higher level of education changed from 0.79 in age- and race-adjusted analysis to 0.89 in fully adjusted analysis. The 11% lower risk per 1 category of education remained significant (P for trend=0.006), suggesting that controlling for both novel and traditional risk factors could not explain the protective effect of education. A similar analysis for income revealed that its relationship with CVD events was explained largely by these noted risk factors. CONCLUSIONS: In this prospective analysis, we observed a decrease in incident CVD events with increasing levels of education and income. In contrast to the relationship between income and CVD events, the relationship of CVD events with education was explained only partially by traditional and novel risk factors for CVD.

Adult↗
PubMed2006

Selective prescribing led to overestimation of the benefits of lipid-lowering drugs.

OBJECTIVE: Observational studies have found beneficial effects of lipid-lowering drugs on diverse outcomes, including venous thromboembolism, hip fracture, dementia, and all-cause mortality. Selective use of these drugs in frail people may confound these relationships. STUDY DESIGN AND SETTING: We measured 1-year mortality in two cohorts of New Jersey residents, aged 65-99 years, enrolled in state-sponsored drug benefits programs: 112,463 persons hospitalized during the years 1991-1994 and 106,838 nonhospitalized enrollees. Use of lipid-lowering drugs and other medications, as well as diagnoses, were evaluated before follow-up. RESULTS: In age- and sex-adjusted analyses, users of lipid-lowering drugs had a 43% reduced death rate relative to nonusers among hospitalized enrollees and a 56% reduction in the nonhospitalized sample. Available markers of frailty and comorbidity predicted decreased use of these drugs. Control for the propensity to use lipid-lowering drugs attenuated but did not eliminate these effects. After such adjustment, users had a 30% reduction in death rate (95% confidence interval [CI]: 25%-35%) among hospitalized enrollees and a 41% reduction (95% CI: 35%-47%) in the nonhospitalized sample. Unmeasured frailty associated with a 26%-33% reduced odds of receiving lipid-lowering therapy could explain this effect. CONCLUSION: Frailty and comorbidity that influence use of preventive therapies can substantially confound apparent benefits of lipid-lowering drugs on outcomes.

Age Factors↗
PubMed2006

Valsartan, blood pressure reduction, and C-reactive protein: primary report of the Val-MARC trial.

Increased levels of high-sensitivity C-reactive protein (hsCRP) are associated with incident hypertension as well as cardiovascular events, and angiotensin II is a potent proinflammatory mediator. However, whether angiotensin receptor blockade lowers hsCRP is uncertain. We performed a randomized trial in which 1668 patients with stage 2 hypertension were treated with 160 mg valsartan or 160/12.5 mg valsartan/hydrochlorothiazide (HCTZ) once daily for 2 weeks with forced titration to 320 mg valsartan or 320/12.5 mg valsartan/HCTZ for an additional 4 weeks. After 6 weeks, systolic blood pressure (-25 versus -18 mm Hg; P<0.001) and diastolic blood pressure (-14 versus -9 mm Hg; P<0.001) were reduced to a greater degree among those allocated to valsartan/HCTZ than to valsartan monotherapy. The median change in hsCRP was -0.12 mg/L among those allocated to valsartan compared with +0.05 mg/L among those allocated to valsartan/HCTZ, a 13.3% difference (P<0.001); this difference between valsartan and valsartan/HCTZ was present in all subgroups evaluated despite the fact that blood pressure reduction was greater in the combined therapy group. No relationship was observed between hsCRP reduction and blood pressure; in all analyses, the proportion of variation in change in hsCRP with valsartan monotherapy explained by change in blood pressure was <2%. Thus, in this prospective trial, valsartan reduced hsCRP levels in a manner independent of degree of blood pressure reduction. These data raise the hypothesis that angiotensin receptor blockade may have anti-inflammatory effects in addition to blood pressure-lowering effects.

Adult↗
PubMed2006

Variable selection for propensity score models.

Despite the growing popularity of propensity score (PS) methods in epidemiology, relatively little has been written in the epidemiologic literature about the problem of variable selection for PS models. The authors present the results of two simulation studies designed to help epidemiologists gain insight into the variable selection problem in a PS analysis. The simulation studies illustrate how the choice of variables that are included in a PS model can affect the bias, variance, and mean squared error of an estimated exposure effect. The results suggest that variables that are unrelated to the exposure but related to the outcome should always be included in a PS model. The inclusion of these variables will decrease the variance of an estimated exposure effect without increasing bias. In contrast, including variables that are related to the exposure but not to the outcome will increase the variance of the estimated exposure effect without decreasing bias. In very small studies, the inclusion of variables that are strongly related to the exposure but only weakly related to the outcome can be detrimental to an estimate in a mean squared error sense. The addition of these variables removes only a small amount of bias but can increase the variance of the estimated exposure effect. These simulation studies and other analytical results suggest that standard model-building tools designed to create good predictive models of the exposure will not always lead to optimal PS models, particularly in small studies.

Confounding Factors, Epidemiologic↗
PubMed2006

Value of an Endpoints Committee versus the use of nosologists for validating cause of death.

PURPOSE: Few studies have directly compared the use of nosologists versus other sources of mortality information deemed a gold standard, including the use of an Endpoints Committee (EC), which is commonly utilized in clinical studies. METHODS: We conducted a study of 421 participants in the Physicians' Health Study (PHS), known to have died of confirmed causes during the period of April 1982 to January 1988. Classification of cause of death was compared when coded by certified nosologists directly from the death certificate without the availability of full hospital and medical records versus determinations made by the PHS Endpoints Committee (EC). RESULTS: The sensitivity of the nosologists, using the PHS EC as the gold standard, was 90% for total cardiovascular death, 89% for cancer and 89% for other deaths. However, when considering more specific causes of death, sensitivity for acute MI, sudden cardiac deaths and deaths from other cardiovascular causes were lower. Specificity was generally excellent for all endpoints, ranging from 90% to 100%. In analyses stratified by age, nosologists tended to overestimate the frequency of cardiovascular deaths in the elderly. CONCLUSIONS: Mortality endpoints classified by trained nosologists versus the PHS EC indicate that nosologists can review death certificates to reasonably and quickly classify broad categories of causes of death in men, whereas an EC remains the preferable strategy when more specific causes of death must be ascertained by reviewing medical records and other accompanying information.

Adult↗
PubMed2006

Cardiovascular outcomes in new users of coxibs and nonsteroidal antiinflammatory drugs: high-risk subgroups and time course of risk.

OBJECTIVE: Controversy persists regarding the cardiovascular risks of treatment with selective cyclooxygenase 2 inhibitors (coxibs) and nonselective nonsteroidal antiinflammatory drugs (NSAIDs). This study was undertaken to examine, in a large group of new users of coxibs and NSAIDs, the rate of cardiovascular events, their time course, and whether baseline cardiovascular risk modified the rate ratios (RRs) for future events. METHODS: This cohort study included Medicare beneficiaries who enrolled in a state-run prescription drug plan that fully covered NSAIDs and coxibs without restriction. All study patients started use of a coxib or NSAID after January 1, 1999. The primary composite end point was a hospital admission for either myocardial infarction or ischemic stroke. Predefined exposure groups included the 3 coxibs available in the US during the study period (celecoxib, rofecoxib, and valdecoxib), as well as oral formulations of diclofenac, ibuprofen, naproxen, and a composite of all other NSAIDs. We compared the rate of cardiovascular events associated with each of these agents with that in a reference group of patients who did not use NSAIDs or coxibs, but started other medications unrelated to cardiovascular risk. Daily exposure to all study drugs was assessed based on filled prescription data. A Cox proportional hazards model stratified on calendar year that included other baseline cardiovascular risk factors constituted the primary analysis. RESULTS: We identified 74,838 users of NSAIDs or coxibs, and 23,532 comparable users of other drugs comprised the reference group. Adjusted models demonstrated a significant elevation in the event rate for rofecoxib (RR 1.15, 95% confidence interval [95% CI] 1.06-1.25) and a significant reduction in the rate for naproxen (RR 0.75, 95% CI 0.62-0.92). No other coxib or NSAID was associated with a significant increase or decrease in cardiovascular event rate. The increased rate associated with rofecoxib was seen in the first 60 days of use (adjusted RR 1.14, 95% CI 1.01-1.29) and thereafter (adjusted RR 1.14, 95% CI 1.02-1.28). Kaplan-Meier event curves showed a similar pattern of risk (early and persistent separation of the event curves) among long-term rofecoxib users at low or high baseline cardiovascular risk. CONCLUSION: We found an increased cardiovascular event rate among users of rofecoxib, and a decreased rate with naproxen use. Other coxibs and NSAIDs did not appear to be associated with a difference in event rate compared with users of other drugs. The increase in rate associated with rofecoxib was seen within the first 60 days and persisted. There was no important modification of the event rate based on the patient's baseline cardiovascular risk.

Aged↗
PubMed2006

Tumor necrosis factor alpha antagonist use and cancer in patients with rheumatoid arthritis.

OBJECTIVE: Concerns persist about a possible association between tumor necrosis factor alpha (TNFalpha) antagonist treatment and development of cancers in patients with rheumatoid arthritis (RA). This study was undertaken to estimate the association between treatment with biologic disease-modifying antirheumatic drugs (DMARDs) and development of cancer in patients with RA. METHODS: We conducted a cohort study pooling administrative databases from 2 US states and 1 Canadian province. A cohort of patients who had received a diagnosis of RA on > or =1 occasion and had been prescribed DMARDs was identified. We categorized patients with a prescription for a biologic DMARD as biologic DMARD users, and those with a prescription for methotrexate (MTX) but no biologic DMARD as MTX users. We used time-varying propensity scores to adjust for the large number of possible confounders and stratified proportional hazards regression to estimate the effects of biologic DMARDs on cancer. The primary end points were hematologic malignancies (lymphoma, multiple myeloma, and leukemia) and common solid tumors (colorectal, lung, stomach, breast, prostate, uterine, ovarian, urinary tract/bladder, and melanoma). RESULTS: The pooled cohort included 1,152 biologic DMARD users and 7,306 MTX users. We identified 11 hematologic malignancies and 46 solid tumors during 2,940 person-years of biologic DMARD use, and 88 hematologic malignancies and 558 solid tumors during 30,300 person-years of MTX use. Comparing biologic DMARD users with MTX users, the propensity score-adjusted pooled hazard ratio was 1.37 (95% confidence interval 0.71-2.65) for hematologic malignancies and 0.91 (95% confidence interval 0.65-1.26) for solid tumors. CONCLUSION: Our results indicate that users of biologic agents are unlikely to have a substantial increase in the risk of hematologic malignancies and solid tumors as compared with MTX users. Despite the use of large combined data sets, studying the effect of an infrequent exposure (biologic DMARDs) on rare diseases (hematologic malignancies) remains a challenge.

Aged↗
PubMed2006

Insights into different results from different causal contrasts in the presence of effect-measure modification.

PURPOSE: Both propensity score (PS) matching and inverse probability of treatment weighting (IPTW) allow causal contrasts, albeit different ones. In the presence of effect-measure modification, different analytic approaches produce different summary estimates. METHODS: We present a spreadsheet example that assumes a dichotomous exposure, covariate, and outcome. The covariate can be a confounder or not and a modifier of the relative risk (RR) or not. Based on expected cell counts, we calculate RR estimates using five summary estimators: Mantel-Haenszel (MH), maximum likelihood (ML), the standardized mortality ratio (SMR), PS matching, and a common implementation of IPTW. RESULTS: Without effect-measure modification, all approaches produce identical results. In the presence of effect-measure modification and regardless of the presence of confounding, results from the SMR and PS are identical, but IPTW can produce strikingly different results (e.g., RR = 0.83 vs. RR = 1.50). In such settings, MH and ML do not estimate a population parameter and results for those measures fall between PS and IPTW. CONCLUSIONS: Discrepancies between PS and IPTW reflect different weighting of stratum-specific effect estimates. SMR and PS matching assign weights according to the distribution of the effect-measure modifier in the exposed subpopulation, whereas IPTW assigns weights according to the distribution of the entire study population. In pharmacoepidemiology, contraindications to treatment that also modify the effect might be prevalent in the population, but would be rare among the exposed. In such settings, estimating the effect of exposure in the exposed rather than the whole population is preferable.

Bias↗
PubMed2006

Colorectal cancer after start of nonsteroidal anti-inflammatory drug use.

PURPOSE: Nonsteroidal anti-inflammatory drugs (NSAIDs), including aspirin, have been consistently shown to reduce the risk of colorectal cancer (CRC) in non-experimental studies, but little is known of the factors associated with starting and continuing regular NSAID use and their effect on the NSAID and CRC association. SUBJECTS AND METHODS: We performed a prospective cohort study of 22,071 healthy male physicians aged 40 to 84 years without indications or contraindications to regular NSAID use at baseline. Annual questionnaires assessed quantity of NSAID use, occurrence of cancer, and risk factors for CRC. Propensity for regular NSAID use (>60 days/year) was estimated using generalized estimating equations. We used a time-varying Cox proportional hazards model to estimate the association between duration since initiation of regular NSAID use and risk for CRC. RESULTS: Regular non-aspirin and any NSAID use increased from 0% to 12% and 1% to 56% over time, respectively and was predicted by age, body mass index, alcohol consumption, medication use, coronary artery disease, gastrointestinal diseases, arthritis, hypertension, and headaches. Over a median follow-up of 18 years, 495 physicians were diagnosed with CRC. There was no trend of CRC risk with increased duration of regular NSAID use. Five or more years of regular use of any NSAID were associated with a relative risk for CRC of 1.0 (95% confidence interval: 0.7-1.5), after adjustment for predictors of regular NSAID use. CONCLUSION: Regular NSAID use was not associated with a substantial risk reduction of CRC after controlling for time-varying predictors of both NSAID use and CRC.

Adult↗
PubMed2006

Impact of two sequential drug cost-sharing policies on the use of inhaled medications in older patients with chronic obstructive pulmonary disease or asthma.

BACKGROUND: Evaluations of drug cost-sharing policies within the same population are needed for a fair comparison of different options. OBJECTIVE: The aim of this work was to analyze the impact of 2 changes in a public drug insurance plan on the use of inhaled medications in British Columbia (BC), Canada. METHODS: Data for the period from 1997 to 2004 were used to assess whether changes in the use of steroid, beta2-agonist, and anticholinergic inhalers were associated with insurance-plan changes in a large, natural experiment involving all BC residents aged>or=65 years. The 3 sequential policies included full coverage, fixed copayments at the beginning of 2002, and 25% coinsurance with an income-based deductible beginning May 1, 2003. Linkable prescription, physician billing, hospitalization, and mortality records were obtained from the BC Ministry of Health Services. From the total population of residents aged>or=65 years, we extracted data for all patients to whom inhaled steroids, beta2-agonists, or anticholinergics were dispensed on or after January 1, 1997. Multivariable linear regression was used to estimate inhaler use during a 60-month baseline period and during implementation of the subsequent copayment and coinsurance plus deductible policies. We used logistic regression to identify predictors of initiation and cessation use of inhaled medications among older patients. RESULTS: Use declined for inhaled steroids (-12.3%; P<0.001), inhaled anticholinergics (-12.2%; P<0.001), and inhaled beta2-agonists (-5.8%; P<0.001). Patients with new diagnoses of asthma or chronic obstructive pulmonary disease were 25% (95% CI, 14%-31%) less likely to initiate treatment with inhaled steroids when covered by the copayment or coinsurance plus deductible policies than when they had full coverage. Chronic users of inhaled steroids were 47% (95% CI, 40%-55%) more likely to cease treatment when they were covered by the copayment policy and 22% (95% CI, 15%-29%) more likely to cease when covered by the coinsurance plus deductible policy than when they had full coverage. CONCLUSIONS: The copayment and coinsurance plus deductible policies were associated with significant reductions in use of inhaled medications, mostly due to decreased initiation and increased cessation rates. However, the consequences of these policies on health outcomes have not yet been determined.

Administration, Inhalation↗
PubMed2006

A therapeutic substitution policy for proton pump inhibitors: clinical and economic consequences.

OBJECTIVE: With the growing need to provide prescription drug benefits to older patients and to contain costs, it will be necessary to direct that coverage so as to make expenditures as efficient as possible. We evaluated the clinical and economic consequences of coverage restriction for 3 leading proton pump inhibitors (PPIs) in a large-scale natural experiment. METHODS: The study design was a time-trend analysis in the setting of a provincial drug benefits program in British Columbia, Canada. We studied all British Columbia residents aged 66 or older (N = 501,104) using linked data on all prescription drug dispensings, physician services, and hospitalizations between January 2002 and June 2004. The new policy restricted coverage to rabeprazole and required treatment failure with a histamine H2 blocker. More widely used PPIs (omeprazole, pantoprazole, and lansoprazole) had to be paid for out of pocket, unless the physician requested an exemption. The main outcome measures were utilization of PPIs, drug discontinuation rates, gastrointestinal hemorrhage rates, and drug expenditures. RESULTS: Utilization of the restricted PPIs declined sharply after the policy change (-14,850 daily doses per month per 10,000 residents, P < .0001), whereas use of the covered PPI increased sharply (+19,300, P < .0001), with 45% of all PPI users switching to the covered agent within 6 months. We found no increased use of H2 blockers or stopping of gastroprotective drugs. There was no increase in the monthly rate of hospitalization for gastrointestinal hemorrhage after the PPI restriction (P = .35) even though the study had the power to detect increases of 24 events per 10,000 residents with 95% confidence. There was a slight increase in physician visits 3 months after the policy change (P = .01) for a 2-month period when 9% of new rabeprazole users were switched back to a restricted PPI. In the first 6 months of the policy change, the provincial health plan saved at least 2.9 million Canadian dollars as a result of the policy change. CONCLUSIONS: Coverage restriction of 3 leading PPIs led to substantial utilization changes and savings, without increased noncompliance or clinical complication.

2-Pyridinylmethylsulfinylbenzimidazoles↗
PubMed2006

The Wilcoxon signed rank test for paired comparisons of clustered data.

The Wilcoxon signed rank test is a frequently used nonparametric test for paired data (e.g., consisting of pre- and posttreatment measurements) based on independent units of analysis. This test cannot be used for paired comparisons arising from clustered data (e.g., if paired comparisons are available for each of two eyes of an individual). To incorporate clustering, a generalization of the randomization test formulation for the signed rank test is proposed, where the unit of randomization is at the cluster level (e.g., person), while the individual paired units of analysis are at the subunit within cluster level (e.g., eye within person). An adjusted variance estimate of the signed rank test statistic is then derived, which can be used for either balanced (same number of subunits per cluster) or unbalanced (different number of subunits per cluster) data, with an exchangeable correlation structure, with or without tied values. The resulting test statistic is shown to be asymptotically normal as the number of clusters becomes large, if the cluster size is bounded. Simulation studies are performed based on simulating correlated ranked data from a signed log-normal distribution. These studies indicate appropriate type I error for data sets with > or =20 clusters and a superior power profile compared with either the ordinary signed rank test based on the average cluster difference score or the multivariate signed rank test of Puri and Sen. Finally, the methods are illustrated with two data sets, (i) an ophthalmologic data set involving a comparison of electroretinogram (ERG) data in retinitis pigmentosa (RP) patients before and after undergoing an experimental surgical procedure, and (ii) a nutritional data set based on a randomized prospective study of nutritional supplements in RP patients where vitamin E intake outside of study capsules is compared before and after randomization to monitor compliance with nutritional protocols.

Humans↗
PubMed2006

Extension of the rank sum test for clustered data: two-group comparisons with group membership defined at the subunit level.

The Wilcoxon rank sum test is widely used for two-group comparisons for nonnormal data. An assumption of this test is independence of sampling units both between and within groups. In ophthalmology, data are often collected on two eyes of an individual, which are highly correlated. In ophthalmological clinical trials, randomization is usually performed at the subject level, but the unit of analysis is the eye. If the eye is used as the unit of analysis, then a modification to the usual Wilcoxon rank sum variance formula must be made to account for the within-cluster dependence. For some clustered data designs, where the unit of analysis is the subunit, group membership may be defined at the subunit level. For example, in some randomized ophthalmologic clinical trials, different treatments may be applied to fellow eyes of some patients, while the same treatment may be applied to fellow eyes of other patients. In general, binary eye-specific covariates may be present (scored as exposed or unexposed) and one wishes to compare nonnormally distributed outcomes between exposed and unexposed eyes using the Wilcoxon rank sum test while accounting for the clustering. In this article, we present a corrected variance formula for the Wilcoxon rank sum statistic in the setting of eye (subunit)-specific covariates. We apply it to compare ocular itching scores in ocular allergy patients between eyes treated with active versus placebo eye drops, where some patients receive the same eye drop in both eyes, while other patients receive different eye drops in fellow eyes. We also present comparisons between the clustered Wilcoxon test and each of the signed rank tests and mixed model approaches and show dramatic differences in power in favor of the clustered Wilcoxon test for some designs.

Biometry↗
PubMed2006

Indications for propensity scores and review of their use in pharmacoepidemiology.

Use of propensity scores to identify and control for confounding in observational studies that relate medications to outcomes has increased substantially in recent years. However, it remains unclear whether, and if so when, use of propensity scores provides estimates of drug effects that are less biased than those obtained from conventional multivariate models. In the great majority of published studies that have used both approaches, estimated effects from propensity score and regression methods have been similar. Simulation studies further suggest comparable performance of the two approaches in many settings. We discuss five reasons that favour use of propensity scores: the value of focus on indications for drug use; optimal matching strategies from alternative designs; improved control of confounding with scarce outcomes; ability to identify interactions between propensity of treatment and drug effects on outcomes; and correction for unobserved confounders via propensity score calibration. We describe alternative approaches to estimate and implement propensity scores and the limitations of the C-statistic for evaluation. Use of propensity scores will not correct biases from unmeasured confounders, but can aid in understanding determinants of drug use and lead to improved estimates of drug effects in some settings.

Bias↗
PubMed2006

Metabolic abnormalities and risk for colorectal cancer in the physicians' health study.

BACKGROUND: Obesity and diabetes are established risk factors for colorectal cancer but have mainly been assessed independently. There are few data about whether the metabolic syndrome, which refers to a clustering of cardiovascular disease risk factors thought to be related to insulin resistance, including obesity, type 2 diabetes, hyperlipidemia, and hypertension, is associated with colorectal cancer risk. METHODS: During and after the randomized trial of aspirin and beta-carotene, 22,071 healthy male physicians, initially ages 40 to 84 years, reported overweight (body mass index, >or=27 kg/m2), diabetes, elevated blood pressure (>or=130/85 mmHg or use of antihypertensive medication), hypercholesterolemia (>or=240 mg/dL or use of lipid-lowering medication), and occurrence of cancer on annual questionnaires. Adjusted relative risks (RR) and 95% confidence intervals (95% CI) for time-varying metabolic abnormalities and colorectal cancer were estimated using a multivariable proportional hazards model. RESULTS: During 369,966 person-years of follow-up (median, 19 years), 494 physicians developed colorectal cancer. With aging of the cohort, the prevalence of having two or more abnormalities increased from 13% to 35%. Overweight (RR, 1.4; 95% CI, 1.1-1.7) and diabetes (RR, 1.5; 95% CI, 1.1-2.0) were associated with increased risk for colorectal cancer, whereas elevated blood pressure (RR, 1.1; 95% CI, 0.9-1.3) and hypercholesterolemia (RR, 0.9; 95% CI, 0.7-1.1) were not. This model assessing metabolic abnormalities independently was more predictive for colorectal cancer than a model based on the number of abnormalities (each additional abnormality RR, 1.16; 95% CI, 1.05-1.29). CONCLUSIONS: Out of the markers of the metabolic syndrome assessed, overweight and diabetes are risk factors for colorectal cancer, whereas, in contrast to their role in cardiovascular disease, elevated blood pressure and hypercholesterolemia are not.

Adult↗