PubMed Health⌕ Search

Biomedical subjects

Sander Greenland

Publications and source records attributed to Sander Greenland.

At least 19 recordsLinked to original sources

Prior data for non-normal priors.

Data augmentation priors facilitate contextual evaluation of prior distributions and the generation of Bayesian outputs from frequentist software. Previous papers have presented approximate Bayesian methods using 2x2 tables of 'prior data' to represent lognormal relative-risk priors in stratified and regression analyses. The present paper describes extensions that use the tables to represent generalized-F prior distributions for relative risks, which subsume lognormal priors as a limiting case. The method provides a means to increase tail-weight or skew the prior distribution for the log relative risk away from normality, while retaining the simple 2x2 table form of the prior data. When prior normality is preferred, it also provides a more accurate lognormal relative-risk prior in for the 2x2 table format. For more compact representation in regression analyses, the prior data can be compressed into a single data record. The method is illustrated with historical data from a study of electronic foetal monitoring and neonatal death.

Bayes Theorem↗

The Hanford Thyroid Disease Study: an alternative view of the findings.

The Hanford Thyroid Disease Study (HTDS) is one of the largest and most complex epidemiologic studies of the relation between environmental exposures to I and thyroid disease. The study detected no dose-response relation using a 0.05 level for statistical significance. The results for thyroid cancer appear inconsistent with those from other studies of populations with similar exposures, and either reflect inadequate statistical power, bias, or unique relations between exposure and disease risk. In this paper, we explore these possibilities, and present evidence that the HTDS statistical power was inadequate due to complex uncertainties associated with the mathematical models and assumptions used to reconstruct individual doses. We conclude that, at the very least, the confidence intervals reported by the HTDS for thyroid cancer and other thyroid diseases are too narrow because they fail to reflect key uncertainties in the measurement-error structure. We recommend that the HTDS results be interpreted as inconclusive rather than as evidence for little or no disease risk from Hanford exposures.

Cohort Studies↗

Exposure-measurement error is frequently ignored when interpreting epidemiologic study results.

INTRODUCTION: One important source of error in study results is error in measuring exposures. When interpreting study results, one should consider the impact that exposure-measurement error (EME) might have had on study results. METHODS: To assess how often this consideration is made and the form it takes, journal articles were randomly sampled from original articles appearing in the American Journal of Epidemiology and Epidemiology in 2001, and the International Journal of Epidemiology between December 2000 and October 2001. RESULTS: Twenty-two (39%) of the 57 articles surveyed mentioned nothing about EME. Of the 35 articles that mentioned something about EME, 16 articles described qualitatively the effect EME could have had on study results. Only one study quantified the impact of EME on study results; the investigators used a sensitivity analysis. Few authors discussed the measurement error in their study in any detail. CONCLUSIONS: Overall, the potential impact of EME on error in epidemiologic study results appears to be ignored frequently in practice.

Bias↗

Sensitivity analysis of misclassification: a graphical and a Bayesian approach.

PURPOSE: Misclassification can produce bias in measures of association. Sensitivity analyses have been suggested to explore the impact of such bias, but do not supply formally justified interval estimates. METHODS: To account for exposure misclassification, recently developed Bayesian approaches were extended to incorporate prior uncertainty and correlation of sensitivity and specificity. Under nondifferential misclassification, a contour plot is used to depict relations among the corrected odds ratio, sensitivity, and specificity. RESULTS: Methods are illustrated by application to a case-control study of cigarette smoking and invasive pneumococcal disease while varying the distributional assumptions about sensitivity and specificity. Results are compared with those of conventional methods, which do not account for misclassification, and a sensitivity analysis, which assumes fixed sensitivity and specificity. CONCLUSION: By using Bayesian methods, investigators can incorporate uncertainty about misclassification into probabilistic inferences.

Bayes Theorem↗

Multiple-imputation for measurement-error correction.

BACKGROUND: There are many methods for measurement-error correction. These methods remain rarely used despite the ubiquity of measurement error. METHODS: Treating measurement error as a missing-data problem, the authors show how multiple-imputation for measurement-error (MIME) correction can be done using SAS software and evaluate the approach with a simulation experiment. RESULTS: Based on hypothetical data from a planned cohort study of 600 children with chronic kidney disease, the estimated hazard ratio for end-stage renal disease from the complete data was 2.0 [95% confidence limits (95% CL) 1.4, 2.8] and was reduced to 1.5 (95% CL 1.1, 2.1) using a misclassified exposure of low glomerular filtration rate at study entry (sensitivity of 0.9 and specificity of 0.7). The MIME correction hazard ratio was 2.0 (95% CL 1.2, 3.3), the regression calibration (RC) hazard ratio was 2.0 (95% CL 1.1, 3.7), and restriction to a 25% validation substudy yielded a hazard ratio of 2.0 (95% CL 1.0, 3.7). Based on Monte Carlo simulations across eight scenarios, MIME was approximately unbiased, had approximately correct coverage, and was sometimes more powerful than misclassified or RC analyses. Using root mean squared error as a criterion, the MIME bias correction is sometimes outweighed by added imprecision. CONCLUSION: The choice between MIME and RC depends on performance, ease, and objectives. The usefulness of MIME correction in specific applications will depend upon the sample size or the proportion validated. MIME correction may be valuable in interpreting imperfectly measured epidemiological data.

Bias↗

Accounting for independent nondifferential misclassification does not increase certainty that an observed association is in the correct direction.

Researchers sometimes argue that their exposure-measurement errors are independent of other errors and are nondifferential with respect to disease, resulting in estimation bias toward the null. Among well-known problems with such arguments are that independence and nondifferentiality are harder to satisfy than ordinarily appreciated (e.g., because of correlation of errors in questionnaire items, and because of uncontrolled covariate effects on error rates); small violations of independence or nondifferentiality may lead to bias away from the null; and, if exposure is polytomous, the bias produced by independent nondifferential error is not always toward the null. The authors add to this list by showing that, in a 2 x 2 table (for which independent nondifferential error produces bias toward the null), accounting for independent nondifferential error does not reduce the p value even though it increases the point estimate. Thus, such accounting should not increase certainty that an association is present.

Bias↗

Associations of maternal age- and parity-related factors with trends in low-birthweight rates: United States, 1980 through 2000.

OBJECTIVES: We assessed the effects of changes in the maternal age-parity distribution and age-and parity-specific low-birthweight rates on low-birthweight trends in the United States. METHODS: We used natality file data from 1980 through 2000 to assess very-low-birthweight and low-birthweight rates among singleton live-born infants. RESULTS: Changes in age-and parity-specific low-birthweight rates were the main contributor to the overall trend in rates. However, changes in the age-parity distribution, primarily delayed childbearing, had a smaller but noticeable impact. The very-low-birthweight rate increased 27% among Black women, and changes in the age-parity distribution were associated with, on average, more than 20% of the increased rate during the 1990s. Among Hispanic and non-Hispanic White women, on average, more than 10% of the rate increase observed during the 1990s was associated with changes in the age-parity distribution. CONCLUSIONS: Assuming minimal changes in age-specific rates, delayed childbearing may play an increasingly important role in low-birthweight trends in the United States.

Adolescent↗

Polymorphism of Xeroderma Pigmentosum group G and the risk of lung cancer and squamous cell carcinomas of the oropharynx, larynx and esophagus.

We investigated the effects of XPG His1104Asp polymorphism (rs17655) on the risk of lung cancer and squamous cell carcinomas of the oropharynx, larynx and esophagus (SCCOLE). This population-based case-control study involves 611 new cases of lung cancer, 601 new cases of oropharyngeal, laryngeal and esophageal cancers, and 1,040 cancer-free controls. The XPG polymorphism was assayed by PCR-RFLP method for 497 lung cancer cases, 443 cases of oropharyngeal, laryngeal and esophageal cancers and 912 controls. Binary and polytomous unconditional logistic regression models were fitted to assess the main effects and the effect modifications between the polymorphism and environmental exposures. With the adjustment for potential confounders, the XPG Asp1104Asp genotype was inversely associated with lung cancer (odds ratio [OR] = 0.62, 95% confidence limits [CL] = 0.38, 1.0) and SCCOLE (OR = 0.47, 95% CL = 0.27, 0.82), with the combined His1104His and His1104Asp genotypes as the referent. With subjects having genotype Asp1104Asp and no tobacco smoking exposure as the common referent, the ORs on lung cancer were 13 (95% CL = 4.4, 37) for heavy tobacco smoking (>20 pack-years), 1.9 (95% CL = 0.78, 4.5) for having at least one copy of 1104His, and 23 (95% CL = 9.5, 56) for the joint effect, respectively. Compared to non-smokers with the Asp1104Asp genotype, the adjusted OR on SCCOLE for heavy smokers (>20 pack-years) having at least one copy of 1104His was 8.0 (95% CL = 2.7, 24). Similarly, compared to non-drinkers with the Asp1104Asp genotype, the adjusted OR on SCCOLE for heavy drinkers (> or =3 drinks/day) with at least one copy of 1104His was 10 (95% CL = 2.7, 38). In conclusion, our study suggests that the XPG Asp1104Asp genotype may be associated with decreased susceptibility to lung cancer and SCCOLE.

Adenocarcinoma↗

Bayesian perspectives for epidemiological research: I. Foundations and basic methods.

One misconception (of many) about Bayesian analyses is that prior distributions introduce assumptions that are more questionable than assumptions made by frequentist methods; yet the assumptions in priors can be more reasonable than the assumptions implicit in standard frequentist models. Another misconception is that Bayesian methods are computationally difficult and require special software. But perfectly adequate Bayesian analyses can be carried out with common software for frequentist analysis. Under a wide range of priors, the accuracy of these approximations is just as good as the frequentist accuracy of the software--and more than adequate for the inaccurate observational studies found in health and social sciences. An easy way to do Bayesian analyses is via inverse-variance (information) weighted averaging of the prior with the frequentist estimate. A more general method expresses the prior distributions in the form of prior data or 'data equivalents', which are then entered in the analysis as a new data stratum. That form reveals the strength of the prior judgements being introduced and may lead to tempering of those judgements. It is argued that a criterion for scientific acceptability of a prior distribution is that it be expressible as prior data, so that the strength of prior assumptions can be gauged by how much data they represent.

Bayes Theorem↗

Curious phenomena in Bayesian adjustment for exposure misclassification.

Many epidemiologic investigations involve some discussion of exposure misclassification, but rarely is there an attempt to adjust for misclassification formally in the statistical analysis. Rather, investigators tend to rely on intuition to comment qualitatively on how misclassification might impact their findings. We point out several ways in which intuition might fail, in the context of unmatched case-control analysis with non-differential exposure misclassification. Particularly, we focus on how intuition can conflict with the results of a Bayesian analysis that accounts for the various uncertainties at hand. First, the Bayesian adjustment for misclassification can weaken the evidence about the direction of an exposure-disease association. Second, admitting uncertainty about the misclassification parameters can lead to narrower interval estimates concerning the association. We focus on the simple setting of unmatched case-control analysis with binary exposure and without adjustment for confounders, though much of our discussion should be relevant more generally.

Anti-Bacterial Agents↗

Childhood leukemia, electric and magnetic fields, and temporal trends.

During the past 25 years concern has been raised about the possible health effects of extremely low frequency (ELF) electric and magnetic fields (EMFs), particularly regarding childhood leukemia. Comparison of changes in electricity consumption (a surrogate for exposure) to changes in childhood-leukemia rates, known as ecologic correlation, have been used to argue both for and against the association between magnetic fields and childhood leukemia. In this paper we explore what can be learned from such an ecologic approach. We first examine separately the evidence on trends in exposure to EMFs and on trends in leukemia rates, and then compare the two. Both incidence rates and exposures have increased, but there are so many approximations and assumptions involved in connecting the two trends that we cannot regard the ecologic evidence as providing any meaningful evidence for or against a causal link.

Acute Disease↗

Longitudinal associations between dietary protein intake and survival in hemodialysis patients.

BACKGROUND: Decreased dietary protein intake may be associated with increased mortality risk in individuals with kidney failure undergoing maintenance hemodialysis (MHD). We hypothesized that longitudinal changes in dietary protein intake have independent associations with survival in MHD patients. METHODS: The relation between urea kinetic-based normalized protein nitrogen appearance (nPNA) and all-cause and cardiovascular mortality was examined in a 2-year (July 2001 to June 2003) cohort of 53,933 MHD patients from virtually all DaVita dialysis clinics in the United States, using both conventional and time-dependent (repeated-measure) Cox models to estimate death hazard ratios for quarterly averaged nPNA categories controlled for case-mix, comorbidity, dialysis dose (Kt/V), and available markers of malnutrition-inflammation complex syndrome (MICS). RESULTS: The best survival was associated with nPNA between 1.0 and 1.4 g/kg/d, whereas nPNA less than 0.8 or greater than 1.4 g/kg/d was associated with greater mortality in almost all models. Adjustment for MICS mitigated the associations substantially. A decrease in protein intake during the first 6 months in patients with an nPNA in the 0.8- to 1.2-g/kg/d range was associated incrementally with greater death risks in the subsequent 18 months, whereas an increase in nPNA tended to correlate with reduced death risk. CONCLUSION: Low daily protein intake or decrease in its magnitude over time is associated with increased risk for death in MHD patients. Whether the association between time-varying protein intake and survival is causal or a consequence of anorexia secondary to MICS or other factors needs to be explored further in interventional trials.

Aged↗

Leukemia attributable to residential magnetic fields: results from analyses allowing for study biases.

Nearly every epidemiologic study of residential magnetic fields and childhood leukemia has exhibited a positive association. Nonetheless, because these studies suffer from various methodologic limitations and there is no known plausible mechanism of action, it remains uncertain as to how much, if any, of these associations are causal. Furthermore, because the observed associations are small and involve only the highest and most infrequent levels of exposure, it is believed that the public health impact of an effect would be small. We present some formal analyses of the impact of power-frequency residential magnetic-field exposure (as measured by attributable fractions), accounting for our uncertainties about study biases as well as uncertainties about exposure distribution. These analyses support the idea that the public health impact of residential fields is likely to be limited, but both no impact and a substantial impact remain possibilities in light of the available data.

Bayes Theorem↗

The performance of random coefficient regression in accounting for residual confounding.

Greenland (2000, Biometrics 56, 915-921) describes the use of random coefficient regression to adjust for residual confounding in a particular setting. We examine this setting further, giving theoretical and empirical results concerning the frequentist and Bayesian performance of random coefficient regression. Particularly, we compare estimators based on this adjustment for residual confounding to estimators based on the assumption of no residual confounding. This devolves to comparing an estimator from a nonidentified but more realistic model to an estimator from a less realistic but identified model. The approach described by Gustafson (2005, Statistical Science 20, 111-140) is used to quantify the performance of a Bayesian estimator arising from a nonidentified model. From both theoretical calculations and simulations we find support for the idea that superior performance can be obtained by replacing unrealistic identifying constraints with priors that allow modest departures from those constraints. In terms of point-estimator bias this superiority arises when the extent of residual confounding is substantial, but the advantage is much broader in terms of interval estimation. The benefit from modeling residual confounding is maintained when the prior distributions employed only roughly correspond to reality, for the standard identifying constraints are equivalent to priors that typically correspond much worse.

Bayes Theorem↗

Marijuana use and the risk of lung and upper aerodigestive tract cancers: results of a population-based case-control study.

BACKGROUND: Despite several lines of evidence suggesting the biological plausibility of marijuana being carcinogenic, epidemiologic findings are inconsistent. We conducted a population-based case-control study of the association between marijuana use and the risk of lung and upper aerodigestive tract cancers in Los Angeles. METHODS: Our study included 1,212 incident cancer cases and 1,040 cancer-free controls matched to cases on age, gender, and neighborhood. Subjects were interviewed with a standardized questionnaire. The cumulative use of marijuana was expressed in joint-years, where 1 joint-year is equivalent to smoking one joint per day for 1 year. RESULTS: Although using marijuana for > or =30 joint-years was positively associated in the crude analyses with each cancer type (except pharyngeal cancer), no positive associations were observed when adjusting for several confounders including cigarette smoking. The adjusted odds ratio estimate (and 95% confidence limits) for > or =60 versus 0 joint-years was 1.1 (0.56, 2.1) for oral cancer, 0.84 (0.28, 2.5) for laryngeal cancer, and 0.62 (0.32, 1.2) for lung cancer; the adjusted odds ratio estimate for > or =30 versus 0 joint-years was 0.57 (0.20, 1.6) for pharyngeal cancer, and 0.53 (0.22, 1.3) for esophageal cancer. No association was consistently monotonic across exposure categories, and restriction to subjects who never smoked cigarettes yielded similar findings. CONCLUSIONS: Our results may have been affected by selection bias or error in measuring lifetime exposure and confounder histories; but they suggest that the association of these cancers with marijuana, even long-term or heavy use, is not strong and may be below practically detectable limits.

Adenocarcinoma↗

Associations between changes in hemoglobin and administered erythropoiesis-stimulating agent and survival in hemodialysis patients.

Although treating anemia of chronic kidney disease by erythropoiesis-stimulating agents (ESA) may improve survival, most studies have examined associations between baseline hemoglobin values and survival and ignored variations in clinical and laboratory measures over time. It is not clear whether longitudinal changes in hemoglobin or administered ESA have meaningful associations with survival after adjustment for time-varying confounders. With the use of time-dependent Cox regression models, longitudinal associations were examined between survival and quarterly (13-wk averaged) hemoglobin values and administered ESA dose in a 2-yr (July 2001 to June 2003) cohort of 58,058 maintenance hemodialysis patients from a large dialysis organization (DaVita) in the United States. After time-dependent and multivariate adjustment for case mix, quarterly varying administered intravenous iron and ESA doses, iron markers, and nutritional status, hemoglobin levels between 12 and 13 g/dl were associated with the greatest survival. Among prevalent patients, the lower range of the recommended Kidney Disease Quality Outcomes Initiative hemoglobin target (11 to 11.5 g/dl) was associated with a higher death risk compared with the 11.5- to 12-g/dl range. A decrease or increase in hemoglobin over time was associated with higher or lower death risk, respectively, independent of baseline hemoglobin. Administration of any dose of ESA was associated with better survival, whereas among those who received ESA, requiring higher doses were surrogates of higher death risk. In this observational study, greater survival was associated with a baseline hemoglobin between 12 and 13 g/dl, treatment with ESA, and rising hemoglobin. Falling hemoglobin and requiring higher ESA doses were associated with decreased survival. Randomized clinical trials are required to examine these associations.

Aged↗