Two-stage case-control studies: precision of parameter estimates and considerations in selecting sample size

Association between sex hormones regulation-related SNP rs12233719 and lung cancer risk among never-smoking Chinese women

BACKGROUND

The mechanism of rapidly increased non-small cell lung cancer (NSCLC) among never-smoking Chinese women has not been elucidated. Ovarian sex steroid hormones have been suggested to counteract lung cancer development, and sex hormone-binding globulin (SHBG) is essential in sex hormones regulation. This study aims to exploring single nucleotide polymorphisms (SNPs) in genomic regions associated with SHBG concentrations that contributed to never-smoking female NSCLC.

METHODS

Candidate genes were selected by a genome-wide association (GWAS) meta-analysis and gene expression profiles of never-smoking NSCLC of Chinese women. The candidate SNPs limited to common minor allele frequency (MAF), missense variant, ethnic heterogeneous distribution, and SNPs were genotyped using the TaqMan method. A two-stage case-control design was adopted for exploration and validation of associations between candidate SNPs and risk of NSCLC. All participants were never-smoking Chinese women. Chi-square test and multivariate logistic regression were applied.

RESULTS

Beginning with 12 genomic regions associated with circulating SHBG concentrations and gene expression profiles from never-smoking NSCLC in Chinese women, candidate SNP rs12233719 and rs7439366 both located in candidate gene UGT₂ B_7, which may be related to circulating SHBG concentrations and cancer risk, were identified. A two-stage case-control study was conducted in Shenyang and Tianjin represented as the training stage and validation stage, respectively. Under the dominant model, compared to individuals with the wild G/G genotype, the adjusted OR of those with the T allele was 1.58 (95% CI: 1.15-2.16) in Chinese Shenyang training set, and was 1.49 (95% CI: 1.02-2.18) in Chinese Tianjin validation set, both accompanied with a significant trend relationship consistently. UGT2B7 was upregulated in female NSCLC patients' tumor tissues and was associated with a poor prognosis in NSCLC.

CONCLUSION

Our findings indicated that a sex hormones regulation-related SNP rs12233719 was associated with never-smoking female lung cancer risk, which might partially explain NSCLC-susceptibility in Chinese women.

Two-stage g-computation: Evaluating Treatment and Intervention Impacts in Observational Cohorts When Exposure Information Is Partly Missing

Illustrations of the g-computation algorithm to evaluate population average treatment and intervention effects have been predominantly implemented in settings with complete exposure information. Thus, worked examples of approaches to handle missing data in this causal framework are needed to facilitate wider use of these estimators. We illustrate two-stage g-computation estimators that leverage partially observed information on the full study sample and complete exposure information on a subset to estimate causal effects. In a hypothetical cohort of 1,623 human immunodeficiency virus (HIV)-positive women with 30% complete opioid prescription information, we illustrate a two-stage extrapolation g-computation estimator for the average treatment effect of shorter or longer duration opioid prescriptions; we further illustrate two-stage inverse probability weighting and imputation g-computation estimators for the average intervention effect of shortening the duration of prescriptions relative to the status quo. Two-stage g-computation estimators approximated the true risk differences for the population average treatment and intervention effects while g-computation fit to the subset of complete cases was biased. In 10,000 Monte Carlo simulations, two-stage approaches considerably reduced bias and mean squared error and improved the coverage of 95% confidence limits. Although missing data threaten validity and precision, two-stage g-computation designs offer principled approaches to handling missing information.

Association Between Bacillus Calmette-Guérin Vaccination and Childhood Asthma in the Quebec Birth Cohort on Immunity and Health

We estimated the association between bacillus Calmette-Guérin (BCG) vaccination and childhood asthma in a birth cohort using administrative databases, and we determined the impact of adjusting for potential confounders collected from a subset of the cohort members. Data were collected in 2 waves: 1) Administrative data for 76,623 individuals (stage 1) was gathered from the Quebec Birth Cohort on Immunity and Health (1974-1994), including BCG vaccination status, perinatal and sociodemographic characteristics, and use of health services for asthma; and 2) self-reported asthma risk factors were collected in 2012 by telephone interviews with 1,643 participants (stage 2) using a balanced 2-stage sampling design. We estimated odds ratios and 95% confidence intervals for asthma using logistic regression and correcting for the known sampling fractions from stage 1 to stage 2, overall and sex-stratified. In total, 35,612 (46.5%) individuals were BCG vaccinated, and 5,870 (7.7%) had asthma. The final odds ratio, integrating results from both stages of sampling, was 0.95 (95% confidence interval: 0.87, 1.04). Results did not differ according to sex (P for interaction = 0.327). To our knowledge, this is the largest study ever conducted on this topic, and using the best possible comprehensive adjustment for confounders, we found no association between BCG vaccination and asthma.

Evaluating the Validity of a Two-stage Sample in a Birth Cohort Established from Administrative Databases

BACKGROUND

When using administrative databases for epidemiologic research, a subsample of subjects can be interviewed, eliciting information on undocumented confounders. This article presents a thorough investigation of the validity of a two-stage sample encompassing an assessment of nonparticipation and quantification of the extent of bias.

METHODS

Established through record linkage of administrative databases, the Québec Birth Cohort on Immunity and Health (n = 81,496) aims to study the association between Bacillus Calmette-Guérin vaccination and asthma. Among 76,623 subjects classified in four Bacillus Calmette-Guérin-asthma strata, a two-stage sampling strategy with a balanced design was used to randomly select individuals for interviews. We compared stratum-specific sociodemographic characteristics and healthcare utilization of stage 2 participants (n = 1,643) with those of eligible nonparticipants (n = 74,980) and nonrespondents (n = 3,157). We used logistic regression to determine whether participation varied across strata according to these characteristics. The effect of nonparticipation was described by the relative odds ratio (ROR = ORparticipants/ORsource population) for the association between sociodemographic characteristics and asthma.

RESULTS

Parental age at childbirth, area of residence, family income, and healthcare utilization were comparable between groups. Participants were slightly more likely to be women and have a mother born in Québec. Participation did not vary across strata by sex, parental birthplace, or material and social deprivation. Estimates were not biased by nonparticipation; most RORs were below one and bias never exceeded 20%.

CONCLUSIONS

Our analyses evaluate and provide a detailed demonstration of the validity of a two-stage sample for researchers assembling similar research infrastructures.

Effect of maternal moderate to severe asthma on perinatal outcomes

BACKGROUND/OBJECTIVES

It has been reported that adverse fetal outcomes are more prevalent in pregnant women with asthma than they are in women without asthma. In our study, we investigated the effect that the severity of asthma during pregnancy has on the risk of a small for gestational age (SGA) infant, low birth weight (LBW), and preterm birth.

METHODS

A population-based cohort of 13,007 pregnancies from asthmatic women was reconstructed through the linking of three of Quebec's (Canada) administrative databases covering the period between 1990 and 2002. A two-stage sampling cohort design was used to collect additional information on the selected women's life-style habits via a mailed questionnaire. Asthma severity during pregnancy was measured with a validated database index. A logistic regression model was used to obtain the adjusted odds ratios of SGA, LBW and preterm birth as a function of the level of asthma severity.

RESULTS

The proportions of women with mild, moderate and severe asthma were 82.5%, 12.5% and 5.0%, respectively. We sent 3,168 questionnaires to selected women, with a 40.2% (n=1274) response rate. Final estimates showed that the risk of SGA was significantly higher among severe (OR:1.48, 95%CI: 1.15-1.91) and moderate asthmatic women (OR: 1.30, 95%CI:1.10-1.55) than mild asthmatic women. No significant associations were found between asthma severity, preterm birth and LBW.

CONCLUSIONS

Mothers with severe and moderate asthma during pregnancy have a higher risk of SGA babies than those with mild asthma.

Statistical methods in cancer epidemiological studies

In this chapter, we discuss statistical methods for various study designs that are commonly used in epidemiological research and particularly in cancer epidemiological research. After a brief review of basic concepts in epidemiological studies, statistical methods for case-control studies and cohort studies are discussed. Statistical methods for nested case-control and case-cohort studies, which have been increasingly used in cancer epidemiology, also are discussed. This chapter is designed for cancer epidemiologists who understand basic statistical methods for commonly used epidemiological study designs and are able to initiate power and sample size calculations. Therefore, this chapter emphasizes newly developed statistical methods for epidemiological studies as well as study planning.

Efficient sampling approaches to address confounding in database studies

Administrative and other population-based databases are widely used in pharmacoepidemiology to study the unintended effects of medications. They allow investigators to study large case series, and they document prescription medication exposure without having to contact individuals or medical charts, or rely on human recall. However, such databases often lack information on potentially important confounding variables. This review describes some of the sampling approaches and accompanying data-analysis methods that can be used to assess, and deal efficiently with, such confounding.

Adjustments for unmeasured confounders in pharmacoepidemiologic database studies using external information

BACKGROUND

Nonexperimental studies of drug effects in large automated databases can provide timely assessment of real-life drug use, but are prone to confounding by variables that are not contained in these databases and thus cannot be controlled.

OBJECTIVES

To describe how information on additional confounders from validation studies can help address the problem of unmeasured confounding in the main study.

RESEARCH DESIGN

Review types of validation studies that allow adjustment for unmeasured confounding and illustrate these with an example.

SUBJECTS

Main study: New Jersey residents age 65 years or older hospitalized between 1995 and 1997, who filled prescriptions within Medicaid or a pharmaceutical assistance program. Validation study: representative sample of Medicare beneficiaries.

MEASURES

Association between nonsteroidal antiinflammatory drugs (NSAIDs) and mortality.

RESULTS

Validation studies are categorized as internal (ie, additional information is collected on participants of the main study) or external. Availability of information on disease outcome will affect choice of analytic strategies. Using an external validation study without data on disease outcome to adjust for unmeasured confounding, propensity score calibration (PSC) leads to a plausible estimate of the association between NSAIDs and mortality in the elderly, if the biases caused by measured and unmeasured confounders go in the same direction.

CONCLUSIONS

Estimates of drug effects can be adjusted for confounders that are not available in the main, but can be measured in a validation study. PSC uses validation data without information on disease outcome under a strong assumption. The collection and integration of validation data in pharmacoepidemiology should be encouraged.

Efficiency of two-phase methods with focus on a planned population-based case-control study on air pollution and stroke

BACKGROUND

We plan to conduct a case-control study to investigate whether exposure to nitrogen dioxide (NO2) increases the risk of stroke. In case-control studies, selective participation can lead to bias and loss of efficiency. A two-phase design can reduce bias and improve efficiency by combining information on the non-participating subjects with information from the participating subjects. In our planned study, we will have access to individual disease status and data on NO2 exposure on group (area) level for a large population sample of Scania, southern Sweden. A smaller sub-sample will be selected to the second phase for individual-level assessment on exposure and covariables. In this paper, we simulate a case-control study based on our planned study. We develop a two-phase method for this study and compare the performance of our method with the performance of other two-phase methods.

METHODS

A two-phase case-control study was simulated with a varying number of first- and second-phase subjects. Estimation methods: Method 1: Effect estimation with second-phase data only. Method 2: Effect estimation by adjusting the first-phase estimate with the difference between the adjusted and unadjusted second-phase estimate. The first-phase estimate is based on individual disease status and residential address for all study subjects that are linked to register data on NO2-exposure for each geographical area. Method 3: Effect estimation by using the expectation-maximization (EM) algorithm without taking area-level register data on exposure into account. Method 4: Effect estimation by using the EM algorithm and incorporating group-level register data on NO2-exposure.

RESULTS

The simulated scenarios were such that, unbiased or marginally biased (< 7%) odds ratio (OR) estimates were obtained with all methods. The efficiencies of method 4, are generally higher than those of methods 1 and 2. The standard errors in method 4 decreased further when the case/control ratio is above one in the second phase. For all methods, the standard errors do not become substantially reduced when the number of first-phase controls is increased.

CONCLUSION

In the setting described here, method 4 had the best performance in order to improve efficiency, while adjusting for varying participation rates across areas.

A planning tool for two-phase case-control studies

We present a software tool for planning two-phase case-control studies assuming categorical covariates. Two-phase designs, in which validation or complete information data are sampled stratified both on a dichotomous outcome and a covariate from a first-phase study with incomplete data, result in efficient estimates compared to standard designs. Efficiency and power depend on sample size, sampling fractions within each stratum x outcome cell, distributional assumptions and the regression model. Our software, called Two-Phase Planning Tool (or 2P Planning Tool), offers a graphical user interface (GUI) to organize and input the relevant anticipated entities and calculates a normed, expected two-phase case-control study. The 2P Planning Tool is especially helpful in selecting a stratification. The data are output into an Excel-sheet, which in turn can be read into a standard statistics package to perform "experimental" power calculations. Its use is illustrated by an example from epidemiology. Software for analyzing logistic two-phase studies is also provided.