Search

Nested case-control sampled event time data under a highly stratified proportional hazards model, in which the number of strata increases proportional to sample size, is described and analyzed. The data can be characterized as stratified sampling from the event time risk sets and the analysis approach of Borgan et al. (Ann Stat 23:1749-1778, 1995) is adapted to accommodate both the stratification and case-control sampling from the stratified risk sets. Conditions for the consistency and asymptotic normality of the maximum partial likelihood estimator are provided and the results are used to compare the efficiency of the stratified analysis to an unstratified analysis when the baseline hazards can be semi-parametrically modeled in two special cases. Using the stratified sampling representation of the stratified analysis, methods for absolute risk estimation described by Borgan et al. (1995) for nested case-control data are used to develop methods for absolute risk estimation under the stratified model. The methods are illustrated by a year of birth stratified analysis of radon exposure and lung cancer mortality in a cohort of uranium miners from the Colorado Plateau.

While the spatial resolution of exposure surfaces has greatly improved, our ability to locate people in space remains a limiting factor in accurate exposure assessment. In this case‐control study, two approaches to geocoding participant locations were used to study the impact of geocoding uncertainty on the estimation of ambient pesticide exposure and breast cancer risk among women living in California's Central Valley. Residential and occupational histories were collected and geocoded using a traditional point‐based method along with a novel area‐based method. The standard approach to geocoding uses centroid points to represent all geocoded locations, and is unable to adapt exposure areas based on geocode quality, except through the exclusion of low‐certainty locations. In contrast, area‐based geocoding retains the complete area to which an address matched (the same area from which the centroid is returned), and therefore maintains the appropriate level of precision when it comes to assessing exposure by geography. Incorporating the total potential exposure area for each geocoded location resulted in different exposure classifications and resulting odds ratio estimates than estimates derived from the centroids of those same areas (using a traditional point‐based geocoder). The direction and magnitude of these differences varied by pesticide, but in all cases odds ratios differed by at least 6% and up to 35%. These findings demonstrate the importance of geocoding in exposure estimation and suggest it is important to consider geocode certainty and quality throughout exposure assessment, rather than simply using the best available point geocodes., Key Points Area‐based geocoding includes the precise area of uncertainty on an address‐by‐address basis, which is not possible using current methodsGeocode positional error can bias exposure and associated risk estimates in unpredictable waysArea‐based geocoding incorporates heterogeneity of the exposure surface that likely plays a role in the magnitude and direction of biases

A standard approach to analysis of case-cohort data involves fitting log-linear models. In this paper, we describe how standard statistical software can be used to fit a broad class of general relative rate models to case-cohort data and derive confidence intervals. We focus on a case-cohort design in which a roster has been assembled and events ascertained but additional information needs to be collected on explanatory variables. The additional information is ascertained just for persons who experience the event of interest and for a sample of the cohort members enumerated at study entry. One appeal of such a case-cohort design is that this sample of the cohort may be used to support analyses of several outcomes. The ability to fit general relative rate models to case-cohort data may allow an investigator to reduce model misspecification in exposure-response analyses, fit models in which some factors have effects that are additive and others multiplicative, and facilitate estimation of relative excess risk due to interaction. We address model fitting for simple random sampling study designs as well as stratified designs. Data on lung cancer among radon-exposed men (Colorado Plateau uranium miners, 1950–1990) are used to illustrate these methods.

Cost-efficient sampling schemes for population-based case-control studies are necessary for sampling subjects in geographically dispersed populations where in-house surveys are expensive to conduct due to high interviewer travel costs that may be associated with simple random sampling. Motivated by the original study conducted by Cockburn et al. (2011) that investigated the relationship between exposure to pesticides and prostate carcinogenesis, a set of cluster-based individually matched case-control designs is presented for cost-efficient sampling of additional controls. Based on cluster sampling from the field of survey sampling, the case-control study designs presented, where one case is individually matched to three controls, use case-control status in the sampling of the primary sampling clusters. In the secondary stage, interviewer travel costs are reduced by subsampling additional controls within primary sampling clusters as opposed to selecting additional controls purely at random, which would be highly inefficient from a cost perspective. Compared to the simple random sampling (SRS) 1:1 and SRS 1:3 (one case matched to: n SRS control(s)) designs, computer simulations demonstrate that these cluster-based designs provide unbiased rate ratio estimation and statistical efficiencies that are no worse than the SRS 1:1 design and moderately less than the SRS 1:3 design. Even under situations where the intracluster correlation for the exposure variable is extremely high for the exposure of interest, the cluster-based designs have statistical efficiencies that are comparable to that of the SRS 1:1 design. Furthermore, a cost-efficiency analysis is presented that demonstrates that the cluster-based designs are more cost-efficient compared to the SRS 1:3 design.

Geolocation uncertainty is common in epidemiological studies that depend on addresses to determine exposure. We developed a spatial construct and statistical framework by which to characterize geolocation uncertainty, develop analysis methods, and compare those methods. Exposure is represented by a three-dimensional step function over the partitioned spatial surface and a person’s geolocation boundary is defined as the union of partition elements that cover the possible index location. Disease rates are defined by exposures from the partitioned surface. Standard process theory was used for analytic results and an empirical evaluation computer simulation method was used to compare methods. A case-control study of pesticide exposure and childhood cancer was used to illustrate the problem. Pesticide exposure was derived from geolocations determined from birth residences, where about half the reference addresses resolved to ZIP Codes while the rest resolve to smaller areas. We found that the centroid method has much worse power (0.35) to detect pesticide–disease effects than using either whole area proportion exposed (0.52) or when using an induced intensity approach (0.79). The latter approach properly accounted for the geolocation uncertainty even if only ZIP Code address information was used. ZIP Code address data had twice the variance compared to using the actual geolocation boundaries which had twice the variance compared to if there were no geolocation uncertainty. Our area based analytic approach confirms that geolocation ambiguity should be considered in the context of exposure–disease investigations that rely on address data for determining exposure.

The purpose of this study was to evaluate the effects of a hypnotically-based intervention for pain and fear in women undergoing labor who are about to receive an epidural catheter. A group of 155 women received interventions that included either (a) patient rocking, gentle touching, and hypnotic communication or (b) patient rocking, gentle touching, and standard communication. The authors found that the hypnotic communication intervention was more effective than the standard communication intervention for reducing both pain intensity and fear. The results support the use of hypnotic communication just before and during epidural placement for women who are in labor and also indicate that additional research to evaluate the benefits and mechanism of this treatment is warranted.

Because lifetime estrogen exposure is a key factor in breast cancer development, exposure to endocrine-disrupting pesticides might contribute to breast cancer development.1–3 While the estrogenic properties of some pesticides have been established,4–6 results from previous studies of pesticide exposure on breast cancer risk are conflicting: some studies show a positive association,7–11 while others are null.12–20 Many previous epidemiologic studies relied upon self-reported exposure based on pesticide usage, occupation, or living on a farm.11,14,17–20 Aside from a cohort study which did find some associations for specific pesticides,7,13 the majority of these studies grouped together pesticides with varying toxicologic effects, likely leading to nondifferential exposure misclassification and reported null effects that may have obscured associations with specific chemicals. Other studies used ecologic designs,16 spatial regression,8,9 or proximity to aggregated pesticide data at only one residential location.10,15,20 Studies that use biomarkers to measure pesticide metabolites in serum samples taken near the time of cancer diagnosis may not reflect previous exposures that are most relevant for breast cancer etiology and do not reflect long-term exposure.21 There is a need for research methodologies that can reconstruct exposures occurring decades before diagnosis and evaluate pesticide-specific exposures on breast cancer risk.22,23 We conducted a case–control study of breast cancer risk from exposure to pesticides using a Geographical Information Systems (GIS)–based method that combines geocoded residential and occupational histories with state pesticide use reports and land use data24 in California’s highest-ranking counties (Fresno, Tulare, and Kern) for agricultural density and commercial pesticide use in the United States.25 In highly agricultural regions, pesticide drift from neighboring application sites presents a major source of nonoccupational exposure.26–30 We evaluated a group of structurally and toxicologically similar31 organochlorine pesticides with known estrogenic effects that are most likely related to breast carcinogenesis because of their ability to accumulate in adipose tissue and potential to act synergistically (aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene).3,5,32–36 We also assessed breast cancer risk from exposure to three commonly applied pesticides in the region (chlorpyrifos, diazinon, and 1,3-dichloropropene) detected at levels of concern to human health in air monitoring conducted by the California Department of Pesticide Regulation (CDPR) in 2006 in a Fresno County farming community.37 Because no regulatory ambient air standards exist for most pesticides, CDPR developed health-based screening levels for 35 pesticides and found that diazinon exceeded its screening level, chlorpyrifos approached its screening level and was frequently detected, and 1,3-dichloropropene exceeded its cancer potency value.

To investigate whether the impact of postdiagnosis exercise on breast cancer outcomes in women diagnosed with early-stage breast cancer differs on the basis of tumor clinicopathologic and molecular features. Using a prospective design, 6,211 patients with early-stage breast cancer from two large population-based cohort studies were studied. Age-adjusted and multivariable Cox regression models were performed to determine the relationship between exercise exposure (total MET-hours/week) and recurrence and breast cancer–related death for: (i) all patients (“unselected” cohort), and on the basis of (ii) classic clinicopathologic features, (iii) clinical subtypes, (iv) PAM50-based molecular intrinsic subtypes, and (v) individual PAM50 target genes. After a median follow-up of 7.2 years, in the unselected cohort (n = 6,211) increasing exercise exposure was not associated with a reduction in the risk of recurrence (adjusted Ptrend = 0.60) or breast cancer–related death (adjusted Ptrend = 0.39). On the basis of clinicopathologic features, an exercise-associated reduction in breast cancer–related death was apparent for tumors

Purpose To determine whether insurance status, race, and ethnicity correlate with increased retinoblastoma invasiveness as a marker of both risk and time to diagnosis. Design Retrospective case-control study. Participants All 203 patients from the United States enrolled in the Children's Oncology Group (COG) trial ARET0332, a study of patients with unilateral retinoblastoma requiring enucleation. Main Outcome Measures All surgical specimens underwent pathologic review to determine the presence of well-defined histopathologic features correlating with a higher risk of disease progression. Insurance status, race, and ethnicity were compiled from the study record for each patient. Results On institutional pathologic review, nonprivate insurance, nonwhite race, and Hispanic ethnicity all correlated significantly with a greater rate of high-risk pathologic findings. Hispanic ethnicity remained a significant predictor on multivariate analysis. On central pathologic review, these correlations remained but did not reach statistical significance. The differences in results from institutional versus central pathologic reviews appeared to be due to a higher likelihood of patients in minority groups of being misclassified as high risk by institutional pathologists. Conclusions In this controlled study population of patients with retinoblastoma who had central pathologic review, our findings suggest a higher rate of more advanced disease associated with nonprivate insurance, nonwhite race, and Hispanic ethnicity; these findings may be due to delays in diagnosis for these groups. Future work should use direct methods to study the impact of other variables, including English-language proficiency and socioeconomic status. Further effort also should focus on where in the diagnostic process potential delays exist, so that interventions can be designed to overcome barriers to care for these groups. In addition, potential systematic differences in pathologic reads based on demographic variables deserve further study.

Positive detection of minimal residual disease (MRD) by multichannel flow cytometry (MFC) prior to hematopoietic cell transplantation (HCT) of patients with acute lymphoblastic leukemia (ALL) identifies patients at high risk for relapse, but many pre-HCT MFC-MRD negative patients also relapse, and the predictive power MFC-MRD early post-HCT is poor. To test whether the increased sensitivity of next-generation sequencing (NGS)-MRD better identifies pre- and post-HCT relapse risk, we performed immunoglobulin heavy chain (IgH) variable, diversity, and joining (V[D]J) DNA sequences J NGS-MRD on 56 patients with B-cell ALL enrolled in Children's Oncology Group trial ASCT0431. NGS-MRD predicted relapse and survival more accurately than MFC-MRD (P < .0001), especially in the MRD negative cohort (relapse, 0% vs 16%; P = .02; 2-year overall survival, 96% vs 77%; P = .003). Post-HCT NGS-MRD detection was better at predicting relapse than MFC-MRD (P < .0001), especially early after HCT (day 30 MFC-MRD positive relapse rate, 35%; NGS-MRD positive relapse rate, 67%; P = .004). Any post-HCT NGS positivity resulted in an increase in relapse risk by multivariate analysis (hazard ratio, 7.7; P = .05). Absence of detectable IgH-V(D)J NGS-MRD pre-HCT defines good-risk patients potentially eligible for less intense treatment approaches. Post-HCT NGS-MRD is highly predictive of relapse and survival, suggesting a role for this technique in defining patients early who would be eligible for post-HCT interventions. The trial was registered at www.clinicaltrials.gov as #NCT00382109.

Background: The incidence of melanoma is increasing worldwide. Guidelines for clinical skin exam for improving early diagnosis of melanoma remain inconsistent, and current data on factors associated with regular skin screening on a population basis are limited. Methods: We used self-reported data from 50,044 members of the California Twin Program, a population-based cohort of twins born in California between 1908 and 1982, to identify prevalence and determinants of recent clinical screening for skin cancer. Results: Prevalence of skin examination was higher than national estimates, with 32% of respondents of all ages reporting ever having skin examination. Sociodemographic and constitutional risk factors including white race, educational attainment, marital status, and number of large moles were strongly associated with recent screening, as were individual and family history of skin cancer. Lower socioeconomic status, racial/ethnic minority status, and paradoxically, frequent UV-related risk behaviors in adulthood were associated with a lower likelihood of recent screening. Conclusions: As the evidence concerning the efficacy of skin examination continues to evolve, attention should be paid to motivators and barriers of screening, particularly in high-risk subgroups where lack of screening may contribute to disparate rates of thicker melanomas and lower survival. Impact: Our results demonstrate the need for prevention strategies targeted to specific at-risk groups to increase earlier detection leading to improved outcomes. Cancer Epidemiol Biomarkers Prev; 24(8); 1–9. ©2015 AACR.

We previously showed that minimal residual disease (MRD) detection pre-hematopoietic cell transplant (HCT) and acute GvHD (aGvHD) independently predicted risk of relapse in pediatric ALL. In this study we further define risk by assessing timing of relapse and the effects of leukemia risk category and post-HCT MRD. By multivariate analysis, pre-HCT MRD

The incidence of melanoma is rising among Hispanic populations in the United States. The purpose of this study is to evaluate the impact of a pilot sun safety educational intervention conducted from 2006 to 2012 on Hispanic early adolescents in a high ultraviolet environment. Nineteen schools with high Hispanic enrollment were recruited from urban neighborhoods in Los Angeles. The analytic sample was restricted to students identifying as Hispanic or Latino (n = 777). A mixed effects linear model was used to test mean changes from pre- to posttest on students' sun protection knowledge, attitudes and behaviors. Significant improvements were observed across several cognitive outcomes related to sun protection, including knowledge of and attitudes toward sun protection and self-efficacy to wear sunscreen. However, changes in sun protective behaviors were not achieved. Although some improvements were observed, future studies should identify the factors that motivate sun protection in this population and develop tailored prevention strategies, as improving the sun safe behaviors of Hispanic youths may aid in reducing the risk of melanoma in adulthood in this population.

Idiopathic pneumonia syndrome (IPS) is an acute, noninfectious lung disorder associated with high morbidity and mortality after hematopoietic cell transplantation. Previous studies have suggested a role for TNFα in the pathogenesis of IPS. We report a multicenter phase II trial investigating a soluble TNF-binding protein, etanercept (Enbrel, Amgen, Thousand Oaks, CA), for the treatment of pediatric patients with IPS. Eligible patients were < 18 years old, within 120 days after transplantation, and with radiographic evidence of a diffuse pneumonitis. All patients underwent a pretherapy broncho-alveolor lavage (BAL) to establish the diagnosis of IPS. Systemic corticosteroids (2.0 mg/kg/day) plus etanercept (.4 mg/kg twice weekly × 8 doses) were administered. Response was defined as survival and discontinuation of supplemental oxygen support by day 28 of study. Thirty-nine patients (median age, 11 years; range, 1 to 17) were enrolled, with 11 of 39 patients nonevaluable because of identification of pathogens from their pretherapy BAL. In the remaining 28 patients, the median fraction of inspired oxygen at study entry was 45%, with 17 of 28 requiring mechanical ventilation. Complete responses were seen in 20 (71%) patients, with a median time to response of 10 days (range, 1 to 24). Response rates were higher for patients not requiring mechanical ventilation at study entry (100% versus 53%, P = .01). Overall survival at 28 days and 1 year after therapy were 89% (95% confidence interval [CI], 70% to 96%) and 63% (95% CI, 42% to 79%), respectively. Plasma levels of proinflammatory cytokines were significantly increased at onset of therapy, subsequently decreasing in responding patients. The addition of etanercept to high-dose corticosteroids was associated with high response rates and survival in children with IPS.

Sirolimus has activity against acute lymphoblastic leukemia (ALL) in xenograft models and efficacy in preventing acute graft-versus-host disease (aGVHD). We tested whether addition of sirolimus to GVHD prophylaxis of children with ALL would decrease aGVHD and relapse. Patients were randomized to tacrolimus/methotrexate (standard) or tacrolimus/methotrexate/sirolimus (experimental). The study met futility rules for survival after enrolling 146 of 259 patients. Rate of Grade 2-4 aGVHD was 31% vs 18% (standard vs experimental, P = .04), however, grade 3-4 aGVHD was not different (13% vs 10%, P = .28). Rates of veno-occlusive disease (VOD) and thrombotic microangiopathy (TMA) were lower in the nonsirolimus arm (9% vs 21% VOD, P = .05; 1% vs 10% TMA, P = .06). At 2 years, event free survival (EFS) and overall survival (OS) were 56% vs 46%, and 65% vs 55% (standard vs experimental), respectively (P = .28 and .23). Multivariate analysis showed increased relapse risk in children with ≥0.1% minimal residual disease (MRD) pretransplant, and decreased risk in patients with grades 1-3 aGVHD (P = .04). Grades 1-3 aGVHD were associated with improved EFS (P = .02), whereas grade 4 aGVHD and extramedullary disease at diagnosis led to inferior OS. Although addition of sirolimus decreased aGVHD, survival was not improved. This study is registered with ClinicalTrials.gov as #NCT00382109.

The role of SCT in pediatric oncology has continued to evolve with the introduction of new therapeutic agents and immunological insights into cancer. COG has focused its efforts on the study of hematopoietic stem cell transplantation in the treatment of pediatric malignancies in several major multi-institutional Phase II and Phase III studies. These studies include addressing the impact of allogenicity in ALL (ASCT0431), and establishing autologous stem cell transplant as the standard of care in neuroblastoma. Reducing transplant-associated toxicity was addressed in the ASCT0521 study, where the TNFα inhibitor etanercept was tested for the treatment of idiopathic pneumonia syndrome. Impact of cell dose was explored in the single versus tandem umbilical cord blood study CTN-0501, in close collaboration with the BMT-CTN.

The Children's Oncology Group conducted a multicenter Phase III trial for chronic graft-versus-host disease (cGVHD). The double-blind, placebo-controlled, randomized study evaluated hydroxychloroquine added to standard therapy for children with newly diagnosed cGVHD. The study also used a novel grading and response scoring system and evaluated clinical laboratory correlates of cGVHD. The primary endpoint was complete response (CR) after 9 months of therapy. Fifty-four patients (27 on each arm) were enrolled before closure because of slow accrual. The CR rate was 28% in the hydroxychloroquine arm versus 33% in the placebo arm (odds ratio [OR] = 0.77, 95% confidence interval [CI]: 0.20-2.93, P = .75) for 42 evaluable patients. For 41 patients with severity assessment at enrollment, 20 (49%) were severe and 18 (44%) moderate according to the National Institutes of Health Consensus Conference global scoring system. The CR rate was 15% for severe cGVHD and 44% for moderate cGVHD (OR = 0.24, 95% CI: 0.05-1.06, P = .07). Although the study could not resolve the primary question, it provided important information for future cGVHD study design in this population.

Purpose To evaluate a chemoreduction regimen using systemic vincristine and carboplatin (VC) and local ophthalmic therapies to avoid external-beam radiotherapy (EBRT) or enucleation in patients with Group B intraocular retinoblastoma. Patients and methods Twenty-one patients (25 eyes) were treated with six cycles of VC, accompanied by local ophthalmic therapies after cycle 1. The primary study objective was to determine the 2-year event-free survival (EFS) where an event was defined as the use of systemic chemotherapy in addition to vincristine or carboplatin, EBRT, and/or enucleation. Results All patients had tumor regression after the first cycle of VC and only two patients had progression during therapy. There were seven treatment failures within 2 years of study enrollment, resulting in 2-year EFS of 65% and early study closure in accordance with the statistical design. The 2-year cumulative incidence of enucleation was 15%; for external beam radiation therapy, it was 10%; and for chemotherapy to control progressive disease, it was 10%. All patients sustaining a treatment failure were salvaged with additional therapy. Conclusions For the majority of patients with Group B intraocular retinoblastoma, chemoreduction with VC, without etoposide, in conjunction with local therapy provides excellent opportunity for ocular salvage. Local therapy given with every chemotherapy cycle and incorporation of etoposide may provide improved ocular salvage rates. Central review of group at diagnosis is critical in assigning appropriate therapies.

Background Dactinomycin (AMD) and vincristine (VCR) have been used for the treatment of childhood cancer over the past 40 years but evidence-based dosing guidance is lacking.

Resume Objectif Observer les effets du balancement sur la douleur pendant le travail et durant la mise en place de l’analgesie peridurale. Type d’etude Etude clinique observationnelle. Patients et methodes Les scores ENS (0–10) de la douleur des contractions uterines ont ete releves chez 50 parturientes en position allongee, assise immobile, puis assise et se balancant d’avant en arriere. La satisfaction globale egalement mesuree par ENS et les commentaires libres sur le balancement ont ensuite ete recueillis. Resultats Une femme a refuse de se balancer, cinq ont alterne des phases d’immobilite et de balancement. La douleur des contractions uterines etait similaire entre la position immobile couchee (8,1 ± 1,8) et assise (8,0 ± 1,8), alors qu’elle a significativement diminue lors du balancement en position assise (6,6 ± 1,7 ; p Discussion Dans les limites que comporte une etude observationnelle et preliminaire, nous suggerons que le balancement est associe a une diminution significative de la douleur liee aux contractions uterines. La satisfaction des femmes vis-a-vis de cette methode a aussi ete importante. Plusieurs hypotheses sont proposees pour expliquer ces effets.

Purpose The severity of endometrial hyperplasia (EH)—simple (SH), complex (CH), or atypical (AH)—influences clinical management, but valid estimates of absolute risk of clinical progression to carcinoma are lacking. Materials and Methods We conducted a case-control study nested in a cohort of 7,947 women diagnosed with EH (1970-2002) at one prepaid health plan who remained at risk for at least 1 year. Patient cases (N = 138) were diagnosed with carcinoma, on average, 6 years later (range, 1 to 24 years). Patient controls (N = 241) were matched to patient cases on age at EH, date of EH, and duration of follow-up, and they were counter-matched to patient cases on EH severity. After we independently reviewed original slides and medical records of patient controls and patient cases, we combined progression relative risks (AH v SH, CH, or disordered proliferative endometrium [ie, equivocal EH]) from the case-control analysis with clinical censoring information (ie, hysterectomy, death, or left the health plan) on all cohort members to estimate interval-specific (ie, 1 to 4, 5 to 9, and 10 to 19 years) and cumulative (ie, through 4, 9, and 19 years) progression risks. Results For nonatypical EH, cumulative progression risk increased from 1.2% (95% CI, 0.6% to 1.9%) through 4 years to 1.9% (95% CI, 1.2% to 2.6%) through 9 years to 4.6% (95% CI, 3.3% to 5.8%) through 19 years after EH diagnosis. For AH, cumulative risk increased from 8.2% (95% CI, 1.3% to 14.6%) through 4 years to 12.4% (95% CI, 3.0% to 20.8%) through 9 years to 27.5% (95% CI, 8.6% to 42.5%) through 19 years after AH. Conclusion Cumulative 20-year progression risk among women who remain at risk for at least 1 year is less than 5% for nonatypical EH but is 28% for AH.

Given the greatly elevated risks of contralateral breast cancer (CBC) observed in breast cancer patients who carry mutations in BRCA1 and BRCA2, it is critical to determine the effectiveness of standard adjuvant therapies in preventing CBC in mutation carriers. The WECARE study is a matched, case-control study of 708 women with CBC as cases and 1,399 women with unilateral breast cancer (UBC) as controls, including 181 BRCA1/BRCA2 mutation carriers. Interviews and medical record reviews provided detailed information on risk factors and breast cancer therapy. All study participants were screened for BRCA1 and BRCA2 mutations using denaturing high-performance liquid chromatography (DHPLC) to detect genetic variants in the coding and flanking regions of the genes. Conditional logistic regression was used to compare the risk of CBC associated with chemotherapy and tamoxifen in BRCA1/BRCA2 mutation carriers and non-carriers. Chemotherapy was associated with lower CBC risk both in non-carriers (RR = 0.6 [95% CI: 0.5-0.7]) and carriers (RR = 0.5 [95% CI: 0.2-1.0]; P value = 0.04). Tamoxifen was associated with a reduced CBC risk in non-carriers (RR = 0.7 [95% CI: 0.6-1.0]; P value = 0.03). We observed a similar but non-significant reduction associated with tamoxifen in mutation carriers (RR = 0.7 [95% CI: 0.3-1.8]). The tests of heterogeneity comparing carriers to non-carriers did not provide evidence for a difference in the associations with chemotherapy (P value = 0.51) nor with tamoxifen (P value = 0.15). Overall, we did not observe a difference in the relative risk reduction associated with adjuvant treatment between BRCA1/BRCA2 mutation carriers and non-carriers. However, given the higher absolute CBC risk in mutation carriers, the potentially greater impact of adjuvant therapy in reducing CBC risk among mutation carriers should be considered when developing treatment plans for these patients.

Reproductive factors, such as early age at menarche, late age at menopause, and nulliparity are known risk factors for breast cancer. Previously, we reported these factors to be associated with risk of developing contralateral breast cancer (CBC). In this study, we evaluated the association between these factors and CBC risk among BRCA1 and BRCA2 (BRCA1/2) mutation carriers and non-carriers.The WECARE Study is a population-based multi-center case-control study of 705 women with CBC (cases) and 1,397 women with unilateral breast cancer (controls). All participants were screened for BRCA1/2 mutations and 181 carriers were identified. Conditional logistic regression models were used to evaluate associations between reproductive factors and CBC for mutation carriers and non-carriers.None of the associations between reproductive factors and CBC risk differed between mutation carriers and non-carriers. The increase in risk with younger age at menarche and decrease in risk in women with more than two full-term pregnancies seen in non-carriers were not significantly different in carriers (adjusted RRs = 1.31, 95% CI 0.65-2.65 and 0.53, 95% CI 0.19-1.51, respectively). No significant associations between the other reproductive factors and CBC risk were observed in mutation carriers or non-carriers.For two reproductive factors previously shown to be associated with CBC risk, we observed similar associations for BRCA1/2 carriers. This suggests that reproductive variables that affect CBC risk may have similar effects in mutation carriers and non-carriers.

In individually matched case–control studies, when some covariates are incomplete, an analysis based on the complete data may result in a large loss of information both in the missing and completely observed variables. This usually results in a bias and loss of efficiency. In this article, we propose a new method for handling the problem of missing covariate data based on a missing-data-induced intensity approach when the missingness mechanism does not depend on case–control status and show that this leads to a generalization of the missing indicator method. We derive the asymptotic properties of the estimates from the proposed method and, using an extensive simulation study, assess the finite sample performance in terms of bias, efficiency, and 95% confidence coverage under several missing data scenarios. We also make comparisons with complete-case analysis (CCA) and some missing data methods that have been proposed previously. Our results indicate that, under the assumption of predictable missingness, the suggested method provides valid estimation of parameters, is more efficient than CCA, and is competitive with other, more complex methods of analysis. A case–control study of multiple myeloma risk and a polymorphism in the receptor Inter-Leukin-6 (IL-6-α) is used to illustrate our findings.

The potential effects of oral contraceptive (OC) and post-menopausal hormone (PMH) use are not well understood among BRCA1 or BRCA2 deleterious mutation carriers with a history of breast cancer. In this study, we investigated the association between OC and PMH use and risk of contralateral breast cancer (CBC) in the WECARE (Women’s Environment, Cancer and Radiation Epidemiology) Study. The WECARE Study is a population-based case-control study of 705 women with asynchronous CBC and 1,398 women with unilateral breast cancer, including 181 BRCA1/2 mutation carriers. Risk factor information was assessed by telephone interview. Mutation status was measured using denaturing high-performance liquid chromatography followed by direct sequencing in all participants. Outcomes, treatment and tumor characteristics were abstracted from medical records. Ever use of OCs was not associated with risk among non-carriers (RR=0.87; 95% CI=0.66–1.15) or BRCA2 carriers (RR=0.82; 95% CI=0.21–3.13). BRCA1 carriers who used OCs had a non-significant greater risk than non-users (RR=2.38; 95% CI=0.72–7.83). Total duration of OC use and at least five years of use before age 30 were associated with a non-significant increased risk among mutation carriers, but not among non-carriers. Few women had ever used PMH and we found no significant associations between lifetime use and CBC risk among carriers and non-carriers. In conclusion, the association between OC/PMH use and risk of CBC does not differ significantly between carriers and non-carriers; however, because carriers have a higher baseline risk of second primaries even a potential small increase in risk as a result of OC use may be clinically relevant.

It has been recently asserted that the nested case-control study design, in which case-control sets are sampled from cohort risk sets, can introduce bias ("study design bias") when there are lagged exposures. The bases for this claim include a theoretical and an "empirical evaluation" argument. We examined both of these arguments and found them to be incorrect. We describe an appropriate empirical evaluation method to explore the performance of nested case-control study designs and analysis methods from an existing cohort. This empirical evaluation approach relies on simulating case-control outcomes from risk sets in the cohort from which the case-control study is to be performed. Because it is based on the underlying cohort structure, the empirical evaluation can provide an assessment that is tailored to the specific characteristics of the study under consideration. The methods are illustrated using samples from the Colorado Plateau uranium miners cohort.

An adequate depiction of exposure-time-response relationships is important in assessing public health implications of an occupational or environmental exposure. Recent advances have focused on flexible modeling of the overall shape of latency. Methods are needed to allow for varying shapes of latency under different exposure profiles. A tensor product spline model is proposed for describing exposure-response relationships for protracted time-dependent occupational exposure histories in epidemiologic studies. The methods use flexible multi-dimensional techniques to jointly model age, latency and exposure-response effects. In analyzing data from the Colorado Plateau Uranium Miners cohort, a model that allows for varying exposure-dependent latency shapes is found to be superior to models that only allowed for an overall latency curve. Specifically, the model suggests that, at low exposure levels risk increased at short latencies followed by a slow decline for longer latency periods. On the other hand, risk was higher but did not change much by latency for higher exposure levels. The proposed methodology has the advantage of allowing for latency functions that vary by exposure levels and, conversely, exposure-response relationships that are influenced by the latency structure.

Methods for the analysis of individually matched case-control studies with location-specific radiation dose and tumor location information are described. These include likelihood methods for analyses that just use cases with precise location of tumor information and methods that also include cases with imprecise tumor location information. The theory establishes that each of these likelihood based methods estimates the same radiation rate ratio parameters, within the context of the appropriate model for location and subject level covariate effects. The underlying assumptions are characterized and the potential strengths and limitations of each method are described. The methods are illustrated and compared using the WECARE study of radiation and asynchronous contralateral breast cancer.

Results from randomized trials indicate that treatment with tamoxifen or chemotherapy for primary breast cancer reduces the risk for contralateral breast cancer. However, less is known about how long the risk is reduced and the impact of factors such as age and menopausal status.The study included 634 women with contralateral breast cancer (case patients) and 1158 women with unilateral breast cancer (control subjects) from the Women's Environment, Cancer and Radiation Epidemiology Study. The women were younger than age 55 when they were first diagnosed with breast cancer during 1985-1999. Rate ratios (RRs) and 95% confidence intervals (CIs) for contralateral breast cancer after treatment with chemotherapy or tamoxifen were assessed by multivariable adjusted conditional logistic regression analyses.Chemotherapy was associated with a lower risk for contralateral breast cancer (RR = 0.57, 95% CI = 0.42 to 0.75) than no chemotherapy. A statistically significant association between chemotherapy and reduced risk for contralateral breast cancer persisted up to 10 years after the first breast cancer diagnosis and was stronger among women who became postmenopausal within 1 year of the first breast cancer diagnosis (RR = 0.28, 95% CI = 0.11 to 0.76). Tamoxifen use was also associated with reduced risk for contralateral breast cancer (RR = 0.66, 95% CI = 0.50 to 0.88) compared with no use, and the association was statistically significant for 5 years after the first diagnosis.The associations between chemotherapy and tamoxifen treatment and reduced risk for contralateral breast cancer appear to continue for 10 and 5 years, respectively, after the initial breast cancer is diagnosed. Ovarian suppression may have a role in the association between chemotherapy and reduced risk for contralateral breast cancer.

Classifying endometrial hyperplasia (EH) according to the severity of glandular crowding (simple hyperplasia (SH) vs complex hyperplasia (CH)) and nuclear atypia (simple atypical hyperplasia (SAH) vs complex atypical hyperplasia (CAH)) should predict subsequent endometrial carcinoma risk, but data on progression are lacking. Our nested case–control study of EH progression included 138 cases, who were diagnosed with EH and then with carcinoma (1970–2003) at least 1 year (median, 6.5 years) later, and 241 controls, who were individually matched on age, date, and follow-up duration and counter-matched on EH classification. After centralised pathology panel and medical record review, we generated rate ratios (RRs) and 95% confidence intervals (CIs), adjusted for treatment and repeat biopsies. With disordered proliferative endometrium (DPEM) as the referent, AH significantly increased carcinoma risk (RR=14, 95% CI, 5–38). Risk was highest 1–5 years after AH (RR=48, 95% CI, 8–294), but remained elevated 5 or more years after AH (RR=3.5, 95% CI, 1.0–9.6). Progression risks for SH (RR=2.0, 95% CI, 0.9–4.5) and CH (RR=2.8, 95% CI, 1.0–7.9) were substantially lower and only slightly higher than the progression risk for DPEM. The higher progression risks for AH could foster management guidelines based on markedly different progression risks for atypical vs non-atypical EH.

Computational methods, which can be implemented using standard Cox regression software, are given for fitting ''exact'' pseudolikehood estimates and robust and asymptotic variance estimators from case-cohort data. These methods are based on the computational approach of Therneau and Li [1999. Computing the Cox model for case cohort designs. Lifetime Data Anal. 5, 99-112] but will be less subject to small sample bias. Further, it is shown how to accommodate time-dependent covariates and estimate absolute risk. Extensions to stratified case-cohort sampled data are also provided. The methods are illustrated in analyses of case-cohort samples from a study of radiation exposure from fluoroscopy and breast cancer using SAS software.

Four case studies are presented to illustrate how information available on cohort members can be used to inform the control selection in epidemiologic case-control studies. The basic framework is the nested case-control paradigm and accompanying analysis methods. Empha- sis is on development of intuition for choosing study design candidates, the form of the estimators, and extensions of the basic theory to solve design and analysis problems.

In many epidemiological studies, disease occurrences and their rates are naturally modelled by counting processes and their intensities, allowing an analysis based on martingale methods. Applied to the Mantel-Haenszel estimator, these methods lend themselves to the analysis of general control selection sampling designs and the accommodation of time-varying exposures.

Informative sampling based on counter-matching risk set subjects on exposure correlated with a variable of interest has been shown to be an efficient alternative to simple random sampling; however, the opposite is true when correlation between the two covariates is absent. Thus, the counter-matching design will entail substantial gains in statistical efficiency compared to simple random sampling at a first stage of analyses focused by design on variables correlated with the counter-matching variable but will lose efficiency at a second stage of analyses aimed at variables independent of the counter-matching variable and not conceived as a part of the initial study. In an effort to recover efficiency in such second stage of analyses scenarios, we considered a naive analysis of the effect of a dichotomous covariate on the disease rates in the population that ignores the underlying counter-matching sampling design. We derive analytical expressions for the bias and variance and show that when the counter-matching and the new dichotomous variable of interest are uncorrelated and a multiplicative main effects model holds, such an analysis is advantageous over the standard “weighted” approach, especially when the counter-matching variable is rare and in such scenarios the efficiency gains exceeds 80%. Moreover, we consider all possible conceptual violations of the required assumptions and show that moderate departures from the above-mentioned requirements lead to negligible levels of bias; numerical values for the bias under common scenarios are provided. The method is illustrated via an analysis of BRCA1/2 deleterious mutations in the radiation treatment counter-matched WECARE study of second breast cancer.

Numerous epidemiologic studies have reported associations between breastfeeding and breast cancer risk (1–9). Breastfeeding is hypothesized to decrease breast cancer risk by reducing a woman’s lifetime number of menstrual cycles, and thus her cumulative exposure to endogenous hormones (10), and increasing the differentiation of ductal cells, making them less susceptible to carcinogens (11). Additionally, breastfeeding may facilitate excretion of carcinogens via the milk ducts (12,13). History of breastfeeding could also have a beneficial effect on breast cancer prognosis and survival and might be an important prognostic factor to consider when evaluating a newly diagnosed patient. However, studies to date have reported mixed results for mortality (14–16), and to our knowledge no studies have examined recurrence. Breast cancer is a heterogeneous disease with distinct molecular subtypes that appear to have different risk factor profiles, require different treatment regimens, and have considerable variability in prognosis. For example, history and duration of breastfeeding have been found to be associated with decreased risk of triple-negative or basal-like breast cancer, although not with ER+ or luminal A breast cancer (2,4,6,7,9,17). Recently, the PAM50 gene expression assay has been used to subtype breast tumors into luminal A, luminal B, human epidermal growth factor receptor 2 (HER2)–enriched, basal-like, and normal-like intrinsic subtypes (18). The assay has been shown to add valuable clinical information beyond standard clinical and pathological factors (estrogen receptor [ER], progesterone receptor [PR], HER2) used for risk assessment and therapy decision-making in breast cancer (19,20). Our study aims were to examine the associations of lifetime breastfeeding with breast cancer intrinsic subtype, as defined by the PAM50 assay, and the associations of lifetime breastfeeding with breast cancer recurrence and mortality overall and by subtype.

The case–cohort design includes a randomly sampled subcohort that is selected without regard to disease status or disease event times from the entire cohort, plus those cohort members who develop disease (cases) and who are not members of the subcohort. Selection of the subcohort offers several advantages, including: ability to gather data on subcohort members from the very beginning of the study and to implement quality control procedures immediately; ability to serve as a comparison group for a variety of disease endpoints; flexibility of analysis that comes from selection of the subcohort without matching on event times, including choice in the type of analysis model (e.g. semiparametric proportional hazards or Poisson regression models) and straightforward estimation of cumulative risk. We review analytical approaches for the case–cohort design and discuss its advantages and disadvantages, including issues of statistical efficiency, in comparison with the nested case–control design. Keywords: subcohort; case–cohort design; nested case–control design; cohort study

Objective: To investigate the associations of maternal and grandmaternal smoking before, during, and after pregnancy with childhood asthma. Design, setting, and participants: We conducted a case-control study nested within the Children’s Health Study in southern California. The case patients consisted of 338 children with asthma that had been diagnosed in the first 5 years of life, and 570 control subjects were countermatched on in utero exposure to maternal smoking within grade, sex, and community of residence. Measurements: Detailed maternal and household smoking histories and other asthma risk factor information was obtained by telephone interview. Results: The participation rates were 72.3% and 82.5%, respectively, for control subjects and case patients. In utero exposure to maternal smoking was associated with increased risk for asthma diagnosed in the first 5 years of life (odds ratio [OR], 1.5; 95% confidence interval [CI], 1.0 to 2.3), and for persistent asthma (OR, 1.5; 95% CI, 1.0 to 2.3). The associations did not differ in children with early transient asthma compared to those with early persistent asthma. Relative to never-smokers, children whose mothers smoked throughout the pregnancy had an elevated risk of asthma in the first 5 years of life (OR, 1.6; 95% CI, 1.0 to 2.6). Children of mothers who quit smoking prior to the pregnancy showed no increased risk (OR, 0.9; 95% CI, 0.5 to 1.5). We were unable to assess the association of smoking cessation during pregnancy because very few mothers were reported to have done so (15%). Asthma risk did not increase in a monotonic pattern with smoking intensity during pregnancy. Postnatal secondhand smoke exposure was not independently associated with asthma. Grandmaternal smoking during the mother’s fetal period was associated with increased asthma risk in her grandchildren (OR, 2.1; 95% CI, 1.4 to 3.2). Conclusions: Maternal and grandmaternal smoking during pregnancy may increase the risk of childhood asthma. (CHEST 2005; 127:1232–1241)

Early-life experiences and environmental exposures have been associated with childhood asthma. To investigate further whether the timing of such experiences and exposures is associated with the occurrence of asthma by 5 years of age, we conducted a prevalence case-control study nested within the Children's Health Study, a population-based study of > 4,000 school-aged children in 12 southern California communities. Cases were defined as physician-diagnosed asthma by age 5, and controls were asthma-free at study entry, frequency-matched on age, sex, and community of residence and countermatched on in utero exposure to maternal smoking. Telephone interviews were conducted with mothers to collect additional exposure and asthma histories. Conditional logistic regression models were fitted to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Asthma diagnosis before 5 years of age was associated with exposures in the first year of life to wood or oil smoke, soot, or exhaust (OR = 1.74; 95% CI, 1.02-2.96), cockroaches (OR = 2.03; 95% CI, 1.03-4.02), herbicides (OR = 4.58; 95% CI, 1.36-15.43), pesticides (OR = 2.39; 95% CI, 1.17-4.89), and farm crops, farm dust, or farm animals (OR = 1.88; 95% CI, 1.07-3.28). The ORs for herbicide, pesticide, farm animal, and crops were largest among children with early-onset persistent asthma. The risk of asthma decreased with an increasing number of siblings (ptrend = 0.01). Day care attendance within the first 4 months of life was positively associated with early-onset transient wheezing (OR = 2.42; 95% CI, 1.28-4.59). In conclusion, environmental exposures during the first year of life are associated with childhood asthma risk.

Some experimental and epidemiologic evidence suggests that residential exposure to power-frequency magnetic fields can increase breast cancer risk. This association was investigated in a nested case-control study of female breast cancer within a cohort of African Americans, Latinas, and Caucasians in Los Angeles County, California. Incident breast cancer was ascertained from 1993 to 1999 by linkage to county and state tumor registries. Controls were selected from a random sample of cohort members without breast cancer at baseline. Exposure was assessed in 1995-2001 by means of wiring configuration coding (an indirect measure of magnetic field exposure that has been associated with increased risk of childhood leukemia in Los Angeles and elsewhere in North America) in all homes occupied over the previous 10 years for 743 cases and 699 controls and by measurement of magnetic fields in the bedroom over a 7-day period for 347 cases and 286 controls. The estimated risk of breast cancer was not higher among women with wiring configuration codes associated with the highest magnetic fields (for a very high current configuration relative to very low, the adjusted odds ratio was 0.76 (95% confidence interval: 0.49, 1.18)). Stronger measured fields were not significantly associated with increased risk. These data suggest that residential magnetic field exposures commonly experienced by US women do not influence risk of breast cancer.

The asymptotic variance and robust variance estimators of rate ratios estimated using conditional logistic regression from individually-matched case-control data are derived when the presumed proportional hazards model ismisspecified. The robust variance estimators are easily computed using Schoenfeld residuals generated from standard partial likelihood estimation software for failure time data. Simulation studies indicate that the robust variance estimators perform well for typical sizes and that the 'rare disease' version should be adequate for all practical purposes. It was also found that model misspecification must be quite extreme before the model-based, i.e. inverse information, variance is significantly biased and that the robust variance estimators are somewhat more variable than the model-based. We conclude that the model-based variance estimator can be used when model misspecification is not severe. The robust estimator should be used when the presumed model clearly fits the data poorly.

In a family-matched case-control study, a population-based sample of cases is selected from a well-defined geographic region over a fixed period of time. For diseases of adult onset, the control is generally a sibling or cousin who is matched on sex and age without regard to location of residence. Such a design can lead to biased estimates of environmental relative risk if the prevalence of an environmental risk factor varies by the geographic region from which the cases and controls are drawn. However, assuming the independence of genotype and environmental exposure, the estimators for the gene and gene-environment interaction effects are consistent. This suggests that we must use caution in interpreting parameters that estimate environmental main effects from a family-based case-control study if controls are selected from outside the case-ascertainment region.

The purpose of this study was to evaluate the outcome of radical prostatectomy alone and compare it with that of surgery followed by planned adjuvant radiotherapy in patients with pT3N0 prostate cancer (CaP). A total of 402 patients with CaP were treated with prostatectomy, including 311 (77%) who received a planned course of adjuvant radiotherapy (RT) (surgery [S] + RT) to the prostatic fossa (median dose: 48 Gy) and 91 (23%) who had surgery alone. Patients in the former group had worse risk factors than those in the latter group, such as a higher clinical and pathologic stage (p = 0.001), higher Gleason score (p = 0.09), and higher preoperative prostate-specific antigen (PSA) level (p = 0.0001). PSA failure was defined as more than 0.05 ng/ml. Median follow-up was 59 months. The 5- and 10-year overall survival for the 311 S+RT patients was 91% and 81%, respectively, and it was similar for those 91 in the surgery-alone group, p = 0.59. The 5- and 10-year probability of freedom from PSA and/or clinical failure for the former group was 70% and 53%, respectively, whereas it was 66% and 46%, respectively, for the latter group, p = 0.72. Any recurrence developed in a total of 96 (31%) patients in the S+RT group as compared with 23 (25%) in the surgery-alone group. Local recurrence was noted in 10 (3.2%) S+RT and in 6 (6.6%) surgery-alone patients (N.S.). The time to clinical or chemical recurrence was also similar for both treatment groups (median time: 3.0 versus 3.8 years). Patients with pT3b tumors had relatively poor 5- and 10-year disease-free survival (53% and 32%, respectively, for S+RT and 38% and 0%, respectively, for surgery alone, p = 0.82). In multivariate analyses, pathologic stage and Gleason score were independent predictors of recurrence, each with p0.001 after controlling for the other. The worst prognostic category included patients with pT3bN0, Gleason score 7-10 disease who had 5.0 times the risk of recurrence as compared with pT3aN0, Gleason score 2-6 patients. No significant difference in disease-free survival by the treatment group was seen in Cox regression analysis controlling for pathologic stage (p = 0.59), Gleason score (p = 0.99), and PSA (p = 0.28). S+RT patients were predicted to have disease recurrence at 83% the rate of surgery-alone patients, p = 0.42. Preoperative PSA (25 ng/ml) was predictive of recurrence (2.0 x risk) in univariate analysis, but it was not a significant predictor in multivariate analysis. It appears that moderate-dose, localized fields postoperative irradiation reduced the incidence of local recurrence in patients who were at a higher risk of recurrence as compared with those treated with surgery alone. New treatment strategies need to be developed to manage pT3bN0, Gleason score 7-10 patients whose 10-year disease-free survival was poor.

BACKGROUND Different approaches have been proposed to investigate latency in epidemiologic studies where detailed exposure histories are available. METHODS We demonstrate the application of a flexible, yet parsimonious, spline function model to investigate latency patterns for radon progeny exposure and lung cancer in the Colorado Plateau uranium miners cohort. The model extends a previously proposed bilinear model. RESULTS The excess relative risk (ERR) reached a maximum of 0.6 per 100 working level months, for exposures received 14 years previously. The ERR then declined, and was estimated to approach zero for exposures received 35 years and more in the past. The point-wise 95% confidence intervals supported ERRs > 0 for the period 9-32 years before the event. The estimated latency curve was homogeneous across categories of attained age, duration of exposure, rate of exposure, and smoking. CONCLUSIONS The proposed spline model is a flexible tool for latency analyses, and extends previously used methods.

The interest in studying gene-environment interaction is increasing for complex diseases. However, most methods of detecting gene-environment interactions may not be appropriate for the study of interactions involving rare genes (G:) or uncommon environmental exposures (E:), because of poor statistical power. To increase this power, the authors propose the counter-matching design. This design increases the number of subjects with the rare factor without increasing the number of measurements that must be performed. In this paper, the efficiency and feasibility (required sample sizes) of counter-matching designs are evaluated and discussed. Counter-matching on both G: and E: appears to be the most efficient design for detecting gene-environment interaction. The sensitivity and specificity of the surrogate measures, the frequencies of G: and E:, and, to a lesser extent, the value of the interaction effect are the most important parameters for determining efficiency. Feasibility is also more dependent on the exposure frequencies and the interaction effect than on the main effects of G: and E: Although the efficiency of counter-matching is greatest when the risk factors are very rare, the study of such rare factors is not realistic unless one is interested in very strong interaction effects. Nevertheless, counter-matching appears to be more appropriate than most traditional epidemiologic methods for the study of interactions involving rare factors.

We explore the question of how to characterize a factor that would explain the patterns of wire code-childhood leukemia association observed in three major United States case-control studies. In positive studies, such a factor needs to be more prevalent in "high" wire codes than "low" whereas in negative studies, the factor should not be correlated to wire code. These concepts are quantified and we use these results to characterize potential factors that might explain the wire code-childhood leukemia association. We then re-evaluate findings from a survey of correlates of wire code in Connecticut with respect to potential for explaining the wire code-leukemia association. Under the presumption that the Connecticut correlations apply to the populations in which a wire code-leukemia association has been observed, only age and type of home are correlated enough to be potential explanatory factors. In particular, it appears very unlikely that traffic density can explain a significant portion of the wire code association. We suggest that correlation studies can be useful for identifying potential explanatory factors, but they must be done in regions where a wire code-childhood leukemia association has been observed.

We compare the asymptotic relative efficiency (ARE) of different study designs for estimating gene and gene-environment interaction effects using matched case-control data. In the sampling schemes considered, cases are selected differentially based on their family history of disease. Controls are selected either from unrelated subjects or from among the case's unaffected siblings and cousins. Parameters are estimated using weighted conditional logistic regression, where the likelihood contributions for each subject are weighted by the fraction of cases sampled sharing the same family history. Results showed that compared to random sampling, over-sampling cases with a positive family history increased the efficiency for estimating the main effect of a gene for sib-control designs (103-254% ARE) and decreased efficiency for cousin-control and population-control designs (68-94% ARE and 67-84% ARE, respectively). Population controls and random sampling of cases were most efficient for a recessive gene or a dominant gene with an relative risk less than 9. For estimating gene-environment interactions, over-sampling positive-family-history cases again led to increased efficiency using sib controls (111-180% ARE) and decreased efficiency using population controls (68-87% ARE). Using case-cousin pairs, the results differed based on the genetic model and the size of the interaction effect; biased sampling was only slightly more efficient than random sampling for large interaction effects under a dominant gene model (relative risk ratio = 8, 106% ARE). Overall, the most efficient study design for studying gene-environment interaction was the case-sib-control design with over-sampling of positive-family-history-cases.

Many epidemiologists and statisticians believe that the odds ratio is the only measure that can be reliably estimated from case-control studies. I have received many reviews of case-control study papers instructing us to change “rate ratio” to “odds ratio” when, in fact, it was the rate ratio that we had estimated. Further, in discussions about methods we have developed to estimate absolute risk measures from nested case-control studies, I have found that many are surprised to learn that absolute risk can be estimated systematically and reliably in the case-control setting. While case-control studies are best suited for estimation of relative measures, and I do not wish to minimize the challenges of estimation on other scales from case-control data generally, there are reliable methods for doing so. Reporting of absolute risk estimates seems desirable to supplement the usual case-control relative measure analyses, 1 but this is only rarely done even when it is feasible. So why is there an odds-ratio fixation? I believe the core problem is that epidemiologists often think of case-control studies as a “retrospective model.” According to this view, we start with a set of cases and controls. Then the covariates (exposures and other factors) occur as independent realizations with distribution dependent on disease status. This is in contrast to the “prospective model” of cohort data, in which we start with a group of subjects with given covariates, and disease status is the result of independent realizations with probability dependent on the covariate values. Students learn that the odds ratio parameters in the retrospective logistic model are same as the odds ratio parameters in the corresponding prospective logistic model, but that other measures do not translate. The matter is further confused because valid estimation of odds ratio parameters from retrospective model casecontrol data may be obtained using the corresponding prospective cohort data logistic regression, with the estimated “baseline odds” a nuisance parameter to be ignored. 2,3 But case-control studies are not backwards cohort (“trohoc”) studies. 4 A more realistic way to represent case-control designs is as sampling from a prospective cohort that depends on disease outcomes and other information available on cohort subjects— what we have called the nested case-control model. 5 While this representation is certainly not new, and is often used in epidemiology textbooks as a “conceptual framework” to think about basic sampling and bias issues, almost invariably the retrospective model approach is used to develop the analysis methods. The alternative approach my colleagues and I have taken is to develop case-control study methods based completely on the nested case-control model, and to provide a unifying framework across cohort and case-control analysis methods, as well as across individually matched and unmatched case-control study designs. While there are still some important gaps to be filled, we have made progress. After