Stratified Case-cohort Design Research Articles

Abstract 188: Association Between Uric Acid and Stroke in the REgards Case-cohort Study

Introduction: The association between hyperuricemia and stroke has been inconsistent especially after adjustment for cardiovascular risk factors. Previous studies were limited in the number of events and did not measure these associations in subgroups. Methods: We used a stratified case-cohort design within the REasons for Geographic and Racial Differences in Stroke study. Clinically adjudicated stroke events (n=903) were defined as focal neurologic deficit lasting >24 hours or non-focal neurological symptoms with brain imaging consistent with stroke. Controls (n=951) were a random stratified sample selected from baseline to ensure representation of high-risk groups. Uric acid was assayed using the baseline sample and categorized into three groups: < 6 (referent), 6-6.8, ≥ 6.8 mg/dl (hyperuricemia). Cox-proportional-hazard-models adjusted for demographic and clinical variables were fit to examine the association between uric acid levels and stroke. We repeated the analysis stratified by race, gender, and age (</≥ 65years) and finally, since uric acid is associated with hypertension, explored the count of antihypertensive treatment classes (nHTN) as a potential mediator. Results: Hyperuricemic individuals were more likely to be male and black. Hyperuricemia versus the referent category was significantly associated with stroke after adjustment for race, sex, age, and age*race interaction [model 1 HR(95%CI)= 1.42(1.12-1.80)], and after further adjustment for systolic and diastolic blood pressure [HR (95%CI)= 1.42(1.12-1.80)]. The associations was attenuated after full adjustment [HR (95%CI)= 1.22(0.91-1.63)]. Incremental adjustment by clinical variables suggested nHTN attenuated the association between hyperuricemia and stroke. Results were similar for hyperuricemia within subgroups defined by age, gender and race, except among men <65 years [HR (95%CI)= 2.72(1.25-5.93)]. Mediation analysis estimated that nHTN accounted for 43%(95%CI: 15 to 158%) of the relationship. Conclusion: We observed a significant association between hyperuricemia and stroke that was partially dependent upon hypertension severity (count of antihypertensive treatment classes). The association was strongest among males aged <65 years.

Ultraviolet radiation and skin cancer risk in offshore workers.

Excess skin cancer mortality and incidence have been reported among both land-based and offshore petroleum workers. The association between skin cancer and ultraviolet radiation (UVR) exposure has not been examined in these workers, although they have long off-duty periods and high average income that may allow travelling to sunny destinations. In addition, they have access to solariums free of charge on many accommodation platforms. To prospectively examine risk of incident cutaneous melanoma (CM) and non-melanoma skin cancer (NMSC), according to sun-tanning habits with adjustment for aromatic hydrocarbon exposure. A cohort of men employed offshore from 1965 to 1999 was linked through the Cancer Registry of Norway 1999-2012. Cox regression adapted to a stratified case-cohort design was used to estimate hazard ratios with 95% confidence intervals. The cohort included 24917 men. During 13.5 years of follow-up, 112 CMs and 70 NMSCs occurred. A positive dose-response relationship was seen between sunburn frequency and risk of CM (Ptrend < 0.05) and NMSC (Ptrend < 0.01). Solarium use both before and after age 20 was related to increased risk of NMSC. Sunscreen use was associated with increased risk of NMSC (Ptrend < 0.001). UVR exposure seems to be a significant contributor to the elevated risk of skin cancer observed in North Sea offshore workers. The positive association between solarium use and NMSC risk adds to the growing body of literature on artificial UV devices as carcinogenic.

A regularized variable selection procedure in additive hazards model with stratified case-cohort design

Case-cohort designs are commonly used in large epidemiological studies to reduce the cost associated with covariate measurement. In many such studies the number of covariates is very large. An efficient variable selection method is needed for case-cohort studies where the covariates are only observed in a subset of the sample. Current literature on this topic has been focused on the proportional hazards model. However, in many studies the additive hazards model is preferred over the proportional hazards model either because the proportional hazards assumption is violated or the additive hazards model provides more relevent information to the research question. Motivated by one such study, the Atherosclerosis Risk in Communities (ARIC) study, we investigate the properties of a regularized variable selection procedure in stratified case-cohort design under an additive hazards model with a diverging number of parameters. We establish the consistency and asymptotic normality of the penalized estimator and prove its oracle property. Simulation studies are conducted to assess the finite sample performance of the proposed method with a modified cross-validation tuning parameter selection methods. We apply the variable selection procedure to the ARIC study to demonstrate its practical use.

Recent progresses in outcome-dependent sampling with failure time data.

An outcome-dependent sampling (ODS) design is a retrospective sampling scheme where one observes the primary exposure variables with a probability that depends on the observed value of the outcome variable. When the outcome of interest is failure time, the observed data are often censored. By allowing the selection of the supplemental samples depends on whether the event of interest happens or not and oversampling subjects from the most informative regions, ODS design for the time-to-event data can reduce the cost of the study and improve the efficiency. We review recent progresses and advances in research on ODS designs with failure time data. This includes researches on ODS related designs like case-cohort design, generalized case-cohort design, stratified case-cohort design, general failure-time ODS design, length-biased sampling design and interval sampling design.

Improving the efficiency of estimation in the additive hazards model for stratified case-cohort design with multiple diseases

Improving the efficiency of estimation in the additive hazards model for stratified case-cohort design with multiple diseases

Abstract 13677: Left Atrial Structure and Function in Relation to Incident Stroke/TIA and Dementia: The Multi-ethnic Study of Atherosclerosis

Introduction: Increased left atrial (LA) size and reduced LA function are known predictors of atrial fibrillation (AF). Evidence also links increased LA size and stroke. We sought to examine the association of LA function, measured with cardiac magnetic resonance imaging (CMR) and incident stroke/TIA and dementia. Methods: This case-cohort study compared LA size and function in 180 MESA participants with incident Stroke/TIA and 95 with incident dementia, over a median follow up of 10.7 years, to 550 participants randomly selected from the whole MESA cohort. All individuals were free of clinical cardiovascular and major valvular heart disease. Phasic LA volumes, LA emptying fractions (LAEF) and peak strain were quantified in sinus rhythm using tissue-tracking CMR. Vascular neurologists adjudicated stroke events by medical record review. Dementia cases were identified using ICD9 codes from hospital records. Modified Cox proportional hazard models weighted for the stratified case-cohort sampling design was used to examine the associations. Results: Incident Stroke/TIA (age 67 ± 9 years, 45% male) and dementia cases (age 74 ± 6 years, 58% male) were older than the subcohort population (age: 61 ± 10, 49% male). In multivariable analysis adjusted for standard cardiovascular risk factors and left ventricular ejection fraction, lower peak LA strain and passive LAEF were associated with both incident stroke/TIA and dementia. After further adjustment for incident AF, the associations were attenuated for peak LA strain but remained significant for passive LAEF (HR for incident Stroke/TIA: 0.72 per SD 95% CI: 0.55-0.93, p=0.013 and for incident dementia: 0.59 per SD 95% CI: 0.37-0.95, p=0.031). Conclusions: Reduced LA conduit function was associated with incident Stroke/TIA and dementia independent of other cardiovascular risk factors and incident AF. Assessment of LA function may add further information in risk stratifying individuals at risk for stroke and dementia.

No increase in risk of hip fracture at high serum retinol concentrations in community-dwelling older Norwegians: the Norwegian Epidemiologic Osteoporosis Studies

Norway has the highest hip fracture rates worldwide and a relatively high vitamin A intake. Increased fracture risk at high intakes and serum concentrations of retinol (s-retinol) have been observed in epidemiologic studies. We aimed to study the association between s-retinol and hip fracture and whether high s-retinol may counteract a preventive effect of vitamin D. We conducted the largest prospective analysis of serum retinol and hip fracture to date in 21,774 men and women aged 65-79 y (mean age: 72 y) who attended 4 community-based health studies during 1994-2001. Incident hip fractures occurring up to 10.7 y after baseline were retrieved from electronic hospital discharge registers. Retinol determined by high-pressure liquid chromatography with ultraviolet detection in stored serum was available in 1154 incident hip fracture cases with valid body mass index (BMI) data and in a subcohort defined as a sex-stratified random sample (n = 1418). Cox proportional hazards regression weighted according to the stratified case-cohort design was performed. There was a modest increased risk of hip fracture in the lowest compared with the middle quintile of s-retinol (HR: 1.41; 95% CI: 1.09, 1.82) adjusted for sex and study center. The association was attenuated after adjustment for BMI and serum concentrations of α-tocopherol (HR: 1.16; 95% CI: 0.88, 1.51). We found no increased risk in the upper compared with the middle quintile. No significant interaction between serum concentrations of 25-hydroxyvitamin D and s-retinol on hip fracture was observed (P = 0.68). We found no evidence of an adverse effect of high serum retinol on hip fracture or any interaction between retinol and 25-hydroxyvitamin D. If anything, there tended to be an increased risk at low retinol concentrations, which was attenuated after control for confounders. We propose that cod liver oil, a commonly used food supplement in Norway, should not be discouraged as a natural source of vitamin D for fracture prevention.

Carboxymethyl lysine, an advanced glycation end product, and incident diabetes: a case-cohort analysis of the ARIC Study.

To verify whether elevated fasting levels of circulating carboxymethyl lysine (CML), an advanced glycation end product, predict the development of diabetes in middle-age adults. Using a stratified case-cohort design, we followed 543 middle-aged individuals who developed diabetes and 514 who did not over a median 9 years in the Atherosclerosis Risk in Communities Study. Weighted Cox proportional hazards analyses were used to account for the design. In weighted analyses, correlation between CML levels and anthropometric, inflammatory or metabolic variables was minimal (Pearson correlations usually < 0.10). CML, when modelled as a continuous variable and after adjustment for age, sex, race, centre, parental history of diabetes, BMI, waist-to-hip ratio, non-esterified fatty acids, oxidized LDL-cholesterol, GFR, smoking, an inflammation score, adiponectin, leptin, insulin and glucose levels, was associated with an increased risk of diabetes [Hazard ratio (HR) = 1.35; 95% confidence interval (CI) 1.09-1.67, for each 100 ng/ml CML increment]. Baseline glucose level and race each modified the association (P < 0.05 for interaction), which was present only among those with impaired fasting glucose (≥ 5.6 mmol/l, HR = 1.61, 95% CI 1.26-2.05) and among white participants (HR = 1.50, 95% CI 1.13-1.99). Elevated fasting CML, after adjustment for multiple risk factors for diabetes, predicts the development of incident diabetes, the association being present among those with impaired fasting glucose and in white participants. These prospective findings suggest that advanced glycation end products might play a role in the development of diabetes.

Improving the efficiency of estimation in the additive hazards model for stratified case-cohort design with multiple diseases.

The case-cohort study design has often been used in studies of a rare disease or for a common disease with some biospecimens needing to be preserved for future studies. A case-cohort study design consists of a random sample, called the subcohort, and all or a portion of the subjects with the disease of interest. One advantage of the case-cohort design is that the same subcohort can be used for studying multiple diseases. Stratified random sampling is often used for the subcohort. Additive hazards models are often preferred in studies where the risk difference, instead of relative risk, is of main interest. Existing methods do not use the available covariate information fully. We propose a more efficient estimator by making full use of available covariate information for the additive hazards model with data from a stratified case-cohort design with rare (the traditional situation) and non-rare (the generalized situation) diseases. We propose an estimating equation approach with a new weight function. The proposed estimators are shown to be consistent and asymptotically normally distributed. Simulation studies show that the proposed method using all available information leads to efficiency gain and stratification of the subcohort improves efficiency when the strata are highly correlated with the covariates. Our proposed method is applied to data from the Atherosclerosis Risk in Communities study.

Semiparametric Accelerated Failure Time Modeling for Clustered Failure Times From Stratified Sampling

Clustered failure times often arise from studies with stratified sampling designs where it is desired to reduce both cost and sampling error. Semiparametric accelerated failure time (AFT) models have not been used as frequently as Cox relative risk models in such settings due to lack of efficient and reliable computing routines for inferences. The challenge roots in the nonsmoothness of the rank-based estimating functions, and for clustered data, the asymptotic properties of the estimator from the weighted version have not been available. The recently proposed induced smoothing approach, which provides fast and accurate rank-based inferences for AFT models, is generalized to incorporate weights to accommodate stratified sampling designs. The estimator from the induced smoothing weighted estimating equations are shown to be consistent and have the same asymptotic distribution as that from the nonsmooth version, which has not been developed before. The variance of the estimator is estimated by computationally efficient sandwich estimators aided by a multiplier bootstrap. The proposed method is assessed in extensive simulation studies where the estimators appear to provide valid and efficient inferences. A stratified case-cohort design with clustered times to tooth extraction in a dental study illustrates the usefulness of the method.

Sample size/power calculation for stratified case-cohort design.

The case-cohort (CC) study design usually has been used for risk factor assessment in epidemiologic studies or disease prevention trials for rare diseases. The sample size/power calculation for a stratified CC (SCC) design has not been addressed before. This article derives such result based on a stratified test statistic. Simulation studies show that the proposed test for the SCC design utilizing small sub-cohort sampling fractions is valid and efficient for situations where the disease rate is low. Furthermore, optimization of sampling in the SCC design is discussed and compared with proportional and balanced sampling techniques. An epidemiological study is provided to illustrate the sample size calculation under the SCC design.

Intrinsic Subtypes from the PAM50 Gene Expression Assay in a Population-Based Breast Cancer Survivor Cohort: Prognostication of Short- and Long-term Outcomes

The PAM50, a gene expression assay to categorize breast tumors into intrinsic subtypes, has not been previously used to examine short- and long-term prognostication in a population-based cohort where treatment patterns and time of initial follow-up vary. In a stratified case-cohort design of 1,691 women from the LACE and Pathways breast cancer survivor cohorts, we used PAM50 to categorize tumors into Luminal A (LumA), Luminal B (LumB), HER2-enriched (HER2-E), Basal-like and Normal-like, and to examine risk of early and late recurrence and mortality by Cox proportional hazards regression. Compared with LumA, cumulative risk of recurrence and breast cancer death was higher for LumB, HER2-E, and Basal-like tumors at 2, 5, and 10 years. However, HR of breast cancer death varied over time [<5 years (early) vs. > 5 years (late)] for both Basal-like (HR, 6.23 early vs. HR, 0.63 late) and HER2-E tumors (HR, 2.97 early vs. HR, 0.73 late) but not for LumB tumors where risk was elevated consistently (HR, 2.67 early vs. HR, 1.47 late). The contrast between LumB, HER2-E, and Basal-like compared with LumA on early recurrence was stronger when subtype was defined by PAM50 than by immunohistochemistry (IHC) markers. The PAM50 categorized intrinsic subtypes in a manner that more accurately predicts recurrence and survival, especially for luminal tumors, compared with commonly used methods that rely on traditional IHC clinical markers. The PAM50 is robust for use in epidemiologic studies and should be considered when archived tumor tissues are available.

Race and breast cancer survival by intrinsic subtype based on PAM50 gene expression.

To evaluate whether differences in PAM50 breast cancer (BC) intrinsic (Luminal A, Luminal B, Basal-like, and HER2-enriched) subtypes help explain the Black-White BC survival disparity. Utilizing a stratified case-cohort design, this study included 1,635 women from the Pathways and Life After Cancer Epidemiology cohorts, selecting women with tumors based upon IHC classification, recurrences, and deaths.One millimeter punches were obtained from tumor tissue, and expression of the PAM50 genes for molecular subtype was determined by RT-qPCR of extracted RNA. Cox proportional hazards models were used to analyze associations between race and BC outcomes, adjusted for PAM50 BC subtype. All tests of statistical significance were two-sided. Black women had a higher prevalence of the Basal-like BC subtype. Adjusted for potential confounding variables and disease characteristics at diagnosis, Black women had higher risks of recurrence (HR 1.65, 95 % CI 1.06-2.57) and breast cancer-specific mortality (HR 1.71, 95 % CI 1.02-2.86) than White women, but adjusting further for subtype did not attenuate survival disparities. By contrast, Hispanic women had a lower risk of recurrence (HR 0.54, 95 % CI 0.30-0.96) than Whites. Among those with the Basal-like subtype, Black women had a similar recurrence risk as women in other race groups but a higher recurrence risk for all other subtypes. Hispanic women had a lower recurrence risk within each subtype, though associations were not significant, given limited power. Although Black women had a higher risk of the Basal-like subtype, which has poor prognosis, this did not explain the Black-White BC survival disparity.

Sample size/power calculation for stratified case-cohort design

Sample size/power calculation for stratified case-cohort design

Low Serum Levels of 25-Hydroxyvitamin D Predict Hip Fracture in the Elderly: A NOREPOS Study

Despite considerable interest, the relationship between circulating 25-hydroxyvitamin D and the risk of hip fracture is not fully established. The objective of the study was to study the association between serum 25-hydroxyvitamin D concentrations [s-25(OH)D] and the risk of hip fracture in Norway, a high-latitude country that has some of the highest hip fracture rates worldwide. A total of 21 774 men and women aged 65-79 years attended 4 community-based health studies during 1994-2001. Information on subsequent hip fractures was retrieved from electronic hospital discharge registers, with a maximum follow-up of 10.7 years. Using a stratified case-cohort design, s-25(OH)D was determined by HPLC-atmospheric pressure chemical ionization-mass spectrometry in stored serum samples in hip fracture cases (n = 1175; 307 men, 868 women) and in gender-stratified random samples (n = 1438). Cox proportional hazards regression adapted for the case-cohort design was performed. We observed an inverse association between s-25(OH)D and hip fracture; those with s-25(OH)D in the lowest quartile (<42.2 nmol/L) had a 38% [95% confidence interval (CI) 9-74%] increased risk of hip fracture compared with the highest quartile (≥67.9 nmol/L) in a model accounting for age, gender, study center, and body mass index. The association was stronger in men than in women: hazard ratio 1.65 (95% CI 1.04-2.61) vs hazard ratio 1.25 (95% CI 0.95-1.65). In this prospective case-cohort study of hip fractures, the largest ever reported, we found an increased risk of hip fracture in subjects in the lowest compared with the highest quartile of serum 25-hydroxyvitamin D. In accordance with the findings of previous community-based studies, low vitamin D status was a modest risk factor for hip fracture.

Asymptotic results for fitting marginal hazard models from stratified case-cohort studies with multiple disease outcomes

In stratified case-cohort designs, samplings of case-cohort samples are conducted via a stratified random sampling based on covariate information available on the entire cohort members. In this paper, we extended the work of Kang and Cai (2009) to a generalized stratified case-cohort study design for failure time data with multiple disease outcomes. Under this study design, we developed weighted estimating procedures for model parameters in marginal multiplicative intensity models and for the cumulative baseline hazard function. The asymptotic properties of the estimators are studied using martingales, modern empirical process theory, and results for finite population sampling.

Censored linear regression for case-cohort studies

SUMMARY Right-censored data from a classical case-cohort design and a stratified case-cohort design are considered. In the classical case-cohort design the subcohort is obtained as a simple random sample of the entire cohort, whereas in the stratified design this subcohort is elected by independent Bernoulli sampling with arbitrary selection probabilities. For each design and under a linear regression model, methods for estimating the regression parameters are proposed and analysed. These methods are derived by modifying the linear ranks tests and estimating equations that arise from full-cohort data using methods that are similar to the pseudolikelihood estimating equation that has been used in relative risk regression for these models. The estimators so obtained are shown to be consistent and asymptotically normal. Variance estimation and numerical illustrations are also provided.

Polymorphisms in Interleukin -2, -6, and -10 Are Not Associated with Gastric Cardia or Esophageal Cancer in a High-Risk Chinese Population

Polymorphisms in Interleukin -2, -6, and -10 Are Not Associated with Gastric Cardia or Esophageal Cancer in a High-Risk Chinese Population

Prospective study of serum retinol, beta-carotene, beta-cryptoxanthin, and lutein/zeaxanthin and esophageal and gastric cancers in China.

This study examined the relationship between pretrial serum concentrations of retinol, beta-carotene, beta-cryptoxanthin, and lutein/zeaxanthin and the subsequent risk of developing esophageal squamous cell carcinoma and gastric cardia or non-cardia adenocarcinoma in subjects selected from a randomized nutritional intervention trial in Linxian, China, a region with epidemic rates of esophageal and gastric cardia cancer. We used a stratified case-cohort design to select cohort members for inclusion in this study. In all we measured serum concentrations of the above vitamins in 590 esophageal, 395 gastric cardia, and 87 gastric non-cardia case subjects as well as in 1053 control subjects. Relative risks (RRs) were estimated using Cox proportional hazards models. Median values in our cohort were low for serum retinol (33.6 microg/dl), beta-carotene (4.3 microg/dl), and beta-cryptoxanthin (3.5 microg/dl), but were high for lutein/zeaxanthin (40.0 microg/dl). Gastric cardia cancer incidence fell 10% for each quartile increase in serum retinol (RR = 0.90, 95% CI = 0.83-0.99). For esophageal cancer, an inverse association with retinol levels was found only in male non-smokers (RR = 0.79 per quartile increase, 95% CI = 0.63-0.99). For gastric non-cardia cancer, an inverse association was limited to subjects 50 years old or younger (RR = 0.58 per quartile, 95% CI = 0.31-0.96). For beta-cryptoxanthin there was a borderline significant protective association for gastric non-cardia cancer (RR = 0.88 per quartile, 95% CI = 0.76-1.0). In contrast, we found the incidence of gastric non-cardia cancer increased (RR = 1.2 per quartile, 95% CI = 1.0-1.3) with increasing concentration of serum lutein/zeaxanthin. In this population, we found that low retinol and high lutein/zeaxanthin concentrations increased the risks of gastric cardia and gastric non-cardia cancer respectively. We found that there were no strong associations between any of the other analytes and any of the cancer sites.