Robustness of risk-based allocation of resources for disease prevention.

Risk models for disease incidence can be useful for allocating resources for disease prevention if risk assessment is not too expensive. Assume there is a preventive intervention that should be given to everyone, but preventive resources are limited. We optimize risk-based prevention strategies and investigate robustness to modeling assumptions. The optimal strategy defines the proportion of the population to be given risk assessment and who should be offered intervention. The optimal strategy depends on the ratio of available resources to resources needed to intervene on everyone, and on the ratio of the costs of risk assessment to intervention. Risk assessment is not recommended if it is too expensive. Preventive efficiency decreases with decreasing compliance to risk assessment or intervention. Risk measurement error has little effect nor does misspecification of the risk distribution. Ignoring population substructure has small effects on optimal prevention strategy but can lead to modest over- or under-spending. We give conditions under which ignoring population substructure has no effect on optimal strategy. Thus, a simple one-population model offers robust guidance on prevention strategy but requires data on available resources, costs of risk assessment and intervention, population risk distribution, and probabilities of acceptance of risk assessment and intervention.

The Need for Combined Assessment of Multiple Outcomes in Noninferiority Trials in Oncology.

Noninferiority trials in oncology assess novel therapies with the potential for slightly worse recurrence or death outcomes (ie, the margin of noninferiority) than standard therapies. This poses a dilemma because, in the absence of potential health outcome advantages, these trials may not provide the treatment equipoise required for an ethical study. Any new treatment with the potential for slightly worse recurrence or death outcomes should have countervailing health outcome advantages, but these are rarely taken into account in the design of noninferiority trials. This article presents the argument that not only the potentially worse health outcomes but also the potential benefits of the novel therapy should be considered when designing, analyzing, and reporting noninferiority trials. Some approaches to study design and analysis that consider both primary and secondary end points are discussed, and reporting the joint distributions of end points for the novel and standard treatments is recommended.

Associations between cancer and Parkinson's disease in U.S. elderly adults.

BACKGROUND: Several studies suggest that cancer is reduced before and after a Parkinson's disease (PD) diagnosis. However, determining relationships among diseases of ageing is challenging due to possible biases in ascertaining disease. This study evaluates the PD and cancer relationship, addressing potential biases. METHODS: Using Surveillance, Epidemiology, and End Results-Medicare linked data (1992-2005) of adults ≥ 65 years, we assessed PD risk after cancer comparing PD in 743 779 cancer patients with PD in a non-cancer group (n = 419 432) in prospective cohort analyses. We also conducted a case-control study of 836 947 cancer cases and 142 869 controls to assess cancer following PD. We applied Cox proportional hazards models to estimate hazards ratios (HRs) for PD after cancer and unconditional logistic regression to estimate odds ratios (ORs) for PD preceding cancer, controlling for physician visits and other factors. To explore biases in ascertaining cancer, we examined relationships between cancer and automobile accident injuries, which we expected to be null. RESULTS: No association was observed between cancer and subsequent PD [HR = 0.97; 95% confidence interval (CI) = 0.92-1.01] nor between cancer and subsequent automobile injuries (HR = 1.03; 95% CI = 0.98-1.07). One site, lung cancer, was associated with subsequent reduced PD, which may reflect confounding by smoking. In the case-control analysis, PD was associated with reduced subsequent cancer, overall (OR = 0.77; 95% CI = 0.71-0.82) and for several cancer sites. However, the automobile injury/ subsequent cancer association was similar (OR = 0.83; 95% CI = 0.78-0.88), suggesting a cancer detection bias after serious health outcomes. CONCLUSIONS: In totality, our data do not support a biological relationship between PD and cancer.

Piecewise exponential survival times and analysis of case-cohort data.

Case-cohort designs select a random sample of a cohort to be used as control with cases arising from the follow-up of the cohort. Analyses of case-cohort studies with time-varying exposures that use Cox partial likelihood methods can be computer intensive. We propose a piecewise-exponential approach where Poisson regression model parameters are estimated from a pseudolikelihood and the corresponding variances are derived by applying Taylor linearization methods that are used in survey research. The proposed approach is evaluated using Monte Carlo simulations. An illustration is provided using data from the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study of male smokers in Finland, where a case-cohort study of serum glucose level and pancreatic cancer was analyzed.

Assessment of the human faecal microbiota: II. Reproducibility and associations of 16S rRNA pyrosequences.

BACKGROUND: We conducted a pilot study of reproducibility and associations of microbial diversity and composition in faecal microbial DNA. METHODS AND RESULTS: Participants (25 men and 26 women, aged 17-65 years) provided questionnaire data and multiple samples of one stool collected with two Polymedco and two Sarstedt devices preloaded with RNAlater. 16S rRNA genes in each faecal DNA aliquot were amplified, sequenced (Roche/454 Life Sciences) and assigned to taxa. Devices were compared for ease of use and reproducibility [intraclass correlation coefficient (ICC)] between duplicate aliquots on diversity and taxonomic assignment. Associations were tested by linear regression. Both collection devices were easy to use. Both alpha diversity (Shannon index) and beta diversity (UniFrac) were higher between than within duplicates (P ≤ 10(-8) ) and did not differ significantly by device (P ≥ 0·62). Reproducibility was good (ICC≥0·77) for alpha diversity and taxonomic assignment to the most abundant phyla, Firmicutes and Bacteroidetes (71·5% and 25·0% of sequences, respectively), but reproducibility was low (ICC≤0·48) for less abundant taxa. Alpha diversity was lower with nonantibiotic prescription medication (P = 0·02), with younger age (P = 0·03) and marginally with higher body mass index (P = 0·08). CONCLUSIONS: With sampling from various parts of a stool, both devices provided good reproducibility on overall microbial diversity and classification for the major phyla, but not for minor phyla. Implementation of these methods should provide insights into how broad microbial parameters, but not necessarily rare microbes, affect risk of various conditions.

Assessment of the human faecal microbiota: I. Measurement and reproducibility of selected enzymatic activities.

BACKGROUND: The intestinal microbial community has major effects on human health, but optimal research methods are unsettled. To facilitate epidemiologic and clinical research, we sought to optimize conditions and to assess reproducibility of selected core functions of the distal gut microbiota, ß-glucuronidase and ß-glucosidase bioactivities. METHODS AND RESULTS: A colorimetric kinetic method was optimized and used to quantify activities of ß-glucuronidase and ß-glucosidase in human faeces. Enzyme detection was optimal with neutral pH, snap freezing in liquid nitrogen and rapid thawing to 37 °C before protein extraction. Enzymatic stability was assessed by delayed freezing for 2-48 h to mimic field settings. Activities decayed approximately 20% within 2 h and 40% within 4 h at room temperature. To formally assess reproducibility, 51 volunteers (25 men; mean age 39) used two devices to self-collect and rapidly chill four replicates of a stool. Devices were compared for mean enzymatic activities and intraclass correlation coefficients (ICC) in paired replicates of the self-collected specimens. Reproducibility was excellent with both devices for ß-glucuronidase (ICC 0·92). The larger collection device had significantly higher reproducibility for ß-glucosidase (ICC 0·92 vs. 0·76, P < 0·0001) and higher mean activities for both enzymes (P < 0·0001). CONCLUSIONS: Optimal measurement of these core activities of the microbiota required a sufficient quantity of rapidly chilled or frozen specimens collected in phosphate buffered saline at pH7·0. Application of these methods to clinical and epidemiologic research could provide insights on how the intestinal microbiota affects human health.

Benefit/risk assessment for breast cancer chemoprevention with raloxifene or tamoxifen for women age 50 years or older.

PURPOSE: The Study of Tamoxifen and Raloxifene (STAR) demonstrated that raloxifene was as effective as tamoxifen in reducing the risk of invasive breast cancer (IBC) in postmenopausal women and had lower risks of thromboembolic events, endometrial cancer, and cataracts but had a nonstatistically significant higher risk of noninvasive breast cancer. There is a need to summarize the risks and benefits of these agents. PATIENTS AND METHODS: Baseline incidence rates of IBC and other health outcomes, absent raloxifene and tamoxifen, were estimated from breast cancer chemoprevention trials; the Surveillance, Epidemiology and End Results Program; and the Women's Health Initiative. Effects of raloxifene and tamoxifen were estimated from STAR and the Breast Cancer Prevention Trial. We assigned weights to health outcomes to calculate the net benefit from raloxifene compared with placebo and tamoxifen compared with placebo. RESULTS: Risks and benefits of treatment with raloxifene or tamoxifen depend on age, race, breast cancer risk, and history of hysterectomy. Over a 5-year period, postmenopausal women with an intact uterus had a better benefit/risk index for raloxifene than for tamoxifen. For postmenopausal women without a uterus, the benefit/risk ratio was similar. The benefits and risks of raloxifene and tamoxifen are described in tables that can help identify groups of women for whom the benefits outweigh the risks. CONCLUSION: We developed a benefit/risk index to quantify benefits from chemoprevention with tamoxifen or raloxifene. This index can complement clinical evaluation in deciding whether to initiate chemoprevention and in comparing the benefits and risks of raloxifene versus tamoxifen.

Confirmation of family cancer history reported in a population-based survey.

BACKGROUND: Knowledge of family cancer history is essential for estimating an individual's cancer risk and making clinical recommendations regarding screening and referral to a specialty cancer genetics clinic. However, it is not clear if reported family cancer history is sufficiently accurate for this purpose. METHODS: In the population-based 2001 Connecticut Family Health Study, 1019 participants reported on 20 578 first-degree relatives (FDR) and second-degree relatives (SDR). Of those, 2605 relatives were sampled for confirmation of cancer reports on breast, colorectal, prostate, and lung cancer. Confirmation sources included state cancer registries, Medicare databases, the National Death Index, death certificates, and health-care facility records. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for reports on lung, colorectal, breast, and prostate cancer and after stratification by sex, age, education, and degree of relatedness and used to estimate report accuracy. Pairwise t tests were used to evaluate differences between the two strata in each stratified analysis. All statistical tests were two-sided. RESULTS: Overall, sensitivity and positive predictive value were low to moderate and varied by cancer type: 60.2% and 40.0%, respectively, for lung cancer reports, 27.3% and 53.5% for colorectal cancer reports, 61.1% and 61.3% for breast cancer reports, and 32.0% and 53.4% for prostate cancer reports. Specificity and negative predictive value were more than 95% for all four cancer types. Cancer history reports on FDR were more accurate than reports on SDR, with reports on FDR having statistically significantly higher sensitivity for prostate cancer than reports on SDR (58.9% vs 21.5%, P = .002) and higher positive predictive value for lung (78.1% vs 31.7%, P < .001), colorectal (85.8% vs 43.5%, P = .004), and breast cancer (79.9% vs 53.6%, P = .02). CONCLUSIONS: General population reports on family history for the four major adult cancers were not highly accurate. Efforts to improve accuracy are needed in primary care and other health-care settings in which family history is collected to ensure appropriate risk assessment and clinical care recommendations.

Cancer burden in the HIV-infected population in the United States.

BACKGROUND: Effective antiretroviral therapy has reduced the risk of AIDS and dramatically prolonged the survival of HIV-infected people in the United States. Consequently, an increasing number of HIV-infected people are at risk of non-AIDS-defining cancers that typically occur at older ages. We estimated the annual number of cancers in the HIV-infected population, both with and without AIDS, in the United States. METHODS: Incidence rates for individual cancer types were obtained from the HIV/AIDS Cancer Match Study by linking 15 HIV and cancer registries in the United States. Estimated counts of the US HIV-infected and AIDS populations were obtained from Centers for Disease Control and Prevention surveillance data. We obtained estimated counts of AIDS-defining (ie, Kaposi sarcoma, non-Hodgkin lymphoma, and cervical cancer) and non-AIDS-defining cancers in the US AIDS population during 1991-2005 by multiplying cancer incidence rates and AIDS population counts, stratified by year, age, sex, race and ethnicity, transmission category, and AIDS-relative time. We tested trends in counts and standardized incidence rates using linear regression models. We multiplied overall cancer rates and HIV-only (HIV infected, without AIDS) population counts, available from 34 US states during 2004-2007, to estimate cancers in the HIV-only population. All statistical tests were two-sided. RESULTS: The US AIDS population expanded fourfold from 1991 to 2005 (96,179 to 413,080) largely because of an increase in the number of people aged 40 years or older. During 1991-2005, an estimated 79 656 cancers occurred in the AIDS population. From 1991-1995 to 2001-2005, the estimated number of AIDS-defining cancers decreased by greater than threefold (34,587 to 10,325 cancers; P(trend) < .001), whereas non-AIDS-defining cancers increased by approximately threefold (3193 to 10,059 cancers; P(trend) < .001). From 1991-1995 to 2001-2005, estimated counts increased for anal (206 to 1564 cancers), liver (116 to 583 cancers), prostate (87 to 759 cancers), and lung cancers (875 to 1882 cancers), and Hodgkin lymphoma (426 to 897 cancers). In the HIV-only population in 34 US states, an estimated 2191 non-AIDS-defining cancers occurred during 2004-2007, including 454 lung, 166 breast, and 154 anal cancers. CONCLUSIONS: Over a 15-year period (1991-2005), increases in non-AIDS-defining cancers were mainly driven by growth and aging of the AIDS population. This growing burden requires targeted cancer prevention and treatment strategies.

Using cases to strengthen inference on the association between single nucleotide polymorphisms and a secondary phenotype in genome-wide association studies.

Case-control genome-wide association studies provide a vast amount of genetic information that may be used to investigate secondary phenotypes. We study the situation in which the primary disease is rare and the secondary phenotype and genetic markers are dichotomous. An analysis of the association between a genetic marker and the secondary phenotype based on controls only (CO) is valid, whereas standard methods that also use cases result in biased estimates and highly inflated type I error if there is an interaction between the secondary phenotype and the genetic marker on the risk of the primary disease. Here we present an adaptively weighted (AW) method that combines the case and control data to study the association, while reducing to the CO analysis if there is strong evidence of an interaction. The possibility of such an interaction and the misleading results for standard methods, but not for the AW or CO approaches, are illustrated by data from a case-control study of colorectal adenoma. Simulations and asymptotic theory indicate that the AW method can reduce the mean square error for estimation with a prespecified SNP and increase the power to discover a new association in a genome-wide study, compared to CO analysis. Further experience with genome-wide studies is needed to determine when methods that assume no interaction gain precision and power, thereby can be recommended, and when methods such as the AW or CO approaches are needed to guard against the possibility of nonzero interactions.

Applying the Lorenz curve to disease risk to optimize health benefits under cost constraints.

This paper shows how the Lorenz curve can be used, together with models of disease risk, to allocate scarce resources so as to optimize a health benefit. Consider the example of breast cancer mortality. If there were sufficient resources to provide all women with mammograms, a certain maximal number of lives could be saved. Suppose, however, that only a fraction of that amount of money is available for prevention activities. Suppose that a questionnaire could be given to assess a woman's risk of dying of breast cancer. Depending on the amount of money available, on the ratio of the cost of a questionnaire to the cost of a mammogram, and on the Lorenz curve of the distribution of risks of breast cancer mortality, I calculate the proportion of women who should be given questionnaires, the proportion of women given the questionnaires who should be given mammograms because they have high risks, and the proportion of women not given questionnaires who should be assigned to receive mammograms at random so as to maximize the number of lives saved.

Cancer risk prediction models: a workshop on development, evaluation, and application.

Cancer researchers, clinicians, and the public are increasingly interested in statistical models designed to predict the occurrence of cancer. As the number and sophistication of cancer risk prediction models have grown, so too has interest in ensuring that they are appropriately applied, correctly developed, and rigorously evaluated. On May 20-21, 2004, the National Cancer Institute sponsored a workshop in which experts identified strengths and limitations of cancer and genetic susceptibility prediction models that were currently in use and under development and explored methodologic issues related to their development, evaluation, and validation. Participants also identified research priorities and resources in the areas of 1) revising existing breast cancer risk assessment models and developing new models, 2) encouraging the development of new risk models, 3) obtaining data to develop more accurate risk models, 4) supporting validation mechanisms and resources, 5) strengthening model development efforts and encouraging coordination, and 6) promoting effective cancer risk communication and decision-making.

Estimates of the number of US women who could benefit from tamoxifen for breast cancer chemoprevention.

BACKGROUND: The Breast Cancer Prevention Trial demonstrated that tamoxifen treatment produced a 49% reduction in the risk of invasive breast cancer among women at elevated risk for the disease. The U.S. Food and Drug Administration (FDA) subsequently approved tamoxifen for women aged 35 years or older with a 5-year breast cancer risk of 1.67% or higher for breast cancer chemoprevention. However, tamoxifen use has been associated with adverse outcomes, and not all eligible women have a positive benefit/risk ratio. METHODS: We used weighted data from the year 2000 National Health Interview Survey Cancer Control Module to estimate the total number of U.S. women, aged 35-79 years, who were eligible for tamoxifen chemoprevention based on the FDA eligibility criteria. We also estimated the numbers of white and black women who would benefit from tamoxifen chemoprevention on the basis of a positive benefit/risk index developed by Gail et al. RESULTS: Of the 65,826,074 women aged 35-79 years without reported breast cancer in the United States in 2000, 10,232 816 women (15.5%, 95% confidence interval [CI] = 14.7% to 16.3%) would be eligible for tamoxifen chemoprevention. The percentage of U.S. women who would be eligible varied dramatically by race, with 18.7% (95% CI = 17.8% to 19.7%) of white women, 5.7% (95% CI = 4.3% to 7.5%) of black women, and 2.9% (95% CI = 2.1% to 3.9%) of Hispanic women being eligible. Of the 50,104,829 white U.S. women aged 35-79 years, 2,431,911 (4.9%, 95% CI = 4.3% to 5.4%) would have a positive benefit/risk index for tamoxifen chemoprevention. Of the 7,481,779 black U.S. women aged 35-79 years, only 42,768 (0.6%, 95% CI = 0.2% to 1.3%) would have a positive benefit/risk index. Among white women, 28,492 (95% CI = 24,693 to 32,292) breast cancers would be prevented or deferred if those women who have a positive net benefit index took tamoxifen over the next 5 years. CONCLUSION: A substantial percentage of U.S. women would be eligible for tamoxifen chemoprevention according to FDA criteria, but a much smaller percentage would have an estimated net benefit. Nevertheless, this latter percentage corresponds to more than two million women.

Robustness of inference on measured covariates to misspecification of genetic random effects in family studies.

Family studies to identify disease-related genes frequently collect only families with multiple cases. It is often desirable to determine if risk factors that are known to influence disease risk in the general population also play a role in the study families. If so, these factors should be incorporated into the genetic analysis to control for confounding. Pfeiffer et al. [2001 Biometrika 88: 933-948] proposed a variance components or random effects model to account for common familial effects and for different genetic correlations among family members. After adjusting for ascertainment, they found maximum likelihood estimates of the measured exposure effects. Although it is appealing that this model accounts for genetic correlations as well as for the ascertainment of families, in order to perform an analysis one needs to specify the distribution of random genetic effects. The current work investigates the robustness of the proposed model with respect to various misspecifications of genetic random effects in simulations. When the true underlying genetic mechanism is polygenic with a small dominant component, or Mendelian with low allele frequency and penetrance, the effects of misspecification on the estimation of fixed effects in the model are negligible. The model is applied to data from a family study on nasopharyngeal carcinoma in Taiwan.