Use of quantitative bias analysis to evaluate single-arm trials with real-world data external controls.

PURPOSE: Use of real-world data (RWD) for external controls added to single-arm trials (SAT) is increasingly prevalent in regulatory submissions. Due to inherent differences in the data-generating mechanisms, biases can arise. This paper aims to illustrate how to use quantitative bias analysis (QBA). METHODS: Advanced non-small cell lung cancer (NSCLC) serves as an example, where many small subsets of patients with molecular tumor subtypes exist. First, some sources of bias that may occur in oncology when comparing RWD to SAT are described. Second, using a hypothetical immunotherapy agent, a dataset is simulated based on expert input for survival analysis of advanced NSCLC. Finally, we illustrate the impact of three biases: missing confounder, misclassification of exposure, and outcome evaluation. RESULTS: For each simulated scenario, bias was induced by removing or adding data; hazard ratios (HRs) were estimated applying conventional analyses. Estimating the bias-adjusted treatment effect and uncertainty required carefully selecting the bias model and bias factors. Although the magnitude of each biased and bias-adjusted HR appeared moderate in all three hypothetical scenarios, the direction of bias was variable. CONCLUSION: These findings suggest that QBA can provide an intuitive framework for bias analysis, providing a key means of challenging assumptions about the evidence. However, the accuracy of bias analysis is itself dependent on correct specification of the bias model and bias factors. Ultimately, study design should reduce bias, but QBA allows us to evaluate the impact of unavoidable bias to assess the quality of the evidence.

Methodologic considerations for noninterventional studies of switching from reference biologic to biosimilars.

PURPOSE: As more biosimilars become available in the United States, postapproval noninterventional studies describing biosimilar switching and comparing effectiveness and/or safety between switchers and nonswitchers will play a key role in generating real-world evidence to inform clinical practices and policy decisions. Ensuring sound methodology is critical for making valid inferences from these studies. METHODS: The Biologics and Biosimilars Collective Intelligence Consortium (BBCIC) convened a workgroup consisting of academic researchers, industry scientists, and practicing clinicians to establish best practice recommendations for the conduct of noninterventional studies of biosimilar and reference biologic switching. The workgroup members participated in eight teleconferences between August 2017 and February 2018 to discuss specific topics and build consensus. RESULTS: This report provides workgroup recommendations covering five main considerations relating to noninterventional studies describing reference biologic to biosimilar switching and comparing reference biologic to biosimilars for safety and effectiveness in the presence of switching at treatment initiation and during follow-up: (a) selecting appropriate data sources from a range of available options including insurance claims, electronic health records, and registries; (b) study designs; (c) outcomes of interest including health care utilization and clinical endpoints; (d) analytic approaches including propensity scores, disease risk scores, and instrumental variables; and (e) special considerations including avoiding designs that ignore history of biologic use, avoiding immortal time bias, exposure misclassification, and accounting for postindex switching. CONCLUSION: Recommendations provided in this report provide a framework that may be helpful in designing and critically evaluating postapproval noninterventional studies involving reference biologic to biosimilar switching.

Usability evaluation and adaptation of the e-health Personal Patient Profile-Prostate decision aid for Spanish-speaking Latino men.

BACKGROUND: The Personal Patient Profile-Prostate (P3P), a web-based decision aid, was demonstrated to reduce decisional conflict in English-speaking men with localized prostate cancer early after initial diagnosis. The purpose of this study was to explore and enhance usability and cultural appropriateness of a Spanish P3P by Latino men with a diagnosis of prostate cancer. METHODS: P3P was translated to Spanish and back-translated by three native Spanish-speaking translators working independently. Spanish-speaking Latino men with a diagnosis of localized prostate cancer, who had made treatment decisions in the past 24 months, were recruited from two urban clinical care sites. Individual cognitive interviews were conducted by two bilingual research assistants as each participant used the Spanish P3P. Notes of user behavior, feedback, and answers to direct questions about comprehension, usability and perceived usefulness were analyzed and categorized. RESULTS: Seven participants with a range of education levels identified 25 unique usability issues in navigation, content comprehension and completeness, sociocultural appropriateness, and methodology. Revisions were prioritized to refine the usability and cultural and linguistic appropriateness of the decision aid. CONCLUSIONS: Usability issues were discovered that are potential barriers to effective decision support. Successful use of decision aids requires adaptation and testing beyond translation. Our findings led to revisions further refining the usability and linguistic and cultural appropriateness of Spanish P3P.

Comprehensive evaluation of the incidence of late effects in 5-year survivors of breast cancer.

Late effects of breast cancer affect the quality of survivorship. Using administrative data, we compared the occurrence of almost all ICD9 codes among older breast cancer survivors to that among a matched comparison cohort to generate new hypotheses. Breast cancer patients 65 years or older diagnosed 1990-1994 in 6 integrated care settings and who survived at least 5 years were matched with a cohort of women without a history of breast cancer on care setting, age, and calendar time. We collected data on the occurrence of incident ICD9 codes beginning 6 years after the breast cancer diagnosis date and continuing to year 15, and comparable data for the matched woman. We calculated hazard ratios (HRs) and 95 % confidence intervals associating breast cancer survivorship with incidence of each ICD9 code. We used semi-Bayes methods to address multiple comparisons. Older breast cancer survivors had about the same occurrence of diseases and conditions 6-15 years after breast cancer diagnosis as comparable women. The median of 564 adjusted HRs equaled 1.06, with interquartile range 0.92-1.3. The distribution of HRs pertaining to cancer-related ICD codes was shifted toward positive associations, and the distribution pertaining to cardiovascular-related ICD codes was shifted toward negative associations. In this hypothesis-scanning study, we observed little difference in the occurrence of non-breast cancer-related diseases and conditions among older, long-term breast cancer survivors, and comparable women without a history of breast cancer.

Long-term surveillance mammography and mortality in older women with a history of early stage invasive breast cancer.

Annual surveillance mammograms in older long-term breast cancer survivors are recommended, but this recommendation is based on little evidence and with no guidelines on when to stop. Surveillance mammograms should decrease breast cancer mortality by detecting second breast cancer events at an earlier stage. We examined the association between surveillance mammography beyond 5 years after diagnosis on breast cancer-specific mortality in a cohort of women aged ≥ 65 years diagnosed 1990-1994 with early stage breast cancer. Our cohort included women who survived disease free for ≥ 5 years (N = 1,235) and were followed from year 6 through death, disenrollment, or 15 years after diagnosis. Asymptomatic surveillance mammograms were ascertained through medical record review. We used Cox proportional hazards regression stratified by follow-up year to calculate the association between time-varying surveillance mammography and breast cancer-specific and other-than-breast mortality adjusting for site, stage, primary surgery type, age and time-varying Charlson Comorbidity Index. The majority (85 %) of the 1,235 5-year breast cancer survivors received ≥ 1 surveillance mammogram in years 5-9 (yearly proportions ranged from 48 to 58 %); 82 % of women received ≥ 1 surveillance mammogram in years 10-14. A total of 120 women died of breast cancer and 393 women died from other causes (average follow-up 7.3 years). Multivariable models and lasagna plots suggested a modest reduction in breast cancer-specific mortality with surveillance mammogram receipt in the preceding year (IRR 0.82, 95 % CI 0.56-1.19, p = 0.29); the association with other-cause mortality was 0.95 (95 % CI 0.78-1.17, p = 0.64). Among older breast cancer survivors, surveillance mammography may reduce breast cancer-specific mortality even after 5 years of disease-free survival. Continuing surveillance mammography in older breast cancer survivors likely requires physician-patient discussions similar to those recommended for screening, taking into account comorbid conditions and life-expectancy.

Fracture risk in older, long-term survivors of early-stage breast cancer.

OBJECTIVES: To examine the effect of breast cancer and its treatment on fracture risk in older breast cancer survivors. DESIGN: A 10-year prospective cohort study beginning 5 years after a diagnosis of breast cancer for survivors and match date for comparison women. SETTING: Six integrated healthcare systems. PARTICIPANTS: Women aged 65 and older (1,286 survivors, 1,286 comparison women, mean age 77.7 in both groups, white, non-Hispanic: survivors, 81.6%; comparison women, 85.2%) who were alive and recurrence free 5 years after a diagnosis of early-stage breast cancer and matched on age, study site, and enrollment year to a comparison cohort without breast cancer. MEASUREMENTS: Cox proportional hazards models were used to estimate the association between fracture risk and survivor-comparison status, adjusting for drugs and risk factors associated with bone health. A subanalysis was used to evaluate the association between tamoxifen exposure and fracture risk. RESULTS: No difference was observed in fracture rates between groups (hazard ratio (HR) = 1.1, 95% confidence interval (CI) = 0.9-1.3). The protective effect of tamoxifen was not statistically significant (HR = 0.9, 95% CI = 0.6-1.2). CONCLUSION: Long-term survivors of early-stage breast cancer diagnosed at age 65 and older are not at greater risk of osteoporotic fractures than age-matched women without breast cancer. There appears to be no long-term protection from fractures with tamoxifen use.

Patient self-appraisal of change and minimal clinically important difference on the European organization for the research and treatment of cancer quality of life questionnaire core 30 before and during cancer therapy.

BACKGROUND: Clinical interpretation of health related quality of life (HRQOL) scores is challenging. The purpose of this analysis was to interpret score changes and identify minimal clinically important differences (MCID) on the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-Core 30 (QLQ-C30) before (T1) and during (T2) cancer treatment. METHODS: Patients (N = 627) in stem cell transplant (SCT) and medical (MED) or radiation (RAD) oncology at two comprehensive cancer centers, enrolled in the Electronic Self-Report Assessment-Cancer study and completed the QLQ-C30 at T1 and T2. Perceived changes in five QOL domains, physical (PF), emotional (EF), social (SF), cognitive functioning (CF) and global quality of life (QOL), were reported using the Subject Significance Questionnaire (SSQ) at T2. Anchored on SSQ ratings indicating "improvement", "the same", or "deterioration", means and effect sizes were calculated for QLQ-C30 score changes. MCID was calculated as the mean difference in QLQ-C30 score changes reflecting one category change on SSQ rating, using a two-piece linear regression model. RESULTS: A majority of SCT patients (54%) perceived deteriorating global HRQOL versus improvement (17%), while approximately equal proportions of MED/RAD patients perceived improvement (25%) and deterioration (26%). Global QOL decreased 14.2 (SCT) and 2.0 (MED/RAD) units, respectively, among patients reporting "the same" in the SSQ. The MCID ranged 5.7-11.4 (SCT) and 7.2-11.8 (MED/RAD) units among patients reporting deteriorated HRQOL; ranged 2.7-3.4 units among MED/RAD patients reporting improvement. Excepting for the global QOL (MCID =6.9), no meaningful MCID was identified among SCT patients reporting improvement. CONCLUSIONS: Cancer treatment has greater impact on HRQOL among SCT patients than MED/RAD patients. The MCID for QLQ-C30 score change differed across domains, and differed for perceived improvement and deterioration, suggesting different standards for self-evaluating changes in HRQOL during cancer treatment. Specifically, clinical attention can be focused on patients who report at least a 6 point decrease, and for patients who report at least a 3 point increase on QLQ-C30 domains. TRIAL REGISTRATION: The trial was registered with ClinicalTrials.gov: NCT00852852.

Incident malignancies among older long-term breast cancer survivors and an age-matched and site-matched nonbreast cancer comparison group over 10 years of follow-up.

BACKGROUND: Of the approximately 2.4 million American women with a history of breast cancer, 43% are aged ≥ 65 years and are at risk for developing subsequent malignancies. METHODS: Women from 6 geographically diverse sites included 5-year breast cancer survivors (N = 1361) who were diagnosed between 1990 and 1994 at age ≥ 65 years with stage I or II disease and a comparison group of women without breast cancer (N = 1361). Women in the comparison group were age-matched and site-matched to breast cancer survivors on the date of breast cancer diagnosis. Follow-up began 5 years after the index date (survivor diagnosis date or comparison enrollment date) until death, disenrollment, or through 15 years after the index date. Data were collected from medical records and electronic sources (cancer registry, administrative, clinical, National Death Index). Analyses included descriptive statistics, crude incidence rates, and Cox proportional hazards regression models for estimating the risk of incident malignancy and were adjusted for death as a competing risk. RESULTS: Survivors and women in the comparison group were similar: >82% were white, 55% had a Charlson Comorbidity Index of 0, and ≥ 73% had a body mass index ≤ 30 kg/m(2) . Of all 306 women (N = 160 in the survivor group, N = 146 in the comparison group) who developed a first incident malignancy during follow-up, the mean time to malignancy was similar (4.37 ± 2.81 years vs 4.03 ± 2.76 years, respectively; P = .28), whereas unadjusted incidence rates were slightly higher in survivors (1882 vs 1620 per 100,000 person years). The adjusted hazard of developing a first incident malignancy was slightly elevated in survivors in relation to women in the comparison group, but it was not statistically significant (hazard ratio, 1.17; 95% confidence interval, 0.94-1.47). CONCLUSIONS: Older women who survived 5 years after an early stage breast cancer diagnosis were not at an elevated risk for developing subsequent incident malignancies up to 15 years after their breast cancer diagnosis.

The Personal Patient Profile-Prostate decision support for men with localized prostate cancer: a multi-center randomized trial.

OBJECTIVE: The purpose of this trial was to compare usual patient education plus the Internet-based Personal Patient Profile-Prostate, vs. usual education alone, on conflict associated with decision making, plus explore time-to-treatment, and treatment choice. METHODS: A randomized, multi-center clinical trial was conducted with measures at baseline, 1-, and 6 months. Men with newly diagnosed localized prostate cancer (CaP) who sought consultation at urology, radiation oncology, or multi-disciplinary clinics in 4 geographically-distinct American cities were recruited. Intervention group participants used the Personal Patient Profile-Prostate, a decision support system comprised of customized text and video coaching regarding potential outcomes, influential factors, and communication with care providers. The primary outcome, patient-reported decisional conflict, was evaluated over time using generalized estimating equations to fit generalized linear models. Additional outcomes, time-to-treatment, treatment choice, and program acceptability/usefulness, were explored. RESULTS: A total of 494 eligible men were randomized (266 intervention; 228 control). The intervention reduced adjusted decisional conflict over time compared with the control group, for the uncertainty score (estimate -3.61; (confidence interval, -7.01, 0.22), and values clarity (estimate -3.57; confidence interval (-5.85,-1.30). Borderline effect was seen for the total decisional conflict score (estimate -1.75; confidence interval (-3.61,0.11). Time-to-treatment was comparable between groups, while undecided men in the intervention group chose brachytherapy more often than in the control group. Acceptability and usefulness were highly rated. CONCLUSION: The Personal Patient Profile-Prostate is the first intervention to significantly reduce decisional conflict in a multi-center trial of American men with newly diagnosed localized CaP. Our findings support efficacy of P3P for addressing decision uncertainty and facilitating patient selection of a CaP treatment that is consistent with the patient values and preferences.

Personal preferences and discordant prostate cancer treatment choice in an intervention trial of men newly diagnosed with localized prostate cancer.

BACKGROUND: Men diagnosed with localized prostate cancer (LPC) can choose from multiple treatment regimens and are faced with a decision in which medical factors and personal preferences are important. The Personal Patient Profile-Prostate (P3P) is a computerized decision aid for men with LPC that focuses on personal preferences. We determined whether the P3P intervention improved the concordance of treatment choice with self-reported influential side-effects compared with a control group. METHODS: English/Spanish-speaking men diagnosed with LPC (2007-2009) from four US cities were enrolled into a randomized trial and followed through 6-months via mailed or online questionnaire. Men were randomized to receive the P3P intervention or standard education plus links to reputable websites. We classified choice as concordant if men were concerned with (a) sexual function and chose external beam radiotherapy or brachytherapy, (b) bowel function and chose prostatectomy, (c) sex, bowel, and/or bladder function and chose active surveillance, or (d) not concerned with any side effect and chose any treatment. Using logistic regression, we calculated odds ratios (OR) and 95% confidence intervals (CI) for the association between the P3P intervention and concordance. RESULTS: Of 448 men, most were <65 years, non-Hispanic white, had multiple physician consultations prior to enrollment, and chose a treatment discordant with concerns about potential side effects. There was no significant difference in concordance between the intervention (45%) and control (50%) group (OR = 0.82; 95%CI = 0.56, 1.2). CONCLUSIONS: The P3P intervention did not improve concordance between potential side effects and treatment choice. Information and/or physician consultation immediately after diagnosis was likely to influence decisions despite concerns about side effects. The intervention may be more effective before the first treatment options consultation. TRIAL REGISTRATION: NCT00692653 http://clinicaltrials.gov/ct2/show/NCT00692653.

Cardiometabolic factors and breast cancer risk in U.S. black women.

Previous studies have suggested that metabolic syndrome may be associated with an increased risk of breast cancer, particularly in postmenopausal women, but U.S. black women have not been assessed. We examined the associations of abdominal obesity, type 2 diabetes, hypertension, and high cholesterol individually and in combination with breast cancer incidence in the Black Women's Health Study. By means of Cox regression models, we estimated incidence rate ratios (IRR) and 95 % confidence intervals (CI) for the associations of baseline and time-dependent values of self-reported abdominal obesity, type 2 diabetes, hypertension, and high cholesterol with breast cancer incidence. During 516,452 person years of follow-up (mean years = 10.5; standard deviation = 2.9) from 1995 to 2007, 1,228 breast cancer cases were identified. After adjustment for age, education, body mass index at age 18, physical activity, and individual cardiometabolic factors, neither individual nor combinations of cardiometabolic factors were associated with breast cancer incidence overall; the multivariable IRR was 1.04 (95 % CI 0.86-1.25) for the combination of ≥3 factors relative to the absence of all factors, and 1.17 (0.85-1.60) for having all four factors. Among postmenopausal women, however, the comparable IRRs were 1.23 (0.93-1.62) and 1.63 (1.12-2.37), respectively. Our findings provide some support for an association between cardiometabolic factors and breast cancer incidence among postmenopausal U.S. black women.

Age, comorbidity, and breast cancer severity: impact on receipt of definitive local therapy and rate of recurrence among older women with early-stage breast cancer.

BACKGROUND: The definitive local therapy options for early-stage breast cancer are mastectomy and breast-conserving surgery followed by radiation therapy. Older women and those with comorbidities frequently receive breast-conserving surgery alone. The interaction of age and comorbidity with breast cancer severity and their impact on receipt of definitive therapy have not been well-studied. STUDY DESIGN: In a cohort of 1,837 women aged 65 years and older receiving treatment for early-stage breast cancer in 6 integrated health care delivery systems in 1990-1994 and followed for 10 years, we examined predictors of receiving nondefinitive local therapy and assessed the impact on breast cancer recurrence within levels of severity, defined as level of risk for recurrence. RESULTS: Age and comorbidity were associated with receipt of nondefinitive therapy. Compared with those at low risk, women at the highest risk were less likely to receive nondefinitive therapy (odds ratio = 0.32; 95% CI, 0.22-0.47), and women at moderate risk were about half as likely (odds ratio = 0.54; 95% CI, 0.35-0.84). Nondefinitive local therapy was associated with higher rates of recurrence among women at moderate (hazard ratio = 5.1; 95% CI, 1.9-13.5) and low risk (hazard ratio = 3.2; 95% CI, 1.1-8.9). The association among women at high risk was weak (hazard ratio = 1.3; 95% CI, 0.75-2.1). CONCLUSIONS: Among these older women with early-stage breast cancer, decisions about therapy partially balanced breast cancer severity against age and comorbidity. However, even among women at low risk, omitting definitive local therapy was associated with increased recurrence.

Use of antiepileptic medications in pregnancy in relation to risks of birth defects.

PURPOSE: To evaluate use of specific antiepileptic drugs (AEDs) in pregnancy in relation to specific birth defects. METHODS: Using data from the National Birth Defects Prevention Study, we assessed use of AEDs and the risk of neural tube defects (NTDs), oral clefts (OCs), heart defects (HDs), hypospadias, and other major birth defects, taking specific agent, timing, and indication into consideration. RESULTS: Drug-specific increased risks were observed for valproic acid in relation to NTDs [adjusted odds ratio (aOR), 9.7;, 95% confidence interval (CI), 3.4-27.5], OCs (aOR, 4.4; 95% CI, 1.6-12.2), HDs (aOR, 2.0; 95% CI, 0.78-5.3), and hypospadias (aOR. 2.4; 95% CI, 0.62-9.0), and for carbamazapine in relation to NTDs (aOR, 5.0; 95% CI, 1.9-12.7). Epilepsy history without AED use did not seem to increase risk. CONCLUSIONS: Valproic acid, which current guidelines suggest should be avoided in pregnancy, was most notable in terms of strength and breadth of its associations. Carbamazapine was associated with NTDs, even after controlling for folic acid use. Sample sizes were still too small to adequately assess risks of less commonly used AEDs, but our findings support further study to identify lower risk options for pregnant women.

Regular aspirin use and breast cancer risk in US Black women.

BACKGROUND: Epidemiologic studies have yet to provide consistent evidence to support a protective effect of aspirin and other non-steroidal anti-inflammatory drugs (NSAID) on the incidence of breast cancer. OBJECTIVE: We evaluated the relations of current use of aspirin, non-aspirin NSAIDs, and acetaminophen with breast cancer incidence in the Black Women's Health Study. METHODS: Biennial follow-up of 59,000 study participants began in 1995. During 558,600 person-years of follow-up through 2007, 1,275 breast cancer cases were identified. Using Cox proportional hazards regression, we estimated incidence rate ratios (IRR) and 95% confidence intervals (CI) for associations of current and former use of aspirin, other NSAIDs, and acetaminophen with incident breast cancer. RESULTS: After adjustment for age, education, body mass index at age 18, physical activity, female hormone use, current smoking, and other NSAID use, the IRRs were 0.79 (95% CI = 0.66, 0.95) for current aspirin use and 0.68 (95% CI = 0.50, 0.92) for ≥5 years of aspirin use. Similar associations were observed for acetaminophen use. CONCLUSIONS: Both aspirin and acetaminophen use were inversely associated with breast cancer incidence in the present study. Reasons for the association with acetaminophen use are unclear, given that acetaminophen has very weak anti-inflammatory effects.

Metformin and incident breast cancer among diabetic women: a population-based case-control study in Denmark.

BACKGROUND: Preliminary evidence suggests that metformin may decrease breast cancer risk by decreasing insulin levels and reducing cell proliferation. We evaluated the effect of metformin medication on the risk of incident breast cancer among peri- and postmenopausal women. METHODS: We used Danish medical registries to conduct a nested case-control study among type 2 diabetic women 50 years or older who resided in northern Denmark from 1989 to 2008 (n = 4,323). We identified 393 diabetic cases and used risk-set sampling to select 10 diabetic controls per case (n = 3,930) matched on county of residence. Odds ratios (OR) and 95% CIs were estimated by conditional logistic regression associating metformin use with breast cancer occurrence. RESULTS: Ninety-six cases (24%) and 1,154 controls (29%) used metformin for at least 1-year duration. Cases were slightly older on average than controls, but they were similar in distribution for parity, use of hormone replacement therapy, and history of diabetes complications. Metformin users were less likely with a diagnosis of breast cancer (OR = 0.77; 95% CI = 0.61-0.99) than nonmetformin users. Adjustment for diabetes complications, clinically diagnosed obesity, and important predictors of breast cancer did not substantially alter the association (OR = 0.81; 95% CI = 0.63-0.96). CONCLUSION: Our results suggest that metformin may protect against breast cancer in type 2 diabetic peri- or postmenopausal women. IMPACT: This study supports the growing evidence of a role for metformin in breast cancer chemoprevention.

Reproducibility of reported nutrient intake and supplement use during a past pregnancy: a report from the Children's Oncology Group.

Maternal diet and nutrition have been thought to play a role in many childhood conditions. Studies using food frequency questionnaires (FFQ) have reported associations with maternal diet, but these findings are difficult to interpret because the reliability and validity of the FFQs for diet during a past pregnancy are not known. We determined the reproducibility of reported diet and supplement use during a past pregnancy in a subset of mothers interviewed for a case-control study of maternal diet in relation to the risk of childhood brain tumours. Cases were Children's Oncology Group patients, diagnosed at age <6 with medulloblastoma or primitive neuroectodermal tumour from 1991 to 1997. Area code, race/ethnicity, and birth date matched controls were selected by random-digit-dialling. Case and control mothers completed a modified Willett FFQ a mean of 5 years after the index child's birth. A mean of 3.6 months later, a subset of mothers consisting of 52 case and 51 control mothers repeated the interview; these comprise the reproducibility study population. The mean intra-class correlation was 0.59 (range 0.41, 0.69) for energy-adjusted nutrients from dietary sources only; it was 0.41 (range 0.06, 0.70) when supplements were included. Agreement for reporting multivitamin use during pregnancy by time period and pattern was good to very good (kappa = 0.66-0.85). Overall, the reproducibility of nutrient estimates and supplement use in pregnancy was good and similar to that reported for adult diet.

A most stubborn bias: no adjustment method fully resolves confounding by indication in observational studies.

OBJECTIVE: To evaluate the effectiveness of methods that control for confounding by indication, we compared breast cancer recurrence rates among women receiving adjuvant chemotherapy with those who did not. STUDY DESIGN AND SETTING: In a medical record review-based study of breast cancer treatment in older women (n=1798) diagnosed between 1990 and 1994, our crude analysis suggested that adjuvant chemotherapy was positively associated with recurrence (hazard ratio [HR]=2.6; 95% confidence interval [CI]=1.9, 3.5). We expected a protective effect, so postulated that the crude association was confounded by indications for chemotherapy. We attempted to adjust for this confounding by restriction, multivariable regression, propensity scores (PSs), and instrumental variable (IV) methods. RESULTS: After restricting to women at high risk for recurrence (n=946), chemotherapy was not associated with recurrence (HR=1.1; 95% CI=0.7, 1.6) using multivariable regression. PS adjustment yielded similar results (HR=1.3; 95% CI=0.8, 2.0). The IV-like method yielded a protective estimate (HR=0.9; 95% CI=0.2, 4.3); however, imbalances of measured factors across levels of the IV suggested residual confounding. CONCLUSION: Conventional methods do not control for unmeasured factors, which often remain important when addressing confounding by indication. PS and IV analysis methods can be useful under specific situations, but neither method adequately controlled confounding by indication in this study.

Breast cancer recurrence in older women five to ten years after diagnosis.

Little is known about the risk of recurrence >5 years after diagnosis among older breast cancer survivors. A community-based population of women >or=65 years diagnosed with early-stage breast cancer who survived disease free for 5 years was followed for 5 additional years or until a diagnosis of breast cancer recurrence, second primary, death, or loss to follow-up. These 5-year disease-free survivors (N = 1,277) had primary breast cancers that were node negative (77%) and estrogen receptor positive or unknown (86%). Five percent (n = 61) developed a recurrence between 5 and 10 years after diagnosis: 25% local, 9.8% regional, and 66% distant. Women who were node positive [hazard ratio (HR), 3.9; 95% confidence interval (95% CI), 1.5-10], had poorly differentiated tumors (HR, 2.5; 95% CI, 0.9-6.6), or who received breast conserving surgery without radiation therapy (HR, 2.4; 95% CI, 1.0-5.8) had higher recurrence rates compared with node negative, well differentiated, and receipt of mastectomy, respectively. Not receiving adjuvant tamoxifen, compared with receiving adjuvant tamoxifen, was also positively associated with late recurrence among women with estrogen receptor-positive/unknown tumors. Although relatively few women experience a late recurrence, most recurrences present as advanced disease, which is difficult to treat in older women. This study of late recurrence emphasizes that the risk, although small, is not negligible even in this group at high risk of death due to competing causes.