Abstract P1-08-35: Stromal tumor infiltrating lymphocytes analysis by race and ethnicity in triple negative breast cancers from 2 phase III randomized adjuvant breast cancer trials: ECOG-ACRIN E2197 and E1199

Background: Black patients with triple negative breast cancer (TNBC) have worse survival outcomes, even after adjusting for stage at diagnosis, income, insurance status and other socioeconomic factors. Little is known regarding anti-tumor immune responses in Black patients and how these differences affect responses to treatment in TNBC. Limited data exists regarding the stromal tumor infiltrating lymphocytes (sTILs, which are strongly prognostic in TNBC) distribution based on race and ethnicity. Here we evaluate the prevalence, distribution, and prognostic impact of sTILs in TNBC by race/ethnicity from 2 prospective clinical trials of adjuvant anthracycline/taxane-based chemotherapy (E2197 and E1199). Methods: Full-face hematoxylin and eosin-stained sections of 481 tumors from ECOG-ACRIN trials E2197 and E1199 were previously evaluated for density of sTILs and shown to be associated with disease-free survival (DFS), distant recurrence-free interval (DRFI), and overall survival (OS) (Adams, et al JCO 2014). Further analyses were undertaken to evaluate the impact of race/ethnicity. Results: The majority of the 481 TNBC were from White patients (82.3%, n=403); with 12.3% (n=59) Black patients, 1.6% (n=14) other (9 Hispanic, 3 Asian, 2 Other), and 0.5% (n=5) unknown race. Age distribution (mean 49.2 for White and 49.2 for Black) and node negative disease (White 68/403 (42%), Black 24/59 (41%)) were similar. However, tumor size ≤2cm was seen more commonly in White patients (34%, 137/403) compared with Black patients (20%, 12/59). Black patients had a higher proportion of high sTILs (≥30%) with 23.7% (14/59) compared to White patients (11.4%, 46/403). The association of continuous stromal TILs with DFS (hazard ratio for a 10-point difference) was 0.84 (95% CI 0.72, 0.98) for White patients and 0.94 (95% CI 0.73, 1.20) for Black patients [159 DFS events for Whites, 26 DFS events for Blacks]. Conclusions: This is the first dataset from prospective clinical trials evaluating sTILs in TNBC in Black patients. Prevalence of high sTILs was greater in Black patients compared to White patients. The association between increasing sTILs and improved invasive disease-free survival across racial/ethnic groups must be investigated in larger datasets.

Table 1.Race/EthnicityTotal (n=481)White (n=403)Black (n=59)Other (n=19)Mean age49.049.249.245.6T1 (tumor <=2cm)157(32.6%)137 (34.0%)12 (20.3%)8 (42.1%)T2 (tumor >2 and <=5cm)283(58.8%)232 (57.6%)41 (69.5%)10 (52.6%)T3 and T441 (8.5%)34 (8.4%)6 (10.2%)1 (5.3%)Node negative197 (41.0%)168 (41.7%)24 (40.7%)5 (26.3%)Median sTILs (Quartiles)10 (10, 20)10 (10, 20)10 (10,20)20 (10, 30)sTILs = 095 (19.8%)83 (20.6%)10 (16.9%)2 (10.5%)sTILs 10-29%319 (66.3%)274 (68.0%)35 (59.3%)10 (52.6%)sTILs ≥30%67 (13.9%)46 (11.4%)14 (23.7%)7 (36.8%)—sTIL 30-49%,46 (9.6%)32 (7.9%)11 (18.6%)3 (15.8%)—sTIL 50-74%,17 (3.5%)11 (2.7%)3 (5.1%)3 (15.8%)—sTIL 75-100%4 (0.8%)3 (0.7%)01 (5.2%)iDFS (HR for 10% sTIL increase)0.86 (95% CI 0.76, 0.98)0.84 (95% CI 0.72, 0.98)0.94 (95% CI 0.73, 1.20)0.97 (95% CI 0.68, 1.40)DRFI (HR for 10% sTIL increase)0.82 (95% CI 0.68, 0.99)0.79 (95% CI 0.63, 1.00)1.08 (95% CI 0.82, 1.44)0.54 (95% CI 0.32, 0.90)OS (HR for 10% sTIL increase)0.81 (95% CI 0.69, 0.95)0.76 (95% CI 0.62, 0.94)1.01 (95% CI 0.76, 1.35)0.83 (95% CI 0.54, 1.29)

Citation Format: Natalie Klar, Robert J Gray, Sylvia Adams, Joseph A Sparano, Lori J Goldstein, Angela M DeMichele, Antonio C Wolff, Nancy E Davidson, George W Sledge, Sunil S Badve. Stromal tumor infiltrating lymphocytes analysis by race and ethnicity in triple negative breast cancers from 2 phase III randomized adjuvant breast cancer trials: ECOG-ACRIN E2197 and E1199 [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P1-08-35.

POS0531 FACTORS ASSOCIATED WITH BASELINE HYPERTENSION IN EARLY RHEUMATOID ARTHRITIS: DATA FROM A REAL-WORLD LARGE INCIDENT COHORT

Background:

It is not well understood why hypertension (HTN) is so common in rheumatoid arthritis (RA) patients. Reported prevalence of HTN in RA patients ranges from 4-73%.(1)

Objectives:

This study explored the prevalence of HTN at time of RA diagnosis and which demographic, behavioural and clinical factors were associated with HTN.

Methods:

Data from the Canadian Early Arthritis Cohort (CATCH), a prospective inception cohort of patients with RA <1 year duration, were used to analyze baseline demographic, behavioural and clinical characteristics associated with HTN, which was reported by physicians. Univariate logistic regression models were created to explore associations with baseline HTN. A multivariate logistic regression model was built based on goodness of fit indicated by likelihood ratio tests. Variables included in the model were age, sex, race, body mass index (BMI), education, smoking, alcohol servings, seropositivity, disease activity and comorbidities.

Results:

In total, 2052 subjects were included with mean (±SD) age of 55 (±14) years and symptom duration 5.60(5.47, 5.73) months, 71% of subjects were female and 85% were Caucasian. HTN was reported in 26% of subjects at baseline. Hypertensive subjects were older and more likely to be male. Other factors significantly associated with HTN at baseline were lower education, ever smoking, high BMI, diabetes, hyperlipidemia, worse RA disease activity, longer duration of RA symptoms, being seropositive, as well as the use of NSAIDs and/or corticosteroids (Table 1). In multivariable analysis HTN was associated with older age, overweight and obese BMI, diabetes, and hyperlipidemia. Expression of anti-citrullinated protein antibodies was inversely associated with HTN (Table 1). Other RA disease factors and treatments were not significantly associated with HTN on multivariable analysis.

Table 1.

Results of univariate and multivariate logistic regression analyses exploring the association between baseline characteristics and HTN in early RA.

Univariate Logistic RegressionMultivariable Logistic RegressionVariableCrude OR (95% CI)Adjusted OR (95% CI)Socio-Demographic20-39 years old0.15 (0.07, 0.26)0.14(0.05, 0.34)40-59 years oldReference60-79 years old2.81 (2.26, 3.50)2.26(1.65, 3.11)80-99 years old5.87 (3.36,10.25)3.80(1.53, 9.41)Female0.55 (0.45, 0.68)1.10(0.78, 1.54)Lifestyle/BehaviouralNormal weight (18.5- 24.9kg/m2)ReferenceOverweight (25-29.9 kg/m2)2.33(1.74, 3.11)1.63(1.10, 2.43)Obese (30+ kg/m2)3.19(2.38, 4.27)2.84(1.91, 4.23Ever-smoking1.41(1.15, 1.73)1.02(0.75, 1.40)Post-secondary education0.58(0.47, 0.71)0.88(0.65, 1.20)Clinical CharacteristicsSymptom duration0.99(0.99, 0.99)1.00(1.00, 1.00)DAS-281.09(1.09, 1.17)1.02(0.92, 1.13)ACPA+0.68(0.56, 0.85)0.64(0.44, 0.92)Corticosteroid use pre-baseline1.37(1.04, 1.81)OmittedNSAID use at baseline0.68(0.55, 0.84)OmittedDiabetes5.62(4.09, 7.73)3.20(1.99, 5.15)Hyperlipidemia4.75(3.74, 6.03)2.80(1.94, 4.02),CVD15.59(3.35, 72.64)Omitted

DAS-28; Disease activity score 28, ACPA; Anti-citrullinated protein antibody, CVD; Cardiovascular disease. Pre-baseline is 29 to 365 days before entering the cohort. Baseline is within 28 days before entering the cohort. Omitted variables either failed likelihood ratio test or were colinear. Additional variables tested but found insignificant: race, alcohol servings, depression, RF+, and use of DMARDs.

Conclusion:

Approximately 1 in 4 diagnosed with RA had HTN reported by their rheumatologists, which is similar to that of the general population. This suggests that increased risk of HTN in RA patients may develop as RA disease or treatment time progresses. Factors that may be predictive of this excess risk will be explored in further analysis.

References:

[1]Panoulas VF, Metsios GS, Pace AV, et al. Hypertension in rheumatoid arthritis. Rheumatology (Oxford) 2008;47:1286-98.

Acknowledgements:

The CATCH study was designed and implemented by the investigators and financially supported through unrestricted research grants from: Amgen and Pfizer Canada - Founding sponsors since January 2007; AbbVie Corporation and Hoffmann-LaRoche since 2011; Medexus Inc. since 2013;, Merck Canada since 2017, Sandoz Canada, Biopharmaceuticals since 2019,Gilead Sciences Canada since 2020 and Fresenius Kabi Canada Ltd. since 2021. Previously funded by Janssen Biotech from 2011-2016, UCB Canada and Bristol-Myers Squibb Canada from 2011-2018, Sanofi Genzyme from 2016-2017, and Eli Lilly Canada from 2016-2020.

Disclosure of Interests:

Brook Hadwen: None declared, Saverio Stranges: None declared, Neil Klar: None declared, Kuriya Bindee: None declared, Janet Pope Speakers bureau: UCB, Consultant of: AbbVie, Actelion, Amgen, Bayer, BMS, Eicos Sciences, Eli Lilly & Company, Emerald, Gilead, Janssen, Merck, Novartis, Pfizer, Roche, Sandoz, Sanofi, UCB;, Grant/research support from: Abbvie, BMS, Eli Lilly & Company, Merck, Roche, Seattle Genetics, UCB, Susan J. Bartlett Consultant of: Pfizer, UCB, Lilly, Novartis, Merck, Janssen, Abbvie, Gilles Boire Speakers bureau: Merck, BMS, Pfizer, Janssen, Grant/research support from: Amgen, Abbvie, BMS, Eli Lilly, Merck, Novartis, Pfizer, Sandoz, Louis Bessette Speakers bureau: Amgen, BMS, Janssen, Roche, UCB, AbbVie, Pfizer, Merck, Celgene, Sanofi, Lilly, Novartis, Consultant of: Amgen, BMS, Janssen, Roche, UCB, AbbVie, Pfizer, Merck, Celgene, Sanofi, Lilly, Novartis., Grant/research support from: Amgen, BMS, Janssen, Roche, UCB, AbbVie, Pfizer, Merck, Celgene, Sanofi, Lilly, Novartis., Carol Hitchon Grant/research support from: Pfizer and UCB Canada, Glen Hazlewood: None declared, Edward Keystone Speakers bureau: Amgen, AbbVie, Bristol-Myers Squibb, F. Hoffmann-La Roche Inc., Janssen Inc., Merck, Pfizer Pharmaceuticals, Sanofi Genzyme, UCB, Consultant of:: AbbVie, Amgen, AstraZeneca Pharma, Bristol-Myers Squibb Company, Celltrion, Myriad Autoimmune, F. Hoffmann-La Roche Inc, Genentech Inc, Gilead, Janssen Inc, Lilly Pharmaceuticals, Merck, Pfizer Pharmaceuticals, Sandoz, Sanofi-Genzyme, Samsung Bioepsis, Grant/research support from: AbbVie, Amgen, Gilead Sciences, Lilly Pharmaceuticals, Merck, Pfizer Pharmaceuticals, PuraPharm, Sanofi, Orit Schieir: None declared, Carter Thorne Speakers bureau: Medexus/Medac, Consultant of: Abbvie, Centocor, Janssen, Lilly, Medexus/Medac, Pfizer, Grant/research support from: Amgen, Pfizer, Abbvie, Celgene, CaREBiodam, Novartis, Diane Tin: None declared, Marie-France Valois: None declared, Vivian Bykerk Consultant of: Amgen, BMS, Gilead, Sanofi-Genzyme/Regeneron, Scipher, Pfizer Pharmaceuticals, UCB, NIH, Lillian Barra: None declared

Serial single-cell profiling analysis of metastatic TNBC during Nab-paclitaxel and pembrolizumab treatment

PURPOSE

Immunotherapy has recently been shown to improve outcomes for advanced PD-L1-positive triple-negative breast cancer (TNBC) in the Impassion130 trial, leading to FDA approval of the first immune checkpoint inhibitor in combination with taxane chemotherapy. To further develop predictive biomarkers and improve therapeutic efficacy of the combination, interrogation of the tumor immune microenvironment before therapy as well as during each component of treatment is crucial. Here we use single-cell RNA sequencing (scRNA-seq) on tumor biopsies to assess immune cell changes from two patients with advanced TNBC treated in a prospective trial at predefined serial time points, before treatment, on taxane chemotherapy and on chemo-immunotherapy.

METHODS

Both patients (one responder and one progressor) received the trial therapy, in cycle 1 nab-paclitaxel given as single agent, in cycle 2 nab-paclitaxel in combination with pembrolizumab. Tumor core biopsies were obtained at baseline, 3 weeks (after cycle 1, chemotherapy alone) and 6 weeks (after cycle 2, chemo-immunotherapy). Single-cell RNA sequencing (scRNA-seq) of both cancer cells and infiltrating immune cells isolated were performed from fresh tumor core biopsy specimens by 10 × chromium sequencing.

RESULTS

ScRNA-seq analysis showed significant baseline heterogeneity of tumor-infiltrating immune cell populations between the two patients as well as modulation of the tumor microenvironment by chemotherapy and immunotherapy. In the responding patient there was a population of PD-1high-expressing T cells which significantly decreased after nab-paclitaxel plus pembrolizumab treatment as well as a presence of tissue-resident memory T cells (TRM). In contrast, tumors from the patient with rapid disease progression showed a prevalent and persistent myeloid compartment.

CONCLUSIONS

Our study provides a deep cellular analysis of on-treatment changes during chemo-immunotherapy for advanced TNBC, demonstrating not only feasibility of single-cell analyses on serial tumor biopsies but also the heterogeneity of TNBC and differences in on-treatment changes in responder versus progressor.

Bortezomib, C1-inhibitor and plasma exchange do not prolong the survival of multi-transgenic GalT-KO pig kidney xenografts in baboons

Galactosyl-transferase KO (GalT-KO) pigs represent a potential solution to xenograft rejection, particularly in the context of additional genetic modifications. We have performed life supporting kidney xenotransplantation into baboons utilizing GalT-KO pigs transgenic for human CD55/CD59/CD39/HT. Baboons received tacrolimus, mycophenolate mofetil, corticosteroids and recombinant human C1 inhibitor combined with cyclophosphamide or bortezomib with or without 2-3 plasma exchanges. One baboon received a control GalT-KO xenograft with the latter immunosuppression. All immunosuppressed baboons rejected the xenografts between days 9 and 15 with signs of acute humoral rejection, in contrast to untreated controls (n = 2) that lost their grafts on days 3 and 4. Immunofluorescence analyses showed deposition of IgM, C3, C5b-9 in rejected grafts, without C4d staining, indicating classical complement pathway blockade but alternate pathway activation. Moreover, rejected organs exhibited predominantly monocyte/macrophage infiltration with minimal lymphocyte representation. None of the recipients showed any signs of porcine endogenous retrovirus transmission but some showed evidence of porcine cytomegalovirus (PCMV) replication within the xenografts. Our work indicates that the addition of bortezomib and plasma exchange to the immunosuppressive regimen did not significantly prolong the survival of multi-transgenic GalT-KO renal xenografts. Non-Gal antibodies, the alternative complement pathway, innate mechanisms with monocyte activation and PCMV replication may have contributed to rejection.

Developments in cluster randomized trials and Statistics in Medicine

The design and analysis of cluster randomized trials has been a recurrent theme in Statistics in Medicine since the early volumes. In celebration of 25 years of Statistics in Medicine, this paper reviews recent developments, particularly those that featured in the journal. Issues in design such as sample size calculations, matched paired designs, cohort versus cross-sectional designs, and practical design problems are covered. Developments in analysis include modification of robust methods to cope with small numbers of clusters, generalized estimation equations, population averaged and cluster specific models. Finally, issues on presenting data, some other clustering issues and the general problem of evaluating complex interventions are briefly mentioned.

Breast screening outcomes in women with and without a family history of breast and/or ovarian cancer

OBJECTIVES

To compare breast screening outcomes between women with a moderate or strong family history of breast and/or ovarian cancer with those without such a history.

SETTING

The Ontario Breast Screening Programme (OBSP) is a population-based programme offering mammography and clinical breast examination to Ontario women of 50 and older.

METHODS

Data from a cohort of 143,574 women screened by the OBSP from 1996 to 1997 were included. Referral rates, cancer detection rates, positive predictive values and the histological features of screen-detected cancers were examined within family history groups, age groups and screening modalities. Logistic regression analysis of cancer detection was conducted to adjust for potential confounding variables; subgroup analysis by hormone replacement therapy (HRT) use was also undertaken.

RESULTS

Compared with women with no family history, women with a moderate or strong family history of breast and/or ovarian cancer were more likely to have their cancer detected (odds ratio [OR]=1.44, 95% confidence interval [CI] 1.20-1.74 and OR=1.42, 95% CI 1.10-1.83, respectively). Among women using HRT, however, there was no association observed between family history and cancer detection (moderate: OR=0.98, 95% CI 0.65-1.48; strong: OR=1.17, 95% CI 0.68-2.02) with history. The histological features of invasive tumours were similar among family history groups.

CONCLUSIONS

Greater cancer detection rates and high proportions of invasive tumours with good prognosis indicate that women aged 50 and over with a family history may have the potential to benefit from regular breast cancer screening. Further studies are required to identify optimal screening guidelines and to examine whether HRT reduces the ability to detect cancer in these women.

Lymph node retrieval and assessment in stage II colorectal cancer: a population-based study

BACKGROUND

Adjuvant chemotherapy for patients with stage III (node-positive) colorectal cancer (CRC) reduces mortality by one third. Retrieval of an inadequate number of lymph nodes in the surgical specimen may result in incorrectly designating some patients as stage II (node negative), and consequently, such patients may not be offered appropriate chemotherapy. Recent National Cancer Institute guidelines suggest that a minimum of 12 nodes should be examined to ensure accurate staging.

METHODS

This population-based study identified stage II (T3N0 and T4N0) CRC cases by using CRC pathology reports (1997-2000) from the Ontario Cancer Registry. Patients aged 19 to 75 years were identified, and demographic, surgical, pathologic, and hospital data were extracted. Factors relating to the number of lymph nodes assessed were examined.

RESULTS

A total of 8848 CRC cases were reviewed, and 1789 stage II cases were identified. Seventy-three percent of cases were designated as node negative on the basis of assessment of <12 lymph nodes. Multivariate analysis showed that age, tumor size, specimen length, use of a pathology template, and academic status of the hospital were significant predictors of the number of lymph nodes assessed.

CONCLUSIONS

A subset of patients with CRC in Ontario were assigned stage II disease on the basis of examination of relatively few lymph nodes.

Current and future challenges in the design and analysis of cluster randomization trials

Randomized trials in which the unit of randomization is a community, worksite, school or family are becoming widely used in the evaluation of life-style interventions for the prevention of disease. The increasing interest in adopting a cluster randomization design is being matched by rapid methodological developments. In this paper we describe several of these developments. Brief mention is also made of issues related to economic analysis and to the planning and conduct of meta-analyses for cluster randomization trials. Recommendations for reporting are also discussed.

Inference procedures for assessing interobserver agreement among multiple raters

We propose a new procedure for constructing inferences about a measure of interobserver agreement in studies involving a binary outcome and multiple raters. The proposed procedure, based on a chi-square goodness-of-fit test as applied to the correlated binomial model (Bahadur, 1961, in Studies in Item Analysis and Prediction, 158-176), is an extension of the goodness-of-fit procedure developed by Donner and Eliasziw (1992, Statistics in Medicine 11, 1511-1519) for the case of two raters. The new procedure is shown to provide confidence-interval coverage levels that are close to nominal over a wide range of parameter combinations. The procedure also provides a sample-size formula that may be used to determine the required number of subjects and raters for such studies.

Nipple fluid carcinoembryonic antigen and prostate-specific antigen in cancer-bearing and tumor-free breasts

PURPOSE

Mammograms and breast examinations are established methods for early breast cancer detection. Routine mammography screening reduces breast cancer mortality among women ages > or = 50 years, but additional screening methods are needed. We and others have found high levels of carcinoembryonic antigen (CEA) and prostate-specific antigen (PSA) in nipple aspirate fluids (NAFs), but the usefulness for these bio-markers for early breast cancer detection is unknown.

PATIENTS AND METHODS

NAFs from one or both breasts of 388 women were analyzed for CEA, PSA, and albumin levels. The study included 44 women with newly diagnosed invasive breast cancers, 67 women with proliferative breast lesions (ductal and lobular carcinoma in situ and atypical ductal hyperplasia), and 277 controls without these breast lesions. Analyses were conducted using the log(10)-transformed CEA and PSA levels to normalize the distributions of these tumor markers.

RESULTS

Nipple fluid CEAs are significantly higher for cancerous breasts than tumor-free breasts (median 1,830 and 1,400 ng/mL, respectively; P <.01). However, at 90% specificity of the assay (CEA = 11,750 ng/mL), the corresponding sensitivity for cancer detection is 32%. CEA levels are not significantly different for breasts with proliferative lesions compared with tumor-free breasts. Nipple fluid PSAs do not differ by tumor status. Analyses of NAF albumin-standardized CEAs and PSAs yield similar results. Nipple fluid CEA and PSA titers are correlated in the affected and unaffected breast of women with unilateral lesions.

CONCLUSION

Nipple fluid CEAs are higher for breasts with untreated invasive cancers, but the test sensitivity is low. Nipple fluid PSA titers do not seem to be useful for breast cancer detection.

Disclosure of familial genetic information: perceptions of the duty to inform

BACKGROUND

The familial implications of genetic information can lead to a conflict between a physician's duties to maintain patient confidentiality and to inform at-risk relatives about susceptibility to genetic diseases. As genes are discovered that can identify patients at risk of adverse outcomes, this conflict has become the subject of discussion and debate.

METHODS

We performed a one-time telephone survey of a population-based sample of 200 Jewish women to assess knowledge and attitudes about genetic testing. Attitudes toward sharing genetic test results with family members were evaluated using three hypothetical scenarios that described an easily preventable disease, a disease (breast cancer) in which the only option for prevention was prophylactic mastectomies, and a nonpreventable disease.

RESULTS

Nearly all respondents believed that a patient should inform at-risk family members when the disease was preventable (100% and 97% in the relevant scenarios), compared with only 85% who felt a duty to inform at-risk family members about a nonpreventable disease (P <0.001). The proportions of respondents who believed that physicians should seek out and inform at-risk family members against a patient's wishes was much lower: only 18% of respondents to the easily preventable disease scenario, 22% of respondents to the breast cancer scenario, and 16% of respondents to the nonpreventable disease scenario.

CONCLUSIONS

Most women surveyed believed that genetic information should be shared within families, unless it violated a patient's wishes. These sorts of opinions should be considered in the debate over the confidentiality of genetic information.

Understanding of prognosis among parents of children who died of cancer: impact on treatment goals and integration of palliative care

CONTEXT

Parents' understanding of prognosis or decision making about palliative care for children who die of cancer is largely unknown. However, a more accurate understanding of prognosis could alter treatment goals and expectations and lead to more effective care.

OBJECTIVES

To evaluate parental understanding of prognosis in children who die of cancer and to assess the association of this factor with treatment goals and the palliative care received by children.

DESIGN, SETTING, AND PARTICIPANTS

Survey, conducted between September 1997 and August 1998, of 103 parents of children who received treatment at the Dana-Farber Cancer Institute and Children's Hospital, Boston, Mass, and who died of cancer between 1990 and 1997 (72% of those eligible and those located) and 42 pediatric oncologists.

MAIN OUTCOME MEASURE

Timing of parental understanding that the child had no realistic chance for cure compared with the timing of physician understanding of this prognosis, as documented in the medical record.

RESULTS

Parents first recognized that the child had no realistic chance for cure a mean (SD) of 106 (150) days before the child's death, while physician recognition occurred earlier at 206 (330) days before death. Among children who died of progressive disease, the group characterized by earlier recognition of this prognosis by both parents and physicians had earlier discussions of hospice care (odds ratio [OR], 1.03; 95% confidence interval [CI], 1.01-1.06; P =.01), better parental ratings of the quality of home care (OR, 3.31; 95% CI, 1.15-9.54; P =.03), earlier institution of a do-not-resuscitate order (OR, 1.03; 95% CI, 1.00-1.06; P =.02), less use of cancer-directed therapy during the last month of life (OR, 2.80; 95% CI, 1.05-7.50; P =.04), and higher likelihood that the goal of cancer-directed therapy identified by both physician and parent was to lessen suffering (OR, 5.17; 95% CI, 1.86-14.4; P =.002 for physician and OR, 6.56; 95% CI, 1.54-27.86; P =.01 for parents).

CONCLUSION

Considerable delay exists in parental recognition that children have no realistic chance for cure, but earlier recognition of this prognosis by both physicians and parents is associated with a stronger emphasis on treatment directed at lessening suffering and greater integration of palliative care. JAMA. 2000;284:2469-2475.

Comparison of methods for analyzing binary data arising from two-sample twin studies

The study of twins is widely used for research into genetic and environmental influences on human outcome measurements. For the study design in which independent samples of monozygotic and dizygotic twins are compared with respect to their similarity on a binary trait, several statistical methods have been proposed. Using a Monte Carlo simulation, we compare the five following procedures: 1) goodness-of-fit method based on the common correlation model, 2) normal approximation of the maximum likelihood estimators of the common correlation coefficients, 3) Ramakrishnan et al. [(1992) Genet Epidemiol 9:273-282] method of odds ratio comparison, 4) generalized estimating equations method of odds ratio estimation, and 5) tetrachoric correlation method. The results show that the goodness-of-fit approach has similar or better performance in both type-one error rates and power than the other methods in all parameter settings. Its advantage with respect to type-one error rates is particularly clear under conditions of small sample sizes, extreme prevalences, or high values of the intraclass correlation coefficients. Therefore, the goodness-of-fit method is recommended for the two-sample twin study design.

Cognitive sequelae in children treated for acute lymphoblastic leukemia with dexamethasone or prednisone

PURPOSE

The cognitive sequelae of treatment for childhood acute lymphoblastic leukemia (ALL) were compared in a group of patients who received dexamethasone during the intensification and maintenance phases of therapy with those in a historical control group for whom antileukemia therapy was similar, except that the corticosteroid component of therapy was prednisone.

METHODS

Patients treated for ALL on Dana-Farber Cancer Institute protocols 87-01 (n = 44) and 91-01 (n = 23) were evaluated by standard cognitive and achievement tests. Corticosteroid therapy was delivered in 5-day pulses given every 3 weeks during intensification and continuation phases of therapy for a total of 2 years.

RESULTS

Children treated on protocol 87-01 received prednisone at a dose of 40 mg/m2/d (standard risk, SR) or 120 mg/ m2/d (high risk, HR); those treated on protocol 91-01 received dexamethasone at a dose of 6 mg/m2 per day (SR) or 18 mg/m2 per day (HR). Children treated on protocol 91-01 performed less well on cognitive testing. Subsample analysis indicated that cranial radiation therapy and methotrexate dose did not account for differences in cognitive outcomes.

CONCLUSIONS

The findings of this preliminary study are consistent with the hypothesis that dexamethasone therapy can increase risk for neurocognitive late effects in children treated for ALL and indicate that further investigation of this question is warranted.

Analysis of dichotomous outcome data for community intervention studies

Community intervention trials are becoming increasingly popular as a tool for evaluating the effectiveness of health education and intervention strategies. Typically, units such as households, schools, towns, counties, are randomized to receive either intervention or control, then outcomes are measured on individuals within each of the units of randomization. It is well recognized that the design and analysis of such studies must account for the clustering of subjects within the units of randomization. Furthermore, there are usually both subject level and cluster level covariates that must be considered in the modelling process. While suitable methods are available for continuous outcomes, data analysis is more complicated when dichotomous outcomes are measured on each subject. This paper will compare and contrast several of the available methods that can be applied in such settings, including random effects models, generalized estimating equations and methods based on the calculation of 'design effects', as implemented in the computer package SUDAAN. For completeness, the paper will also compare these methods of analysis with more simplistic approaches based on the summary statistics. All the methods will be applied to a case study based on an adolescent anti-smoking intervention in Australia. The paper concludes with some general discussion and recommendations for routine design and analysis.

Testing the equality of two dependent kappa statistics

Procedures are developed and compared for testing the equality of two dependent kappa statistics in the case of two raters and a dichotomous outcome variable. Such problems may arise when each of a sample of subjects are rated under two distinct settings, and it is of interest to compare the observed levels of inter-observer and intra-observer agreement. The procedures compared are extensions of previously developed procedures for comparing kappa statistics computed from independent samples. The results of a Monte Carlo simulation show that adjusting for the dependency between samples tends to be worthwhile only if the between-setting correlation is comparable in magnitude to the within-setting correlations. In this case, a goodness-of-fit procedure that takes into account the dependency between samples is recommended.

Symptoms and suffering at the end of life in children with cancer

BACKGROUND

Cancer is the second leading cause of death in children, after accidents. Little is known, however, about the symptoms and suffering at the end of life in children with cancer.

METHODS

In 1997 and 1998, we interviewed the parents of children who had died of cancer between 1990 and 1997 and who were cared for at Children's Hospital, the Dana-Farber Cancer Institute, or both. Additional data were obtained by reviewing medical records.

RESULTS

Of 165 eligible parents, we interviewed 103 (62 percent), 98 by telephone and 5 in person. The interviews were conducted a mean (+/-SD) of 3.1+/-1.6 years after the death of the child. Almost 80 percent died of progressive disease, and the rest died of treatment-related complications. Forty-nine percent of the children died in the hospital; nearly half of these deaths occurred in the intensive care unit. According to the parents, 89 percent of the children suffered "a lot" or "a great deal" from at least one symptom in their last month of life, most commonly pain, fatigue, or dyspnea. Of the children who were treated for specific symptoms, treatment was successful in 27 percent of those with pain and 16 percent of those with dyspnea. On the basis of a review of the medical records, parents were significantly more likely than physicians to report that their child had fatigue, poor appetite, constipation, and diarrhea. Suffering from pain was more likely in children whose parents reported that the physician was not actively involved in providing end-of-life care (odds ratio, 2.6; 95 percent confidence interval, 1.0 to 6.7).

CONCLUSIONS

Children who die of cancer receive aggressive treatment at the end of life. Many have substantial suffering in the last month of life, and attempts to control their symptoms are often unsuccessful. Greater attention must be paid to palliative care for children who are dying of cancer.

Predicting costs of stem-cell transplantation

PURPOSE

Few studies have formally evaluated the relationship between costs, baseline patient characteristics, and major complications of stem-cell transplantation. We sought (1) to determine whether obtaining baseline information enabled identification of patients whose treatments would be the most costly and (2) to estimate inpatient costs for managing specific transplantation complications.

PATIENTS AND METHODS

We collected inpatient costs and clinical information for 236 consecutive patients undergoing transplantation at a single institution between July 1, 1994, and February 20, 1997. Multivariable linear regression was used to evaluate the associations between baseline patient characteristics and costs of hospitalization for initial transplantation and between clinical events and such costs.

RESULTS

The median initial inpatient cost in 1997 dollars was $55,500 for autologous transplantation (range, $28,200 to $148,200) and $105,300 for allogeneic transplantation (range, $32,500 to $338,000). When only baseline variables were considered, use of a mismatched allogeneic donor and year of transplantation were significant predictors of costs. No characteristics predicted which patients would incur the highest 10% of costs. When clinical events were considered, infection and in-hospital death were associated with higher costs in autologous transplant recipients ($18,400 and $20,500, respectively), whereas infection, veno-occlusive disease, acute graft-versus-host disease, and death were predicted to add between $15,300 and $28,100 each to allogeneic transplantation costs.

CONCLUSION

We were not able to identify before transplantation the patients whose treatments would be the most costly. However, the association between clinical complications and higher costs suggests that prevention may have significant economic benefits. Interventions that decrease these complications may have favorable cost-benefit ratios even if they do not affect overall survival.

Behavioral risk factors among women presenting for genetic testing

Considerable research attention has been given to the impact of genetic testing on psychological outcomes. Participation in genetic testing also may impact on health behaviors that increase the risk of cancer and other chronic diseases. The purpose of this study is to describe behavioral cancer risk factors of women who requested genetic testing for breast and ovarian cancer susceptibility (BRCA1, BRCA2). Before participation in a genetic testing program, 119 women completed a series of questionnaires designed to assess their health behaviors, perception of risk, and depressive symptomatology. Eight percent of participants were current smokers, 27% did not engage in at least moderate exercise, 46% did not regularly protect themselves from the sun, 39% did not consume at least five servings of fruits and vegetables per day, and 9% drank at least one alcoholic beverage per day. Poisson regression analysis revealed that age was the only predictor of behavioral risk profiles, with older women having fewer cancer risk behaviors. These patients who presented for genetic testing generally had better health behaviors than the general population. However, given their possible high-risk status, these patients should consider further improving their preventable cancer risk factors and, in particular, their diet, sun protection, and physical activity levels. Inclusion of behavioral risk factor counseling in the context of the genetic testing process may be an important opportunity to reach this at-risk population.

Testing the equality of twin correlations with multinomial outcomes

Twin studies are widely used to study genetic and environmental influences on human measurements. Correlations are often used in such studies to compare the levels of similarity between monozygotic and dizygotic twins with respect to a specified trait. In this paper, we compare three procedures for testing the equality of twin correlations when the outcome variable of interest is multinominal. One method is a likelihood ratio test based on an underlying Dirichlet-multinomial distribution. The second method is based on the estimated large sample variance of the estimated correlation, and the third method is based on a chi 2 goodness-of-fit test. The results of a Monte Carlo simulation show that the three methods have similar properties if the number of twin pairs is large (> 100), and the prevalence of the underlying trait is not extreme. Otherwise, the goodness-of-fit approach is to be preferred. We illustrate the methods by analyzing data from a previously published smoking study.

The merits of matching in community intervention trials: a cautionary tale

Concern about potential imbalance on risk factors in community intervention trials often prompts researchers to adopt a pair-matched design in which similar clusters of individuals are paired and one member of each matched pair is then randomly assigned to the intervention group. It is known that if there are few clusters in trial, it becomes increasingly difficult to obtain close matches on all potential risk factors. One may thus offset any gain in precision with loss in degrees of freedom due to matching. We shown in this paper that there are also several analytic limitations with pair-matched designs. These include: the restriction of prediction models to cluster-level baseline risk factors (for example, cluster size), the inability to test for homogeneity of odds ratios, and difficulties in estimating the intracluster correlation coefficient. These limitations lead us to present arguments that favour stratified designs in which there are more than two clusters in each stratum.

The statistical analysis of kappa statistics in multiple samples

Methods are presented for assessing and comparing the results of k > or = 2 independent samples of measured agreement or concordance, where in each sample a given member of a pair of observations is classified according to the presence or absence of a binary trait. Examples include the assessment of interobserver agreement across different groups of patients in a clinical study, investigations of sibling concordance across different genetic groups, and meta-analyses of observer agreement across different studies. The methodology described is based on application of goodness-of-fit theory to testing hypotheses concerning kappa statistics. Partitioning methods allow a variety of hypotheses to be tested, including an assessment of the degree of agreement within each sample, a testing procedure based on the pooled data, and a test of heterogeneity that may be used to assess the validity of pooling across samples. Three examples are given.

Statistical considerations in the design and analysis of community intervention trials

Community intervention trials are often characterized by the allocation of intact social units to different intervention groups. The assessment of adequate sample size for such trials must take into account the statistical dependencies among responses observed within an allocated unit. However, the small numbers of units typically involved in such trials imply that many methods of analysis that have been proposed for analyzing correlated data, particularly in the case of a dichotomous outcome variable, are not applicable to such designs. In this article we investigate this issue and determine the minimum number of units required per group, for the case of both a dichotomous and a continuous outcome variable, needed to provide adequate statistical power for detecting various levels of treatment effect. The use of significance testing as a method of detecting intracluster correlation is also investigated, and, in general, discouraged.

Testing the homogeneity of kappa statistics

Procedures are compared for testing the homogeneity of k > or = 2 independent kappa statistics in the case of two raters and a dichotomous outcome. One of the procedures is based on the estimated large sample variance derived under a model frequently adopted for inferences concerning interobserver agreement. The other is based on a goodness-of-fit approach to this model. The results of a Monte Carlo simulation show that the two approaches have similar properties if the number of subjects in each sample is large (> 100), and the prevalence of the underlying trait of interest is not extreme, while the goodness-of-fit approach is recommended for comparisons involving smaller numbers of subjects or in which the prevalence of the underlying trait is small (< 0.3).

Accounting for cluster randomization: a review of primary prevention trials, 1990 through 1993

OBJECTIVES

This methodological review aims to determine the extent to which design and analysis aspects of cluster randomization have been appropriately dealt with in reports of primary prevention trials.

METHODS

All reports of primary prevention trials using cluster randomization that were published from 1990 to 1993 in the American Journal of Public Health and Preventive Medicine were identified. Each article was examined to determine whether cluster randomization was taken into account in the design and statistical analysis.

RESULTS

Of the 21 articles, only 4 (19%) included sample size calculations or discussions of power that allowed for clustering, while 12 (57%) took clustering into account in the statistical analysis.

CONCLUSIONS

Design and analysis issues associated with cluster randomization are not recognized widely enough. Reports of cluster randomized trials should include sample size calculations and statistical analyses that take clustering into account, estimates of design effects to help others planning trials, and a table showing the baseline distribution of important characteristics by intervention group, including the number of clusters and average cluster size for each group.

Cluster randomization trials in tropical medicine: a case study

Field trials in tropical medicine are often designed so that intact social units (e.g., families, schools, communities) rather than independent individuals are randomized to an intervention group. Reasons are diverse, but include administrative convenience, a desire to reduce the effect of treatment contamination, and the need to avoid ethical issues that might otherwise arise. Dependencies among cluster members typical of such designs must be considered when determining sample size and analysing the resulting data. Failure to do so can result in false conclusions that the treatment is effective. The purpose of this paper is to compare different methods which can be used to construct tests of the effect of treatment when outcomes are binary (e.g., infected/uninfected). The discussion will be illustrated using data from a trial which randomly assigned families to either a control group or a screening and treatment programme for imported intestinal parasites.

Statistical methodology for estimating twin similarity with respect to a dichotomous trait

Statistical methodology is presented for the estimation of twin similarity with respect to a dichotomous trait. The methodology focuses on the intraclass correlation as the parameter of interest and is analogous to methodology commonly applied to continuous outcome data. For inference problems involving a single sample, confidence interval construction and significance testing are discussed. For two sample problems, test procedures are provided that are an alternative to an approach recently presented by Ramakrishnan et al. [(1992) Genet Epidemiol 9:273-287 (1)]. Two examples based on published data sets are given to illustrate the proposed techniques.

Methods for comparing event rates in intervention studies when the unit of allocation is a cluster

The aim of many research investigations is to compare the proportion of individuals in each of several groups that have a certain characteristic. The unit of allocation for such investigations is often an intact social unit, as in randomizing families, medical practices, schools, or entire communities, to different intervention groups. Standard statistical methods are not appropriate for these designs, since they do not take into account the dependencies among individuals within the same cluster. The authors review the strengths and weaknesses of several approaches for dealing with this problem, using data from a school-based smoking cessation trial. A principal conclusion is that the choice of method should depend on whether or not random allocation is used in the assignment of interventions.

A comparison of methods for testing homogeneity of proportions in teratologic studies

We consider teratologic studies in which the aim is to compare the survival rate of animals in a treatment group to the corresponding rate in a control group. The design of such studies often involves the allocation of intact litters of animals to treatment, invalidating the application of standard statistical methods. We review the strengths and weaknesses of several approaches for dealing with this problem including methodology recently developed for the analysis of clustered binary data. A simulation study is conducted in which litter sizes are generated from a distribution having specified mean and degree of imbalance. It is recommended on the basis of this study and on theoretical considerations that the choice of method should depend on whether the comparison of interest is experimental or observational. For experimental comparisons, involving the random assignment of litters to different treatment groups, methods based on the adjustment of standard chi-square statistics are recommended unless the number of litters in each group is very large.

Initial nonfunction in cadaveric renal transplantation

Patients who receive a kidney transplant that does not function immediately have more complications and decreased graft survival than patients whose allografts function immediately. To determine the causes of initial nonfunction (INF), the authors reviewed 188 consecutive cadaveric kidney transplants performed between 1985 and 1988 at the University Hospital, London, Ont. Data were collected on 16 putative risk factors for INF, which were divided into three categories: donor, recipient and technical. INF was defined as the need for dialysis within 7 days of transplantation. Forty-eight (26%) of the 188 allografts had INF, 6 of which never functioned. Univariate analysis identified five variables associated with increased risk of INF: no donor dopamine use, back-table flush, single-organ retrieval, exchanged kidney and prolonged cold ischemic time. Multivariate analysis, however, identified only three variables associated with INF: cerebrovascular accident as the cause of donor death, no donor dopamine use and single-organ retrieval. The authors recommend (a) low-dose dopamine therapy for all donors and (b) multiorgan retrieval to produce quality kidneys for transplantation.

Confidence interval construction for effect measures arising from cluster randomization trials

Methods of confidence interval construction are provided for summary measures of treatment effect arising from designs randomizing clusters to one of two treatment groups. Three basic designs are considered for the case of continuous and dichotomous variables: completely randomized, pair-matched and stratified.

Enhanced Ig production by human peripheral lymphocytes induced by aggregated C1q

Because B cells express receptors for C1q, we have investigated the role of C1q in the stimulation of B cells. When B cells were cultured in the presence of C1q that had been frozen, T cells, and suboptimal concentrations of PWM, there was a dose-dependent enhancement of IgM, IgG, and IgA by the B cells. No significant enhancement of Ig production by B cells was seen in the absence of T cells or PWM. The contribution of T cells or PWM could be replaced by supernatants of PMA and Con A-activated PBMC (T cell growth factor). C1q that had been frozen, in contrast with freshly isolated C1q, was at least 3 times more active in enhancement of the production of Ig by B cells in culture in the presence of suboptimal concentrations of T cell growth factor. The capability of C1q to stimulate B cells could be ascribed to aggregates of C1q. Monomeric C1q was only marginally active to stimulate B cell Ig production, whereas dimeric and tetrameric C1q were able to enhance Ig production by B cells in relation to their size. Furthermore, aggregation of C1q on soluble aggregates of rabbit IgM also increased its potential to enhance B cell Ig production. The interaction of C1q with the B cells occurs via the collagenous tail of C1q, as suggested by inhibition experiments with purified collagenous tails and globular heads of C1q. These results indicate that triggering of C1qR on B cells positively regulates Ig production in vitro.

Enhanced Ig production by human peripheral lymphocytes induced by aggregated C1q

Because B cells express receptors for C1q, we have investigated the role of C1q in the stimulation of B cells. When B cells were cultured in the presence of C1q that had been frozen, T cells, and suboptimal concentrations of PWM, there was a dose-dependent enhancement of IgM, IgG, and IgA by the B cells. No significant enhancement of Ig production by B cells was seen in the absence of T cells or PWM. The contribution of T cells or PWM could be replaced by supernatants of PMA and Con A-activated PBMC (T cell growth factor). C1q that had been frozen, in contrast with freshly isolated C1q, was at least 3 times more active in enhancement of the production of Ig by B cells in culture in the presence of suboptimal concentrations of T cell growth factor. The capability of C1q to stimulate B cells could be ascribed to aggregates of C1q. Monomeric C1q was only marginally active to stimulate B cell Ig production, whereas dimeric and tetrameric C1q were able to enhance Ig production by B cells in relation to their size. Furthermore, aggregation of C1q on soluble aggregates of rabbit IgM also increased its potential to enhance B cell Ig production. The interaction of C1q with the B cells occurs via the collagenous tail of C1q, as suggested by inhibition experiments with purified collagenous tails and globular heads of C1q. These results indicate that triggering of C1qR on B cells positively regulates Ig production in vitro.

Binding and catabolism of aggregated immunoglobulins bearing C3b or iC3b by U937 cells

Mononuclear cells play an important role in the elimination of immune complexes (IC). In the presence of complement (C) the binding and degradation of IC by mononuclear cells is enhanced at least two-fold. The enhancement of binding is caused by a synergistic interaction of the IC with cellular Fc and complement receptors (R). In the present study we have investigated the contribution of the complement receptors CR1 and CR3 of human monocyte cell line U937 on the complement-mediated binding and degradation of immune complexes and soluble aggregates of IgG (AIgG) bearing C3b or iC3b. It was found that deposition of C3b on AIgG enhanced the binding of AIgG to U937 cells at least two-fold. The C3b-mediated enhancement of binding was abolished by anti-CR1. iC3b-bound to AIgG also enhanced the binding of AIgG to the cells. This binding was only partially reduced by anti-CR3 antibodies, but the combination of anti-CR1 and anti-CR3 fully abolished the iC3b-mediated enhancement of binding. These results suggest that both CR1 and CR3 contribute to the complement-mediated binding and degradation of soluble IC by mononuclear phagocytes.

The pathophysiology of protein-overload proteinuria

Alterations in glomerular function and structure were studied in protein-overload nephrosis in the rat induced by intraperitoneal administration of bovine serum albumin (BSA). Fractional clearance (C/GFR) studies using inulin and tracer proteins of different molecular size and charge revealed in proteinuric rats 1) unchanged glomerular filtration rate and renal plasma flow; 2) a 34-fold increase in C/GFR of rat serum albumin, reaching values similar to BSA; 3) a 2-fold increase in C/GFR for anionic horse radish peroxidase (HRP), but normal values for neutral and cationic HRP, and 4) an 11- and 3-fold increase for heterologous IgG and IgM, respectively. Glomerular epithelial cells showed degenerative changes, but the distribution of anionic sites in the glomerular basement membrane was found to be unaltered, as determined by polyethyleneimine binding studies. In summary, an elevation of serum albumin concentration resulted in an increased transcapillary albumin transport. This was found to lead to degenerative changes of glomerular epithelial cells with development of large pore defects, which were completely reversible.