Ascertaining prognosis for breast cancer in node-negative patients with innovative survival analysis

Development and validation of a nomogram for survival benefit of lymphadenectomy in resected gallbladder cancer

Background

Due to absence of large, prospective, randomized, clinical trial data, the potential survival benefit of lymphadenectomy with different number of regional lymph nodes (LNs) remains controversial. We aim to create a predicting model to help estimate individualized potential survival benefit of lymphadenectomy with more regional LNs for patients with resected gallbladder cancer (GBC).

Methods

Patients with resected GBC were selected from the Surveillance, Epidemiology, and End Results database who were diagnosed between 2004 and 2014. Covariates included age, race, sex, grade, histological stage, tumor sizes and receipt of non-primary surgery. Two types of multivariate survival regression models were constructed and compared. The best model performance was tested by the external validation data from our hospital.

Results

A total of 1,669 patients met the inclusion criteria for this study. The lognormal survival model showed the best performance and was tested by the external validation data, including 193 patients with resected GBC from our hospital. Nomograms, which based on the accelerated failure time parametric survival model, were built to estimate individualized survival. C-index, was up to 0.754 and 0.710 in internal validation for more and less regional LNs removed, respectively. Both of internal and external calibration curves showed good agreement between predicted and observed outcomes in the 1-, 3-, and 5-year overall survival (OS).

Conclusions

A predicting model can be used as a decision model to predict which patients may obtain benefit from lymphadenectomy with more regional LNs.

Relapse-free survival of statistically standardized continuous RT-PCR estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2): NCIC CTG MA.14

Recent ASCO/CAP guidelines focus on decision making associated with the presence/absence of continuous breast biomarkers. Statistical standardization (SS) is demonstrated as a method to evaluate the effects of continuous RT-PCR biomarker expression levels on breast cancer outcomes. MA.14 allocated 667 postmenopausal patients to tamoxifen based on locally determined ER/PR. Of 299 available patient tumor samples, 292 passed internal quality control. All tumors were centrally assessed by RT-PCR ER/PR/HER2 with each biomarker's z-scores categorized: ≥1.0 standard deviation (SD) below mean; <1.0 SD below mean; ≤1.0 SD above mean; >1.0 SD above mean. Log-rank statistics tested univariate differences in breast cancer relapse-free survival (RFS). Continuous SS-ER/PR/HER2 were assessed in multivariate Cox step-wise forward regression, adding a factor if p ≤ 0.05. Sensitivity analyses examined an external HER2+ cut-point of 1.32. Patients whose tumors were tested were representative of the MA.14 population (p values = 0.18-0.90). At 9.8 years median follow-up, SS-ER did not univariately impact RFS (p = 0.31). SS-PR values above the mean (z ≥ 0.0) had the best univariate RFS (p = 0.03). SS-HER2 also univariately impacted RFS (p = 0.004) with lowest (z-scores ≤ -1.0) and highest (z-scores > 1.0) having shortest RFS. Multivariate stratified/unstratified Cox models indicated patients with T1 tumors (p = 0.02/p = 0.0002) and higher SS-PR (p = 0.02/p = 0.01) had longer RFS; node-negative patients had better RFS (in unstratified analysis, p < 0.0001). Local ER/PR status did not impact RFS (p > 0.05). Patients with SS HER2+ ≥ 1.32 had worse RFS (univariate, p = 0.05; multivariate, p = 0.06). We demonstrated that higher SS-PR, and SS HER2 levels, measured by RT-PCR impacted breast cancer RFS outcomes. Evaluation in other trials may provide support for this methodology.

Effect of continuous statistically standardized measures of estrogen and progesterone receptors on disease-free survival in NCIC CTG MA.12 Trial and BC Cohort

INTRODUCTION

We hypothesized improved inter-laboratory comparability of estrogen receptor (ER) and progesterone receptor (PgR) across different assay methodologies with adjunctive statistical standardization, akin to bone mineral density (BMD) z-scores. We examined statistical standardization in MA.12, a placebo-controlled pre-menopausal trial of adjuvant tamoxifen with locally assessed hormone receptor +/- tumours, and in a cohort of post-menopausal British Columbia (BC) tamoxifen-treated patients.

METHODS

ER and PgR were centrally assessed for both patient groups with real time quantitative reverse transcription polymerase chain reaction (qPCR) and immunohistochemistry (IHC). Effects on disease-free survival (DFS) were investigated separately for 345 MA.12 and 673 BC patients who had both qPCR and IHC assessments. Comparisons utilized continuous laboratory units and statistically standardized z-scores. Univariate categorization of ER/PgR was by number of standard deviations (SD) above or below the mean (z-score ≥1.0 SD below mean; z-score <1.0 SD below mean; z-score ≤1.0 SD above mean; z-score >1.0 SD above mean). Exploratory multivariate examinations utilized step-wise Cox regression.

RESULTS

Median follow-up for MA.12 was 9.7 years; for BC patients, 11.8 years. For MA.12, 101 of 345 (29%) patients were IHC ER-PgR-. ER was not univariately associated with DFS (qPCR, P = 0.19; IHC, P = 0.08), while PgR was (qPCR, P = 0.09; IHC, P = 0.04). For BC patients, neither receptor was univariately associated with DFS: for ER, PCR, P = 0.36, IHC, P = 0.24; while for PgR, qPCR, P = 0.17, IHC, P = 0.31. Multivariately, MA.12 patients randomized to tamoxifen had significantly better DFS (P = 0.002 to 0.005) than placebo. Meanwhile, jointly ER and PgR were not associated with DFS whether assessed by qPCR or by IHC in all patients, or in the subgroup of patients with IHC positive stain, for pooled or separate treatment arms. Different results by type of continuous unit supported the concept of ER level being relevant for medical decision-making. For postmenopausal BC tamoxifen patients, higher qPCR PgR was weakly associated with better DFS (P = 0.06).

CONCLUSIONS

MA.12 pre-menopausal patients in a placebo-controlled tamoxifen trial had similar multivariate prognostic effects with statistically standardized hormone receptors when tumours were assayed by qPCR or IHC, for hormone receptor +/- and + tumours. The BC post-menopausal tamoxifen cohort did not exhibit a significant prognostic association of ER or PgR with DFS. Adjunctive statistical standardization is currently under investigation in other NCIC CTG endocrine trials.

Mean overall survival gain with aflibercept plus FOLFIRI vs placebo plus FOLFIRI in patients with previously treated metastatic colorectal cancer

BACKGROUND

Mean survival in cancer trials can be estimated with statistical techniques to extrapolate study survival curves. This methodology was applied to data from the VELOUR trial, where use of the novel biologic aflibercept (ziv-aflibercept in the United States) in combination with fluorouracil+leucovorin+irinotecan (FOLFIRI), had significantly increased median overall survival (OS) by 1.44 months, vs placebo plus FOLFIRI in patients with metastatic colorectal cancer (mCRC) resistant to, or that had progressed following, an oxaliplatin-containing regimen.

METHODS

Parametric survival analyses were used to identify distributions with the best fit to the empirical VELOUR data. Mean OS for the two treatment groups (and pre-defined subgroups) was calculated from the fitted curves over a 15-year survival period.

RESULTS

Overall, the log-logistic distribution was the best-fitting for both treatment arms and, with it, the estimated difference in mean OS over 15 years between aflibercept+FOLFIRI and placebo+FOLFIRI was 4.7 months. In addition, the survival advantage with aflibercept was at least 3 months for the ITT population, whichever distribution was used to extrapolate survival.

CONCLUSION

Extrapolation of survival curves suggests the mean OS difference for aflibercept in the VELOUR trial is at least 3 months in the ITT population and selected subgroups.

Applying conventional and saturated generalized gamma distributions in parametric survival analysis of breast cancer

BACKGROUND

The generalized gamma distribution statistics constitute an extensive family that contains nearly all of the most commonly used distributions including the exponential, Weibull and log normal. A saturated version of the model allows covariates having effects through all the parameters of survival time distribution. Accelerated failure-time models assume that only one parameter of the distribution depends on the covariates.

METHODS

We fitted both the conventional GG model and the saturated form for each of its members including the Weibull and lognormal distribution; and compared them using likelihood ratios. To compare the selected parameter distribution with log logistic distribution which is a famous distribution in survival analysis that is not included in generalized gamma family, we used the Akaike information criterion (AIC; r=l(b)-2p). All models were fitted using data for 369 women age 50 years or more, diagnosed with stage IV breast cancer in BC during 1990-1999 and followed to 2010.

RESULTS

In both conventional and saturated parametric models, the lognormal was the best candidate among the GG family members; also, the lognormal fitted better than log-logistic distribution. By the conventional GG model, the variables "surgery", "radiotherapy", "hormone therapy", "erposneg" and interaction between "hormone therapy" and "erposneg"are significant. In the AFT model, we estimated the relative time for these variables. By the saturated GG model, similar significant variables are selected. Estimating the relative times in different percentiles of extended model illustrate the pattern in which the relative survival time change during the time.

CONCLUSIONS

The advantage of using the generalized gamma distribution is that it facilitates estimating a model with improved fit over the standard Weibull or log- normal distributions. Alternatively, the generalized F family of distributions might be considered, of which the generalized gamma distribution is a member and also includes the commonly used log-logistic distribution.

Modelling survival in hepatocellular carcinoma

OBJECTIVES

To identify the pattern of the risk of death over long-term in unresectable hepatocellular carcinoma by determining the appropriate distribution to extrapolate overall survival and to assess the role of the Weibull distribution as the standard survival model in oncology.

RESEARCH DESIGN AND METHODS

To select the appropriate distribution, three types of data sources have been analysed. Patient level data from two randomized controlled trials and published Kaplan-Meier curves from a systematic literature review provided short term follow-up data. They were supplemented with patient level data, with long-term follow-up from the Cancer Institute New South Wales, Australia. Published Kaplan-Meier curves were read in and a time-to-event dataset was created. Distributions were fitted to the data from the different sources separately. Their fit was assessed visually and compared using statistical criteria based on log-likelihood, the Akaike information criterion (AIC), and the Bayesian information criterion (BIC).

RESULTS

Based on both published and patient-level, and both short- and long-term follow-up data, the Weibull distribution, used very often in cost-effectiveness models in oncology, does not seem to offer a good fit in hepatocellular carcinoma among the different survival models. The best fitting distribution appears to be the lognormal, with loglogistic as the second-best fitting function. Results were consistent between the different sources of data.

CONCLUSIONS

In unresectable hepatocellular carcinoma, the Weibull model, which is often treated at the gold standard, does not appear to be appropriate based on different sources of data (two clinical trials, a retrospective database and published Kaplan-Meier curves). Lognormal distribution seems to be the most appropriate distribution for extrapolating overall survival.

Nomogram for predicting the benefit of adjuvant chemoradiotherapy for resected gallbladder cancer

PURPOSE

Although adjuvant chemoradiotherapy for resected gallbladder cancer may improve survival for some patients, identifying which patients will benefit remains challenging because of the rarity of this disease. The specific aim of this study was to create a decision aid to help make individualized estimates of the potential survival benefit of adjuvant chemoradiotherapy for patients with resected gallbladder cancer.

METHODS

Patients with resected gallbladder cancer were selected from the Surveillance, Epidemiology, and End Results (SEER) -Medicare database who were diagnosed between 1995 and 2005. Covariates included age, race, sex, stage, and receipt of adjuvant chemotherapy or chemoradiotherapy (CRT). Propensity score weighting was used to balance covariates between treated and untreated groups. Several types of multivariate survival regression models were constructed and compared, including Cox proportional hazards, Weibull, exponential, log-logistic, and lognormal models. Model performance was compared using the Akaike information criterion. The primary end point was overall survival with or without adjuvant chemotherapy or CRT.

RESULTS

A total of 1,137 patients met the inclusion criteria for the study. The lognormal survival model showed the best performance. A Web browser-based nomogram was built from this model to make individualized estimates of survival. The model predicts that certain subsets of patients with at least T2 or N1 disease will gain a survival benefit from adjuvant CRT, and the magnitude of benefit for an individual patient can vary.

CONCLUSION

A nomogram built from a parametric survival model from the SEER-Medicare database can be used as a decision aid to predict which gallbladder patients may benefit from adjuvant CRT.

Considerations in the statistical analysis of hemodialysis patient survival

The association of hemodialysis dosage with patient survival is controversial. Here, we tested the hypothesis that methods for survival analysis may influence conclusions regarding dialysis dosage and mortality. We analyzed all-cause mortality by proportional hazards and accelerated failure time regression models in a cohort of incident hemodialysis patients who were followed for 9 yr. Both models identified age, race, heart failure, physical functioning, and comorbidity scores as important predictors of patient survival. Using proportional hazards, there was no statistically significant association between mortality and Kt/V (hazard ratio 0.72; 95% confidence interval 0.45 to 1.14). In contrast, using accelerated failure time models, each 0.1-U increment of Kt/V improved adjusted median patient survival by 3.50% (95% confidence interval 0.20 to 7.08%). Proportional hazard models also yielded less accurate estimates for median survival. These findings are consistent with an additive damage model for the survival of patients who are on hemodialysis. In this conceptual model, the assumptions of the proportional hazard model are violated, leading to underestimation of the importance of dialysis dosage. These results suggest that future studies of dialysis adequacy should consider this additive damage model when selecting methods for survival analysis. Accelerated failure time models may be useful adjuncts to the Cox model when studying outcomes of dialysis patients.

Avoiding Pitfalls in the Statistical Analysis of Heterogeneous Tumors

Information about tumors is usually obtained from a single assessment of a tumor sample, performed at some point in the course of the development and progression of the tumor, with patient characteristics being surrogates for natural history context. Differences between cells within individual tumors (intratumor heterogeneity) and between tumors of different patients (intertumor heterogeneity) may mean that a small sample is not representative of the tumor as a whole, particularly for solid tumors which are the focus of this paper. This issue is of increasing importance as high-throughput technologies generate large multi-feature data sets in the areas of genomics, proteomics, and image analysis. Three potential pitfalls in statistical analysis are discussed (sampling, cut-points, and validation) and suggestions are made about how to avoid these pitfalls.

Competing causes of death from a randomized trial of extended adjuvant endocrine therapy for breast cancer

BACKGROUND

Older women with early-stage breast cancer experience higher rates of non-breast cancer-related death. We examined factors associated with cause-specific death in a large cohort of breast cancer patients treated with extended adjuvant endocrine therapy.

METHODS

In the MA.17 trial, conducted by the National Cancer Institute of Canada Clinical Trials Group, 5170 breast cancer patients (median age = 62 years; range = 32-94 years) who were disease free after approximately 5 years of adjuvant tamoxifen treatment were randomly assigned to treatment with letrozole (2583 women) or placebo (2587 women). The median follow-up was 3.9 years (range 0-7 years). We investigated the association of 11 baseline factors with the competing risks of death from breast cancer, other malignancies, and other causes. All statistical tests were two-sided likelihood ratio criterion tests.

RESULTS

During follow-up, 256 deaths were reported (102 from breast cancer, 50 from other malignancies, 100 from other causes, and four from an unknown cause). Non-breast cancer deaths accounted for 60% of the 252 known deaths (72% for those > or = 70 years and 48% for those < 70 years). Two baseline factors were differentially associated with type of death: cardiovascular disease was associated with a statistically significant increased risk of death from other causes (P.002), and osteoporosis was associated with a statistically significant increased risk of death from other malignancies (P.05). An increased risk of breast cancer-specific death was associated with lymph node involvement (P < .001). Increased risk of death from all three causes was associated with older age (P < .001).

CONCLUSIONS

Non-breast cancer-related deaths were more common than breast cancer-specific deaths in this cohort of 5-year breast cancer survivors, especially among older women.

Disease-specific survival for limited-stage small-cell lung cancer affected by statistical method of assessment

BACKGROUND

In general, prognosis and impact of prognostic/predictive factors are assessed with Kaplan-Meier plots and/or the Cox proportional hazard model. There might be substantive differences from the results using these models for the same patients, if different statistical methods were used, for example, Boag log-normal (cure-rate model), or log-normal survival analysis.

METHODS

Cohort of 244 limited-stage small-cell lung cancer patients, were accrued between 1981 and 1998, and followed to the end of 2005. The endpoint was death with or from lung cancer, for disease-specific survival (DSS). DSS at 1-, 3- and 5-years, with 95% confidence limits, are reported for all patients using the Boag, Kaplan-Meier, Cox, and log-normal survival analysis methods. Factors with significant effects on DSS were identified with step-wise forward multivariate Cox and log-normal survival analyses. Then, DSS was ascertained for patients with specific characteristics defined by these factors.

RESULTS

The median follow-up of those alive was 9.5 years. The lack of events after 1966 days precluded comparison after 5 years. DSS assessed by the four methods in the full cohort differed by 0-2% at 1 year, 0-12% at 3 years, and 0-1% at 5 years. Log-normal survival analysis indicated DSS of 38% at 3 years, 10-12% higher than with other methods; univariate 95% confidence limits were non-overlapping. Surgical resection, hemoglobin level, lymph node involvement, and superior vena cava (SVC) obstruction significantly impacted DSS. DSS assessed by the Cox and log-normal survival analysis methods for four clinical risk groups differed by 1-6% at 1 year, 15-26% at 3 years, and 0-12% at 5 years; multivariate 95% confidence limits were overlapping in all instances.

CONCLUSION

Surgical resection, hemoglobin level, lymph node involvement, and superior vena cava (SVC) obstruction all significantly impacted DSS. Apparent DSS for patients was influenced by the statistical methods of assessment. This would be clinically relevant in the development or improvement of clinical management strategies.