Performance of binary prediction models in high-correlation low-dimensional settings: a comparison of methods.

BACKGROUND: Clinical prediction models are developed widely across medical disciplines. When predictors in such models are highly collinear, unexpected or spurious predictor-outcome associations may occur, thereby potentially reducing face-validity of the prediction model. Collinearity can be dealt with by exclusion of collinear predictors, but when there is no a priori motivation (besides collinearity) to include or exclude specific predictors, such an approach is arbitrary and possibly inappropriate. METHODS: We compare different methods to address collinearity, including shrinkage, dimensionality reduction, and constrained optimization. The effectiveness of these methods is illustrated via simulations. RESULTS: In the conducted simulations, no effect of collinearity was observed on predictive outcomes (AUC, R2, Intercept, Slope) across methods. However, a negative effect of collinearity on the stability of predictor selection was found, affecting all compared methods, but in particular methods that perform strong predictor selection (e.g., Lasso). Methods for which the included set of predictors remained most stable under increased collinearity were Ridge, PCLR, LAELR, and Dropout. CONCLUSIONS: Based on the results, we would recommend refraining from data-driven predictor selection approaches in the presence of high collinearity, because of the increased instability of predictor selection, even in relatively high events-per-variable settings. The selection of certain predictors over others may disproportionally give the impression that included predictors have a stronger association with the outcome than excluded predictors.

Pre- and Postoperative Capecitabine Without or With Oxaliplatin in Locally Advanced Rectal Cancer: PETACC 6 Trial by EORTC GITCG and ROG, AIO, AGITG, BGDO, and FFCD.

PURPOSE: The PETACC 6 trial investigates whether the addition of oxaliplatin to preoperative capecitabine-based chemoradiation and postoperative capecitabine improves disease-free survival (DFS) in locally advanced rectal cancer. METHODS: Between November 2008 and September 2011, patients with rectal adenocarcinoma within 12 cm from the anal verge, T3/4 and/or node positive, were randomly assigned to 5 weeks preoperative capecitabine-based chemoradiation (45-50.4 Gy) followed by six cycles of adjuvant capecitabine, both without (control arm, 1) or with (experimental arm, 2) oxaliplatin. The primary end point was improvement of 3-year DFS by oxaliplatin from 65% to 72% (hazard ratio [HR], 0.763). RESULTS: A total of 1,094 patients were randomly assigned (intention to treat), and 1,068 eligible patients started their allocated treatment (arm 1, 543; arm 2, 525), with completion of protocol treatment in 68% (arm 1) v 54% (arm 2). A higher rate of grade 3/4 adverse events was reported in the experimental arm (14.4% v 37.3% and 23.4% v 46.6% for neoadjuvant and adjuvant treatment, respectively). At a median follow-up of 68 months (interquartile range, 58-74 months), 157 and 156 DFS events were observed in arms 1 and 2, respectively (adjusted HR, 1.02; 95% CI, 0.82 to 1.28; P = .835). Three-year DFS rate was not different, with 76.5% (95% CI, 72.7% to 79.9%) in arm 1, which is higher than anticipated, and 75.8% (95% CI, 71.9% to 79.3%) in arm 2. The 7-year DFS and overall survival (OS) rates were not different as well, with DFS of 66.1% v 65.5% (HR, 1.02) and OS of 73.5% v 73.7% (HR, 1.19) in arms 1 and 2, respectively. Subgroup analyses revealed heterogeneity in treatment effect according to German versus non-German site location, without detectable confounding factors in multivariable analysis. CONCLUSION: The addition of oxaliplatin to preoperative capecitabine-based chemoradiation and postoperative adjuvant chemotherapy impairs tolerability and feasibility and does not improve efficacy.

Lapatinib with ECF/X in the first-line treatment of metastatic gastric cancer according to HER2neu and EGFR status: a randomized placebo-controlled phase II study (EORTC 40071).

PURPOSE: HER2-targeted therapy with trastuzumab and (CF/X) prolonged overall survival (OS) in metastatic HER2neu+ gastric carcinoma (GC). Lapatinib inhibits both EGFR and HER2neu. We investigated the efficacy and safety of lapatinib with epirubicin (E) + CF/X in GC according to HER2neu and EGFR status. METHODS: Tumors from chemotherapy-naïve patients were screened centrally by fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC). Patients with EGFR and/or HER2neu expression or amplification were allocated to three strata based on EGFR/HER2neu status and were randomized to lapatinib (arm A) or placebo (arm B), with 6 cycles of ECF or ECX (investigator-selected). The primary endpoint was progression-free survival (PFS) in stratum 3. RESULTS: 29 of 72 screened patients were randomized to strata 1 (HER2neu+: by FISH and IHC, n = 6), 2 (HER2neu-: by FISH/+ by IHC, n = 5) and 3 (HER2neu-/EGFR+, n = 18), of which 28 patients were eligible (14 per arm). Enrollment was curtailed after announcement of the negative LOGiC trial results. Median PFS was 8.0 versus 5.9 months (HR = 0.86, 95% CI 0.37-1.99) in the per protocol population, and 8.0 versus 6.3 months (HR = 0.85, 95% CI 0.30-2.46) for stratum 3, in the lapatinib versus placebo arm respectively. Median OS was 13.8 versus 10.1 months, respectively (HR = 0.90, 95% CI 0.35-2.27). There were no safety concerns. CONCLUSIONS: Central EGFR and HER2neu stratification by IHC and FISH can be used for further pan-HER strategies. Lapatinib with ECF/X was well tolerated, but did not show clear activity in patients with metastatic GC.

Neoadjuvant chemotherapy compared with surgery alone for locally advanced cancer of the stomach and cardia: European Organisation for Research and Treatment of Cancer randomized trial 40954.

PURPOSE: Patients with locally advanced gastric cancer benefit from combined pre- and postoperative chemotherapy, although fewer than 50% could receive postoperative chemotherapy. We examined the value of purely preoperative chemotherapy in a phase III trial with strict preoperative staging and surgical resection guidelines. PATIENTS AND METHODS: Patients with locally advanced adenocarcinoma of the stomach or esophagogastric junction (AEG II and III) were randomly assigned to preoperative chemotherapy followed by surgery or to surgery alone. To detect with 80% power an improvement in median survival from 17 months with surgery alone to 24 months with neoadjuvant, 282 events were required. RESULTS: This trial was stopped for poor accrual after 144 patients were randomly assigned (72:72); 52.8% patients had tumors located in the proximal third of the stomach, including AEG type II and III. The International Union Against Cancer R0 resection rate was 81.9% after neoadjuvant chemotherapy as compared with 66.7% with surgery alone (P = .036). The surgery-only group had more lymph node metastases than the neoadjuvant group (76.5% v 61.4%; P = .018). Postoperative complications were more frequent in the neoadjuvant arm (27.1% v 16.2%; P = .09). After a median follow-up of 4.4 years and 67 deaths, a survival benefit could not be shown (hazard ratio, 0.84; 95% CI, 0.52 to 1.35; P = .466). CONCLUSION: This trial showed a significantly increased R0 resection rate but failed to demonstrate a survival benefit. Possible explanations are low statistical power, a high rate of proximal gastric cancer including AEG and/or a better outcome than expected after radical surgery alone due to the high quality of surgery with resections of regional lymph nodes outside the perigastic area (celiac trunc, hepatic ligament, lymph node at a. lienalis; D2).

Duration of androgen suppression in the treatment of prostate cancer.

BACKGROUND: The combination of radiotherapy plus long-term medical suppression of androgens (> or = 2 years) improves overall survival in patients with locally advanced prostate cancer. We compared the use of radiotherapy plus short-term androgen suppression with the use of radiotherapy plus long-term androgen suppression in the treatment of locally advanced prostate cancer. METHODS: We randomly assigned patients with locally advanced prostate cancer who had received external-beam radiotherapy plus 6 months of androgen suppression to two groups, one to receive no further treatment (short-term suppression) and the other to receive 2.5 years of further treatment with a luteinizing hormone-releasing hormone agonist (long-term suppression). An outcome of noninferiority of short-term androgen suppression as compared with long-term suppression required a hazard ratio of more than 1.35 for overall survival, with a one-sided alpha level of 0.05. An interim analysis showed futility, and the results are presented with an adjusted one-sided alpha level of 0.0429. RESULTS: A total of 1113 men were registered, of whom 970 were randomly assigned, 483 to short-term suppression and 487 to long-term suppression. After a median follow-up of 6.4 years, 132 patients in the short-term group and 98 in the long-term group had died; the number of deaths due to prostate cancer was 47 in the short-term group and 29 in the long-term group. The 5-year overall mortality for short-term and long-term suppression was 19.0% and 15.2%, respectively; the observed hazard ratio was 1.42 (upper 95.71% confidence limit, 1.79; P=0.65 for noninferiority). Adverse events in both groups included fatigue, diminished sexual function, and hot flushes. CONCLUSIONS: The combination of radiotherapy plus 6 months of androgen suppression provides inferior survival as compared with radiotherapy plus 3 years of androgen suppression in the treatment of locally advanced prostate cancer. (ClinicalTrials.gov number, NCT00003026.)

Prophylactic cranial irradiation in extensive disease small-cell lung cancer: short-term health-related quality of life and patient reported symptoms: results of an international Phase III randomized controlled trial by the EORTC Radiation Oncology and Lung Cancer Groups.

PURPOSE: Prophylactic cranial irradiation (PCI) in patients with extensive-disease small-cell lung cancer (ED-SCLC) leads to significantly fewer symptomatic brain metastases and improved survival. Detailed effects of PCI on health-related quality of life (HRQOL) are reported here. PATIENTS AND METHODS: Patients (age, 18 to 75 years; WHO < or = 2) with ED-SCLC, and any response to chemotherapy, were randomly assigned to either observation or PCI. Health-related quality of life (HRQOL) and patient-reported symptoms were secondary end points. The European Organisation for the Research and Treatment of Cancer core HRQOL tool (Quality of Life Questionnaire C30) and brain module (Quality of Life Questionnaire Brain Cancer Module) were used to collect self-reported patient data. Six HRQOL scales were selected as primary HRQOL end points: global health status; hair loss; fatigue; and role, cognitive and emotional functioning. Assessments were performed at random assignment, 6 weeks, 3 months, and then 3-monthly up to 1 year and 6-monthly thereafter. RESULTS: Compliance with the HRQOL assessment was 93.7% at baseline and dropped to 60% at 6 weeks. Short-term results up to 3 months showed that there was a negative impact of PCI on selected HRQOL scales. The largest mean difference between the two arms was observed for fatigue and hair loss. The impact of PCI on global health status as well as on functioning scores was more limited. For global health status, the observed mean difference was eight points on a scale 0 to 100 at 6 weeks (P = .018) and 3 months (P = .055). CONCLUSION: PCI should be offered to all responding ED SCLC patients. Patients should be informed of the potential adverse effects from PCI. Clinicians should be alert to these; monitor their patients; and offer appropriate support, clinical, and psychosocial care.

Evaluating health-related quality of life and symptom burden in brain tumour patients: instruments for use in experimental trials and clinical practice.

PURPOSE OF REVIEW: The evaluation of new treatments in brain cancer should address outcomes such as functional status, symptoms, cognition and health-related quality of life (HRQOL), yet these are infrequently evaluated. This review focuses on instruments for measuring HRQOL in brain tumour patients and the recent developments in the field. RECENT FINDINGS: HRQOL is now well accepted as a trial endpoint by the US Food and Drug Administration and the European Medicines Agency. Efforts to standardize HRQOL use in clinical trials are ongoing. Existing disease-specific instruments for brain tumour patients (e.g. EORTC BN-20, Functional Assessment of Cancer Therapy-Brain) appear to meet US Food and Drug Administration and European Medicines Agency requirements sufficiently. The EORTC BN-20 has recently completed the final validation process. The psychometric properties of single-item linear analogue scale assessments for patients with newly diagnosed high-grade gliomas were investigated. Nevertheless, only a few brain tumour trial results published over the last year reported on patients' HRQOL data or symptoms. The prognostic significance of HRQOL is debated as to whether it may have additional value over other well known prognostic factors. SUMMARY: With an increase in treatment options for brain tumours and also the risk of long-term neurotoxicity, more detailed reporting of patients' HRQOL and symptoms is essential. The use of HRQOL evaluations in brain tumour trials should be strongly encouraged by the clinical and academic community.

Health-related quality of life in patients treated for anaplastic oligodendroglioma with adjuvant chemotherapy: results of a European Organisation for Research and Treatment of Cancer randomized clinical trial.

PURPOSE: Little is known about the health-related quality of life (HRQOL) of patients treated for anaplastic oligodendrogliomas. The impact of combined procarbazine, CCNU (lomustine), and vincristine (PCV) chemotherapy after radiotherapy (RT) compared with RT alone on HRQOL in the randomized European Organisation for Research and Treatment of Cancer (EORTC) 26951 trial was studied. PATIENTS AND METHODS: Adult patients with anaplastic oligodendrogliomas received RT alone or RT plus PCV chemotherapy. HRQOL was assessed with the EORTC Quality of Life Questionnaire C30 and Brain Cancer Module. Seven prespecified HRQOL end points were selected. We hypothesized that chemotherapy would impair HRQOL during treatment but that there would be a similar HRQOL between treatment arms once off treatment. Assessments were performed at randomization, at the end of RT, and then every 3 to 6 months until progression. RESULTS: A total of 368 patients were randomly assigned to one of the two arms; overall, 58% were male, and the median age was 49 years. Compliance with HRQOL was 78% at baseline and dropped to 55% to 72% up to 2.5 years post-RT. Baseline scores demonstrated considerable impairments in HRQOL for both treatment groups. The longitudinal analysis showed a significant increase in nausea/vomiting in the RT plus PCV chemotherapy arm during and shortly after chemotherapy. Because of a difference in baseline scores for fatigue and physical functioning, the differences between treatment arms during PCV did not reach significance. The nonselected scales of appetite loss and drowsiness demonstrated significant differences between treatment arms during chemotherapy in favor of the RT arm. The long-term results showed no difference between arms. CONCLUSION: The major impact of PCV on HRQOL is on nausea/vomiting, loss of appetite, and drowsiness during and shortly after treatment. There are no long-term effects of PCV chemotherapy.

Prognostic value of health-related quality-of-life data in predicting survival in patients with anaplastic oligodendrogliomas, from a phase III EORTC brain cancer group study.

PURPOSE: This is one of a few studies that have explored the value of baseline symptoms and health-related quality of life (HRQOL) in predicting survival in patients with brain cancer. PATIENTS AND METHODS: Baseline HRQOL scores (from the European Organisation for Research and Treatment of Cancer [EORTC] Quality of Life Questionnaire C30 and the EORTC Brain Cancer Module) were examined in 247 patients with anaplastic oligodendrogliomas to determine the relationship with overall survival by using Cox proportional hazards regression models. Refined techniques as the bootstrap resampling procedure and the computation of C indexes and R2 coefficients were used to explore the stability of the models as well as better assess the potential benefit of using HRQOL to predict survival in clinical practice and research. RESULTS: Classical analysis controlled for major clinical prognostic factors selected emotional functioning (P = .0016), communication deficit (P = .0261), future uncertainty (P = .0481), and weakness of legs (P = .0001) as statistically significant prognostic factors of survival. However, several issues question the validity of these findings and no single model was found to be preferable over all others. C indexes, which estimate the probability of a model to correctly predict which patient among a randomly chosen pair of patients will survive longer, and R2 coefficients, which measure the proportion of variability explained by the model, did not exhibit major improvement when adding selected or all HRQOL scores to clinical factors. CONCLUSION: While classical techniques lead to positive results, more refined analyses suggest that baseline HRQOL scores add relatively little to clinical factors to predict survival. These results may have implications for future use of HRQOL as a prognostic factor for patients with cancer.