Recursive Partitioning Analysis of Prognostic Factors for Survival in Patients With Advanced Cancer

A Systematic Review of Prognostic Factors in Patients with Cancer Receiving Palliative Radiotherapy: Evidence-Based Recommendations

(1) Background: Prognostication in patients with cancer receiving palliative radiotherapy remains a challenge. To improve the process, we aim to identify prognostic factors in this population from the literature and offer evidence-based recommendations on prognostication in patients undergoing palliative radiotherapy for non-curable or advanced cancers. (2) Methods: A systematic review was performed on the medical literature from 2005 to 2023 to extract papers on the prognosis of palliative radiotherapy patients with advanced cancer. The initial selection was performed by at least two authors to determine study relevance to the target area. Studies were then classified based on type and evidence quality to determine final recommendations. (3) Results: The literature search returned 57 papers to be evaluated. Clinical and biological prognostic factors were identified from these papers to improve clinical decision making or construct prognostic models. Twenty prognostic models were identified for clinical use. There is moderate evidence supporting (i) evidence-based factors (patient, clinical, disease, and lab) in guiding decision making around palliative radiation; (ii) that certain biological factors are of importance; (iii) prognostication models in patients with advanced cancer; and that (iv) SBRT or re-irradiation use can be guided by predictions of survival by prognostic scores or clinicians. Patients with more favorable prognoses are generally better suited to SBRT or re-irradiation, and the use of prognostic models can aid in this decision making. (4) Conclusions: This evaluation has identified several factors or tools to aid in prognosis and clinical decision making. Future studies should aim to further validate these tools and factors in a clinical setting, including the leveraging of electronic medical records for data availability. To increase our understanding of how causal factors interact with palliative radiotherapy, future studies should also examine and include prediction of response to radiation as an outcome.

Prognostication for Patients Receiving Palliative Radiation Therapy

Estimation of patient prognosis plays a central role in guiding decision making for the palliative management of metastatic disease, and a number of statistical models have been developed to provide survival estimates for patients in this context. In this review, we discuss several well-validated survival prediction models for patients receiving palliative radiotherapy to sites outside of the brain. Key considerations include the type of statistical model, model performance measures and validation procedures, studies' source populations, time points used for prognostication, and details of model output. We then briefly discuss underutilization of these models, the role of decision support aids, and the need to incorporate patient preference in shared decision making for patients with metastatic disease who are candidates for palliative radiotherapy.

Percent of remaining life on palliative radiation treatment: solely a function of fractionation?

Background

This study analyzed the percent of remaining life (PRL) on treatment in patients irradiated for bone metastases. Bone metastases were treated together with other target volumes, if indicated, e.g. a 10-fraction treatment course that included brain and bone metastases. PRL was determined by calculating the time between start and finish of palliative radiotherapy (minimum 1 day in case of a single-fraction regimen) and dividing it by overall survival in days from start of radiotherapy.

Materials and methods

Different baseline parameters were assessed for association with dichotomized PRL (< 5% vs. ≥ 5%). The retrospective study included 219 patients (287 courses of palliative radiotherapy). After univariate analyses, multi-nominal logistic regression was employed.

Results

PRL on treatment ranged from 1-23%. Single-fraction radiotherapy resulted in < 5% PRL on treatment in all cases. All courses with 10 fractions resulted in at least 5% PRL on treatment. Significant associations were found between various baseline parameters and PRL category. With fractionation included in the regression model, 3 parameters retained significant p-values: Karnofsky performance status (KPS), none-bone target volume and fractionation (all with p < 0.001). If analyzed without fractionation, none-bone target volume (p < 0.001), hemoglobin (p < 0.001), KPS (p = 0.01), lack of additional systemic treatment (p = 0.01), and hypercalcemia (p = 0.04) were significant.

Conclusions

Fractionation is an easily modifiable factor with high impact on PRL. Patients with KPS < 70 and those treated for additional target types during the same course are at high risk of spending a larger proportion of their remaining life on treatment.

Prediction of Gastric Gastrointestinal Stromal Tumors before Operation: A Retrospective Analysis of Gastric Subepithelial Tumors

Gastrointestinal stromal tumors (GISTs), leiomyomas, and schwannomas are the most common gastric subepithelial tumors (GSETs) with similar endoscopic findings. Preoperative prediction of GSETs is difficult. This study analyzed and predicted GSET diagnosis through a retrospective review of 395 patients who underwent surgical resection of GISTs, leiomyomas, and schwannomas measuring 2–10 cm. GSETs were divided by size (group 2–5, >2 and ≤5 cm; group 5–10, >5 and ≤10 cm) for analysis. Demographics, clinical symptoms, and images were analyzed. A recursive partitioning analysis (RPA) was used to identify optimal classifications for specific GSET diagnoses. GIST patients were relatively older than other patients. Both groups had higher proportions of UGI bleeding, lower hemoglobin (Hb) levels, and a higher ratio of necrosis on their computed tomography (CT) scans. The RPA tree showed that (a) age ≤ 55, Hb ≥ 10.7, and CT necrosis; (b) age ≤ 55 and Hb < 10.7; (c) age >55 and Hb < 12.9; and (d) age >55 and CT hetero-/homogeneity can predict high GIST risk in group 2–5. Positive or negative CT necrosis, with age >55, can predict high GIST risk in group 5–10. GIST patients were older and presented with low Hb levels and tumor necrosis. In RPA, the accuracy reached 85% and 89% in groups 2–5 and 5–10, respectively.

The LabBM score is an excellent survival prediction tool in patients undergoing palliative radiotherapy

Background and aim

The prognostic assessment of patients referred for palliative radiotherapy can be conducted by site-specific scores. A quick assessment that would cover the whole spectrum could simplify the working day of clinicians who are not specialists for a particular disease site. This study evaluated a promising score, the LabBM (validated for brain metastases), in patients treated for other indications.

Materials and methods

The LabBM score was calculated in 375 patients by assigning 1 point each for C-reactive protein and lactate dehydrogenase above the upper limit of normal, and 0.5 points each for hemoglobin, platelets and albumin below the lower limit of normal. Uni- and multivariate analyses were performed.

Results

Median overall survival gradually decreased with increasing point sum (range 25.1-1.1 months). When grouped according to the original three-tiered model, excellent discrimination was found. Patients with 0-1 points had a median survival of 15.7 months. Those with 1.5-2 points had a median survival of 5.8 months. Finally, those with 2.5-3.5 points had a median survival of 3.2 months (all p-values ≤ 0.001).

Conclusion

The LabBM score, which is derived from inexpensive blood tests and easy to use, stratified patients into three very distinct prognostic groups and deserves further validation.

Prognostic tools for survival prediction in advanced cancer patients: A systematic review

Survival prediction for palliative cancer patients by physicians is often optimistic. Patients with a very short life expectancy (<4 weeks) may not benefit from radiation therapy (RT), as the time to maximal symptom relief after treatment can take 4-6 weeks. We aimed to identify a prognostic tool (or tools) to predict survival of less than 4 weeks and less than 3 months in patients with advanced cancer to guide the choice of radiation dose and fractionation. We searched Embase, Medline (EBSCOhost) and CINAHL (EBSCOhost) clinical databases for literature published between January 2008 and June 2018. Seventeen studies met the inclusion criteria and were included in the review. Prediction accuracy at less than 4 weeks and less than 3 months were compared across the prognostic tools. Reporting of prediction accuracy among the different studies was not consistent: the Palliative Prognostic Score (PaP), Palliative Prognostic Index (PPI) and Number of Risk Factors (NRF) best-predicted survival duration of less than 4 weeks. The PPI, performance status with Palliative Prognostic Index (PS-PPI), NRF and Survival Prediction Score (SPS) may predict 3-month survival. We recommend PPI and PaP tools to assess the likelihood of a patient surviving less than 4 weeks. If predicted to survive longer and RT is justified, the NRF tool could be used to determine survival probability less than 3 months which can then help clinicians select dose and fractionation. Future research is needed to verify the reliability of survival prediction using these prognostic tools in a radiation oncology setting.

Early and Midtreatment Mortality in Palliative Radiotherapy: Emphasizing Patient Selection in High-Quality End-of-Life Care

BACKGROUND

Palliative radiotherapy (RT) is effective, but some patients die during treatment or too soon afterward to experience benefit. This study investigates end-of-life RT patterns to inform shared decision-making and facilitate treatment consistent with palliative goals.

MATERIALS AND METHODS

All patients who died ≤6 months after initiating palliative RT at an academic cancer center between 2015 and 2018 were identified. Associations with time-to-death, early mortality (≤30 days), and midtreatment mortality were analyzed.

RESULTS

In total, 1,620 patients died ≤6 months from palliative RT initiation, including 574 (34%) deaths at ≤30 days and 222 (14%) midtreatment. Median survival was 43 days from RT start (95% CI, 41-45) and varied by site (P<.001), ranging from 36 (head and neck) to 53 days (dermal/soft tissue). On multivariable analysis, earlier time-to-death was associated with osseous (hazard ratio [HR], 1.33; P<.001) and head and neck (HR, 1.45; P<.001) sites, multiple RT courses ≤6 months (HR, 1.65; P<.001), and multisite treatments (HR, 1.40; P=.008), whereas stereotactic technique (HR, 0.77; P<.001) and more recent treatment year (HR, 0.82; P<.001) were associated with longer survival. No difference in time to death was noted among patients prescribed conventional RT in 1 to 10 versus >10 fractions (median, 40 vs 47 days; P=.272), although the latter entailed longer courses. The 30-day mortality group included 335 (58%) inpatients, who were 27% more likely to die midtreatment (P=.031). On multivariable analysis, midtreatment mortality among these inpatients was associated with thoracic (odds ratio [OR], 2.95; P=.002) and central nervous system (CNS; OR, 2.44; P=.002) indications, >5-fraction courses (OR, 3.27; P<.001), and performance status of 3 to 4 (OR, 1.63; P=.050). Conversely, palliative/supportive care consultation was associated with decreased midtreatment mortality (OR, 0.60; P=.045).

CONCLUSIONS

Earlier referrals and hypofractionated courses (≤5-10 treatments) should be routinely considered for palliative RT indications, given the short life expectancies of patients at this stage in their disease course. Providers should exercise caution for emergent thoracic and CNS indications among inpatients with poor prognoses due to high midtreatment mortality.

External Validation of the Bone Metastases Ensemble Trees for Survival (BMETS) Machine Learning Model to Predict Survival in Patients With Symptomatic Bone Metastases

PURPOSE

The Bone Metastases Ensemble Trees for Survival (BMETS) model uses a machine learning algorithm to estimate survival time following consultation for palliative radiation therapy for symptomatic bone metastases (SBM). BMETS was developed at a tertiary-care, academic medical center, but its validity and stability when applied to external data sets are unknown.

PATIENTS AND METHODS

Patients treated with palliative radiation therapy for SBM from May 2013 to May 2016 at two hospital-based community radiation oncology clinics were included, and medical records were retrospectively reviewed to collect model covariates and survival time. The Kaplan-Meier method was used to estimate overall survival from consultation to death or last follow-up. Model discrimination was estimated using time-dependent area under the curve (tAUC), which was calculated using survival predictions from BMETS based on the initial training data set.

RESULTS

A total of 216 sites of SBM were treated in 182 patients. Most common histologies were breast (27%), lung (23%), and prostate (23%). Compared with the BMETS training set, the external validation population was older (mean age, 67 v 62 years; P < .001), had more primary breast (27% v 19%; P = .03) and prostate cancer (20% v 12%; P = .01), and survived longer (median, 10.7 v 6.4 months). When the BMETS model was applied to the external data set, tAUC values at 3, 6, and 12 months were 0.82, 0.77, and 0.77, respectively. When refit with data from the combined training and external validation sets, tAUC remained > 0.79.

CONCLUSION

BMETS maintained high discriminative ability when applied to an external validation set and when refit with new data, supporting its generalizability, stability, and the feasibility of dynamic modeling.

Prediction of breast cancer-related outcomes with the Edmonton Symptom Assessment Scale: A literature review

PURPOSE

The Edmonton Symptom Assessment Scale (ESAS) is a validated tool used in patients with varied cancer diagnoses to measure patient symptoms. The present manuscript will review the literature assessing the ability of the ESAS to predict patient-related outcomes in breast cancer patients.

METHODS

A literature search was conducted of Cochrane Central Register of Controlled Trials databases, Ovid MEDLINE, and Embase for English articles that investigated the use of predictive modelling with the ESAS in the breast cancer population. Study type, publication year, sample size, patient demographics, predicted outcomes, and strongest predictive factors/symptoms were summarized for each study.

RESULTS

A total of nine articles were included in this review. Five articles used the ESAS in predictive models to determine patient time to death. ESAS was also used to predict emergency department visits, determine symptoms associated with decreased quality of life, and generate a Health Utility Score. Lack of appetite was the most common ESAS symptom, as it was reported in five studies to be associated with decreased survival. In four of the nine articles, an additional survey investigating physical functioning was used in combination with ESAS to strengthen the predictive models.

CONCLUSIONS

Included studies support the use of ESAS in predictive models, particularly for predicting survival. Using the ESAS as a predictive tool allows for more accurate time to death predictions, potentially improving symptom management and preventing overtreatment of palliative patients near the end of life.

Patient-Reported Outcomes for Determining Prognostic Groups in Veterans With Advanced Cancer

CONTEXT

Physicians overestimate survival in patients with advanced cancer. Patient-reported outcomes could provide another way to estimate survival. We previously reported four prognostic groups based on Karnofsky Performance Status, Functional Assessment of Cancer Therapy physical well-being subscale, and Memorial Symptom Assessment Scale-Short Form physical symptom distress subscale scores.

OBJECTIVES

To determine the validity of these four prognostic groups.

METHODS

We performed prospective surveys. Data from a total of 880 Veterans Affairs Medical Center patients, 417 in the First Cohort and 463 in the Validation Cohort, were analyzed. Both inpatients and outpatients were prospectively recruited in Institutional Review Board-approved studies from August 1999 to September 2009. Survival was measured from the date of entry until death or December 1, 2009. Patients completed self-assessments with the Functional Assessment of Cancer Therapy and Memorial Symptom Assessment Scale-Short Form. Analysis of variance was used to test differences between groups in continuous variables; a generalized Wilcoxon test was used for differences between groups for survival.

RESULTS

The average age in the Validation Cohort was 66.5 years and 98% were men. The majority of patients had metastatic cancer (90%), with lung (28%) and prostate (26%) cancers being predominant. The median Karnofsky Performance Status was 70. Median survival was 33, 46.5, 124, and 209.5 days for the four prognostic groups (P < 0.0001, all pair-wise comparisons P < 0.02).

CONCLUSION

The four prognostic groups remained distinct in the prospective cohort. Small differences in patient-reported physical well-being can halve survival estimates. Patient-reported outcomes can correct for physician overestimate of prognosis. This study provides a way to use patient-reported outcomes for prognosis in patients with advanced cancer, with important implications for assessment.

Aggressive periodontitis defined by recursive partitioning analysis of immunologic factors

BACKGROUND

The present study aims to extend recent findings of a non-linear model of the progression of periodontitis supporting the notion that aggressive periodontitis (AgP) and chronic periodontitis (CP) are distinct clinical entities. This approach is based on the implementation of recursive partitioning analysis (RPA) to evaluate a series of immunologic parameters acting as predictors of AgP and CP.

METHODS

RPA was applied to three population samples, that were retrieved from previous studies, using 17 immunologic parameters. The mean values of the parameters in control subjects were used as the cut-off points. Leave-one-out cross-validation (LOOCV) prediction errors were estimated in the proposed models, as well as the Kullback-Leibler divergence (DKL) of the distribution of positive results in AgP compared to CP and negative results in CP compared to AgP.

RESULTS

Seven classification trees were derived showing that the relationship of interleukin (IL)-4, IL-1, IL-2 has the highest potential to rule out or rule in AgP. On the other hand, immunoglobulin (Ig)A, IgM used to rule out AgP and cluster of differentiation 4 (CD4)/CD8, CD20 used to rule in AgP showed the least LOOCV cost. Penalizing DKL with LOOCV cost promotes the IL-4, IL-1, IL-2 model for ruling out AgP, whereas the single CD4/CD8 ratio with a lowered discrimination cut-off point was used to rule in AgP.

CONCLUSIONS

Although a test is unlikely to have both high sensitivity and high specificity, the use of immunologic parameters in the right model can efficiently complement a clinical examination for ruling out or ruling in AgP.

Symptoms tell it all: a systematic review of the value of symptom assessment to predict survival in advanced cancer patients

PURPOSE

To determine the prognostic meaning of symptoms in patients with advanced cancer.

DESIGN

Medline, Embase, Cochrane and Cinahl databases were systematically explored. The predicting symptoms were also evaluated in the three stages of palliative care: disease-directed palliation, symptom-oriented palliation and palliation in the terminal stage.

RESULTS

Out of 3167 papers, forty-four papers satisfied all criteria. Confusion, anorexia, fatigue, cachexia, weight loss, cognitive impairment, drowsiness, dyspnea, dysphagia, dry mouth and depressed mood were associated with survival in ≥ 50% of the studies evaluating these symptoms. Multivariate analysis showed confusion, anorexia, fatigue, cachexia, weight loss, dyspnea and dysphagia as independent prognostic factors in 30-56% of the studies. In the stage of disease-directed palliation anorexia, cachexia, weight loss, dysphagia and pain and in the stage of symptom-oriented palliation confusion, fatigue, cachexia, weight loss, dyspnea, dysphagia and nausea were shown to be independent predictors of survival in >30% of the studies.

CONCLUSION

Symptoms with independent predictive value are confusion, anorexia, fatigue, cachexia, weight loss, dyspnea and dysphagia. New insights are added by the variance between the three palliative stages.

Growth kinetics of renal masses: analysis of a prospective cohort of patients undergoing active surveillance

BACKGROUND

Active surveillance (AS) represents a treatment option for renal masses in patients who are not surgical candidates either because of existing comorbidities or patient choice. Among renal masses undergoing AS, some grow rapidly and require treatment or progress to metastatic disease. Patient and tumour characteristics related to this more aggressive behaviour have been poorly studied.

OBJECTIVE

To report the analysis of a multi-institutional cohort of patients undergoing AS for small renal masses.

DESIGN, SETTING, AND PARTICIPANTS

This prospective study included 82 patients with 84 renal masses who underwent AS in three Canadian institutions between July 2001 and June 2009.

INTERVENTION

All patients underwent AS for renal masses presumed to be renal cell carcinoma (RCC) as based on diagnostic imaging.

MEASUREMENTS

Age, sex, symptoms at presentation, maximum diameter at diagnosis (cm), tumour location (central/peripheral), degree of endophytic component (1-100%), and tumour consistency (solid/cystic) were used to develop a predictive model of the tumour growth rate using binary recursive partitioning analysis with a repeated measures outcome.

RESULTS AND LIMITATIONS

With a median follow-up of 36 mo (range: 6-96), the mean annual renal mass growth rate for the entire cohort was 0.25 cm/yr (standard deviation [SD]: 0.49 cm/yr). Only one patient (1.2%) developed metastatic RCC. Amongst all variables, maximum diameter at diagnosis was the only predictor of tumour growth rate, and two distinct growth rates were identified. Masses that are ≥2.45 cm in largest diameter at diagnosis grow faster than smaller masses. This series was limited by its moderate sample size, although it is the largest published prospective series to date.

CONCLUSIONS

We confirm that most renal masses grow slowly and carry a low metastatic potential. Tumour size is a predictor of tumour growth rate, with renal masses <2.45 cm growing more slowly than masses >2.45 cm.

Palliative radiotherapy tailored to life expectancy in end-stage cancer patients: reality or myth?

BACKGROUND

The purpose of the study was to investigate the adequacy of palliative radiation treatment in end-stage cancer patients.

METHODS

Of 216 patients referred for palliative radiotherapy, 33 died within 30 days and constitute the population of the study. Symptoms, Karnofsky Performance Status (KPS), laboratory tests, and survival estimates were obtained. Treatment course was evaluated by medical records. Univariate analyses were performed by using the 2-sided chi-square test. With significant variables, multiple regression analysis was performed.

RESULTS

Median age was 65 years, and median survival was 15 days. Prevailing primary cancer types were lung (39%) and breast (18%). Metastases were present in 94% of patients, brain (36%), bone (24%) and lung (18%). In 91%, KPS was < 0%. KPS, lactate dehydrogenase, dyspnea, leucocytosis, and brain metastases conveyed a poor prognosis. From 85 survival estimates, only 16% were correct, but 21% expected more than 6 months. Radiotherapy was delivered to 91% of patients. In 90% of radiation treatments, regimens of at least 30 Gy with fractions of 2-3 Gy were applied. Half of the patients spent greater than 60% of their remaining lifespan on therapy. In only 58% of patients was radiotherapy completed. Progressive complaints were noted in 52% and palliation in 26%.

CONCLUSIONS

Radiotherapy was not appropriately customized to these patients considering the median treatment time, which resembles the median survival time. About half of the patients did not benefit despite spending most of their remaining lives on therapy. Prolonged irradiation schedules probably reflect overly optimistic prognoses and unrealistic concerns about late radiation damage. Single-fraction radiotherapy was too seldom used.

Recursive partitioning analysis of prognostic factors in WHO grade III glioma patients treated with radiotherapy or radiotherapy plus chemotherapy

BACKGROUND

We evaluated the hierarchical risk groups for the estimated survival of WHO grade III glioma patients using recursive partitioning analysis (RPA). To our knowledge, this is the first study to address the results of RPA specifically for WHO grade III gliomas.

METHODS

A total of 133 patients with anaplastic astrocytoma (AA, n = 56), anaplastic oligodendroglioma (AO, n = 67), or anaplastic oligoastrocytoma (AOA, n = 10) were included in the study. These patients were treated with either radiotherapy alone or radiotherapy followed by PCV chemotherapy after surgery. Five prognostic factors, including histological subsets, age, performance status, extent of resection, and treatment modality were incorporated into the RPA. The final nodes of RPA were grouped according to their survival times, and the Kaplan-Meier graphs are presented as the final set of prognostic groups.

RESULTS

Four risk groups were defined based on the clinical prognostic factors excluding age, and split variables were all incorporated into the RPA. Survival analysis showed significant differences in mean survival between the different groups: 163.4 months (95% CI: 144.9-182.0), 109.5 months (86.7-132.4), 66.6 months (50.8-82.4), and 27.7 months (16.3-39.0), respectively, from the lowest to the highest risk group (p = 0.00).

CONCLUSION

The present study shows that RPA grouping with clinical prognostic factors can successfully predict the survival of patients with WHO grade III glioma.