Scoring and Staging Systems Using Cox Linear Regression Modeling and Recursive Partitioning

Early Adaptive Natural Killer Cell Expansion Is Associated with Decreased Relapse After Autologous Transplantation for Multiple Myeloma

Adaptive natural killer (NK) cells are long-lived and exhibit properties of immunologic memory against cytomegalovirus (CMV). We previously reported that expansion of adaptive NK cells after CMV reactivation in recipients of allogeneic hematopoietic cell transplantation (HCT) was associated with a lower rate of relapse of acute myelogenous leukemia. In the present study, we examined the impact of adaptive NK cell expansion in a cohort of 110 individuals who underwent autologous HCT (AHCT) for a lymphoid malignancy (lymphoma or multiple myeloma [MM]). In this cohort, higher absolute numbers of adaptive NK cells (>1.58/μL) at day 28 post-AHCT were associated with significantly decreased risk of relapse in patients with MM. No significant association was seen in patients with lymphoma. Further stratification of MM patients by CMV serostatus found a strong protective effect of adaptive NK cells only in CMV-seropositive individuals. These findings suggest that strategies to increase adaptive NK cells after AHCT may be a therapeutic option in patients with MM.

Evaluation of head and neck soft tissue sarcoma 8th edition pathologic staging system and proposal of a novel stage grouping system

BACKGROUND

The AJCC 8th edition issued a dedicated staging system for head and neck soft tissue sarcomas (HN-STS) with 2 and 4 cm tumor cut-off points, as well as a T4 classification based on invasion of adjacent structures. Stage groupings were not provided due to a paucity of data.

METHODS

We identified HN-STS patients undergoing primary surgery without neoadjuvant therapy patients in the Surveillance, Epidemiology, and End Results (SEER) database. We used multivariable analysis to examine adverse prognosticators. Then, using, recursive partitioning analysis (RPA), we established a stage grouping system that was externally validated in the National Cancer Database (NCDB).

RESULTS

Multivariable analysis in the SEER cohort (N = 546) demonstrated worsened survival with tumors invading adjacent structures (P < 0.001) and increasing de-differentiation (P < 0.001). There was no prognostic difference based on size for T1-3 tumors; however, when assessed as a continuous variable, a 5 cm tumor size cut-off point was predictive of outcome. RPA generated a stage grouping system with the following five-year overall survival: RPA Stage I (pT1-3N0-1G1-2M0) 71.2%, RPA Stage II (pT4abN0-1G1-2M0/pT1-3N0-1G3-4M0) 53.4%, and RPA Stage III (pT4abN0-1G3-4M0) 17.5%. This was successfully externally validated in the NCDB cohort (P < 0.001).

CONCLUSIONS

We confirm the importance of structural invasion and grade and demonstrate that the currently used size cut-off points are not prognostic. We propose a novel stage grouping system. A 5 cm tumor size cut-off point for tumor stage should be further evaluated.

Stratification of Postoperative Prognosis by Invasive Tumor Thickness in Perihilar Cholangiocarcinoma

BACKGROUND

The pathological tumor classification of distal cholangiocarcinoma in the American Joint Committee on Cancer (AJCC) Cancer Staging Manual 8th edition is based on invasive depth, whereas that of perihilar cholangiocarcinoma (PHCC) continues to be layer-based. We aimed to clarify whether invasive depth measurement based on invasive tumor thickness (ITT) could help determine postoperative prognosis in patients with PHCC.

METHODS

We enrolled 184 patients with PHCC who underwent hepatectomy plus extrahepatic bile duct resection or hepatopancreatoduodenectomy with curative intent. ITT was measured using simple definitions according to the sectioning direction or gross tumor pattern.

RESULTS

The median ITT was 5.8 mm (range 0.7-15.5). Using the recursive partitioning technique, ITT was classified into grades A (ITT < 2 mm, n = 9), B (2 mm ≤ ITT < 5 mm, n = 68), C (5 mm ≤ ITT < 11 mm, n = 81), and D (11 mm < ITT, n = 26). The median survival times (MSTs) in patients with grade B, C, or D were 90.8, 44.6, and 21.1 months, respectively (patients with grade A did not reach the MST). There were significant differences in postoperative prognosis between ITT grades (A vs. B, p = 0.027; B vs. C, p < 0.001; C vs. D, p = 0.004). Through multivariate analysis, regional node metastasis, invasive carcinoma at the resected margin, and ITT grade were determined as independent prognostic factors.

CONCLUSION

ITT could be measured using simple methods and may be used to stratify postoperative prognosis in patients with PHCC.

Decision tree-based diagnosis of coronary artery disease: CART model

BACKGROUND AND OBJECTIVE

As the most common cardiovascular defect, coronary artery disease (CAD), also called ischemic heart disease, is one of the substantial causes of death globally. Several diagnosis approaches such as baseline electrocardiography, echocardiography, magnetic resonance imaging, and coronary angiography are suggested for screening the suspected patients that may suffer from CAD. However, applying such methods may have health side effects and/or expensive costs.

METHODS

As an alternative to the available diagnosis tools/methods, this research involves a decision tree learning algorithm called classification and regression tree (CART) for a simple and reliable diagnosis of CAD. Several CART models are developed based on the recently CAD dataset published in the literature.

RESULTS

Utilizing all the features of the dataset (55 independent parameters), it was found that only 40 independent parameters influence the CAD diagnosis and consequently development of the predictive model. Based on the feature importance obtained from the first CART model, three new CART models are then developed using 18, 10, and 5 selected features. Except for the five-feature CART model, the outcomes of developed CART models demonstrate the maximum achievable accuracy, sensitivity, and specificity for CAD diagnosis (100%), while comparing the predictions with the reported targets. The error analysis reveals that the literature models including sequential minimal optimization (SMO), bagging SMO, Naïve Bayes (NB), artificial neural network (ANN), C4.5, J48, Bagging, and ANN in conjunction with the genetic algorithm (GA) do not outperform the CART methodology in classifying patients as normal or CAD.

CONCLUSIONS

Hence, the robustness of the tree-based algorithm in accurate and fast predictions is confirmed, implying the proposed classification technique can be successfully utilized to develop a coherent decision-making system for the CAD diagnosis.

Alternative statistical methods and their application to research in intensive care setting

In nested case-control studies, sampling of controls is usually done by density of incidence and pairing. With regard to the classic control cases studies, nested ones are more efficient, allow the calculation of the incidence of the disease and they have more internal validity due to the lower presence of bias. Competitive risks techniques can be used if we study different types of events and focus on the time and type of the first event. Recursive partitioning is a type of multivariate analysis whose purpose is the construction of classification algorithms, and it is especially useful when there are a large number of predictive variables with complex relationships with the event.

Intratumoral heterogeneity characterized by pretreatment PET in non-small cell lung cancer patients predicts progression-free survival on EGFR tyrosine kinase inhibitor

Intratumoral heterogeneity has been suggested to be an important resistance mechanism leading to treatment failure. We hypothesized that radiologic images could be an alternative method for identification of tumor heterogeneity. We tested heterogeneity textural parameters on pretreatment FDG-PET/CT in order to assess the predictive value of target therapy. Recurred or metastatic non-small cell lung cancer (NSCLC) subjects with an activating EGFR mutation treated with either gefitinib or erlotinib were reviewed. An exploratory data set (n = 161) and a validation data set (n = 21) were evaluated, and eight parameters were selected for survival analysis. The optimal cutoff value was determined by the recursive partitioning method, and the predictive value was calculated using Harrell’s C-index. Univariate analysis revealed that all eight parameters showed an increased hazard ratio (HR) for progression-free survival (PFS). The highest HR was 6.41 (P<0.01) with co-occurrence (Co) entropy. Increased risk remained present after adjusting for initial stage, performance status (PS), and metabolic volume (MV) (aHR: 4.86, P<0.01). Textural parameters were found to have an incremental predictive value of early EGFR tyrosine kinase inhibitor (TKI) failure compared to that of the base model of the stage and PS (C-index 0.596 vs. 0.662, P = 0.02, by Co entropy). Heterogeneity textural parameters acquired from pretreatment FDG-PET/CT are highly predictive factors for PFS of EGFR TKI in EGFR-mutated NSCLC patients. These parameters are easily applicable to the identification of a subpopulation at increased risk of early EGFR TKI failure. Correlation to genomic alteration should be determined in future studies.

Comparisons of recursive partitioning analysis and conventional methods for selection of uncuffed endotracheal tubes for pediatric patients

BACKGROUND

Numerous studies have investigated the best method of selecting the appropriate size of endotracheal tube (ETT) for children. However, none of the methods or formulae for selection of ETT size have shown better prediction over another, and they have required complex formulae calculation or even use of cumbersome equipment. Recursive partitioning analysis creates a decision tree that is more likely to enable clearer and easier visualization of decision charts compared to other data mining methods.

OBJECTIVES

The aim of the current study was to develop a clinically practical and intuitive chart for prediction of ETT size.

METHODS

Pediatric patients aged 2-9 years undergoing general anesthesia were intubated with uncuffed ETT. The tube size was considered optimal when a tracheal leak was detected at an inflation pressure between 10 and 25 cmH2 O. The observed ETT size was compared with the predicted ETT size calculated using Cole's formula, multivariate regression analysis, ultrasonographic measurements, and recursive partitioning tree structure analysis. Preference among the prediction methods was also investigated by asking physicians about their preference of methods.

RESULTS

Correct prediction rates were 33.3%, 50%, 61.9%, and 59.5%, and close prediction rates were 61.9%, 83.3%, 88.1%, and 93.7% for Cole's formulae, multivariate regression analysis, ultrasonographic measurements, and recursive partitioning tree model, respectively. Fourteen of 16 physicians prefer to use the easy-to-interpret tree model.

CONCLUSIONS

Analysis of the tree model by recursive partitioning structure analysis accomplished a high correct and close prediction rate for selection of an appropriate ETT size. The intuitive and easy-to-interpret tree model would be a quick and helpful tool for selection of an ETT tube for pediatric patients.

Post-metastasis survival in extremity soft tissue sarcoma: a recursive partitioning analysis of prognostic factors

BACKGROUND

Recursive partitioning analysis (RPA) enables grouping of patients into homogeneous prognostic groups in a visually intuitive form and has the capacity to account for complex interactions among prognostic variables. In this study, we employed RPA to generate a prognostic model for extremity soft tissue sarcoma (STS) patients with metastatic disease.

METHODS

A retrospective review was conducted on 135 patients with metastatic STS who had undergone surgical removal of their primary tumours. Patient and tumour variables along with the performance of metastasectomy were analysed for possible prognostic effect on post-metastatic survival. Significant prognostic factors on multivariate analysis were incorporated into RPA to build regression trees for the prediction of post-metastatic survival.

RESULTS

RPA identified six terminal nodes based on histological grade, performance of metastasectomy and disease-free interval (DFI). Based on the median survival time of the terminal nodes, four prognostic groups with significantly different post-metastatic survival were generated: (1) group A: low grade/metastasectomy; (2) group B: low grade/no metastasectomy/DFI ⩾ 12 months or high grade/metastasectomy; (3) group C: low grade/no metastasectomy/DFI < 12 months or high grade/no metastasectomy/DFI ⩾ 12 months; and (4) group D: high grade/no metastasectomy/DFI < 12 months. The 3-year survival rates for each group were: group A, 76.1 ± 9.6%; group B, 42.3 ± 10.3%; group C, 18.8 ± 8.0%; and group D, 0.0 ± 0.0%.

CONCLUSION

Our prognostic model using RPA successfully divides STS patients with metastasis into groups that can be easily implemented using standard clinical parameters.

The skeleton flight apparatus of North American bluebirds (Sialia): phylogenetic thrushes or functional flycatchers?

To better understand ecological traits of organisms, one can study them from two, not necessarily mutually exclusive perspectives: how the traits evolved, and their current adaptive utility. In birds, foraging behavior and associated morphological traits generally are explained by a combination of adaptive and phylogenetic predictors. The avian skeleton and more specifically, the skeletal flight apparatus is under well-known functional and phylogenetic constraints. This is an interesting area to partition the relative contributions of adaptive correlated evolution and phylogenetic constraint to species clustering in morphological space. A prediction of convergent evolution is that nonphylogenetic morphological clustering is a characteristic of ecological similarity. We tested this using representatives of North American birds from two clades, one with a mixture of foraging modes (Turdid thrushes, solitaires, and bluebirds) and one with more canalized foraging behaviors (Tyrannid flycatchers). Nine characters on the skeletal flight apparatus from 19 species were used to characterize the morphological space and test for ecomorphological clustering. When body size and phylogeny are considered, the three bluebird species and Townsend's solitaire cluster with the ecologically similar flycatchers rather than with their phylogenetic close relatives. Furthermore, sit-and-wait foragers tend to exhibit relatively long distal elements and a long keel while active ground foragers have deeper keels and a longer humerus. Distal elements, expected to be relatively shorter and more bowed in the flycatchers and bluebirds, were actually longer and narrower. A reduction of distal element mass may be more important for facilitating maneuverability than surface area for insertion of wing-rotational musculature.

Artificial neural networks versus proportional hazards Cox models to predict 45-year all-cause mortality in the Italian Rural Areas of the Seven Countries Study

BACKGROUND

Projection pursuit regression, multilayer feed-forward networks, multivariate adaptive regression splines and trees (including survival trees) have challenged classic multivariable models such as the multiple logistic function, the proportional hazards life table Cox model (Cox), the Poisson's model, and the Weibull's life table model to perform multivariable predictions. However, only artificial neural networks (NN) have become popular in medical applications.

RESULTS

We compared several Cox versus NN models in predicting 45-year all-cause mortality (45-ACM) by 18 risk factors selected a priori: age; father life status; mother life status; family history of cardiovascular diseases; job-related physical activity; cigarette smoking; body mass index (linear and quadratic terms); arm circumference; mean blood pressure; heart rate; forced expiratory volume; serum cholesterol; corneal arcus; diagnoses of cardiovascular diseases, cancer and diabetes; minor ECG abnormalities at rest. Two Italian rural cohorts of the Seven Countries Study, made up of men aged 40 to 59 years, enrolled and first examined in 1960 in Italy. Cox models were estimated by: a) forcing all factors; b) a forward-; and c) a backward-stepwise procedure. Observed cases of deaths and of survivors were computed in decile classes of estimated risk. Forced and stepwise NN were run and compared by C-statistics (ROC analysis) with the Cox models. Out of 1591 men, 1447 died. Model global accuracies were extremely high by all methods (ROCs > 0.810) but there was no clear-cut superiority of any model to predict 45-ACM. The highest ROCs (> 0.838) were observed by NN. There were inter-model variations to select predictive covariates: whereas all models concurred to define the role of 10 covariates (mainly cardiovascular risk factors), family history, heart rate and minor ECG abnormalities were not contributors by Cox models but were so by forced NN. Forced expiratory volume and arm circumference (two protectors), were not selected by stepwise NN but were so by the Cox models.

CONCLUSIONS

There were similar global accuracies of NN versus Cox models to predict 45-ACM. NN detected specific predictive covariates having a common thread with physical fitness as related to job physical activity such as arm circumference and forced expiratory volume. Future attention should be concentrated on why NN versus Cox models detect different predictors.

Measurement of the invasion depth of extrahepatic bile duct carcinoma: An alternative method overcoming the current T classification problems of the AJCC staging system

Tumor staging of extrahepatic bile duct (EBD) carcinoma is problematic for a number of reasons, including definitional problems with the current T classification of the American Joint Committee on Cancer staging system and the common occurrence of severe desmoplastic stromal reaction around the advancing edges of these tumors. To address these problems we evaluated the depth of invasion in 222 cases of EBD carcinoma by measuring the distance between the basal lamina of the adjacent normal epithelium to the most deeply infiltrating tumor cells, and compared this evaluation to time of survival and other clinical and pathologic parameters. A complex pattern of survival time versus the depth of invasion was observed by censored local regression. The recursive-partitioning technique was coupled with the log-rank test to identify 2 significant cutoff points for the depth of invasion, 5 and 12 mm, which segregated patients into 3 groups with statistically significant decreasing length of median survival (<5 mm, 61 mo; 5 to 12 mm, 23 mo; >12 mm, 17 mo, P < 0.001). On the basis of the present data, we propose that a measurement of the depth of invasion should be performed in cases of EBD carcinoma, and that the T classification of EBD carcinoma should be changed to incorporate this measurement: T1 (<5 mm), T2 (5 to 12 mm), and T3 (>12 mm).