Multivariate P-value Research Articles

SLCO1B1 and ABCG2 genotype-informed phenotypes are related to variation in ramipril exposure.

Ramipril is an angiotensin-converting enzyme inhibitor used for hypertension and heart failure management. To date, scarce literature is available on pharmacogenetic associations affecting ramipril. The goal of this study was to investigate the effect of 120 genetic variants in 34 pharmacogenes (i.e., genes encoding for enzymes like CYPs or UGTs and transporters like ABC or SLC) on ramipril pharmacokinetic variability and adverse drug reaction (ADR) incidence. Twenty-nine healthy volunteers who had participated in a single-dose bioequivalence clinical trial of two formulations of ramipril were recruited. A univariate and multivariate analysis searching for associations between genetic variants and ramipril pharmacokinetics was performed. SLCO1B1 and ABCG2 genotype-informed phenotypes strongly predicted ramipril exposure. Volunteers with the SLCO1B1 decreased function (DF) phenotype presented around 1.7-fold higher dose/weight-corrected area under the curve (AUC/DW) than volunteers with the normal function (NF) phenotype (univariate p-value [puv] <0.001, multivariate p-value [pmv] <0.001, β =0.533, R2=0.648). Similarly, volunteers with ABCG2 DF + poor function (PF) phenotypes presented around 1.6-fold higher AUC/DW than those with the NF phenotype (puv=0.011, pmv<0.001, β =0.259, R2=0.648). Our results suggest that SLCO1B1 and ABCG2 are important transporters to ramipril pharmacokinetics, and their genetic variation strongly alters its pharmacokinetics. Further studies are required to confirm these associations and their clinical relevance.

Enhancing student creativity and scientific literacy through virtual lab learning

The choice of teaching materials and methods used with students is ineffective, which prevents students from receiving meaningful learning experiences. In this learning process, student activity is crucial to ensure that students receive meaningful learning experiences, are able to think critically and creatively, and are especially literate in science regarding the concept of induction. This study uses a virtual lab on the idea of magnetic induction to encourage students' creativity in their scientific literacy learning. This sort of research employs a cooperative learning model in the experimental class and a traditional learning model in the control class in a quasi-experimental design. Data were gathered using the pretest approach for students' creativity in terms of scientific literacy and the posttest method for students' knowledge of science after four meetings, where there were 20 multiple-choice items and 10 isay questions. MANOVA analysis was used to analyze the data. SPSS Multivariate Because the dependent variable in this study is quantitative (numeric) and the independent variable is qualitative, Mancova is used (categorical). The significance value for the multivariate p-value table is therefore 0.000. Ho is automatically rejected since the significant figure at the 95% confidence level is 0.000 0.05. The SPSS output findings reveal a substantial difference between studying electromagnetic induction content with and without the use of a virtual lab. While practical activities give students first-hand experience, learning using virtual lab aids can enhance students' creative thinking and scientific literacy. The induction electric motion practicum helps students better understand the concepts of the physics course. Students learn how to be more active and search for information through practicum activities using the PHET clorado virtual lab (phet clorado). Through this activity, the electromagnetic induction practicum learned will be more focused, and it can improve students' creativity and scientific literacy

Genetic Variation in CYP2D6 and SLC22A1 Affects Amlodipine Pharmacokinetics and Safety.

Amlodipine is an antihypertensive drug with unknown pharmacogenetic biomarkers. This research is a candidate gene study that looked for associations between amlodipine pharmacokinetics and safety and pharmacogenes. Pharmacokinetic and safety data were taken from 160 volunteers from eight bioequivalence trials. In the exploratory step, 70 volunteers were genotyped for 44 polymorphisms in different pharmacogenes. CYP2D6 poor metabolizers (PMs) showed higher half-life (t1/2) (univariate p-value (puv) = 0.039, multivariate p-value (pmv) = 0.013, β = -5.31, R2 = 0.176) compared to ultrarapid (UMs), normal (NMs) and intermediate metabolizers (IMs). SLC22A1 rs34059508 G/A genotype was associated with higher dose/weight-corrected area under the curve (AUC72/DW) (puv = 0.025; pmv = 0.026, β = 578.90, R2 = 0.060) compared to the G/G genotype. In the confirmatory step, the cohort was increased to 160 volunteers, who were genotyped for CYP2D6, SLC22A1 and CYP3A4. In addition to the previous associations, CYP2D6 UMs showed a lower AUC72/DW (puv = 0.046, pmv = 0.049, β = -68.80, R2 = 0.073) compared to NMs, IMs and PMs and the SLC22A1 rs34059508 G/A genotype was associated with thoracic pain (puv = 0.038) and dizziness (puv = 0.038, pmv = 0.014, log OR = 10.975). To our knowledge, this is the first work to report a strong relationship between amlodipine and CYP2D6 and SLC22A1. Further research is needed to gather more evidence before its application in clinical practice.

The Effect of Naloxone Access Laws on Fatal Synthetic Opioid Overdose Fatality Rates.

Increases in fatal synthetic opioid overdoses over the past 8 years have left states scrambling for effective means to curtail these deaths. Many states have implemented policies and increased service capacity to address this rise. To better understand the effectiveness of policy level interventions we estimated the impact of the presence of naloxone access laws (NALs) on synthetic opioid fatalities at the state level. A multivariable longitudinal linear mixed model with a random intercept was used to determine the relationship between the presence of NALs and synthetic opioid overdose death rates, while controlling for, Good Samaritan laws, opioid prescription rate, and capacity for medication for opioid use disorder (MOUD), utilizing a quadratic time trajectory. Data for the study was collected from the National Vital Statistics System using multiple cause-of-death mortality files linked to drug overdose deaths. The presence of an NAL had a significant (univariate P-value = .013; multivariable p-value = .010) negative relationship to fentanyl overdose death rates. Other significant controlling variables were quadratic time (univariate and multivariable P-value < .001), MOUD (univariate P-value < .001; multivariable P-value = .009), and Good Samaritan Law (univariate P-value = .033; multivariable P-value = .018). Naloxone standing orders are strongly related to fatal synthetic opioid overdose reduction. The effect of NALs, MOUD treatment capacity, and Good Samaritan laws all significantly influenced the synthetic opioid overdose death rate. The use of naloxone should be a central part of any state strategy to reduce overdose death rate.

Stem Cell-Derived Gene Expression Scores Predict Survival and Leukemic Transformation in Myelofibrosis

Myelofibrosis (MF) is a myeloproliferative neoplasm (MPN) with survival outcomes ranging from months to years and variable risk of transformation to acute myeloid leukemia. Allogeneic bone marrow transplantation (BMT) can be curative but is associated with high treatment-related morbidity and mortality, therefore accurate risk stratification is important to guide clinical decision making in MF. Current risk prediction models use clinical and/or genomic features but do not consider the properties of the disease-driving stem cell population. Here, we use transcriptomic variation corresponding to both intra- and inter-patient heterogeneity among MF stem cells to generate novel gene expression-based scores predictive of survival and leukemic transformation in MF. To train and validate novel prognostic scores in MF, we identified 358 patients from an MPN registry at the Princess Margaret Cancer Centre (ClinicalTrials.gov Identifier: NCT02760238) from whom peripheral blood (PB) cells were collected near the date of MF diagnosis. All patients were diagnosed with either primary, post-PV, post-ET or pre-fibrotic MF with clinical follow-up of up to 12.2 years. RNA was extracted from unsorted PB mononuclear cells and RNA sequencing (RNAseq) was performed at an average depth of 50 million reads per sample. We randomly split our MF cohort into training (70%; n=250) and test sets (30%; n=108) and utilized a repeated nested cross validation approach together with statistical regression, to generate and assess the performance of models to predict survival within the training set. We tested 36,000 models derived from 36 initial MF-related genesets, ranging from stem-cell specific genesets to the whole transcriptome. The most accurate models by cross validation (median multivariable p-value = 6e-5) were produced from our retrospective identification of highly variable genes in single-cell RNAseq data derived from 82,255 Lin-CD34+ MF stem and progenitor cells across 15 patients (Psaila et al., 2020). Thus, features of intra- and inter-patient heterogeneity among MF stem and progenitor cells proved to be the most relevant for predicting survival. From these features, we derived our final model calculated as the weighted sum of gene expression across 24 genes (termed MPN24). We categorized patients with MPN24 scores above or below the training cohort median as MPN24 high or low, respectively. This model was validated in the test set, with high and low score patients experiencing 5-year survival rates of 71% [95% CI 57-88%] and 21% [95% CI 9%-52%], respectively, when censored at time of BMT (HR = 5.3 [95% CI 2.6-10.5]; p=2.1e-6) (Figure 1A). MPN24 retained independent prognostic value in multivariable analysis incorporating age, sex, DIPSS category, ECOG status, fibrosis grade, constitutional symptoms, and PB blast percentage (adjusted HR = 5.7 [95% CI 2.2-14]; p = 3e-4). Importantly, DIPSS classification remained a significant covariate, indicating that MPN24 and DIPSS capture distinct features of disease. We therefore developed a new three-tier classification scheme integrating both DIPSS and MPN24 scores (Figure 1B). Patients classified as low-, intermediate- or high-risk in this new classification scheme experienced 5-year survival rates of 88.2% [95% CI 77.9% - 99.9%], 39.3% [95% CI 19.9% - 77.7%] and 10.8% [95% CI 2.1% - 55.8%], respectively (likelihood ratio test p = 1e-8). Finally, from the MPN24 genes we derived a 13 gene subscore (MPN13) predictive of leukemic transformation in the training set. Patients in the test set scoring above the 80th percentile from the training set were classified as high risk and the remaining as low risk. Although total leukemic samples were limiting, MPN13 was significantly associated with risk of transformation (p=4.7e-03) with low and high-risk patients experiencing 3-year cumulative incidences of transformation of 5.2% [95% CI 0.2%-10.2%] and 28.6% [95% CI 3.1%-54.0%] respectively, after adjusting for death as a competing risk. Future studies will transfer our scores to the Nanostring platform for testing in independent cohorts. In summary, we used transcriptional variation among MF stem and progenitor cells to derive novel gene expression scores predictive of survival and leukemic transformation and developed a new integrated 3-tier model for predicting risk in MF patients. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Stability and Determinants of the Demand for Money in Kenya

The structural characteristics of the Kenyan economy and its economy raise the issue of the correct monetary policy for the country, money demand takes a key role in macroeconomics in general and monetary economics in particular, which is why a stable feature of money demand is generally considered necessary for the formulation and conduct of an effective monetary policy. The goal of the study was to investigate the stability of the demand for money in Kenya using recent data due to lack of consensus. A clear understanding of the stability of money demand is important given its consequences for the formulation of monetary policy. The study employed Johansen testing approach for co-integration analysis based on the multiple regressions models and used annual data over the period 1982 to 2014. The purpose of the thesis was to investigate empirically the long-term and short-term determinants of demand for money, and to evaluate how demand for money reacts to inflation, treasury bills, gross domestic product, and commercial bank rates. To achieve these goals, it was hypothesized that the demand for money will make a significant contribution to the implementation of monetary policy by the use of the VECM. Empirical analysis was carried out successfully in R Econometric software; the critical values of ADF were non-stationary at-4.15 at 1%,-3.50 at 5% and-3.18 at 10%, suggesting the possible co-integration of the variables. The study utilizes the long-term correlation between the demand for money and its determinants, the multivariate p-value of the test: &lt; 2.2e-16 acts as a way of cross-checking from a medium to long-term perspective. The results indicate that the demands of the broad monetary aggregates are stable. The existing system in which M2 is considered to be the intermediate target variable of monetary policy may therefore be considered acceptable. It is important that monetary authorities turn their attention to the sectoral implications of money demand for successful monetary and fiscal policies in Kenya. In order to achieve the desired policy goals, policy makers should consider the effect of aggregated real income components on the recognition of different inputs and feedback on them against their policy instruments. This eliminates the bias of the aggregate. Researchers can also take the same path and investigate the role of sector money demand in supporting policy design and management. On the basis of these results, stability and monetary demand determinants suggest that monetary targeting is a viable choice for the conduct of monetary policy for the Central Bank of Kenya.

Still careless: findings from a cross-sectional study of young pedestrians\u2019 risky road crossing behaviors

BackgroundPedestrian-vehicle collision is one of the most common traffic injuries worldwide. This study aimed to investigate the determinants of pedestrians’ road crossing beliefs and behaviors in potentially risky situations using the Theory of Planned Behavior among Iranian young adults.MethodsThis was a population-based study on a sample of 562 young adults aged 18 to 25 years living in Tehran, Iran. Data were collected by using a self-administered validated questionnaire including constructs of the theory of planned behavior and items of perceived risk and severity. The data were analyzed using independent t-test, analysis of covariance and multivariate analysis of variance.ResultsFrom all the respondents, 17.8% reported that they had previous experience of vehicle-collision. Among the participants, those who had previous experience of vehicle-collision reported less safety behaviors in crossing the road than those who had not experience an accident. It was found significant differences between participants with and without a history of vehicle-collision for perceived risk (mean difference, adjusted multivariate P-value: − 5.77, 0.027) and perceived severity (− 6.08, 0.003), attitude toward traffic regulations (− 6.34, 0.006), attitude toward behavior (− 7.56, 0.005), perceived behavioral control (− 5.20, 0.018), behavioral intention (− 5.35, 0.046) and road crossing behavior in potentially risky situations (− 5.37, 0.004).ConclusionsPrevious unpleasant experience of vehicle-collision is not the only determinant of self-protective behaviors in road- crossing which indicate the role of cognitive and motivational factors such as, subjective norms, attitudes towards risk, feelings of invulnerability in case of facing with vehicle collision.

Abstract 1340: High NK cell number predicts poor overall survival in de novo treatment naive acute myeloid leukemia

Abstract Introduction Acute myeloid leukemia (AML) is a deadly disease associated with high mortality and morbidity. Previous studies report defects in NK cell maturation and function in AML patients leading to innate immune evasion. Considering that the number of NK cells may be an important indicator of innate immune regulation of AML, our group sought to understand the prognostic significance of NK cell number in de novo AML patients. Methods Patient samples were obtained from the Oregon Health and Science University (OHSU) Pathology Department between September 2010 and March 2016. This cohort study includes 82 newly diagnosed adult AML patients who had a bone marrow biopsy that was subjected to flow cytometry and a targeted-NGS panel within 30 days prior to induction treatment. Patients were excluded if they had antecedent hematological diseases or therapy-related AML. Bone marrow immunophenotyping was performed using the Ion Torrent PGM platform. Absolute numbers of NK cells were estimated from flow cytometric analysis of the bone marrow samples whereby lymphocytes were identified based on CD45 staining and light side scatter and NK cells were defined as CD3-/CD56+. Results This retrospective cohort of 82 newly diagnosed AML patients were 53.7% male with a median age of 61 years (range 18 - 83). Using 2017 ELN risk stratification guidelines, 36.6% of patients were classified as favorable, 26.8% as intermediate, and 36.6% as adverse. Absolute NK cell numbers in the bone marrow ranged from 1 to 896 cells/μl with a median of 98. When stratifying NK cells by staining intensity to identify subpopulations, the number of CD56 bright NK cells ranged from 0 to 69 cells/μl (median 3) and the number of CD56 dim NK cells ranged from 1 to 585 cells/μl (median 77). Overall survival (OS) was measured from diagnosis to date of death or last contact and varied from 1 day to 5.7 years (median 20.5 months). There were 3 deaths in the first week after diagnosis. Separate Cox proportional hazards regression models were applied to each NK cell group in both the univariable and multivariable settings, with the latter models accounting for patient age and ELN risk. Higher absolute numbers of NK cells (CD3-/CD56+) were significantly correlated with worse OS in both model settings, with a multivariable p-value of 0.001. The number of bright NK cells was a significant predictor of OS in the univariable model but lost significance when controlling for other variables. However, similar to the number of all NK cells and contrary to previous reports, higher counts of dim NK cells were significantly associated with shorter survival times, with a multivariable p-value of 0.001. Conclusions In this study, we show that higher NK cell numbers at diagnosis are associated with worse OS and thus play an important role in prognostication for AML patients. This association with survival is most evident in patients at least 60 years old or those with intermediate or adverse ELN risk. Citation Format: Weiwei Wang, Guang Fan, Andy Kaempf, evan Lind, Tomi Mori, Byung Park, Jennifer Saultz. High NK cell number predicts poor overall survival in de novo treatment naive acute myeloid leukemia [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1340.

Abstract MP035: The Clinical Epidemiology of Fatigue in Newly Diagnosed Heart Failure

Introduction: Fatigue is a common and distressing, but poorly understood symptom among patients with heart failure (HF). The underlying mechanisms of fatigue in HF have not been clearly elucidated, but may lie in both psychologic and physiologic factors. Whether fatigue remains associated with mortality independent of other common clinical attributes also remains unresolved. Accordingly, the current study sought to evaluate the prevalence, predictors, and prognostic value of clinically documented fatigue in newly diagnosed HF patients. Methods: This retrospective cohort study consisted of 12,285 newly diagnosed HF patients receiving health care services through the Geisinger Health System, with passive data collection through electronic medical records (EMR). Incident HF, fatigue, and other study variables were derived from coded data within EMRs. A collection of 87 candidate predictors were evaluated to ascertain the strongest independent predictors of fatigue using logistic regression. The collection of candidate variables was drawn from several domains including demographics, physical examination findings, medical history, laboratory results, and medications. Patients were followed for all-cause mortality for an average of 4.8 years. The associations between fatigue and six-month, 12-month, and overall mortality were evaluated with Cox proportional hazards regression. Results: Clinically documented fatigue was found in 4827 (39%) newly diagnosed HF patients. Among the 87 candidate predictors, 18 were independently associated with fatigue at a multivariable p-value threshold of 0.001. Depression was the strongest predictor of fatigue. Fatigue was often part of a symptom cluster, as other HF symptoms including dyspnea, chest pain, edema, syncope, and palpitations were significant predictors of fatigue. Volume depletion, low body mass index, and abnormal weight loss were also strong predictors of fatigue. Though fatigue was significantly associated with 6-month (HR=1.49, p&lt;0.01), 12-month (HR=1.39, p&lt;0.01), and all-cause mortality (HR=1.20, p&lt;0.01) in unadjusted models, effect sizes were attenuated and non-significant after adjustment for clinical variables with HRs of 1.12 (p=0.16), 1.07 (p=0.26), and 1.00 (p=0.89), for 6-month, 12-month, and overall mortality, respectively. Conclusions: Fatigue is a commonly documented symptom among newly diagnosed HF patients with likely origins in both psychologic and physiologic factors. Though fatigue provided a prognostic signal in the short-term, this was largely explained by physiologic confounders.

Insulin use, adipokine profiles and breast cancer prognosis

BackgroundType-2 diabetes mellitus (T2DM) and breast cancer (BC) share common cytokine signaling changes resultant from adipose tissue dysfunction. This modified adipokine signaling was shown to be directly associated with changes in the body mass index (BMI) and diet and it is expected to also be influenced by T2DM pharmacotherapy. We evaluated the relationship between pre-existing diabetes treatment, circulating adipokine levels at cancer diagnosis, and long-term outcomes. MethodsAll incident BC cases were reviewed (01/01/2003-12/31/2009, N=2194). Each of the subjects with baseline T2DM (cases) was matched with two other subjects without T2DM (controls) based on the following criteria: age, BMI, ethnicity, menopausal status and tumor stage. All cases and controls with available baseline plasma and tumor biopsies, and being surgery and BC treatment naïve, were included (N1=97, N2=194). Clinical history and vital status were documented. Adipokine levels (adiponectin, leptin, TNF-α, CRP, IL-1β, IL-1Ra, IL-6, and C-peptide) were assessed by either ELISA or Luminex® assays. Cancer outcomes were assessed by Kaplan-Meier analysis; associations between categorical variables were assessed by Fisher’s exact test, categorical and continuous variables by Kruskal-Wallis or Wilcoxon Rank-Sum test, where appropriate. Multivariate adjustments (MVP, multivariate p-value) were performed accounting for age, tumor stage, BMI, estrogen receptor (ER) status and cumulative comorbidity. All biomarker correlations were assessed by the Pearson method.Utilization of insulin and insulin secretagogues was associated with ER (−) phenotype (p=0.008, p=0.043) and poorer BC outcomes (p=0.012, p=0.033). Insulin users were found to have lower C-peptide and higher IL-6, TNF-α and CRP levels, of which elevated CRP and TNF-α were associated with poorer BC outcomes (p=0.003, MVP=0.210). Insulin remarked by higher leptin levels as compared to controls (p=0.052), but did not differ significantly from non-users. Although lower adiponectin levels were observed among non-insulin users as compared to controls (p<0.001, MVP=0.006), insulin use seemed to have restored adiponectin production. C-peptide levels were lower among insulin users as compared to non-users (p<0.001, MVP<0.001) and approached levels comparable with those of the controls. In the overall dataset, C-peptide lower than 0.75ng/ml were strongly associated with poorer survival (p=0.007, MVP=0.002). Among insulin users, C-peptide levels were inversely correlated with IL-1β and IL-1Ra levels only after full adjustment (p=0.012, p=0.030); the correlation was unremarkable in other groups. ConclusionInsulin use is associated with elevated leptin, CRP, TNFα, and lower C-peptide and also linked to poor BC outcomes. More research is needed to verify these findings; however, we are among the first to correlate pharmacotherapy use, measures of adipose tissue dysfunction and cancer outcomes.

Prognosis after temporal lobe epilepsy surgery: The value of combining predictors

Although several independent predictors of seizure freedom after temporal lobe epilepsy surgery have been identified, their combined predictive value is largely unknown. Using a large database of operated patients, we assessed the combined predictive value of previously reported predictors included in a single multivariable model. The database comprised a cohort of 484 patients who underwent temporal lobe surgery for drug-resistant epilepsy. Good outcome was defined as Engel class 1, one year after surgery. Previously reported independent predictors were tested in this cohort. To be included in our final prediction model, predictors had to show a multivariable p-value of <0.20. The final multivariable model included predictors obtained from the patient's history (absence of tonic-clonic seizures, absence of status epilepticus), magnetic resonance imaging [MRI; ipsilateral mesial temporal sclerosis (MTS), space occupying lesion], video electroencephalography (EEG; absence of ictal dystonic posturing, concordance between MRI and ictal EEG), and fluorodeoxyglucose positron emission tomography (FDG-PET; unilateral temporal abnormalities), that were related to seizure freedom in our data. The model showed an expected receiver-operating characteristic curve (ROC) area of 0.63 [95% confidence interval (CI) 0.57-0.68] for new patient populations. Intracranial monitoring and surgery-related parameters (including histology) were not important predictors of seizure freedom. Among patients with a high probability of seizure freedom, 85% were seizure-free one year after surgery; however, among patients with a high risk of not becoming seizure-free, still 40% were seizure-free one year after surgery. We could only moderately predict seizure freedom after temporal lobe epilepsy surgery. It is particularly difficult to predict who will not become seizure-free after surgery.

Two simple strategies of analysis to increase the power of experiments with multiple response variables

Summary Analyzing ecological experiments with many response variables is challenging because most statistical tools have low efficiency with such a design. It is widely known that introducing statistical control, i.e. testing one variable for significance while other response variables are held constant, increases the power of the analysis. Here we propose a very simple way to include statistical control to analyze experiments and show through simulations how, at least in simple experiments, the variable power rate (i.e. the power to detect an effect for each response variable) may increase 4× using our approach vs. using ANOVAs following a significant MANOVA. For simple experiments (i.e. one experimental treatment and a control) we propose the use of what we call Inverse Logistic Regression (ILR), named such because relative to the experimental design, the roles of dependent and independent variables become reversed. Using this approach one can obtain a multivariate p-value (i.e. full model) and univarite p-values (single variables) with a single analysis. In addition, our simulations show that when out of many possible response variables only a few correlated variables respond strongly to the treatment, the multivariate statistical power is doubled relative to that of MANOVA by iterative exploratory searching for the best subset of response variables using ILR. We also give an example of ILR with data from a field experiment in which ILR is more likely than MANOVA/ANOVA to uncover the pattern of treatment effects. Our approach can be extended to more complex unifactorial designs by using ordinal and/or multinomial distributions in generalized linear models or linear discriminant analysis when assumptions are met.

Association between Split Selection Instability and Predictive Error in Survival Trees

To evaluate split selection instability in six survival tree algorithms and its relationship with predictive error by means of a bootstrap study. We study the following algorithms: logrank statistic with multivariate p-value adjustment without pruning (LR), Kaplan-Meier distance of survival curves (KM), martingale residuals (MR), Poisson regression for censored data (PR), within-node impurity (WI), and exponential log-likelihood loss (XL). With the exception of LR, initial trees are pruned by using split-complexity, and final trees are selected by means of cross-validation. We employ a real dataset from a clinical study of patients with gallbladder stones. The predictive error is evaluated using the integrated Brier score for censored data. The relationship between split selection instability and predictive error is evaluated by means of box-percentile plots, covariate and cutpoint selection entropy, and cutpoint selection coefficients of variation, respectively, in the root node. We found a positive association between covariate selection instability and predictive error in the root node. LR yields the lowest predictive error, while KM and MR yield the highest predictive error. The predictive error of survival trees is related to split selection instability. Based on the low predictive error of LR, we recommend the use of this algorithm for the construction of survival trees. Unpruned survival trees with multivariate p-value adjustment can perform equally well compared to pruned trees. The analysis of split selection instability can be used to communicate the results of tree-based analyses to clinicians and to support the application of survival trees.