Functional limitations in ankylosing spondylitis (AS) may be due to peripheral joint or axial involvement. To determine if the Bath AS Functional Index (BASFI), an axial-focused measure, can detect limitations related to peripheral joint involvement equally well as the Health Assessment Questionnaire modified for the Spondyloarthropathies (HAQ-S), a peripheral arthritis-focused measure, and vice versa, we compared associations of each questionnaire with spinal and hip range of motion, peripheral arthritis, and enthesitis in patients with AS.
We examined patients every 4 to 6 months in this prospective longitudinal study. We used mixed linear models to examine associations between ten physical examination measures and the BASFI and HAQ-S.
We studied 411 patients for a median of 1.5 years (3 visits). In multivariate analyses, cervical rotation, chest expansion, lateral thoracolumbar flexion, hip motion, tender joint count, and tender enthesis count were equally strongly associated with the BASFI and HAQ-S. Peripheral joint swelling was more strongly associated with the HAQ-S. Individual items of the BASFI were more likely than items of the HAQ-S to be associated with unrelated physical exam measures (e.g. association between difficulty rising from a chair and cervical rotation), which may have diminished the axial/peripheral distinction for the BASFI.
The BASFI and HAQ-S had similar associations with impairments in axial measures, while the HAQ-S had stronger associations with the number of swollen peripheral joints. The HAQ-S should be considered for use in studies focused on spondyloarthritis with peripheral joint involvement.
Ankylosing spondylitis; functional limitations; metrology
Background. Behavioural and psychological symptoms of dementia (BPSD) cause significant patient and caregiver morbidity in vascular cognitive impairment (VCI). Objectives. To study and compare the occurrence and severity of BPSD between multi-infarct dementia (MID), subcortical ischaemic vascular disease (SIVD), and strategic infarct subtypes of poststroke VCI and to evaluate the relationship of these symptoms with the severity of cognitive impairment. Methods. Sixty patients with poststroke VCI were classified into MID, SIVD, and strategic infarct subtypes. BPSD were studied by the neuropsychiatric inventory (NPI). The severity of cognitive impairment was evaluated by the clinical dementia rating scale (CDR). Results. 95% of cases had at least one neuropsychiatric symptom, with depression being the commonest, irrespective of subtype or severity of VCI. Strategic infarct patients had the lowest frequency of all symptoms. SIVD showed a higher frequency and severity of apathy and higher total NPI scores, compared to MID. Apathy and appetite disturbances occurred more commonly with increasing CDR scores. The total NPI score correlated positively with the CDR score. Conclusion. Depression was the commonest neuropsychiatric symptom in VCI. The neuropsychiatric profiles of MID and SIVD were similar. The frequency and severity of apathy and the net burden of BPSD increased with increasing cognitive impairment.
Logistic regression has been the de facto, and often the only, model used in the description and analysis of relationships between a binary outcome and observed features. It is widely used to obtain the conditional probabilities of the outcome given predictors, as well as predictor effect size estimates using conditional odds ratios.
We show how statistical learning machines for binary outcomes, provably consistent for the nonparametric regression problem, can be used to provide both consistent conditional probability estimation and conditional effect size estimates. Effect size estimates from learning machines leverage our understanding of counterfactual arguments central to the interpretation of such estimates. We show that, if the data generating model is logistic, we can recover accurate probability predictions and effect size estimates with nearly the same efficiency as a correct logistic model, both for main effects and interactions. We also propose a method using learning machines to scan for possible interaction effects quickly and efficiently. Simulations using random forest probability machines are presented.
The models we propose make no assumptions about the data structure, and capture the patterns in the data by just specifying the predictors involved and not any particular model structure. So they do not run the same risks of model mis-specification and the resultant estimation biases as a logistic model. This methodology, which we call a “risk machine”, will share properties from the statistical machine that it is derived from.
Consistent nonparametric regression; Logistic regression; Probability machine; Odds ratio; Counterfactuals; Interactions
While rheumatoid arthritis (RA) medications may affect survival in RA, few studies consider the propensity for medication use, which may reflect selection bias in treatment allocation in survival models. We examined the relationship between methotrexate use and mortality in RA, after controlling for individual propensity scores for methotrexate use.
We studied 5626 patients with RA prospectively for 25 years to determine the hazard for death associated with methotrexate use, modeled in time-varying Cox regression models. We used the random forest method to generate individual propensity scores for methotrexate use at study entry and during follow-up in a time-varying fashion; these scores were included in the multivariate model. We also examined if selective discontinuation of methotrexate immediately prior to death altered the hazard for mortality, and examined the association of duration of methotrexate use with survival.
During follow-up 666 patients (12%) died. Methotrexate use was associated with reduced risk of death (adjusted hazard ratio 0.30; 95% confidence interval 0.09, 1.03). Selective methotrexate cessation immediately before death did not account for the protective association of methotrexate use with mortality. Only methotrexate use longer than one year was associated with lower risks of mortality, but associations were not stronger with longer durations of use.
Methotrexate use was associated with a 70% reduction in mortality risk in RA.
The objective of this study was to compare the occurrence and severity of behavioral and psychological symptoms of dementia (BPSD) between vascular dementia (VaD) and vascular cognitive impairment-no dementia (VCI-ND).
Materials and Methods:
Consecutive patients presenting with cognitive impairment at least 3 months after an ischemic stroke and with a Hachinski Ischemic Score ≥4 were included. VaD was diagnosed as per National Institute of Neurological Disorders and Stroke – Association Internationale pour la Recherche et l’Enseignement en Neurosciences criteria for probable VaD and VCI-ND on the lines of the Canadian study of health and aging. The severity of cognitive impairment and the behavioral/psychological symptoms were studied by means of the clinical dementia rating scale and the neuropsychiatric inventory (NPI) respectively.
All patients with VaD and 89% of those with VCI-ND had at least one BPSD. The mean no. of symptoms per patient and the total NPI scores were higher in VaD than in VCI-ND. Apathy and night-time behavior disturbances were significantly more common and severe in VaD.
BPSD are very common both in VCI-ND and in VaD. The profile of BPSD is similar in both groups, albeit more severe in VaD. The net burden of BPSD is higher in VaD as compared to VCI-ND.
Behavioral and psychological symptoms; neuropsychiatric inventory; vascular cognitive impairment; vascular cognitive impairment-no dementia; vascular dementia
Apraxia of speech (AOS) is a rare, but well-defined motor speech disorder. It is characterized by irregular articulatory errors, attempts of self-correction and persistent prosodic abnormalities. Similar to aphasia, AOS is also localized to the dominant cerebral hemisphere. We report a case of Crossed Aphasia with AOS in a 48-year-old right-handed man due to an ischemic infarct in right cerebral hemisphere.
Apraxia of speech; crossed aphasia; crossed apraxia of speech; right-handed
To evaluate clinical profile and short-term outcome of psychogenic non-epileptic seizures (PNES) in Indian adult population.
Setting and Design:
A prospective observational study, conducted at tertiary teaching institute at New Delhi.
Materials and Methods:
Sixty-three patients with confirmed PNES were enrolled. The diagnosis was based on witnessing the event during video-electroencephalography (Video-EEG) monitoring. A detailed clinical evaluation was done including evaluation for coexistent anxiety or depressive disorders. Patients were divided into two groups on the basis of excessive or paucity of movements during PNES attacks. Patients were followed-up to 12 months for their PNES frequency.
Means and standard deviations were calculated for continuous variables. Chi-square and Students t-test were used to compare categorical and continuous variables respectively.
The mean age at onset of PNES was 25.44 years; with F:M ratio of 9.5:1. Coexistent epilepsy was present in 13 (20.63%) cases. Twenty-two patients (44%) with only PNES (n = 50) had received antiepileptic drugs. Out of 63 patients of PNES 24 (38.1%) had predominant motor phenomenon, whereas 39 (61.9%) had limp attacks. The common features observed were pre-ictal headache, ictal eye closure, jaw clenching, resistant behavior, ictal weeping, ictal vocalization, and unresponsiveness during episodes. Comorbid anxiety and depressive disorders was seen in 62.3% and 90.16% patients, respectively. Short-term (6-12 months) outcome of 45 patients was good (seizure freedom in 46.66% and >50% improvement in 24.44% cases).
PNES is common, but frequently misdiagnosed and treated as epileptic seizures. A high index of suspicion is required for an early diagnosis. Proper disclosure of diagnosis and management of the psychiatric comorbidities can improve their outcome.
Limited sample size and change in seizures frequency as the only parameter for the assessment of the outcome are the two major limitations of our study.
Psychogenic non-epileptic seizures; Coexistent epilepsy; video-electroencephalography
To evaluate the prevalence, correlates and subgroups at highest risk for suicidal ideation among adults with arthritis.
We used data on U.S. adults with arthritis, aged ≥40, participating in the 2007–2008 NHANES survey. Suicidal ideation was assessed by item 9 of the Patient Health Questionnaire-9 (PHQ-9). Socio-demographic factors, health behaviors and comorbid conditions were examined as potential correlates. Depression was measured by the PHQ-8 score (range 1–24). We used random forests to identify subgroups at highest risk for suicidal ideation. To determine if any correlates were unique to arthritis, we compared results to those for persons with diabetes mellitus and cancer.
The prevalence (± standard error) of suicidal ideation was 5.6% ± 0.8% among persons with arthritis and 2.4% ± 0.4% among those without. The most important correlates for suicidal ideation in adults with arthritis were depression, anxiety, duration of arthritis, age, income/poverty ratio, number of close friends, pain, alcohol, excessive daytime sleepiness and comorbidities. Eleven of 16 most important contributors for suicidal ideation among adults with arthritis were also important for people with diabetes and cancer. Among persons with arthritis, subgroups at highest risk for suicidal ideation were those with PHQ-8 between 18 and 24 and less than 4.5 years of arthritis (96.5%), and those with PHQ-8 between 7 and 17, ≥1.24 days of binges/month and either income ≥$45,000/year (85.4%) or income <$45,000/year and >3 comorbidities (70.8%).
Depression, short duration of arthritis, binge drinking, income, and >3 comorbidities identified subgroups of adults with arthritis at greatest risk for suicidal ideation.
Suicidal ideation; arthritis; correlates; depression
Genetics Analysis Workshop 17 provided common and rare genetic variants from exome sequencing data and simulated binary and quantitative traits in 200 replicates. We provide a brief review of the machine learning and regression-based methods used in the analyses of these data. Several regression and machine learning methods were used to address different problems inherent in the analyses of these data, which are high-dimension, low-sample-size data typical of many genetic association studies. Unsupervised methods, such as cluster analysis, were used for data segmentation and subset selection. Supervised learning methods, which include regression-based methods (e.g., generalized linear models, logic regression, and regularized regression) and tree-based methods (e.g., decision trees and random forests), were used for variable selection (selecting genetic and clinical features most associated or predictive of outcome) and prediction (developing models using common and rare genetic variants to accurately predict outcome), with the outcome being case-control status or quantitative trait value. We include a discussion of cross-validation for model selection and assessment and a description of available software resources for these methods.
unsupervised learning; supervised learning; cluster analysis; logistic regression; Poisson regression; logic regression; LASSO; ridge regression; decision trees; random forests; cross-validation; software
We have recently proposed a new model of cancer metabolism to explain the role of aerobic glycolysis and L-lactate production in fueling tumor growth and metastasis. In this model, cancer cells secrete hydrogen peroxide (H2O2), initiating oxidative stress and aerobic glycolysis in the tumor stroma. This, in turn, drives L-lactate secretion from cancer-associated fibroblasts. Secreted L-lactate then fuels oxidative mitochondrial metabolism (OXPHOS) in epithelial cancer cells, by acting as a paracrine onco-metabolite. We have previously termed this type of two-compartment tumor metabolism the “reverse Warburg effect,” as aerobic glycolysis takes place in stromal fibroblasts, rather than epithelial cancer cells. Here, we used MCT4 immunostaining of human breast cancer tissue microarrays (TMAs; >180 triple-negative patients) to directly assess the prognostic value of the “reverse Warburg effect.” MCT4 expression is a functional marker of hypoxia, oxidative stress, aerobic glycolysis and L-lactate efflux. Remarkably, high stromal MCT4 levels (score = 2) were specifically associated with decreased overall survival (<18% survival at 10 years post-diagnosis). In contrast, patients with absent stromal MCT4 expression (score = 0), had 10-year survival rates of ∼97% (p-value < 10−32). High stromal levels of MCT4 were strictly correlated with a loss of stromal Cav-1 (p-value < 10−14), a known marker of early tumor recurrence and metastasis. In fact, the combined use of stromal Cav-1 and stromal MCT4 allowed us to more precisely identify high-risk triple-negative breast cancer patients, consistent with the goal of individualized risk-assessment and personalized cancer treatment. However, epithelial MCT4 staining had no prognostic value, indicating that the “conventional” Warburg effect does not predict clinical outcome. Thus, the “reverse Warburg effect” or “parasitic” energy-transfer is a key determinant of poor overall patient survival. As MCT4 is a druggable target, MCT4 inhibitors should be developed for the treatment of aggressive breast cancers, and possibly other types of human cancers. Similarly, we discuss how stromal MCT4 could be used as a biomarker for identifying high-risk cancer patients that could likely benefit from treatment with FDA-approved drugs or existing MCT-inhibitors (such as, AR-C155858, AR-C117977 and AZD-3965).
caveolin-1; oxidative stress; pseudohypoxia; lactate shuttle; MCT4; metabolic coupling; tumor stroma; predictive biomarker; SLC16A3; monocarboxylic acid transporter; two-compartment tumor metabolism
Machine learning approaches are an attractive option for analyzing large-scale data to detect genetic variants that contribute to variation of a quantitative trait, without requiring specific distributional assumptions. We evaluate two machine learning methods, random forests and logic regression, and compare them to standard simple univariate linear regression, using the Genetic Analysis Workshop 17 mini-exome data. We also apply these methods after collapsing multiple rare variants within genes and within gene pathways. Linear regression and the random forest method performed better when rare variants were collapsed based on genes or gene pathways than when each variant was analyzed separately. Logic regression performed better when rare variants were collapsed based on genes rather than on pathways.
G protein-coupled receptor kinase 2 (GRK2), which is upregulated in the failing human myocardium, appears to have a role in heart failure (HF) pathogenesis. In peripheral lymphocytes, GRK2 expression has been shown to reflect myocardial levels. This study represents an attempt to define the role for GRK2 as a potential biomarker of left ventricular function in HF patients. We obtained blood from 24 HF patients before and after heart transplantation and followed them for up to 1 year, also recording hemodynamic data and histological results from endomyocardial biopsies. We determined blood GRK2 protein by Western blotting and enzyme-linked immunosorbent assay. GRK2 levels were obtained before transplant and at first posttransplant biopsy. GRK2 levels significantly declined after transplant and remained low over the course of the study period. After transplantation, we found that blood GRK2 significantly dropped and remained low consistent with improved cardiac function in the transplanted heart. Blood GRK2 has potential as a biomarker for myocardial function in end-stage HF.
heart failure; biomarkers; GRK2; heart transplantation
Background & Aims
Gucy2c is the intestinal cell receptor for the paracrine hormones guanylin and uroguanylin that converts GTP to cyclic (c)GMP. It functions as a tumor suppressor; its loss disrupts intestinal homeostasis and promotes tumorigenesis. We investigated the effects of Gucy2c loss on intestinal cell proliferation, metabolism, signaling, and tumorigenesis in mice.
Intestinal cell proliferation and metabolism were examined in Gucy2c−/− and wild-type mice and human colon cancer cells by microscopy, immunoblot, and functional analyses. AKT regulation and signaling were examined and the role of AKT in Gucy2c-dependent tumorigenesis was defined in Gucy2c−/−Akt1−/− mice. Microarray analyses compared gene expression profiles of intestine of Gucy2c−/− and wild-type mice.
The size and number of intestinal crypts increased in Gucy2c−/− mice; the associated epithelial cells exhibited accelerated proliferation, increased glycolysis, and reduced oxidative phosphorylation, which was reversed by oral administration of cGMP. Conversely, activating GUCY2C in human colon cancer cells delayed cell cycle progression (inhibiting DNA synthesis and colony formation), reduced glycolysis, and increased mitochondrial ATP production. AKT signaling pathways were activated in intestines of Gucy2c−/− mice, associated with increased AKT phosphorylation. Disruption of AKT activity, pharmacologically or genetically, reduced DNA synthesis, proliferation, and glycolysis and increased mitochondrial biogenesis. Intestinal tumorigenesis increased following administration of azoxymethane to Gucy2c−/− mice, compared with wild-type mice, but was eliminated in Gucy2c−/−Akt1−/− mice.
Gucy2c is a tumor suppressor that controls proliferation and survival of intestinal epithelial cells by inactivating AKT signaling. This receptor and its ligands, which are paracrine hormones, might be novel candidates for anti-colorectal cancer therapy.
The expression of protein phosphatase 32 (PP32, ANP32A) is low in poorly differentiated pancreatic cancers and is linked to the levels of HuR (ELAV1), a predictive marker for gemcitabine response. In pancreatic cancer cells, exogenous overexpression of pp32 inhibited cell growth, supporting its long-recognized role as a tumor suppressor in pancreatic cancer. In chemotherapeutic sensitivity screening assays, cells overexpressing pp32 were selectively resistant to the nucleoside analogs gemcitabine and cytarabine (ARA-C), but were sensitized to 5-fluorouracil; conversely, silencing pp32 in pancreatic cancer cells enhanced gemcitabine sensitivity. The cytoplasmic levels of pp32 increased after cancer cells are treated with certain stressors, including gemcitabine. pp32 overexpression reduced the association of HuR with the mRNA encoding the gemcitabine-metabolizing enzyme deoxycytidine kinase (dCK), causing a significant reduction in dCK protein levels. Similarly, ectopic pp32 expression caused a reduction in HuR binding of mRNAs encoding tumor-promoting proteins (e.g., VEGF and HuR), while silencing pp32 dramatically enhanced the binding of these mRNA targets. Low pp32 nuclear expression correlated with high-grade tumors and the presence of lymph node metastasis, as compared to patients' tumors with high nuclear pp32 expression. Although pp32 expression levels did not enhance the predictive power of cytoplasmic HuR status, nuclear pp32 levels and cytoplasmic HuR levels associated significantly in patient samples. Thus, we provide novel evidence that the tumor suppressor function of pp32 can be attributed to its ability to disrupt HuR binding to target mRNAs encoding key proteins for cancer cell survival and drug efficacy.
Here, we investigated the possible predictive value of stromal caveolin-1 (Cav-1) as a candidate biomarker for clinical outcome in triple negative (TN) breast cancer patients. A cohort of 85 TN breast cancer patients was available, with the necessary annotation and nearly 12 years of follow-up data. Our primary outcome of interest in this study was overall survival. Interestingly, TN patients with high-levels of stromal Cav-1 had a good clinical outcome, with >50% of the patients remaining alive during the follow-up period. In contrast, the median survival for TN patients with moderate stromal Cav-1 staining was 33.5 months. Similarly, the median survival for TN patients with absent stromal Cav-1 staining was 25.7 months. A comparison of 5-year survival rates yields a similar pattern. TN patients with high stromal Cav-1 had a good 5-year survival rate, with 75.5% of the patients remaining alive. In contrast, TN patients with moderate or absent stromal Cav-1 levels had progressively worse 5-year survival rates, with 40 and 9.4% of the patients remaining alive. In contrast, in a parallel analysis, the levels of tumor epithelial Cav-1 had no prognostic significance. As such, the prognostic value of Cav-1 immunostaining in TN breast cancer patients is compartment-specific, and selective for an absence of Cav-1 staining in the stromal fibroblast compartment. A recursive-partitioning algorithm was used to assess which factors are most predictive of overall survival in TN breast cancer patients. In this analysis, we included tumor size, histologic grade, whether the patient received surgery, radiotherapy or chemotherapy, CK5/6, EGFR, p53 and Ki67 status, as well as the stromal Cav-1 score. This analysis indicated that stromal loss of Cav-1 expression was the most important prognostic factor for overall survival in TN breast cancer. Virtually identical results were obtained with CK5/6 (+) and/or EGFR (+) TN breast cancer cases, demonstrating that a loss of stromal Cav-1 is also a strong prognostic factor for basal-like breast cancers. Our current findings may have important implications for the close monitoring and treatment stratification of TN and basal-like breast cancer patients.
caveolin-1; mammary tumor stroma; stromal biomarkers; cancer survival; cancer-associated fibroblasts
RNA-binding protein HuR binds U- or AU-rich sequences in the 3′-untranslated regions (UTRs) of target mRNAs, stabilizing them and/or modulating their translation. Given HuR’s links with cancer, we studied the consequences of modulating HuR levels in pancreatic cancer cells. HuR-overexpressing cancer cells, in some instances, are up to 30-fold more sensitive to treatment with gemcitabine (GEM), the main chemotherapeutic component of treatment regimens for pancreatic ductal adenocarcinoma (PDA), compared to control cells. In pancreatic cancer cells, HuR associates with deoxycytidine kinase (dCK) mRNA, which encodes the enzyme that metabolizes and thereby activates GEM. GEM exposure to pancreatic cancer cells, enriches the association between HuR and dCK mRNA and increases cytoplasmic HuR levels. Accordingly, HuR overexpression elevates, while HuR silencing reduces, dCK protein expression in pancreatic cancer cells. In a clinical correlate study of GEM treatment, we found a 7-fold increase in risk of mortality in PDA patients with low cytoplasmic HuR levels compared to patients with high HuR levels, after adjusting for other treatments and demographic variables. These data support the notion that HuR is a key mediator of GEM efficacy in cancer cells, at least in part through its ability to regulate dCK levels post-transcriptionally. We propose that HuR levels in PDA modulate the therapeutic efficacy of GEM, thus serving as a marker of the clinical utility of this common chemotherapeutic agent and a potential target for intervention in pancreatic cancer.
pancreatic ductal adenocarcinoma; pancreatic cancer; HuR; gemcitabine; deoxycytidine kinase; mRNA binding protein; gemcitabine response; RNA binding proteins
Tobacco smoking is responsible for over 90% of lung cancer cases, and yet the precise molecular alterations induced by smoking in lung that develop into cancer and impact survival have remained obscure.
We performed gene expression analysis using HG-U133A Affymetrix chips on 135 fresh frozen tissue samples of adenocarcinoma and paired noninvolved lung tissue from current, former and never smokers, with biochemically validated smoking information. ANOVA analysis adjusted for potential confounders, multiple testing procedure, Gene Set Enrichment Analysis, and GO-functional classification were conducted for gene selection. Results were confirmed in independent adenocarcinoma and non-tumor tissues from two studies. We identified a gene expression signature characteristic of smoking that includes cell cycle genes, particularly those involved in the mitotic spindle formation (e.g., NEK2, TTK, PRC1). Expression of these genes strongly differentiated both smokers from non-smokers in lung tumors and early stage tumor tissue from non-tumor tissue (p<0.001 and fold-change >1.5, for each comparison), consistent with an important role for this pathway in lung carcinogenesis induced by smoking. These changes persisted many years after smoking cessation. NEK2 (p<0.001) and TTK (p = 0.002) expression in the noninvolved lung tissue was also associated with a 3-fold increased risk of mortality from lung adenocarcinoma in smokers.
Our work provides insight into the smoking-related mechanisms of lung neoplasia, and shows that the very mitotic genes known to be involved in cancer development are induced by smoking and affect survival. These genes are candidate targets for chemoprevention and treatment of lung cancer in smokers.
The epidemiologic literature is replete with conceptual discussions about causal inference, but little is known about how the causal criteria are applied in public health practice. The criteria for causal inference in use today by epidemiologists have been shaped substantially by their use over time in reports of the U.S. Surgeon General on Smoking and Health.
We reviewed two classic reports on smoking and health from expert committees convened by the US Surgeon General, in 1964 and 1982, in order to evaluate and contrast how the committees applied causal criteria to the available evidence for the different cancer sites at different time periods. We focus on the evidence for four cancer sites in particular that received detailed reviews in the reports: lung, larynx, esophagus and bladder.
We found that strength of association and coherence (especially dose-response, biological plausibility and epidemiologic sense) appeared to carry the most weight; consistency carried less weight, and temporality and specificity were apparently not applied at all in some cases. No causal claim was made for associations with a summary odds ratio of less than 3.0.
Our findings suggest that the causal criteria as described in textbooks and the Surgeon General reports can have variable interpretations and applications in practice. While the authors of these reports may have considered evidential factors that they did not explicitly cite, such lack of transparency of methods undermines the purpose of the causal criteria to promote objective, evidence-based decision making. Further empirical study and critical examination of the process by which causal conclusions are reached can play an important role in advancing the practice of epidemiology by helping public health scientists to better understand the practice of causal inference.