Simulation enhanced distributed lag models for mortality displacement.

Distributed lag models (DLM) are attractive methods for dealing with mortality displacement, however their estimates can have substantial bias when the data is generated by a multi-state model. In particular DLMs are not valid for mortality displacement. Alternative methods are scarce and lack feasibility and validation. We investigate the breakdown of DLM in three state models by means of simulation and propose simulation enhanced distributed lag models (SEDLM) to overcome the defects. The new method provides simultaneous estimates of the net effect (entry) and the displacement effect (exit). These have improved performance over the singular estimate from a regular DLM. SEDLM entry estimates have negligible bias and their variance is reduced. The exit estimates are unbiased and their variance is one order of magnitude lower with respect to the entry estimates. Applying SEDLM to the original Chicago data, the 95% highest posterior density intervals for both entry and exit contain 0, providing neither evidence for a 'displacement effect' nor for a 'net effect'.

Respiratory medication sales and urban air pollution in Brussels (2005 to 2011).

BACKGROUND: We investigated the associations between daily sales of respiratory medication and air pollutants in the Brussels-Capital Region between 2005 and 2011. METHODS: We used over-dispersed Poisson Generalized Linear Models to regress daily individual reimbursement data of prescribed asthma and COPD medication from the social security database against each subject's residential exposure to outdoor particulate matter (PM10) or NO2 estimated, by interpolation from monitoring stations. We calculated cumulative risk ratios (RR) and their 95% confidence intervals (CI) for interquartile ranges (IQR) of exposure for different windows of past exposure for the entire population and for seven age groups. RESULTS: Median daily concentrations of PM10 and NO2 were 25µg/m(3) (IQR=17.1) and 38µg/m(3) (IQR=20.5), respectively. PM10 was associated with daily medication sales among individuals aged 13 to 64y. For NO2, significant associations were observed among all age groups except >84y. The highest RR were observed for NO2, among adolescents, including three weeks lags (RR=1.187 95%CI: 1.097-1.285). CONCLUSION: The associations found between temporal changes in exposure to air pollutants and daily sales of respiratory medication in Brussels indicate that urban air pollution contributes to asthma and COPD morbidity in the general population.

Short-Term Effect of Pollen and Spore Exposure on Allergy Morbidity in the Brussels-Capital Region.

Belgium is among the European countries that are the most affected by allergic rhinitis. Pollen grains and fungal spores represent important triggers of symptoms. However, few studies have investigated their real link with disease morbidity over several years. Based on aeroallergen counts and health insurance datasets, the relationship between daily changes in pollen, fungal spore concentrations and daily changes in reimbursable systemic antihistamine sales has been investigated between 2005 and 2011 in the Brussels-Capital Region. A Generalized Linear Model was used and adjusted for air pollution, meteorological conditions, flu, seasonal component and day of the week. We observed an augmentation in drug sales despite no significant increase in allergen levels in the long term. The relative risk of buying allergy medications associated with an interquartile augmentation in pollen distributions increased significantly for Poaceae, Betula, Carpinus, Fraxinus and Quercus. Poaceae affected the widest age group and led to the highest increase of risk which reached 1.13 (95% CI [1.11-1.14]) among the 19- to 39-year-old men. Betula showed the second most consistent relationship across age groups. Clear identification of the provoking agents may improve disease management by customizing prevention programmes. This work also opens several research perspectives related to impact of climate modification or subpopulation sensitivity.

Direct cost saving potential in medication costs due to a reduction in outdoor air pollution for the Brussels Capital Region.

INTRODUCTION: The adverse health effects of exposure to air pollution have been well-established and include mortality, hospital admissions, emergency department visits, etc, but also less severe outcomes such as medication use and purchase. The economic impact, an additional motivator for policy, has been studied primarily for the more severe outcomes. METHODS: Purchase data of reimbursed medications typically prescribed for asthma and chronic obstructive pulmonary disease, were obtained through the mandatory Belgian health insurance system. A time series analyses approach was used to model daily sales on daily air pollution concentrations (NO2, PM10 and PM2.5) for residents of the Brussels Capital Region as a whole. In addition, a higher geographical resolution of both sales and pollutant concentrations allowed for a multi-sector approach. Annual savings were estimated for the scenario of a 10% reduction in each of the pollutants. RESULTS: Medication purchase was significantly associated with NO2 concentrations, leading to an annual cost saving potential of € 107,845 [95%CI: € 71,483-€ 143,823] in R03 sales (WHO classification for drugs of obstructive airway diseases). Saving potentials of PM10 and PM2.5 were not significant. Estimates were not sensitive to the geographical resolution, however, higher precision can be obtained with higher resolution data, subject to the condition that the number of sales is sufficiently large.

High natural killer cell number might identify stroke patients at risk of developing infections.

OBJECTIVE: To investigate early changes in leukocyte subsets and autonomic function as predictors of the development of poststroke infections. METHODS: We assessed the time course of leukocyte subsets in the blood of 59 patients with acute ischemic stroke. We divided the patients into 2 groups: those who developed infections during the first 7 days after stroke onset and those who did not. We measured urinary norepinephrine and epinephrine concentrations and pulse rate variability indices within 24 hours of admission. RESULTS: We found that the number of circulating natural killer (NK) cells within the first hours after stroke was higher in stroke patients who developed infections (mean 435 cells/mL; 95% confidence interval [CI] 321-588) than in stroke patients who did not develop infections (mean 236 cells/mL; 95% CI 186-300; p = 0.001). This was followed by a decrease in all lymphocyte subsets from admission to day 1, varying between 22% and 40%, which was not seen in patients without poststroke infection (mean increase varied between 2% and 23%; all p < 0.005). In the group that developed infections, pulse rate variability revealed a decreased high frequency component. These findings all remained significant after adjustment for age and stroke volume. CONCLUSIONS: High circulating NK cell count within the first hours after ischemic stroke onset followed by a drop in all lymphocyte subsets identified patients who developed infections and may be caused by a sympathovagal imbalance with sympathetic overweight. These findings need to be validated in larger studies.

Pulmonary disease in cystic fibrosis: assessment with chest CT at chest radiography dose levels.

PURPOSE: To investigate a computed tomographic (CT) protocol with iterative reconstruction at conventional radiography dose levels for the assessment of structural lung abnormalities in patients with cystic fibrosis ( CF cystic fibrosis ). MATERIALS AND METHODS: In this institutional review board-approved study, 38 patients with CF cystic fibrosis (age range, 6-58 years; 21 patients <18 years and 17 patients >18 years) underwent investigative CT (at minimal exposure settings combined with iterative reconstruction) as a replacement of yearly follow-up posteroanterior chest radiography. Verbal informed consent was obtained from all patients or their parents. CT images were randomized and rated independently by two radiologists with use of the Bhalla scoring system. In addition, mosaic perfusion was evaluated. As reference, the previous available conventional chest CT scan was used. Differences in Bhalla scores were assessed with the χ(2) test and intraclass correlation coefficients ( ICC intraclass correlation coefficient s). Radiation doses for CT and radiography were assessed for adults (>18 years) and children (<18 years) separately by using technical dose descriptors and estimated effective dose. Differences in dose were assessed with the Mann-Whitney U test. RESULTS: The median effective dose for the investigative protocol was 0.04 mSv (95% confidence interval [ CI confidence interval ]: 0.034 mSv, 0.10 mSv) for children and 0.05 mSv (95% CI confidence interval : 0.04 mSv, 0.08 mSv) for adults. These doses were much lower than those with conventional CT (median: 0.52 mSv [95% CI confidence interval : 0.31 mSv, 3.90 mSv] for children and 1.12 mSv [95% CI confidence interval : 0.57 mSv, 3.15 mSv] for adults) and of the same order of magnitude as those for conventional radiography (median: 0.012 mSv [95% CI confidence interval : 0.006 mSv, 0.022 mSv] for children and 0.012 mSv [95% CI confidence interval : 0.005 mSv, 0.031 mSv] for adults). All images were rated at least as diagnostically acceptable. Very good agreement was found in overall Bhalla score ( ICC intraclass correlation coefficient , 0.96) with regard to the severity of bronchiectasis ( ICC intraclass correlation coefficient , 0.87) and sacculations and abscesses ( ICC intraclass correlation coefficient , 0.84). Interobserver agreement was excellent ( ICC intraclass correlation coefficient , 0.86-1). CONCLUSION: For patients with CF cystic fibrosis , a dedicated chest CT protocol can replace the two yearly follow-up chest radiographic examinations without major dose penalty and with similar diagnostic quality compared with conventional CT.

Determinants of medication use in a multi-ethnic population of pregnant women: a cross-sectional study.

OBJECTIVES: To investigate determinants of medication use among multi-ethnic pregnant women. METHODS: A total of 641 pregnant women participated in this cross-sectional study in a Brussels university hospital. A questionnaire was used to obtain data on socio-demographic characteristics and medication use. Chi-squared tests and binary logistic regression analyses were performed using SPSS 19. RESULTS: Medication use during pregnancy (37%) was positively associated with age older than 35 years, Western origin, being born in Belgium, high education and employment status. Highly educated Western women had a greater exposure to drugs with an unknown safety profile than Arab/Turkish and 'Other (non-Western) origins' women. In the latter two groups, low education and nulliparity were the most important determinants of lower drug use. Nulliparous Arab/Turkish women used significantly less medications (17%) during pregnancy than parous women with the same ethnic background (34%; p = 0.024). CONCLUSIONS: Medication use during pregnancy is considerable but differs according to ethnicity. Age, parity, educational level, occupational status and duration of stay in Belgium are important determinants that should be taken into account for risk assessment and preventive measures targeting pregnant women.

High-throughput 16S rRNA gene sequencing reveals alterations of intestinal microbiota in myalgic encephalomyelitis/chronic fatigue syndrome patients.

Human intestinal microbiota plays an important role in the maintenance of host health by providing energy, nutrients, and immunological protection. Intestinal dysfunction is a frequent complaint in myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) patients, and previous reports suggest that dysbiosis, i.e. the overgrowth of abnormal populations of bacteria in the gut, is linked to the pathogenesis of the disease. We used high-throughput 16S rRNA gene sequencing to investigate the presence of specific alterations in the gut microbiota of ME/CFS patients from Belgium and Norway. 43 ME/CFS patients and 36 healthy controls were included in the study. Bacterial DNA was extracted from stool samples, PCR amplification was performed on 16S rRNA gene regions, and PCR amplicons were sequenced using Roche FLX 454 sequencer. The composition of the gut microbiota was found to differ between Belgian controls and Norwegian controls: Norwegians showed higher percentages of specific Firmicutes populations (Roseburia, Holdemania) and lower proportions of most Bacteroidetes genera. A highly significant separation could be achieved between Norwegian controls and Norwegian patients: patients presented increased proportions of Lactonifactor and Alistipes, as well as a decrease in several Firmicutes populations. In Belgian subjects the patient/control separation was less pronounced, however some abnormalities observed in Norwegian patients were also found in Belgian patients. These results show that intestinal microbiota is altered in ME/CFS. High-throughput sequencing is a useful tool to diagnose dysbiosis in patients and could help designing treatments based on gut microbiota modulation (antibiotics, pre and probiotics supplementation).

An analysis of the decision-making process for single implant treatment in general practice.

PURPOSE: To study the frequency of and factors associated with the decision to perform single implant treatment after tooth extraction by general practitioners in a private, fee-for-service setting. MATERIAL AND METHODS: One hundred practitioners with a general dental practice in Ghent were randomly selected. Clinicians were asked to fill in a study form for every single extraction they performed during an 8-week period. The form related to the treatment decision and a number of patient- and clinician-related factors. RESULTS: Ninety-four general dentists (52 men, 42 women; mean age 49; range 24-68) agreed to participate and extracted 1180 single teeth. After exclusion of third molars and cases where the reason for tooth loss would generally prohibit replacement, 900 cases were identified. In 24% of these patients, there was no treatment decision and in 18% replacement was deemed unnecessary. When replacement was necessary (n = 526), removable partial denture (RPD), fixed partial denture (FPD), single implant treatment and resin-bonded bridge were chosen in 54%, 24%, 21% and 1% of the patients, respectively. Multinomial logistic regression was used to evaluate the decision-making process for single implant treatment against RPD and FPD. In relation to RPD, single implant treatment was more likely in highly educated patients with few missing teeth and no bone loss at adjacent teeth. In relation to FPD, single implant treatment was more likely in patients with intact adjacent teeth and when the tooth was extracted by a female dentist. Experience in implant prosthetics was positively associated with single implant treatment in all analyses. CONCLUSIONS: Single implant treatment is not the most common procedure in daily practice to restore a single tooth gap. Patient's education, oral factors and clinician-related factors may affect the decision-making process, whereas medical factors may not.

Health behaviour and health awareness in infant mortality in the Gaza Strip.

BACKGROUND: Infant death rate has declined over the past decades, yet remains high in Palestine. The topic of infant death is well researched and for the first time in Palestine. The objective of our study is to assess the level of awareness and health behaviour of mothers and how this behaviour has affected infant mortality. METHODS: Person to person interviews were done with 550 mothers of infants (275 cases and 275 controls) in the Gaza Strip. Stillbirths were excluded. Binary logistic regression analyses were used to identify the relationship of health behavioural factors and infant mortality. RESULT: The study showed that infant mortality risks were higher in mothers exposed to passive smoking. Infant mortality was lower for infants receiving exclusive breastfeeding. There was also a significant association between the sleeping position of the baby and infant mortality. CONCLUSION: The findings underscore the importance of explicit attention to health education. A well-organized consultation and health promotion approach should focus on couples whose child has died, in order to combat infant mortality.

Determinants of folic acid use in a multi-ethnic population of pregnant women: a cross-sectional study.

AIMS: To investigate the possible differences in folic acid use and to identify the determinants of antenatal folic acid use among multi-ethnic pregnant women. METHODS: Three hundred and fifty pregnant women participated in this cross-sectional study in a university hospital in Brussels, Belgium. A questionnaire was used to obtain data on socio-demographic characteristics and folic acid use. χ(2)-tests and binary logistic regression analyses were performed using SPSS 17. RESULTS: In the overall cohort, 59.2% used folic acid supplements during pregnancy. This supplement use was associated with an age of 26-35 years, being of Western origin, with high education and employment status, understanding physician's instructions, and early booking for antenatal care. Education (odds ratio, OR: 2.24; 95% confidence interval, CI: 1.08-4.63) and early booking for antenatal care OR: 2.45; 95% CI: 1.11-5.40) were the most important determinants. In particular for Arab/Turkish women, a lower employment status OR: 0.42; 95% CI: 0.24-0.73) was associated with a higher risk of not using folic acid supplements. CONCLUSIONS: The lower use of folic acid supplements in Arab/Turkish ethnicities, which may be associated with an increased risk of neural tube defects, is related to socio-economic factors rather than to lower educational attainment. As recommended by other studies, fortification of ethnic minority food may be warranted to reduce the risk of neural tube defects.

The value of a family history of sudden death in patients with diagnostic type I Brugada ECG pattern.

AIMS: We sought to investigate the value of a family history of sudden death (SD) in Brugada syndrome (BS). METHODS AND RESULTS: Two hundred and eighty consecutive patients (mean age: 41 ± 18 years, 168 males) with diagnostic type I Brugada ECG pattern were included. Sudden death occurred in 69 (43%) of 157 families. One hundred and ten SDs were analysed. During follow-up VF (ventricular fibrillation) or SD-free survival rate was not different between patients with or without a family history of SD of a first-degree relative, between patients with or without a family history of multiple SD of a first-degree relative at any age and between patients with or without a family history of SD in first-degree relatives ≤35 years. One patient had family history of SD of two first-degree relative ≤35 years with arrhythmic event during follow-up. In univariate analysis male gender (P = 0.01), aborted SD (P < 0.001), syncope (P = 0.04), spontaneous type I ECG (P < 0.001), and inducibility during electrophysiological (EP) study (P < 0.001) were associated with worse prognosis. The absence of syncope, aborted SD, spontaneous type I ECG, and inducibility during EP study was associated with a significantly better prognosis (P < 0.001). CONCLUSION: Family history of SD is not predictive for future arrhythmic events even if considering only SD in first-degree relatives or SD in first-degree relatives at a young age. The absence of syncope, aborted SD, spontaneous type I ECG, and inducibility during EP study is associated with a good five-year prognosis.

Comparison of the haemostatic properties of conventional monopolar and bipolar transurethral resection of the prostate in patients on oral anticoagulants.

INTRODUCTION: The aim of study is comparing the haemostatic properties of conventional monopolar resection (TURP) and bipolar transurethral resection in saline (TURIS) of the prostate in patients under chronic oral anticoagulants. MATERIAL AND METHODS: Out of a cohort group of 550 endoscopic resections for bladder outlet obstruction, 176 patients on chronic oral anticoagulant therapy required endoscopic resection either by monopolar TURP or bipolar TURIS technology. Changes in haemoglobin, blood transfusion, and clot retention were compared between both groups. RESULTS: Mean postoperative change in haemoglobin level was -1.21 ±0.92 mg/dl in the TURP group compared to -1.29 ±0.99 mg/dl in the TURIS group (p = 0.603). The need for blood transfusions and the mean numbers of units transfused did not significantly differ between the 2 groups. Clot retention appeared in 12 patients (15%) in the TURP group compared to 13 patients (13%) in the TURIS group (p = 0.828). CONCLUSIONS: Despite promising experimental results of better haemostasis and deeper coagulation depth, bipolar technology does not permit one to reduce the amount of blood loss when compared to patients treated by conventional monopolar technology in this study group of patients on oral anticoagulation therapy. Patients on oral anticoagulants suffer more incidents of clot retention, which sometimes results in re-hospitalisation.

Conventional monopolar resection or bipolar resection in saline for the management of large (>60 g) benign prostatic hyperplasia: an evaluation of morbidity.

The objective of this study was to compare the efficacy and safety of bipolar transurethral resection in saline (TURIS) and conventional monopolar transurethral resection (TURP) in men with voiding difficulties due to high volume (>60g) benign prostatic hyperplasia. Between May 2005 and January 2009, 66 men with bladder outlet obstruction due to large benign hyperplasia (BPH) (>60 g) underwent transurethral resection, either by Olympus monopolar or Olympus bipolar TURIS technique. Perioperative and outcome data were recorded and compared. There was no significant difference in patient age, prostate size, PSA, uroflow rate, post voiding residual urine, operation time, catheterization time, and hospital stay. In the monopolar TURP group, an important drop in serum sodium was observed (3.12 mmol/L), statistically significantly different (p = 0.012) from the 1.30 mmol/L drop in the bipolar TURIS group. One case of clinical TUR syndrome was observed in the conventional monopolar TURP group. Early and late complications such as clot retention, urinary retention, bladder neck stenosis, and urethral stricture were identical in both groups. The bipolar TURIS device is a valid treatment option for patients with BPH-related voiding disorders due to high volume BPH. Obviating the risk of TUR syndrome, this minimally invasive procedure may be a good surgical option in the urologist's armamentarium.

Urethral strictures and bipolar transurethral resection in saline of the prostate: fact or fiction?

PURPOSE: To compare the incidence of urethral strictures after monopolar transurethral resection of the prostate (TURP) and bipolar transurethral resection in saline (TURIS) for symptomatic benign prostatic hyperplasia. PATIENTS AND METHODS: Between January 2005 and January 2009, 518 patients with symptomatic benign prostate hyperplasia underwent either monopolar or bipolar resection. In a randomized trial, the incidence of urethral strictures after both treatment modalities was analyzed. RESULTS: Over 48 months, 255 patients were treated with conventional TURP and 263 patients with TURIS. Patient related, operation, and hospitalization characteristics were similar in both groups. After a mean follow-up of 32.1 months (range 50-7 months), the incidence of urethral strictures was 2.4% in the TURP group. After a comparable period of 31.4 months (range 50-7 months), the incidence in the TURIS group was 1.5%. No statistically significant difference was obtained (P = 0.539). These values were compared with the results of other randomized controlled trials with the same or other bipolar technology. No statistically significant difference in urethral structures was noticed between monopolar and bipolar resections (P = 0.739). CONCLUSIONS: With a stricture incidence of 1.5%, bipolar transurethral prostate resection has a low stricture rate, comparable with monopolar TURP (2.4%).

Bipolar transurethral resection in saline: the solution to avoid hyponatraemia and transurethral resection syndrome.

OBJECTIVE: To compare serum sodium changes and the incidence of transurethral resection (TUR) syndrome after monopolar transurethral resection of the prostate (TURP) and bipolar transurethral resection in saline (TURIS) for symptomatic benign prostatic hyperplasia. MATERIAL AND METHODS: Between January 2005 and August 2009, 550 consecutive patients with symptomatic benign prostate hyperplasia underwent TURP, by either a monopolar or bipolar technique. Preoperative and postoperative blood parameters were analysed to compare preoperative and postoperative electrolyte concentrations. RESULTS: Over 56 months, 265 patients underwent a conventional monopolar TURP and 285 patients had a bipolar TURIS (Olympus). Patient profiles regarding age, operation time, resection weight and resection speed were similar in both groups. A drop in sodium of 2.5 mmol/l was measured in the conventional monopolar resection group. The decline of 1.5 mmol/l in the bipolar group, although smaller, was still statistically significant (p = 0.001). With longer operation times (> 1 h) the difference between the groups became more important (p < 0.001). Two clinical TUR syndromes were observed in the monopolar group, while none occurred in the bipolar group. CONCLUSIONS: Bipolar TURP in saline is a safe technique and obviates the risk of TUR syndrome. Repeated serum analysis of electrolytes after TURIS can be omitted.

Bipolar versus monopolar technique for palliative transurethral prostate resection.

INTRODUCTION: The aim was to evaluate the postoperative morbidity and outcome of palliative endoscopic resections for relief of infravesical obstruction in prostate cancer patients with hormone deprivation therapy, and to investigate the added value of bipolar technology over conventional monopolar resections. MATERIAL AND METHODS: A retrospective study was performed on 70 patients with prostate cancer under hormone deprivation therapy undergoing 75 endoscopic procedures, by either monopolar or bipolar technology, between August 2005 and March 2009 at a single institution. The analysis used outpatient, inpatient, and operative records, and observations of electrolyte changes in the serum, postoperative morbidity, and the overall results of palliative endoscopic resections. Preoperative cancer stages and grades were compared with the pathological findings after surgery. Postoperative outcome and complications of conventional monopolar and bipolar technology were compared. RESULTS: Over a period of 44 months, 34 conventional monopolar resections were performed in 32 patients and 41 bipolar resections in 38 patients. Patients' profiles regarding age, initial cancer stage and grade, resection weight, resection speed, catheterization time, and hospital stay were similar in both groups. No statistically significant difference was observed in sodium drop (p = 0.802), clot retention (p = 0.565), or urinary retention (p = 0.292). The overall success rate in relieving obstruction leading to spontaneous voiding was 77%. While 38% of the patients had a high grade tumour at diagnosis, 79% were found to be high grade after the endoscopic resection (p < 0.0001). CONCLUSIONS: Palliative endoscopic transurethral resection is an acceptable and safe adjunctive surgical treatment for voiding disorders in prostate cancer patients. Bipolar technology offers no substantial benefit over conventional monopolar technology.

Association of PPARgamma allelic variation, osteoprotegerin and abdominal aortic aneurysm.

OBJECTIVE: We have previously demonstrated high concentrations of the glycoprotein osteoprotegerin (OPG) in biopsies of abdominal aortic aneurysm (AAA), and demonstrated that ligation of the nuclear receptor peroxisome proliferator-activated receptor gamma (PPARgamma) downregulates OPG in vitro and within a mouse model. The aims of this study were to assess the associations between circulating concentrations of OPG, polymorphisms of the gene encoding PPARgamma (PPARG), AAA presence and growth. DESIGN, PATIENTS AND MEASUREMENTS: Two genetic polymorphisms in PPARG were assessed in 4227 men, 699 of whom had an AAA. For 631 men, who had AAAs, maximum aortic diameter was monitored by yearly ultrasound for a median of 5 years. Plasma OPG was measured in 838 men, 318 of whom had an AAA. RESULTS: Plasma concentrations of OPG were independently associated with AAA (adjusted odds ratio 1.38, 95% CI 1.10-1.72). The PPARG c.1347C > T polymorphism was associated with plasma concentrations of OPG (beta 0.12, P < 0.01). The PPARG c.34G > C polymorphism was weakly associated with AAA (adjusted odds ratio 1.28, 95% CI 1.01-1.61). PPARG c.1347C > T was associated with increased AAA growth (recessive model, P = 0.03). CONCLUSIONS: Circulating concentrations of osteoprotegerin are associated with abdominal aortic aneurysm and with one peroxisome proliferator-activated receptor gamma gene polymorphism. Peroxisome proliferator-activated receptor gamma gene polymorphisms are weakly associated with abdominal aortic aneurysm presence and growth. Confirmation of these findings is required in other cohorts.

Robust cross-validation of linear regression QSAR models.

A quantitative structure-activity relationship (QSAR) model is typically developed to predict the biochemical activity of untested compounds from the compounds' molecular structures. "The gold standard" of model validation is the blindfold prediction when the model's predictive power is assessed from how well the model predicts the activity values of compounds that were not considered in any way during the model development/calibration. However, during the development of a QSAR model, it is necessary to obtain some indication of the model's predictive power. This is often done by some form of cross-validation (CV). In this study, the concepts of the predictive power and fitting ability of a multiple linear regression (MLR) QSAR model were examined in the CV context allowing for the presence of outliers. Commonly used predictive power and fitting ability statistics were assessed via Monte Carlo cross-validation when applied to percent human intestinal absorption, blood-brain partition coefficient, and toxicity values of saxitoxin QSAR data sets, as well as three known benchmark data sets with known outlier contamination. It was found that (1) a robust version of MLR should always be preferred over the ordinary-least-squares MLR, regardless of the degree of outlier contamination and that (2) the model's predictive power should only be assessed via robust statistics. The Matlab and java source code used in this study is freely available from the QSAR-BENCH section of www.dmitrykonovalov.org for academic use. The Web site also contains the java-based QSAR-BENCH program, which could be run online via java's Web Start technology (supporting Windows, Mac OSX, Linux/Unix) to reproduce most of the reported results or apply the reported procedures to other data sets.

Statistical confidence for variable selection in QSAR models via Monte Carlo cross-validation.

A new variable selection wrapper method named the Monte Carlo variable selection (MCVS) method was developed utilizing the framework of the Monte Carlo cross-validation (MCCV) approach. The MCVS method reports the variable selection results in the most conventional and common measure of statistical hypothesis testing, the P-values, thus allowing for a clear and simple statistical interpretation of the results. The MCVS method is equally applicable to the multiple-linear-regression (MLR)-based or non-MLR-based quantitative structure-activity relationship (QSAR) models. The method was applied to blood-brain barrier (BBB) permeation and human intestinal absorption (HIA) QSAR problems using MLR to demonstrate the workings of the new approach. Starting from more than 1600 molecular descriptors, only two (TPSA(NO) and ALOGP) yielded acceptably low P-values for the BBB and HIA problems, respectively. The new method has been implemented in the QSAR-BENCH v2 program, which is freely available (including its Java source code) from www.dmitrykonovalov.org for academic use.