José Labarère

How to derive and validate clinical prediction models for use in intensive care medicine

BackgroundClinical prediction models are formal combinations of historical, physical examination and laboratory or radiographic test data elements designed to accurately estimate the probability that a specific illness is present (diagnostic model), will respond to a form of treatment (therapeutic model) or will have a well-defined outcome (prognostic model) in an individual patient. They are derived and validated using empirical data and used to assist physicians in their clinical decision-making that requires a quantitative assessment of diagnostic, therapeutic or prognostic probabilities at the bedside.PurposeTo provide intensivists with a comprehensive overview of the empirical development and testing phases that a clinical prediction model must satisfy before its implementation into clinical practice.ResultsThe development of a clinical prediction model encompasses three consecutive phases, namely derivation, (external) validation and impact analysis. The derivation phase consists of building a multivariable model, estimating its apparent predictive performance in terms of both calibration and discrimination, and assessing the potential for statistical over-fitting using internal validation techniques (i.e. split-sampling, cross-validation or bootstrapping). External validation consists of testing the predictive performance of a model by assessing its calibration and discrimination in different but plausibly related patients. Impact analysis involves comparative research [i.e. (cluster) randomized trials] to determine whether clinical use of a prediction model affects physician practices, patient outcomes or the cost of healthcare delivery.ConclusionsThis narrative review introduces a checklist of 19 items designed to help intensivists develop and transparently report valid clinical prediction models.

Diagnostic accuracy of combined cardiac troponin and copeptin assessment for early rule-out of myocardial infarction: : a systematic review and meta-analysis

Aims: This systematic review aimed to investigate the diagnostic accuracy of combined cardiac troponin (cTn) and copeptin assessment in comparison to cTn alone for early rule-out of acute myocardial infarction (AMI). Methods: Primary studies were eligible if they evaluated diagnostic accuracy for cTn with and without copeptin in patients with symptoms suggestive of AMI. AMI was defined according to the universal definition, using detection of cTn as a marker for myocardial necrosis. Eligible studies were identified by searching electronic databases (Medline, EMBASE, Science Citation Index Expanded, CINAHL, Pascal, and Cochrane) from inception to March 2013, reviewing conference proceedings and contacting field experts and the copeptin manufacturer. Results: In 15 studies totalling 8740 patients (prevalence of AMI 16%), adding copeptin improved the sensitivity of cTn assays (from 0.87 to 0.96, p=0.003) at the expense of lower specificity (from 0.84 to 0.56, p<0.001). In 12 studies providing data for 6988 patients without ST-segment elevation, the summary sensitivity and specificity estimates were 0.95 (95% CI 0.89 to 0.98) and 0.57 (95% CI 0.49 to 0.65) for the combined assessment of cTn and copeptin. When a high-sensitivity cTnT assay was used in combination with copeptin, the summary sensitivity and specificity estimates were 0.98 (95% CI 0.96 to 1.00) and 0.50 (95% CI 0.42 to 0.58). Conclusion: Despite substantial between-study heterogeneity, this meta-analysis demonstrates that copeptin significantly improves baseline cTn sensitivity. Management studies are needed to establish the effectiveness and safety of measuring copeptin in combination with high-sensitivity cTnT for early rule-out of AMI without serial testing.

Factors and outcomes associated with physicians' adherence to recommendations of infectious disease consultations for inpatients

OBJECTIVES Solicited consultations constitute a substantial workload for infectious disease specialists (IDSs). The impact of physician adherence to recommendations on clinical outcomes following solicited IDS consultations has not been previously studied. The objectives of the study were to identify the factors associated with adherence and to determine whether adherence to recommendations was associated with better clinical outcomes. METHODS A prospective study was conducted on 621 patients, aged > or = 18 years, hospitalized in a university-affiliated hospital in France, who received an IDS consultation between December 2007 and June 2008. The main outcome was early clinical improvement, and the secondary outcomes were length of stay and in-hospital mortality. RESULTS Adherence to the IDSs recommendations was 88.2% (548/621) for antimicrobial treatment and 72.2% (317/439) for diagnostic or monitoring tests. In a multivariable analysis, independent factors of adherence to therapeutic recommendations were a community-acquired infection [adjusted odds ratio (OR), 1.8; 95% confidence interval (CI), 1.1-3.0] and discontinuation or non-use of antibiotic treatment (adjusted OR, 9.7; 95% CI, 1.2-80.3). Adherence to recommendations for antibiotic treatment was associated with a higher rate of early clinical improvement (60.7% versus 43.9%, P = 0.01), shorter median length of stay (20 days versus 23 days, P = 0.03) and comparable in-hospital mortality (7.7% versus 5.5%, P = 0.50). CONCLUSIONS Factors associated with non-adherence must be anticipated by IDSs during consultations, because non-adherence leads to worse clinical outcomes. Further studies are needed to identify the interventions that could improve physician adherence to recommendations made during solicited consultations.

Proadrenomedullin Improves Risk of Early Admission to ICU Score for Predicting Early Severe Community-Acquired Pneumonia

BACKGROUND Whether proadrenomedullin (ProADM) improves the performance of the Risk of Early Admission to ICU (REA-ICU) score in predicting early, severe community-acquired pneumonia (ESCAP) has not been demonstrated. METHODS Secondary analysis was completed of the original data from 877 consecutive patients with community-acquired pneumonia (CAP) enrolled in the Procalcitonin-Guided Antibiotic Therapy and Hospitalization in Patients With Lower Respiratory Tract Infections (ProHOSP) study, a multicenter trial in EDs of six tertiary-care hospitals in Switzerland. ESCAP was defined by either the requirement for mechanical ventilation or vasopressive drugs or occurrence of death within 3 days of ED presentation. RESULTS Eighty patients (9.1%) developed ESCAP (47 required mechanical ventilation, 19 vasopressive drugs, and 16 died) within 3 days of ED presentation. They had a higher median ProADM value (2.18 nmol/L vs 1.15 nmol/L, P < .001). Combining ProADM testing with the REA-ICU score improved the area under the curve (0.81) compared with either parameter (ProADM [0.73] or REA-ICU score [0.76], P < .001) and resulted in a net reclassification improvement of 0.20 (P < .001). A ProADM value ≥ 1.8 nmol/L or assignment to REA-ICU risk classes III-IV predicted ESCAP with a sensitivity of 76.3% and a negative predictive value of 96.7%. Excluding 21 patients with major criteria of severe CAP on presentation showed similar results. CONCLUSION These study findings demonstrate that the addition of ProADM to the REA-ICU score improves the classification of a substantial proportion of patients in the ED at intermediate or high risk for ESCAP, which may translate into better triage decisions.

Risk factors for deep vein thrombosis in older patients: a multicenter study with systematic compression ultrasonography in postacute care facilities in France.

OBJECTIVES: To identify risk factors for deep vein thrombosis (DVT) in older patients with restricted mobility or functional disability.

Efficacy and safety of glycoprotein IIb/IIIa receptor antagonists for patients undergoing percutaneous coronary intervention within twelve hours of fibrinolysis†

Objective: To compare clinical outcomes between glycoprotein IIb/IIIa receptor antagonist recipients and nonrecipients who underwent percutaneous coronary intervention (PCI) within 12 hr of fibrinolysis. Background: Despite limited evidence, glycoprotein IIb/IIIa receptor antagonists are widely used in ST‐elevation myocardial infarction (STEMI) patients undergoing routine early or rescue PCI after fibrinolysis. Methods: We evaluated 87 and 556 glycoprotein IIb/IIIa receptor antagonist recipients and nonrecipients enrolled in a regional registry of STEMI between October 2002 and December 2005. The primary efficacy endpoint was a composite of death from any cause, reinfarction, and stroke at 1 year of follow‐up. The primary safety endpoint was the rate of in‐hospital major bleeding that was not related to coronary artery bypass grafting. Results: The primary efficacy endpoint occurred in 12% (10 of 81) and 13% (72 of 525) of glycoprotein IIb/IIIa receptor antagonist recipients and nonrecipients, respectively (P = 0.74). The corresponding rates of major bleeding during index hospitalization were 4.8% (4 of 84) and 5.1% (28 of 544) (P = 0.88), respectively. Two glycoprotein IIb/IIIa receptor antagonist recipients and five nonrecipients experienced intracranial hemorrhage. After adjusting for propensity score, the odds of primary efficacy (odds ratio, 0.79; 95% confidence interval, 0.34–1.83) and safety (odds ratio, 0.75; 95% confidence interval, 0.22–2.62) endpoints did not differ according to the use of glycoprotein IIb/IIIa receptor antagonists. Conclusion: In this observational cohort study of unselected patients with STEMI, the administration of glycoprotein IIb/IIIa receptor antagonists provided no additional benefit to PCI performed within 12 hr of fibrinolysis, nor did it compromise patient safety.

Overdiagnosis associated with breast cancer screening: A simulation study to compare lead-time adjustment methods.

OBJECTIVE Estimating overdiagnosis associated with breast cancer screening may use annual incidence rates of cancer. We simulated populations invited to screening programmes to assess two lead-time adjustment methods. METHODS Overdiagnosis estimates were computed using the compensatory drop method, which considered the decrease in incidence of cancers among older age groups no longer offered screening, and the method based on the decrease in incidence of late-stage cancers. RESULTS The true value of overdiagnosis was 0% in all the data sets simulated. The compensatory drop method yielded an overdiagnosis estimate of -0.1% (95% credibility interval -0.5% to 0.5%) when participation rates among the population and risk of cancers were constant. However, if participation rates increased with calendar year as well as risk of cancer with birth cohorts, the overdiagnosis estimated was 11.0% (10.5-11.6%). Using the method based on the incidence of early- and late-stage cancers, overdiagnosis estimates were 8.9% (8.5-9.3%) and 17.6% (17.4-17.9%) when participation rates and risks of cancer were constant or increased with time, respectively. CONCLUSION Adjustment for lead time based on the compensatory drop method is accurate only when participation rates and risks of cancer remain constant, whereas the adjustment method based on the incidence of early- and late-stage cancers results in overestimating overdiagnosis regardless of stability of participation rates and breast cancer risk.

Overdiagnosis and overtreatment associated with breast cancer mammography screening: A simulation study with calibration to population-based data

OBJECTIVES The magnitude of overdiagnosis of breast cancer associated with mammography screening remains controversial because of methodological issues. The objective of this study was to quantify overdiagnosis and overtreatment associated with a population-based screening programme, taking into account lead time and uncertainty concerning baseline incidence of breast cancers. MATERIAL AND METHODS A simulation model was developed to replicate incidence and detection rates of breast cancer observed in the Isère Département, France. The parameters of the model were estimated using an approximate Bayesian computation method. RESULTS For women aged 50-74 years during the 2007-2010 period, overdiagnosis of non-progressive breast cancers accounted for 17.0% (95% credibility interval (CI): 2.5%-35.5%) of all in situ cancers diagnosed, 5.5% (95% CI: 0.8%-9.8%) of all invasive cancers diagnosed, and 20.3% (95% CI: 3.0%-38.9%) of in situ and 13.0% (95% CI: 2.2%-23.3%) of invasive screen detected breast cancers. The estimates of overdiagnosis due to competitive causes of death were 1.0% (95% CI: 0.2%-%1.7) and 1.1% (95% CI: 0.6%-1.7%) for all in situ and invasive cancers diagnosed, respectively, and 1.3% (95% CI: 0.2%-2.0%) and 2.6% (95% CI: 1.4%-4.0%) of all in situ and invasive screen detected breast cancers, respectively. Among 1000 screen-detected cancers in 2010, 155 (95% CI: 27-284), 134 (95% CI: 10-242) and 140 (95% CI: 25-254) women underwent breast conserving surgery, lymph node dissection and radiation therapy for overdiagnosed cancers, respectively. CONCLUSION Our estimates of overdiagnosis should be balanced against the reduction of breast cancer mortality to assess the value of breast cancer screening programme.

Pulmonary Capillary Wedge Pressure Measurement: A Challenge for Diagnosis of Pulmonary Arterial Hypertension

Background: International registries report an aging population suffering from PAH (Pulmonary Artery Hypertension) engendering diagnosis difficulties linked with growing cases of group 2 related to left heart failure with preserved Ejection Fraction (PH-HFpE). Pulmonary Capillary Wedge Pressure (PCWP) measurement by right heart catheterization remains a technical challenge for patient classification; many centers still use digital PCWP given by cath lab software. Here, we have tried to demonstrate misclassification impact of this approximation. Methods: We investigated the PCWP-Left Ventricular End Diastolic Pressure (LVEDP) relationship in a prospective series of 31 patients undergoing heart catherization for suspicion of PAH. Digital and end expiration PCWP were measured in right and left pulmonary arteries for comparison with end expiration LVEDP. Results: We explored 31 patients, 65.4 ± 11 years old, 67.7% were female, with LVEF 60.6 ± 5.2%, Diagnosis of HFpEF was found in 62% of cases and PAH in 10% when using end expiration LVEDP. Right end expiration PCWP, left end expiration PCWP, right digital PCWP, left digital PCWP and end expiration LVEDP were respectively 16.2 ± 6.7mmHg, 16,6 ± 6.4mmHg, 12.2 ± 6.1mmHg (p<0.001), 12.7 ± 6.1mmHg, and 15.8 +/- 4.8mmHg, with a significant difference (p<0.001) between the right and left digital PCWP and end expiration LVEDP. Conclusion: Using digital PCWP instead of end expiration PCWP measurement during RHC, results in a significant underestimation of the LVEDP, this translated to 22% of patients with Pulmonary Hypertension (PH) being misclassified as having group1 rather than group2 PH. Misclassified patients are at risk of receiving inadequate therapy and biased therapeutic studies. When in doubt left heart catheterization should be performed for PAH diagnosis.

Estimation du surdiagnostic lié au dépistage du cancer du sein par mammographie : étude en population générale

Le depistage du cancer du sein par mammographie entraine une diminution de la mortalite par cancer du sein mais peut presenter certains effets indesirables comme l’inconfort de la mammographie, la presence de resultats faux positifs, l’exposition aux radiations ionisantes, et le surdiagnostic [1]. Ce dernier correspond aux cancers depistes qui n’auraient jamais ete diagnostiques en l’absence de depistage, en raison de leur faible potentiel evolutif et de l’absence d’apparition de signes cliniques. Les estimations du surdiagnostic sont contradictoires, de moins de 5 % [2] a pres de 36 % [3] des cas de cancers diagnostiques annuellement. Ces resultats peuvent cependant etre biaises si les methodes d’estimation ne prennent pas en compte l’avance au diagnostic, l’evolution de l’incidence sous-jacente du cancer du sein, et l’evolution de la participation au depistage [4].