Median Percentage Error Research Articles

Prediction of Propofol Clearance in Children from an Allometric Model Developed in Rats, Children and Adults versus a 0.75 Fixed-Exponent Allometric Model

For propofol clearance, allometric scaling has been applied successfully for extrapolations between species (rats and humans) and within the human bodyweight range (children and adults). In this analysis, the human bodyweight range is explored to determine for which range an allometric model with a fixed or estimated exponent can be used to predict propofol clearance, without correction for maturation. The predictive value of the allometric equation, clearance (CL) is equal to 0.071 x bodyweight in kg0.78, which was developed from rats, children and adults, and the predictive value of a fixed exponent allometric model derived from the basal metabolic rate, CL is equal to CL standardized to a 70 kg adult x (bodyweight in kg standardized to a 70 kg adult)0.75, were evaluated across five independent patient groups including (i) 25 (pre)term neonates with a postmenstrual age of 27-43 weeks; (ii) 22 postoperative infants aged 4-18 months; (iii) 12 toddlers aged 1-3 years; (iv) 14 adolescents aged 10-20 years; and (v) 26 critically ill adults sedated long term. The median percentage error of the predictions was calculated using the equation %error = (CL(allometric) - CL(i))/CL(i) x 100, where CL(allometric) is the predicted propofol clearance from the allometric equations for each individual and CL(i) is the individual-predicted (post hoc) propofol clearance value derived from published population pharmacokinetic models. In neonates, the allometric model developed from rats, children and adults, and the fixed-exponent allometric model, systematically overpredicted individual propofol clearance, with median percentage errors of 288% and 216%, respectively, whereas in infants, both models systematically underpredicted individual propofol clearance, with median percentage errors of -43% and -55%, respectively. In toddlers, adolescents and adults, both models performed reasonably well, with median percentage errors of -12% and -32%, respectively, in toddlers, 16% and -14%, respectively, in adolescents, and 12% and -18%, respectively, in adults. Both allometric models based on bodyweight alone may be of use to predict propofol clearance in individuals older than 2 years. Approaches that also incorporate maturation are required to predict clearance under the age of 2 years.

Estimation of End Expiratory Volume using Stress Strain

Acute Respiratory Distress Syndrome (ARDS) results in collapse of alveolar units and loss of lung volume at end of expiration. Mechanical ventilation (MV) is used on patients with ARDS or Acute Lung Injury (ALI), with the end objective being to increase the end expiratory volume (EEV) and hence increasing overall functional residual capacity (FRC). Simple methods to estimate EEV at a given positive end expiratory pressure (PEEP) level in patients with ARDS/ ALI currently do not exist. Current viable methods are time consuming, relatively invasive and not model based. Previous studies have found a constant linear relationship between the global stress and strain in the lung independent of lung condition. This study utilizes the constant stress strain ratio and an individual patient's volume responsiveness to PEEP to estimate EEV. The estimation model identifies two global parameters that can be used to estimate EEV in a patient, β and mβ. The parameter β captures physiological parameters of FRC, lung and respiratory elastance, and varies depending on the PEEP level used, and mβ is the gradient of β vs PEEP.EEV was estimated at different PEEP values and compared to the measured EEV for 12 different patients with different levels of lung injury. The median percentage error is 18% (IQR: 6,49) for PEEP = 5 cm H2O, 10% (IQR: 9,18) for PEEP = 7 cm H2O, 28% (IQR: 12,33) for PEEP = 10 cm H2O, 3% (IQR: 2,10) for PEEP = 12 cm H2O and 10% (IQR: 9,11) for PEEP = 15 cm H2O. The results were further validated using a cross correlation test and a linear regression was performed between the estimated and measured EEV with a median R2 of 0.948 (IQR: 0.915,0.968 ; 90% CI: 0.814,0.984). This simple approach can be done at the bedside with minimal intervention to provide an estimate with reasonable accuracy to track patient condition.

Different equations to combine creatinine and cystatin C to predict GFR. Arithmetic mean of existing equations performs as well as complex combinations

Purpose: To test various ways of combining creatinine and cystatin C in equations to predict glomerular filtration rate (GFR). Material and methods: Performance of the following expressions to predict GFR was compared with measured GFR (iohexol clearance, mL/min/1.73 m2) in 857 patients: (i) Lund-Malmö creatinine equation, (ii) Grubb cystatin C equation, (iii) arithmetic mean of (1) and (2), (iv) geometric mean of (1) and (2), (v) linear regression on (1) and (2), (vi) regression on (1) and cystatin C, and (vii) regression on creatinine, cystatin C, age and gender. Results: For the entire cohort median percent error (bias) was <5% for all expressions, though all expressions tended to underestimate (−8.3 to 15.8%) GFR at levels ≥90 mL/min/1.73 m2. The five expressions combining creatinine and cystatin C significantly improved correlation and accuracy (p<0.001) within 15 and 30% of measured GFR compared with the equations based on the separate analytes and with no significant difference between the five expressions. In a subgroup of patients with neurological disease and muscle atrophy the cystatin C equation performed better than the expressions combining creatinine and cystatin C. Conclusion: Simply calculating the arithmetic mean of predicted GFR based on separate creatinine and cystatin C equations performs equally well as more complex equations. Reporting GFR based on separate creatinine and cystatin C equations, and their arithmetic mean also has the definite advantage that the physician can choose the estimated GFR, most appropriate depending on the clinical setting and patient characteristics.

Melanoma histological Breslow thickness predicted by 75-MHz ultrasonography

Ultrasound at 20 MHz tends to overestimate melanoma Breslow thickness due to lymphocytic infiltration or naevus remnant. Objectives To determine the efficacy of 75-MHz ultrasound for estimating melanoma thickness and to assess its reliability to predict surgical requirement. One hundred and twelve suspicious skin lesions were imaged prospectively with the 75-MHz ultrasound in A and B mode. This instrument has a penetration of 3 mm and a lateral resolution of 21 mum. Measurements were performed by two observers and the mean was compared with the histological Breslow thickness. Forty-five of 52 melanomas and 22 of 36 naevi had clear hypoechogenicity boundaries. The median histological Breslow thickness of melanomas was 0.4 mm and 22 were in situ. The median percentage error of the machine was 13% of the histological Breslow thickness, with a high correlation (Pearson's r = 0.908, P < 0.001). Measurement was highly reliable for invasive melanoma, even in the presence of lymphocytic infiltration or naevus. In this series, 75-MHz measurement was highly reliable for predicting invasive melanoma thickness.

Efficient mart-aided modeling for microarchitecture design space exploration and performance prediction

Computer architects usually evaluate new designs by cycle-accurate processor simulation. This approach provides detailed insight into processor performance, power consumption and complexity. However, only configurations in a subspace can be simulated in practice due to long simulation time and limited resource, leading to suboptimal conclusions which might not be applied in a larger design space. In this paper, we propose an automated performance prediction approach which employs state-of-the-art techniques from experiment design, machine learning and data mining. Our method not only produces highly accurate estimations for unsampled points in the design space, but also provides interpretation tools that help investigators to understand performance bottlenecks. According to our experiments, by sampling only 0.02% of the full design space with about 15 millions points, the median percentage errors, based on 5000 independent test points, range from 0.32% to 3.12% in 12 benchmarks. Even for the worst-case performance, the percentage errors are within 7% for 10 out of 12 benchmarks. In addition, the proposed model can also help architects to find important design parameters and performance bottlenecks.

The Lund–Malmö creatinine‐based glomerular filtration rate prediction equation for adults also performs well in children

Objective. To evaluate the clinical performance in a paediatric population of the Lund–Malmö creatinine‐based glomerular filtration rate (GFR) prediction equations, primarily developed for adults. Material and methods. Iohexol clearance was used as the gold standard in 85 paediatric Caucasian patients (0.3–17 years; 37 F/48 M). One Lund–Malmö equation was based on age and gender (LM) and one included lean body mass (LM‐LBM). Comparisons focused on correlation (adjusted R2), bias (median percent error) and accuracy (proportions of predicted GFR differing ⩽30 % from measured GFR) (mL/min/1.73 m2). The performances were compared with those of the Modification of Diet in Renal Disease (MDRD) Study equation, a dedicated paediatric creatinine equation, Counahan–Barratt (CB) and a cystatin C‐based equation. Results. The MDRD equation performed poorly with a median bias of 96 %. Of the remaining equations, only the LM‐LBM produced significant bias (+10 % in median) according to line of identity regression analysis. The LM equation yielded marginally higher accuracy (76 %) than the LM‐LBM equation (74 %) and the CB (73 %), but lower than the cystatin C‐based equation (82 %). However, the estimated accuracy figures for these four equations were generally imprecise and none of the differences compared with the LM equation was statistically significant. Conclusion. In contrast to most creatinine‐based GFR prediction equations, the LM equation performs adequately for both children and adults. This may be due to the unique model‐building principles used when the LM equation was established. Further validation in a larger paediatric population is necessary.

Is the adjustment of serum creatinine level &lt; 0.6 mg/dl to 0.6 mg/dl justified in estimates of carboplatin clearance calculated by the Jelliffe formula?

5072 Background: The Jelliffe formula (JF) does not include body surface area (BSA) or body weight to adjust the body size. We demonstrated that estimates of carboplatin clearance calculated by the JF tend to have greater positive bias compared to other formulae. The JF has been used to estimate carboplatin clearance in Gynecologin Oncology Group studies. In patients with serum creatinine level &lt;0.6 mg/dl, it is adjusted to 0.6 mg/dl in estimates of carboplatin clearance (Adjusted-Jelliffe formula (AJF)). The purpose of this study is to evaluate whether this adjustment is suitable. Methods: Carboplatin clearance was estimated in 115 patients with serum creatinine &lt;0.6 mg/dl who received carboplatin-based chemotherapy for gynecologic malignancies between January 1996 and August 2004. Creatinine clearance was estimated using the Cockroft-Gault formula (CGF), JF, and AJF. The median percent error (MPE) and the median absolute percent error (MAPE) were evaluated by comparing carboplatin clearance. The relationships between BSA and ratios of estimated carboplatin clearance (JF/CGF, AJF/CGF) were evaluated by using simple regression. Results: The estimated carboplatin clearances were: CGF, 126.7 ± 27.7; JF, 148.2 ± 20.5; AJF, 130.5 ± 14.3. Comparing the results of the CGF with the JF, AJF yielded MPEs of +20%, +6%, and MAPEs of 21%, 14%, respectively. There were the linear correlations between ratio of estimated carboplatin clearances and BSA (Y1 = −1.141X + 2.830, Y2 = -1.061 X + 2.581, Y1: ratio of carboplatin clearance (JF/CGF), Y2: ratio of carboplatin clearance (AJF/CGF), X: BSA). Conclusions: As expected, carboplatin clearance was decreased by adjusting serum creatinine to 0.6 mg/ml, but it did not adjust the bias by BSA. Estimates of carboplatin clearance calculated by the AJF tend to have greater negative bias, particularly when the BSA of the patient is large. No significant financial relationships to disclose.

Difference of carboplatin clearance estimated by the Cockroft–Gault, Jelliffe, Modified-Jelliffe, Wright or Chatelut formula

Objective. Although the Calvert formula is the standard method to calculate the dose of carboplatin, there is no consensus how to determine the glomerular filtration rate (GFR) without using [ 51Cr]-ethylenediamine tetraacetic acid ( 51Cr-EDTA). Creatinine clearance (Ccr), calculated using the Cockroft–Gault, Jelliffe, Modified-Jelliffe or Wright formulae, has been used as a substitute for the GFR. In addition to these four formulae, the Chatelut formula has been proposed as a way to calculate carboplatin clearance. Among these formulae, Jelliffe formula does not include body surface area (BSA) or body weight to adjust the body size and thus may have greater bias than the other four formulae in estimating carboplatin clearance. The purpose of this study is to evaluate if these five formulae could equally estimate the carboplatin clearance. Methods. Carboplatin clearance was estimated in 253 patients with gynecologic cancer who received carboplatin-based chemotherapy between January 1996 and August 2004. Ccr was estimated using the Cockroft–Gault, Jelliffe, Modified-Jelliffe or Wright formulae. Carboplatin clearance was also calculated directly by using the Chatelut formula. The five estimations of carboplatin clearance were compared with each other using the post-hoc Wilcoxon signed rank test. The median percent error (MPE) and the median absolute percent error (MAPE) were evaluated by comparing carboplatin clearance. The relationships between BSA and ratios of estimated carboplatin clearance (Jelliffe/Cockroft–Gault, Jelliffe/Modified-Jelliffe, Jelliffe/Wright, Jelliffe/Chatelut) were evaluated by using simple regression. Results. The estimated carboplatin clearances were: Cockroft–Gault formula, 109.8 ± 28.4; Jelliffe formula, 128.5 ± 28.2; Modified-Jelliffe formula, 110.8 ± 25.7; Wright formula, 112.2 ± 24.3; Chatelut formula, 114.1 ± 33.0. A statistically significant difference was observed between carboplatin clearance calculated using Jelliffe formula and that given by each of the other four formulae. Comparing the results of the Cockroft–Gault formula with the Jellife, Modified-Jelliffe, Wright and Chatelut formulae yielded MPEs of +19%, +2%, +3% and +3% and MAPEs of 27%, 5%, 6% and 6%, respectively. There was a significant correlation between BSA and ratio of estimated carboplatin clearance (Jelliffe/Cockroft–Gault, Jelliffe/Modified-Jelliffe, Jelliffe/Wright, Jelliffe/Chatelut), with Pearson correlation coefficients of 0.928, 0.847, 0.965 and 0.839 respectively. As the BSA of patient became smaller, the differences between the carboplatin clearance calculated by the Jelliffe formula from other four formulas became larger. Conclusions. Estimates of carboplatin clearance calculated by the Jelliffe formula tend to have greater positive bias compared to the other four formulae, particularly when the BSA of the patient is small. In order to conduct collaborative international studies, it may be necessary to standardize the formula used to estimate carboplatin clearance to perform international collaboration studies.

A cystatin C‐based formula without anthropometric variables estimates glomerular filtration rate better than creatinine clearance using the Cockcroft‐Gault formula

In 1976, Cockcroft and Gault introduced a widely used formula comprising several anthropometric variables to compensate for the inadequacies of creatinine level as a marker of glomerular filtration rate (GFR). The present work investigates the possibility of introducing cystatin C‐based formulas without anthropometric variables to predict GFR, determined by an invasive “gold standard” procedure (iohexol clearance), and to compare the diagnostic efficiency of such formulas with that of Cockcroft and Gault. All 451 adult patients referred to the University Hospital for determination of GFR by iohexol clearance measurements during a period of 6 months were included in the study. Calculations of bias (median percent error), correlation (adjusted R2), and accuracy (percentage of estimates within 30 and 50% of iohexol clearance) were used in the comparison. The cystatin C‐based formula GFR (ml/min) = 89.12×cystatin C−1.675 had lower bias and higher accuracy in predicting GFR than the Cockcroft–Gault formula. If a cystatin C‐based formula including gender was constructed: GFR (ml/min) = 99.19×cystatin C−1.713×(0.823 for women), an even lower bias and higher accuracy were obtained. It is suggested that measurement of cystatin C should be used for the initial prediction of GFR of a patient.

Estimating blood loss: can teaching significantly improve visual estimation?

Visually estimated blood loss has long been known to be imprecise, inaccurate, and often underestimated, which may lead to delayed diagnosis and treatment. Our purpose is to determine whether a brief didactic course can improve visually estimated blood loss and whether prior clinical experience influences estimation of blood loss. Reconstituted whole blood was obtained from the blood bank, and simulated scenarios with known measured blood loss were created using common surgical materials. Visually estimated blood loss was performed by medical personnel before and after a 20-minute didactic session. Percent errors of estimated blood loss were calculated and comparisons were made before and after the lecture. The effects of actual blood volume and clinical experience on estimation of blood loss were assessed. A total of 53 participants assessed 7 scenarios. There were significant reductions in error for all scenarios. Median percent error in estimated blood loss was not influenced by clinical experience, either before or after the didactic session. Blood loss tends to be overestimated at low volumes and underestimated at high volumes. Error in estimating blood loss is dependent on actual blood loss volume. Medical students and experienced faculty demonstrate similar errors, and both can be improved significantly with limited instruction. This educational process may assist clinicians in everyday practice to more accurately estimate blood loss and recognize patients at risk for hemorrhage-related complications.

Perceived Incremental Submaximal Force Production in F-16 Pilots

0845 PURPOSE: The study examined the force production patterns using perceived stimulus cues from 10% to 90% of maximal force (MF) in a unique sample of F-16 pilots. METHODS: Ten males (age: 33–54 yr) who were not experienced with the exercise protocol performed hand grip contractions on a hand dynamometer. Subjects produced actual forces (AF) perceived to be 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90% MF followed by a MF. A 1-minute rest interval was maintained between each increment. Subjects rested 5 minutes and repeated a second trial of the protocol. Desired forces (DF) were calculated as the required percentage increment of the produced MF. RESULTS: Test-retest results indicated good consistency between the trials for AF and MF (r >.92). The correlations between AF and DF were high (r >.88). On trial 1 AF were below DF for all increments with a median percent error (AF-DF) range of −4% to −33%. On trial 2 median percent errors tended to be smaller. A power function analysis relating the change in AF with DF stimuli produced an exponent (E) of 1.08 (.95CI = .93–.1.2) for trial 1 and .94 (.95CI = .84–1.00) for trial 2. CONCLUSIONS: This study was performed on a sample from a very unique population where force control is essential to the task of flying an F-16 aircraft. The subjects tended to undershoot DF production through perceptual force ranges of 10% to 90% MF. AF grew at linear rate with increased perceptual stimulus cues. Past reports have indicated variable findings with Es >1 (AF growing faster than DF), = 1 (AF linear with DF), or <1 (AF growing slower than DF). The lack of consistent findings may be due to type of muscular contraction, muscle mass involved, and the method of presentation of stimulus cues.

Perceived Incremental Submaximal Force Production in F-16 Pilots

0845 PURPOSE: The study examined the force production patterns using perceived stimulus cues from 10% to 90% of maximal force (MF) in a unique sample of F-16 pilots. METHODS: Ten males (age: 33–54 yr) who were not experienced with the exercise protocol performed hand grip contractions on a hand dynamometer. Subjects produced actual forces (AF) perceived to be 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90% MF followed by a MF. A 1-minute rest interval was maintained between each increment. Subjects rested 5 minutes and repeated a second trial of the protocol. Desired forces (DF) were calculated as the required percentage increment of the produced MF. RESULTS: Test-retest results indicated good consistency between the trials for AF and MF (r >.92). The correlations between AF and DF were high (r >.88). On trial 1 AF were below DF for all increments with a median percent error (AF-DF) range of −4% to −33%. On trial 2 median percent errors tended to be smaller. A power function analysis relating the change in AF with DF stimuli produced an exponent (E) of 1.08 (.95CI = .93–.1.2) for trial 1 and .94 (.95CI = .84–1.00) for trial 2. CONCLUSIONS: This study was performed on a sample from a very unique population where force control is essential to the task of flying an F-16 aircraft. The subjects tended to undershoot DF production through perceptual force ranges of 10% to 90% MF. AF grew at linear rate with increased perceptual stimulus cues. Past reports have indicated variable findings with Es >1 (AF growing faster than DF), = 1 (AF linear with DF), or <1 (AF growing slower than DF). The lack of consistent findings may be due to type of muscular contraction, muscle mass involved, and the method of presentation of stimulus cues.

VISUAL ESTIMATION OF FINGER ANGLES: DO WE NEED GONIOMETERS?

Seventy-one plastic surgeons and therapists, of varying levels of seniority and experience, were asked to examine a resin cast of an adult male hand and use estimation to measure the angles of the metacarpophalangeal and interphalangeal joints of each digit. Visual estimation by all subjects was inaccurate by a mean of approximately 25% (median percentage error 22, range 1–100). Consultants were the most accurate, whilst physiotherapists were the least. Regular goniometer users were no more accurate. However, hand surgery experience correlated with accuracy, as did a stated interest in hand surgery. Although visual accuracy improves with experience, it is still an inaccurate technique. We, therefore, recommend that goniometers should be used for measuring angles in hand surgery patients.

Prediction of Carboplatin Clearance From Standard Morphological and Biological Patient Characteristics

Hematologic toxicity of an antineoplastic drug, carboplatin, is largely dependent on its pharmacokinetics. Its therapeutic efficacy may be related to plasma drug exposure. Dosage adjustment based on isotopic determination of glomerular filtration rate has been proposed, but its ambulatory use is not yet conceivable. The dosage adjustment based on a patient's creatinine clearance relies on accurate measurement of urine volume per unit time and can be done with ease. A population pharmacokinetics study was undertaken to determine a relationship between carboplatin clearance and patient characteristics. A predictive formula was derived that was then prospectively evaluated, and its outcome was compared with that obtained by other methods available to predict carboplatin clearance. Plasma carboplatin pharmacokinetics determined as ultrafilterable platinum in 70 patients (age range, 23-84 years) treated with different combination regimens that included carboplatin at doses ranging from 184 mg to 950 mg (1-hour intravenous infusion) for various tumor types. Data were analyzed using the nonlinear mixed effects model (NONMEM). The data from 34 patients (46 cycles) were utilized to derive the most predictive formula. The reliability of the formula was subsequently evaluated by analyzing the data obtained from 36 other patients (49 cycles). Carboplatin clearance (mL/min) was found to be best predicted by the following formula: 0.134.weight + [218.weight.(1-0.00457.age).(1-0.314.sex)]creatinine expressed in micromolar concentration (with weight in kg, age in years, and sex = 0 if male and sex = 1 if female). Prospectively, this formula predicted the carboplatin clearance with good precision (median absolute percent error of 10% [range, 0% to 30%]) and minimal bias (median percent error of 2% [range, -25% to 30%]). This method of prediction was as accurate as the one derived from the measurement of glomerular filtration rate following the injection of 51 chromium-EDTA. This formula for the determination of carboplatin clearance can permit individualized determination of carboplatin dosage in adults by simply multiplying the calculated carboplatin clearance by the area under the curve for the desired dosage administration.