Application of fuzzy logic to predicting erythropoietic response in hemodialysis patients.

OBJECTIVE: The purpose of this study was to demonstrate how fuzzy sets can be used in a pharmacodynamic model to represent the uncertainty about the classification of an end-stage renal disease patient's response to erythropoietin. METHODS: A pharmacodynamic model was developed to predict future hemoglobin response to administered erythropoietin for a population of 186 patients with end-stage renal failure and anemia. The prediction was performed by a weighted linear combination of past hemoglobin, transferrin saturation, and erythropoietin dose. Patients were classified based on their response to administered erythropoietin into (i) all patients into 1 group (population approach), (ii) all patients into either a poor or normal responder group (subpopulation approach--traditional classification), and (iii) all patients by partial membership into the poor and normal responder groups (subpopulation approach--fuzzy classification). One half of the data set was randomly selected to estimate the model parameters, and the second half was used to test the estimated model. This randomization was repeated 100 times for both males and females. RESULTS: Mean square error decreased significantly through the incorporation of hemoglobin response categorization from the control group (1.32 +/- 0.07), to crisp coding (1.23 +/- 0.07), to fuzzy coding (1.20 +/- 0.07) with an overall p value < 0.001. CONCLUSION: Uncertainty in the categorization of subjects into 2 erythropoietin response groups of poor or normal response has been shown to benefit from the use of fuzzy categories, with a significant improvement in model performance.

Adequacy of peritoneal dialysis: a quality improvement project of The Renal Network, Inc (Illinois, Indiana, Kentucky, and Ohio).

The Renal Network, Inc, (end-stage renal disease [ESRD] Networks 9 and 10) from the states of Illinois, Indiana, Kentucky, and Ohio includes approximately 4,250 peritoneal dialysis patients. To better assess the state of peritoneal dialysis in Networks 9 and 10, we undertook a peritoneal dialysis prescription adequacy quality improvement project. Adequacy and facility practice information were obtained from a 100% sample of peritoneal dialysis patients in the Networks. The mean total Kt/V urea for those facilities with a written policy for adequacy measurement was 2.18 +/- 0.66 versus a mean of 2.13 +/- 0.70 for those facilities without a written policy (P = not significant [NS]). The mean total Kt/V urea for those facilities with a written procedure was 2.17 +/- 0.66 versus a mean of 2.12 +/- 0.70 for those facilities without a written procedure (P = NS). There was a significant positive relationship between the observed total Kt/V urea and the frequency of measurement (P <.001). No relationship existed between the value of the D/P creatinine ratio and the weekly dialysis Kt/V. This study supports the recommendation of frequent measurement of peritoneal dialysis adequacy.

Improved dialyzer reuse after use of a population pharmacodynamic model to determine heparin doses.

Anticoagulation with heparin is performed to prevent clotting during dialysis. However, heparin doses are usually determined empirically, and dialyzer clotting is a common reason for discarding reused dialyzers. We hypothesized that using a population pharmacodynamic model to determine individual heparin doses would improve dialyzer reuse rates. A previously published model was used to determine the loading dose and infusion rate of heparin needed to increase the intradialytic whole-blood clotting time to 150% of the predialysis value. The effectiveness of the model was assessed by comparing dialyzer reuse rates and delivered Kt/V(urea) before and after implementation of the model in 22 chronic hemodialysis patients. As an additional control, a similar group of 22 patients were followed up during the same period without adjustment of their heparin doses. Implementation of the model resulted in no change in the average loading dose (2,382 +/- 628 versus 2,425 +/- 908 IU; P = not significant) or average infusion rate (1,398 +/- 367 versus 1,393 +/- 532 IU/h; P = not significant) of heparin. However, individual patients required changes in loading dose or infusion rate. Dialyzer reuse rates increased significantly over time in the treatment group but remained unchanged in the control group (P < 0.003). Kt/V(urea) remained unchanged throughout the study period in both patient groups. From these data, we conclude that the use of a heparin model can improve dialyzer reuse rates without compromising the delivered dose of dialysis.

Early switch from intravenous to oral antibiotics and early hospital discharge: a prospective observational study of 200 consecutive patients with community-acquired pneumonia.

UNLABELLED: To determine the proportion of patients who can be treated with early switch to oral antibiotics and early discharge, to evaluate clinical outcome and patient satisfaction for patients treated with early switch and early discharge, and to define the factors that interfere with early discharge for some of the patients who underwent early switch to oral antibiotic therapy. DESIGN: Prospective study. PARTICIPANTS: Two hundred consecutive hospitalized patients with community-acquired pneumonia. MAIN OUTCOME MEASURES: Number of days needed to switch to oral therapy and length of hospital stay. Clinical outcome and satisfaction with care were evaluated for those patients treated with early switch and early discharge. RESULTS: Early switch to oral antibiotics (within the first 3 days of hospitalization) was performed in 133 patients (67%). Clinical failure was documented in 1 patient. Early switch and early discharge was performed in 88 patients (44%). The mean length of hospital stay for this group was 3.4 days. The most common reason for prolonged hospitalization after the switch to oral antibiotics was the need for diagnostic workup. More than 95% of patients were satisfied with the care they had received. CONCLUSIONS: Using simple clinical and laboratory criteria, a significant proportion of hospitalized patients with community-acquired pneumonia (44%) can be treated with early switch and early discharge. This model did not affect patient outcome, decreased the length of hospitalization, and was associated with a high level of patient satisfaction.

Prediction of anticoagulation during hemodialysis by population kinetics and an artificial neural network.

All systems currently used for routine hemodialysis require heparin administration to prevent blood clotting in the extracorporeal circuit. We tested the hypothesis that population-based statistical techniques can be used to predict heparin concentrations during routine hemodialysis. Two predictive models were developed, one based on nonlinear mixed effects modeling (NONMEM) and the other on a multilayer perceptron neural network. Serial clotting time data were obtained from forty-nine patients and used to develop the models. The models were used to predict the clotting times of 70 patients in a prospective test. We determined that the neural network provided greater precision, had fewer outliers in its predictions, and did not have the model misspecification in bolus administration that the NONMEM predictions demonstrated. A final NONMEM model was developed using all data from 119 patients to identify important covariates for predicting the heparin pharmacodynamic parameters, volume of distribution, and clearance. Both the volume of distribution and clearance increased following the initiation of dialysis and as the patient's baseline clotting time increased. The volume of distribution also increased as the patient's weight increased but was decreased by smoking and diabetes. Population-based statistical techniques may provide a useful alternative to existing methods for prescribing heparin.

Quality of life during and between hemodialysis treatments: role of L-carnitine supplementation.

End-stage renal disease affects every aspect of a patient's life, including perception of health and quality of life. It is likely that a hemodialysis patient's perceptions of health-related quality of life directly influence compliance with medical, nursing, and nutritional prescriptions. Because L-carnitine supplementation is known to enhance muscle strength and energy in hemodialysis patients, we hypothesized that L-carnitine supplementation would enhance a hemodialysis patient's perception of health-related quality of life. To test this hypothesis, 1 g L-carnitine or placebo was administered orally to 101 patients immediately before and after every hemodialysis treatment for 6 months. To assess health-related quality of life from the patient's perspective, the Medical Outcomes Study Short Form 36 instrument was administered before the study and at 1.5-month intervals for the duration of the study. In addition, a 10-item questionnaire designed to assess common intradialytic symptoms was administered at the end of each dialysis treatment. Other parameters analyzed included Kt/V(urea) and level of nutrition. In the 6-month group, oral L-carnitine supplementation had an early positive effect on general health (P < 0.02) and physical function (P < 0.03), but the perceived effect was not sustained throughout the 6 months of the study. In the 3-month group, L-carnitine supplementation improved vitality (P < 0.02) and general health (P < 0.01). There was no association between Kt/V(urea) and perceived health-related quality of life. Serum albumin concentration was directly correlated to how patients perceived the quality of their lives.

Continuous venovenous hemofiltration and hemodialysis after orthotopic heart transplantation.

Orthotopic heart transplantation (OHT) is an accepted treatment for heart failure refractory to medical management. International registry data reported a 1-year survival after OHT of 79%. The risk factors associated with requiring continuous venovenous hemofiltration or hemodialysis and mortality in patients requiring continuous venovenous hemofiltration (CVVH)/hemodialysis (HD) have not been studied, since improvements have been made in OHT. We tested the hypothesis that requiring CVVH/HD in the immediate posttransplantation period increases mortality after OHT. We studied pretransplantation factors to predict those patients who would need CVVH/HD. Patients undergoing OHT from 1995 to 1996 were studied. Fifty-two patients underwent OHT. Eight patients (15%) needed CVVH/HD. Initial immunosuppression included cyclosporine, azathioprine, and solumedrol. Patients were changed to antilymphocyte therapy if they had oliguric acute renal failure, T or B cell incompatibility, or biopsy-proven rejection. Overall, survival at 1 year was 84.6%. Survival in patients not needing CVVH/HD was 91%, and survival in patients needing CVVH/HD was 36.5%. For each year of age over 55 years, patients had a relative risk of requiring CVVH/HD of 7.

Outcomes of single organism peritonitis in peritoneal dialysis: gram negatives versus gram positives in the Network 9 Peritonitis Study.

The use of the "peritonitis rate" in the management of patients undergoing peritoneal dialysis is assuming importance in comparing the prowess of facilities, care givers and new innovations. For this to be a meaningful outcome measure, the type of infection (causative pathogen) must have less clinical significance than the number of infections during a time interval. The natural history of Staphylococcus aureus, pseudomonas, and fungal peritonitis would not support that the outcome of an episode of peritonitis is independent of the causative pathogen. Could this concern be extended to other more frequently occurring pathogens? To address this, the Network 9 Peritonitis Study identified 530 episodes of single organism peritonitis caused by a gram positive organism and 136 episodes caused by a single non-pseudomonal gram negative (NPGN) pathogen. Coincidental soft tissue infections (exit site or tunnel) occurred equally in both groups. Outcomes of peritonitis were analyzed by organism classification and by presence or absence of a soft tissue infection. NPGN peritonitis was associated with significantly more frequent catheter loss, hospitalization, and technique failure and was less likely to resolve regardless of the presence or absence of a soft tissue infection. Hospitalization and death tended to occur more frequently with enterococcal peritonitis than with other gram positive peritonitis. The outcomes in the NPGN peritonitis group were significantly worse (resolution, catheter loss, hospitalization, technique failure) compared to coagulase negative staphylococcal or S. aureus peritonitis, regardless of the presence or absence of a coincidental soft tissue infection. Furthermore, for the first time, the poor outcomes of gram negative peritonitis are shown to be independent of pseudomonas or polymicrobial involvement or soft tissue infections. The gram negative organism appears to be the important factor. In addition, the outcome of peritonitis caused by S. aureus is worse than that of other staphylococci. Thus, it is clear that all peritonitis episodes cannot be considered equivalent in terms of outcome. The concept of peritonitis rate is only meaningful when specific organisms are considered.

Pharmacokinetics of oral glyburide in subjects with non-insulin-dependent diabetes mellitus and renal failure.

To test the hypothesis that renal failure has no effect on the pharmacokinetics of glyburide, five subjects with non-insulin-dependent diabetes mellitus (NIDDM) and end-stage renal disease requiring hemodialysis, and four NIDDM subjects with normal renal function were studied. On days 0, 1, and 15, subjects consumed 33 carbohydrate grams, and glucose, insulin, and C-peptide were measured for 4 hours. On day 1, subjects received 3 mg glyburide and measured plasma concentrations for 48 hours. On day 3, multiple dosing on 3 mg glyburide daily began. On day 15, plasma concentrations were measured for 48 hours. The pharmacokinetics and pharmacodynamics of glyburide, glucose, insulin, and C-peptide were determined as well as daily fasting blood glucose. Glucose area under the curve (AUC) and daily fasting glucose levels did not change in either controls or hemodialysis subjects. The mean serum glyburide blood levels and pharmacokinetics did not differ after initial or chronic glyburide administration in NIDDM subjects with end-stage renal disease treated with hemodialysis compared with controls. Glyburide half-life averaged 3.3 hours in control subjects and 5.0 hours in hemodialysis subjects. Hemodialysis subjects had increased C-peptide and insulin AUC with chronic dosing. Renal failure does not affect the pharmacokinetics of 3.0 mg oral glyburide.

Hemodynamic, renal, and hormonal responses to lower body positive pressure in human subjects.

Studies in healthy human subjects subjected to lower body positive pressure (LBPP) have failed to elucidate many of the physiologic effects of this maneuver. In 7 healthy, well-hydrated men we studied the following responses to LBPP (35 mm Hg, 1 hour, supine position): systemic and renal hemodynamics; urine volume (UV), urine osmolality (Uosm), and urine sodium level (UNaV); free water (CH20) and osmolar (Cosm) clearances; plasma renin activity (PRA); levels of aldosterone (PA), cortisol (CORT), norepinephrine (NE), atrial natriuretic peptide (ANP), and vasopressin (AVP); osmolality (Posm); and serum sodium level. Subjects were restudied on a control day with zero trouser pressure. The recorded changes (p < 0.05) when comparing the LBPP day with the control day were as follows: fractional Na+ reabsorption increased (98.7% +/- 0.2% to 99.3% +/- 0.1%) and UNaV decreased (0.19 +/- 0.03 mEq/min to 0.10 +/- 0.01 mEq/min), with concomitant increases in PRA (1.7 +/- 0.2 ng/ml/90 min to 4.5 +/- 1.8 ng/ml/90 min), PA (7.7 +/- 0.7 ng/dl to 9.3 +/- 1.5 ng/dl), and CORT (13.0 +/- 2.6 mg/dl to 19.2 +/- 3 mg/dl); the increase in blood pressure with LBPP (96 +/- 3 mm Hg to 112 +/- 4 mm Hg) was greater than that during control conditions. Renal plasma flow tended to display an interactive pattern across days, with a slight decline during LBPP (5%) and a slight elevation under control conditions (9%). On the LBPP day only, filtered Na+ declined (15 +/- I mEq/min to 12 +/- 1 mEq/min) as a function of reduced glomerular filtration rate (112 +/- 5 ml/min to 91 +/- 7 ml/min), blood volume decreased (by 2.7% +/- 0.7%), CO decreased (5.5 +/- 0.3 L/min to 4.7 +/- 0.3 L/min), and stroke volume declined (101 +/- 6 ml to 84 +/- 3 ml). On both days, NE increased (control, 221 +/- 23 pg/ml to 340 +/- 33 pg/ml; LBPP, 236 +/- 17 pg/ml to 369 +/- 31 pg/ml) and ANP increased (control, 47 +/- 7 pg/ml to 97 +/- 21 pg/ml; LBPP, 49 +/- 10 pg/ml to 104 +/- 30 pg/ml). We concluded that LBPP reduces renal sodium excretion. The mechanism for this reduction is not known, although it did occur in association with an increase in plasma renin activity, which in turn results from mechanical reduction of renal perfusion, stress-related CORT stimulation, a reflex-based elevation in peripheral vascular resistance leading to a reflex increase in plasma renin activity, or a combination of these.

Application of artificial neural networks to clinical pharmacology.

Drug dosages and drug choices are determined by a knowledge of the drug's pharmacokinetics and pharmacodynamics. Often, insufficient information is available to determine the pharmacokinetics of a drug or which drug will have a desired effect for an individual patient. We propose that a form of nonlinear regression, an artificial neural network, can be used. We have demonstrated this use with 2 examples. In the first example we use a neural network to predict gentamicin peak and trough concentrations from routine therapeutic drug monitoring. In the second example we predict delayed renal allograft function as a guide for induction of immunosuppression therapy. Predictions were made using a multilayer feedforward perceptron and compared to nonlinear mixed effect modeling (NONMEM) and logistic regression. Neural network peak and trough gentamicin predictions were more precise and less biased than control predictions made using NONMEM. Prediction error for peak serum concentrations averaged 16.5% for the neural networks and 18.6% for NONMEM. Prediction error for trough concentrations were 48.3% for neural networks and 59.0% for NONMEM. When used for the prediction of delayed graft function, the neural network correctly predicted immediate graft function 73% of the time and delayed graft function 65% of the time. For those patients predicted to develop delayed graft function, alternate induction using anti-lymphocyte globulin may be indicated. These 2 examples demonstrate how an artificial neural network may be applied to predictions in clinical pharmacology.

Risk factors for peritonitis in long-term peritoneal dialysis: the Network 9 peritonitis and catheter survival studies. Academic Subcommittee of the Steering Committee of the Network 9 Peritonitis and Catheter Survival Studies.

To determine factors involved in peritoneal dialysis-associated peritonitis and catheter loss, all point prevalent peritoneal dialysis patients in Health Care Finance Administration (HCFA) end-stage renal disease (ESRD) Network 9 were followed throughout 1991 for peritonitis events and throughout 1991 to 1992 for catheter survival. Data were collected by questionnaires compiled by the dialysis facility and validated by network staff. Peritonitis was reported 1,168 times in 729 of the 1,930 patients. By gamma-Poisson regression, a significantly increased risk for peritonitis was observed for patients with previous peritonitis, black race, and those dialyzing with standard connectors or cyclers compared with disconnect systems. Decreased risks were observed for patients with longer ESRD experience and when prophylactic antibiotics were administered before catheter insertion. Postinsertion leakage, diabetes, visual problems, previous or current immunosuppression, and physical activity were not risk factors. Infection of any kind caused the removal of 68% of the 414 catheters lost. Patients with downward-directed tunnels were less likely to experience concomitant exit site/tunnel infections associated with peritonitis. Peritonitis episodes with Staphylococcus epidermidis-like organisms were more likely to resolve with a single course of antibiotics. Perhaps because of their higher infection rate, blacks were more likely than whites to use a disconnect system. In general, the outcome of peritonitis in blacks was similar to that in whites, except that blacks were less likely to be hospitalized and were less likely to die.

Statistical approach to neural network model building for gentamicin peak predictions.

Feed forward neural networks are flexible, nonlinear modeling tools that are an extension of traditional statistical techniques. The hypothesis that feed forward neural network models can be built in a similar fashion as a statistical model was tested. Feed forward neural network models were built using forward and backward variable selection, and zero to five hidden nodes, and tanh and linear transfer functions were used. Gentamicin serum concentrations were predicted as a model drug for testing these methods. Peak observations from 392 patients were used to train, test, and validate the feed forward neural network. Inputs were demographic and drug dosing information. Model selection was performed using the Akaike information criteria (AIC), Bayesian information criteria (BIC), and a method of stopped training. The models with lowest root mean square (rms) error were those with all 10 inputs and five hidden nodes. Average rms error in the validation set was lowest for stopped training (1.46), then AIC (1.51), and finally BIC (1.56). Larger models tended to result in the best predictions. Overfitting can occur in models that are too large, either by using too many nodes in the hidden layer (rms = 1.49) or by using too many inputs with little information associated with them (rms = 1.70). We conclude that neural networks can be built using a large number of parameters that have good predictive performance. Care must be used during training to avoid overfitting the data. A stopped training method resulted in the network with the lowest rms error.

Pharmacodynamics of atrial natriuretic peptide in isolated perfused Dahl rat kidneys.

Atrial natriuretic peptide (ANP) has been implicated in the development of hypertension in Dahl R and S rats. To test the responses of DR and DS kidneys in the absence of the influence of neural and humoral mechanisms, we investigated the pharmacokinetics and pharmacodynamics of ANP in isolated perfused DR and DS kidneys, obtained from rats given a high or low sodium diet, after a bolus injection of ANP (1 microgram) or after a bolus injection plus infusion of ANP to maintain the perfusate concentration at 1000 pg/ml. The elimination rate constant was not different between the groups (DR, 0.044 min-1 vs. DS, 0.050 min-1). Clearance of ANP was 4 times greater than the glomerular filtration rate, indicating that a receptor-mediated peritubular clearance is probably the primary route of elimination. DS kidneys excreted 50% less sodium than DR kidneys. However, ANP caused a 5-fold increase in fractional sodium excretion in both DR and DS. ANP also increased sodium excretion, creatinine clearance, and urine flow. No alteration in ANP kinetics occurred to account for the reportedly increased circulating concentrations of ANP seen in DS rats. We conclude that isolated DR and DS kidneys respond differently to ANP after bolus ANP administration to concentrations of 10,000 pg/ml. This difference in response is due to the sodium excretory defect inherent in the DS kidney and not to an alteration in the DS kidney's ANP responsiveness.

Omeprazole therapy does not affect pharmacokinetics of orally administered ethanol in healthy male subjects.

The purpose of this study was to study the effect of high-dose omeprazole therapy (20 mg twice daily) on kinetics of moderate amounts of orally administered ethanol. Eight healthy men participated in the study. After an overnight fast, they drank 0.5 g/kg body weight ethanol over 20 min. Blood samples were drawn before and then every 20 min after ethanol ingestion for the next 3 h. Subjects then ingested omeprazole 20 mg twice daily for 6 days. On the seventh day, the same dose of oral ethanol was administered as before and blood samples drawn. Blood ethanol concentrations were determined. We fit a one-compartment model with first-order absorption and zero order elimination to the blood ethanol data with PCNONLIN (SCI Software, Lexington, KY, U.S.A.) separately for each subject before as well as after omeprazole therapy. Area under the curve was calculated using the trapezoidal rule. There were no differences in the peak concentration, time to peak concentration, area under the curve, or elimination rate constant for ethanol before and after omeprazole treatment. Omeprazole treatment (20 mg twice daily) does not affect the pharmacokinetics of orally ingested ethanol in healthy male subjects. Our results do not rule out a possible effect on psychomotor function as a result of a pharmacodynamic interaction.

Population pharmacokinetics of the active metabolite of nabumetone in renal dysfunction.

We determined the pharmacokinetics of 6-methoxy-2-naphythylacetic acid (6-MNA), the active metabolite of nabumetone, in normal subjects and in subjects with impaired renal function, including subjects requiring hemodialysis. Subjects received a 1000 mg oral dose of nabumetone either as a single dose or daily for 15 days. We used a noncompartmental pharmacokinetic analysis and compared those results to a population pharmacokinetic analysis performed with nonlinear mixed-effects modeling (NONMEM). The results of the two different analyses were similar. The elimination half-life increased with decreased renal function and ranged from 22 to 44 hours. The area under the curve decreased significantly at steady state, regardless of renal function. The apparent clearance determined by NONMEM analysis increased from 0.68 L/hr after a single dose to 1.13 L/hr at steady state. The apparent volume of distribution was directly proportional to the nonalbumin protein concentration and ranged from 23 to 60 L. We conclude that the pharmacokinetics of 6-MNA in this population are complicated by possible protein binding changes. However, the increased half-life in patients with renal failure is offset by changes in the apparent volume of distribution that prevent the accumulation of 6-MNA in the serum of patients with impaired renal function. Therefore dosage adjustments may not be necessary in patients with decreased renal function.

Pseudomonas peritonitis in peritoneal dialysis patients: the Network #9 Peritonitis Study.

To determine risk factors for the development of Pseudomonas peritonitis (PsP) and outcomes of PsP, the authors compared peritoneal dialysis patients who developed PsP with peritoneal dialysis patients who developed non-Pseudomonas bacterial peritonitis (non-PsP). The authors also sought to determine if there were differences in patients who had resolution of PsP compared with those patients whose PsP did not resolve. The data were derived from the prospective Tristate Renal Network Peritonitis and Catheter Survival Study. Resolution in this study was defined as clearing of peritoneal dialysate on visual inspection, with up to three courses of antibiotic therapy allowed. Catheter removal, switch to hemodialysis, or death were outcomes that were considered separately from resolution because of the study design. There were 31 cases of PsP in 28 patients and 886 cases of non-PsP identified in 667 adult patients. There were no differences in race, gender, age, or incidence of diabetes between the groups. The PsP group had a 25% incidence of previous exposure to immunosuppressive agents, whereas it was 10.6% in the non-PsP group (P = 0.028). PsP infections were more frequently associated with concomitant exit and tunnel infections, higher hospitalization rates, increased incidence of catheter loss, switch to hemodialysis, and a worse rate of resolution when compared with non-PsP (all, P < 0.05). Logistic regression could not identify patients at increased risk of PsP. PsP resolved with antibiotic therapy only in 10 of 31 episodes.(ABSTRACT TRUNCATED AT 250 WORDS)

Neural network predicted peak and trough gentamicin concentrations.

Predictions of steady state peak and trough serum gentamicin concentrations were compared between a traditional population kinetic method using the computer program NONMEM to an empirical approach using neural networks. Predictions were made in 111 patients with peak concentrations between 2.5 and 6.0 micrograms/ml using the patient factors age, height, weight, dose, dose interval, body surface area, serum creatinine, and creatinine clearance. Predictions were also made on 33 observations that were outside the 2.5 and 6.0 micrograms/ml range. Neural networks made peak serum concentration predictions within the 2.5-6.0 micrograms/ml range with statistically less bias and comparable precision with paired NONMEM predictions. Trough serum concentration predictions were similar using both neural networks and NONMEM. The prediction error for peak serum concentrations averaged 16.5% for the neural networks and 18.6% for NONMEM. Average prediction errors for serum trough concentrations were 48.3% for neural networks and 59.0% for NONMEM. NONMEM provided numerically more precise and less biased predictions when extrapolating outside the 2.5 and 6.0 micrograms/ml range. The observed peak serum concentration distribution was multimodal and the neural network reproduced this distribution with less difference between the actual distribution and the predicted distribution than NONMEM. It is concluded that neural networks can predict serum drug concentrations of gentamicin. Neural networks may be useful in predicting the clinical pharmacokinetics of drugs.

Laser indirect ophthalmoscope photocoagulation and scleral depression for rhegmatogenous retinal detachment.

BACKGROUND: Transconjunctival cryotherapy or laser photocoagulation with simultaneous Eisner funnel scleral depression has been used to treat selected cases of rhegmatogenous retinal detachment. There are no studies to date reporting the use of the laser indirect ophthalmoscope coupled with scleral depression for treating retinal detachment. METHODS: 16 consecutive patients (18 retinal detachments in 17 eyes) were enrolled in a prospective, uncontrolled clinical trial using the laser indirect ophthalmoscope with scleral depression as the sole treatment for retinal detachment. The region immediately surrounding the break where subretinal fluid was present was directly treated rather than demarcated. All patients were treated with local anesthesia in an outpatient setting. RESULTS: Complete retinal reattachment was initially achieved in 14 (78%) of 18 eyes after scleral depression and laser alone. Significant postoperative complications of scleral depression with laser indirect ophthalmoscope photocoagulation included macular pucker (2 eyes), late recurrent rhegmatogenous retinal detachment without proliferative vitreoretinopathy (1 eye), and late recurrent rhegmatogenous retinal detachment with proliferative vitreoretinopathy (2 eyes). Failure of initial treatment to flatten the retina, late recurrent retinal detachment, macular pucker, or proliferative vitreoretinopathy led to scleral buckling and/or vitrectomy in 6 (86%) of the 7 eyes with clinical detachment and 3 (30%) of the 10 eyes with localized detachment. Final retinal reattachment at the last follow-up examination was achieved in all 17 eyes with subsequent surgical procedures. CONCLUSION: Although scleral depression with laser indirect ophthalmoscope photocoagulation is a noninvasive outpatient surgical procedure that is capable of flattening selected retinal detachments, its use cannot be recommended because of the relatively high rate of postoperative complications requiring further surgical procedures.