Building Clinical Classifiers Using Incomplete Observations – A Neural Network Ensemble for Hepatoma Detection in Patients with Cirrhosis

One objective of liver transplant evaluation is to identify patients that harbor a hepatoma, but standard screening techniques are not sensitive enough. We trained neural network ensembles to predict the presence of hepatoma in patients with cirrhosis, based on information collected at the time of transplant evaluation. Network architecture and training were modified to handle missing observations. Three ensembles were trained: ensemble A using the subset with no missing observations (528 patients); ensemble B using the complete set, which included missing observations (853 patients); and ensemble C using the smaller subset, originally with complete data, but after a fixed number of observations were deleted (i. e., made “missing”). Ensemble performance on testing sets was very good. The areas under the ROC curves were 0.91, 0.89, and 0.90, for ensembles A. B, and C, respectively. Neural networks can successfully perform this classification task, and strategies can be developed that allow use of incomplete observations.

Recurrent Neural Networks for Predicting Outcomes after Liver Transplantation: Representing Temporal Sequence of Clinical Observations

Objectives: This paper investigates a version of recurrent neural network with the backpropagation through time (BPTT) algorithm for predicting liver transplant graft failure based on a time series sequence of clinical observations. The objective is to improve upon the current approaches to liver transplant outcome prediction by developing a more complete model that takes into account not only the preoperative risk assessment, but also the early postoperative history.

Methods: A 6-fold cross-validation procedure was used to measure the performance of the networks. The data set was divided into a learning set and a test set by maintaining the same proportion of positive and negative cases in the original set. The effects of network complexity on overfitting were investigated by constructing two types of networks with different numbers of hidden units. For each type of network, 10 individual networks were trained on the learning set and used to form a committee. The performance of the networks was measured exhaustively with respect to both the entire training and test sets.

Results: The networks were capable of learning the time series problem and achieved good performances of 90% correct classification on the learning set and 78% on the test set. The prediction accuracy increases as more information becomes progressively available after the operation with the daily improvement of 10% on the learning set and 5% on the test set.

Conclusions: Recurrent neural networks trained with BPTT algorithm are capable of learning to represent temporal behavior of the time series prediction task. This model is an improvement upon the current model that does not take into account postoperative temporal information.

Changes in Serum Electrolytes during Treatment of Patients in Liver Failure with Molecular Adsorbent Recirculating System

PURPOSE

To study the effect of MARS on serum electrolytes during liver failure.

DESIGN

Twenty-three patients admitted to a quaternary health care facility from September 2000 to May 2002, 22 adults and 1 child, 11 males (48%) and 12 females (52%), age 15-70 (median 53), treated with MARS for: 12 acute-on-chronic liver failure (52%); 4 fulminant hepatic failure (17%); 3 intractable pruritus (13%); 2 primary-non-function (9%); 2 following major liver resection (9%).

PROCEDURES

Sodium, potassium, chloride, phosphorus, calcium, and magnesium were measured in the serum, ultrafiltrate and albumin circuit before and after MARS.

STATISTICAL METHODS

A comparison of electrolyte concentrations, before and after MARS, was performed using a paired t test.

MAIN FINDINGS

Serum electrolyte concentrations before and after MARS, while statistically significant in some cases, were very small, and of no clinical relevance.

CONCLUSION

MARS exchanges potassium, chloride, calcium, and magnesium by ultrafiltration; sodium by the albumin dialysis.

Antibody binding to endothelial and epithelial antigens triggers pig-to-rabbit xenograft rejection and its absence results in atypical complement deposition

In pig-to-rabbit kidney xenograft (PRKX), endothelial antigen determinants (EAD) are immediately recognized by IgG and IgA, while IgM does not react with them. The purpose of this study was to investigate the different roles of IgG, IgA, IgM, and complement in the hyperacute rejection of a PRKX model. Nine isolated Landrace pig kidneys were each perfused with 10 ml normal New Zealand rabbit serum. Perfusates (serum A) were collected after discarding the first 0.5 ml. Serum A and rabbit complement were then incubated for 30 min with frozen sections of normal pig kidney. After washing with buffer solution all the specimens were treated for immunohistochemistry. Three frozen sections of normal Landrace pig kidney and three samples of normal New Zealand rabbit serum were used as controls. Immunohistochemical analysis of the nine perfused kidneys demonstrated IgG, IgA and C3 deposition on the peritubular and glomerular vascular endothelium. No IgM reactivity was shown. In the frozen sections exposed to serum A, immunofluorescence showed minimal IgG, IgA and C3 reactivity while IgM deposition was clearly evident on the tubular epithelium. Immunofluorescence of frozen sections exposed to rabbit complement, done by fluorescein-labeled goat anti-rabbit C3 antibodies were positive only in the glomerular endothelium. The same rabbit complement was active in antibody dependent cytotoxicity on human T cells. Our results indicated that in the PRKX model, IgG and IgA acted as preformed antibodies recognizing endothelial EAD. IgM did not bind to any endothelial molecules, but recognized antigens located on the brush border of the tubular epithelium. Furthermore, in this model, absence of antigen-antibody complexes resulted in atypical complement deposition.

Recurrent neural networks for predicting outcomes after liver transplantation: representing temporal sequence of clinical observations

OBJECTIVES

This paper investigates a version of recurrent neural network with the backpropagation through time (BPTT) algorithm for predicting liver transplant graft failure based on a time series sequence of clinical observations. The objective is to improve upon the current approaches to liver transplant outcome prediction by developing a more complete model that takes into account not only the preoperative risk assessment, but also the early postoperative history.

METHODS

A 6-fold cross-validation procedure was used to measure the performance of the networks. The data set was divided into a learning set and a test set by maintaining the same proportion of positive and negative cases in the original set. The effects of network complexity on overfitting were investigated by constructing two types of networks with different numbers of hidden units. For each type of network, 10 individual networks were trained on the learning set and used to form a committee. The performance of the networks was measured exhaustively with respect to both the entire training and test sets.

RESULTS

The networks were capable of learning the time series problem and achieved good performances of 90% correct classification on the learning set and 78% on the test set. The prediction accuracy increases as more information becomes progressively available after the operation with the daily improvement of 10% on the learning set and 5% on the test set.

CONCLUSIONS

Recurrent neural networks trained with BPTT algorithm are capable of learning to represent temporal behavior of the time series prediction task. This model is an improvement upon the current model that does not take into account postoperative temporal information.

Malignant fibrous histiocytoma of the gallbladder: case report and review of the literature

Malignant fibrous histiocytoma is a soft tissue sarcoma of mesenchymal origin. It can rarely present as a primary gallbladder tumor with only five cases having been reported to date in the English literature. Here we report the sixth documented case of malignant fibrous histiocytoma of the gallbladder, and we review all other cases reported. The outcome of the visceral sarcomas is poor when compared with tumors arising from the soft tissues. The treatment of primary malignant fibrous histiocytomas of the gallbladder is surgery. However, tumor recurrence is the norm even if wide clean margins are obtained. In contrast to tumors arising from the extremities the role of adjuvant radiotherapy and chemotherapy is less clear in the case of retroperitoneal and visceral sarcomas. Our patient is still alive and free of disease 46 weeks after surgery. The fact that this is the longest survival reported to date underscores the dismal prognosis of this disease.

The impact of liver dysfunction on outcome in patients with multiple injuries

Multiple organ dysfunction syndrome (MODS) is the leading cause of late deaths after traumatic injury. The relative importance of dysfunction of individual organ systems in determining outcome from MODS has not been clearly defined. Some studies have suggested that hepatic dysfunction associated with MODS increases mortality, whereas others have suggested that it contributes little to outcome in trauma patients. To clarify the role of the hepatic dysfunction after traumatic injury we retrospectively reviewed all trauma patients with an Injury Severity Score > or = 14 admitted from January 1, 1994 through June 30, 1997 for the presence of hepatic dysfunction defined as a serum bilirubin > or = 2.0 mg/dL. Of the 1962 patients who met the entry criteria 154 developed hepatic dysfunction during their hospital stay. Patients with hepatic dysfunction were older (46 +/- 2 versus 41 +/- 1 years), were more severely injured (Injury Severity Score 31.5 +/- 0.9 versus 23.3 + 0.2), and had a lower prehospital blood pressure (102 +/- 3 versus 117 +/- 1 mm Hg) compared with patients who did not develop hepatic dysfunction. Patients with hepatic dysfunction were more likely to present with shock as reflected in a lower initial emergency room blood pressure (109 +/- 3 versus 128 +/- 1 mm Hg) and base deficit (-6.9 +/- 0.6 versus -3.5 +/- 0.1 mEq/L). Patients who developed hyperbilirubinemia had longer lengths of stay in the intensive care unit (15.8 +/- 1.2 versus 3.4 +/- 0.2 days) and the hospital (27.4 +/- 1.7 versus 11.1 +/- 0.2 days) and a higher in-hospital mortality (16.2% versus 2.5%). These data demonstrate that the development of hepatic dysfunction reflects the severity of injury and is associated with a significantly worse outcome after traumatic injury.

Reducing the length of stay after kidney transplantation--the intensive outpatient unit

The need to reduce the costs associated with the initial hospitalization for kidney transplantation has led to the development of outpatient facilities in which patients can be seen on a daily basis. The implementation of a kidney transplant intensive outpatient unit (IOPU) is described. Prior to the opening of the IOPU, the median and mean lengths of stay after kidney transplantation in our program were 14.0 and 18.9 d, respectively. Subsequent to the opening of the IOPU, the median and mean lengths of stay after kidney transplantation have gradually decreased and are currently 5.0 and 7.5 d, respectively. The median inpatient cost of transplantation, excluding organ acquisition charges, has decreased by 54%, from $25516 to $11616. Patient satisfaction has exceeded 80%. The IOPU represents an effective means of reducing the cost associated with transplantation, without sacrificing the quality of care.

Matching donors and recipients

This study identifies the major risk factors associated with outcome after liver transplantation, showing that candidates for this surgery can be stratified into differential risk categories at the time of the actual surgery. All the livers used were flushed with University of Wisconsin solution. The study is a retrospective multivariate analysis of 2376 consecutive transplantations performed on 2019 recipients between November 1, 1987, and December 31, 1993. Donor variables studied were age, sex, blood type, cause of death, intensive care unit length of stay, body mass index, use of pressors (dopamine infusion > 10 micrograms/kg/min or continuous infusion of epinephrine or norepinephrine), use of pitressin, cardiopulmonary resuscitation, terminal transaminase levels, serum sodium level at procurement, and total ischemia time. Recipient variables studied were age; sex; blood type; indication for liver transplantation; history of liver transplantation or upper abdominal surgery; United Network for Organ Sharing urgency status; need for mechanical ventilation; primary immunosuppression; and preoperative bilirubin level, prothrombin time, and creatinine level. The variables independently associated with outcome were donor age, female donor sex, ischemia time, recipient age, prior liver transplant, preoperative mechanical ventilation, preoperative bilirubin level, preoperative creatine level, indication for transplantation and primary immunosuppression used. The results of this study not only give us insight into the probable outcomes of individual patients, but also show that this stratification can be useful when comparing results across different groups or in helping to choose the best donor-recipient combination based on the calculated probability of a favorable outcome.

Multisolutional clustering and quantization algorithm (MCQ)

We have developed a novel clustering and quantization algorithm that allows the user to create multiple one-to-one correspondences between the actual data and its transformed (clustered and quantized) values, based on the user's hypothesis regarding the nature of the classification task. The types of problems for which the algorithm can be beneficial are discussed. We report experiments employing simulated and real data that suggest the proposed algorithm may be useful in neural network analysis of various phenomena in medicine and biology.

Assessing risk in liver transplantation. Special reference to the significance of a positive cytotoxic crossmatch

OBJECTIVE

The authors determined the impact of a positive cytotoxic crossmatch on the outcome of liver transplantation.

SUMMARY BACKGROUND DATA

Liver allografts rarely undergo hyperacute rejection, but transplants performed across a positive cytotoxic crossmatch tend to follow a different clinical course, with higher intraoperative blood use, postoperative graft dysfunction, and, in some cases, graft loss. How this affects overall graft survival has not been determined.

METHODS

The authors provide a retrospective analysis of 1520 liver transplants performed between November 1989 and December 1993, with a minimum follow-up of 1 year. All cases had a cytotoxic crossmatch using serum pretreated with dithiothreitol.

RESULTS

There were 1390 negative crossmatch and 130 positive crossmatch cases. There was no difference in overall graft survival, although early survival rates were lower in the positive crossmatch group, with the maximum difference at 6 months: 0.76 (95% confidence interval, 0.74-0.78) for a negative crossmatch versus 0.68 (95% confidence interval, 0.61-0.77) for a positive crossmatch. These differences become negligible by the 2-year mark. Using stepwise logistic regression, the authors identified seven variables independently associated with outcome: 1) donor age, 2) donor gender, 3) prior liver transplant, 4) medical urgency status, 5) ischemia time, 6) indication for transplantation, and 7) primary immunosuppressant.

CONCLUSIONS

The cytotoxic crossmatch is not statistically associated with overall graft survival after liver transplantation. However, early failure rates are higher in the positive crossmatch cases, a difference that disappears by the second year.

Hepatic Retransplantation--an analysis of risk factors associated with outcome

Hepatic retransplantation is controversial because the results are inferior to primary transplants and organs are so scarce. To determine the factors that are associated with poor outcome within the first year following retransplantation, we performed a multivariate analysis, using stepwise logistic regression, of 418 hepatic retransplantations performed at a single institution from November 1987 to December 1993. The minimum follow-up was 1 year. Seven variables were found to be independently associated with subsequent graft failure (defined as either patient death or retransplantation): donor age (odds ratio 2.2 for each 10-year increase over age 45, 95% CI 1.3 to 3.7), female donor sex (odds ratio 1.7, 95% CI 1.05 to 2.7), recipient age (odds ratio 1.6 for each 10-year increase over age 45,95% CI 1.2 to 2.8), need for preoperative mechanical ventilation (odds ratio 1.8, 95% CI 1.1 to 2.9), pretransplant serum creatinine (odds ratio 1.24 for each increase of 1 mg/dl, 95% CI 1.1 to 1.4), pretransplant total serum bilirubin (odds ratio 1.4 for each 10-mg/dl increase over 15 mg/dl, 95% CI 1.1 to 1.8), and the primary immunosuppressant, using tacrolimus as the reference category (odds ratio for cyclosporine-based immunosuppression 3.9, 95% CI 2.3 to 6.8). Although not part of the logistic regression model, the timing of retransplantation was also found to be important, with the overall probability of failure increasing from 0.58 on day 0 to a peak of 0.8 on day 38 and decreasing slowly after that. The implications of these results regarding the appropriateness of retransplantation are discussed.

Risk factors and predictive indexes of early graft failure in liver transplantation

A retrospective analysis of 462 consecutive liver transplantations has been carried out. These were divided into two groups, according to whether they failed within 90 days (Group I) or survived longer than 90 days (Group II). Twenty-five donor and recipient variables were analyzed. In the univariate analysis, the only donor variable that was significantly different between the two groups was age (45.3 +/- 16.9 years in Group I vs 37.9 +/- 15.4 years in Group II, p < 0.001). There were five recipient variables significantly associated with early graft failure: history of previous liver transplantations (p < 0.0001), United Network for Organ Sharing 4 status (p = 0.003), primary diagnosis (p = 0.001), preoperative serum creatinine (1.97 +/- 1.5 mg/dL in Group I vs 1.46 +/- 1.2 mg/dL in Group II, p = 0.005), and preoperative total serum bilirubin (13.5 +/- 14.4 mg/dL in Group I vs 8.4 +/- 11.4 mg/dL in Group II, p = 0.003). In the multivariate analysis, only three variables were independently associated with outcome: donor age greater than 45 years, abnormal (> 1.5 mg/dL) recipient preoperative creatinine, and a history of previous liver transplantation.

A high incidence of native portal vein thrombosis in veterans undergoing liver transplantation

The incidence of native portal vein thrombosis (PVT) in liver transplant recipients has been reported to range from 2.1 to 13.8%. We have identified an inordinately high incidence of PVT in a consecutive series of U.S. veterans receiving liver transplants. Between October 1989 and February 1994, 88 consecutive U.S. veterans received 99 orthotopic liver transplants under primary Tacrolimus (Prograf, formerly FK506) based immunosuppression. A number of clinical features were examined in an effort to identify risk factors for PVT and outcome was compared to patients without PVT. Native PVT was present in 23/88 (26%) patients. All of these patients were male U.S. veterans with a mean age of 47 years. When compared to the 65 patients without PVT, we found no significant difference with respect to underlying liver disease, age, Childs-Pugh score (mean = 12), UNOS status as defined prior to April 1995 (95% UNOS 3 or 4), previous abdominal surgery, or liver volume. Median blood loss for patients with PVT (21 units of packed red blood cells) was greater than for those without PVT (14 units, P = 0.04). Portal thrombectomy was performed in 11 patients, 11 patients required mesoportal jump grafts, and 1 patient had an interposition graft. Standard veno-venous bypass was used in 10 patients with single bypass utilized for the remainder. Actuarial patient survival for all patients at 1, 2, and 4 years was 88, 85, and 79%, respectively. There was no significant difference in patients with or without PVT. Patients with PVT had poorer graft survival than patients without PVT (86% vs 65%, 1 year; 81% vs 65%, 2 years; 81% vs 61%, 4 years; P = 0.03); however, this was not related to technical problems with the portal venous inflow. PVT occurred in 26% of U.S. veterans undergoing liver transplantation. These patients had significantly higher operative blood loss and poorer graft survival. The high incidence of postnecrotic cirrhosis in a predominantly male group of patients with advanced disease, as is evident by the high mean Childs-Pugh score and UNOS status, perhaps accounts for our observations.

Orthotopic liver transplantation in U.S. veterans under primary tacrolimus immunosupression

The evolution and refinement of surgical techniques, per ioperative patient care, and immunosuppression hav~ estab~ished orthoto~ic li~er transplantation (OLTX) as a ~ighly successful therapeutic modality for patients wrth end-stage hver disease. In February 1989,Tacrohmus (Prograf®, formerly FK 506)was first used successfully at the University of Pittsburgh Medical Center to treat patients with rejection refractory to cyclosporine-based immunosuppression." Clinical trials utilizing Tacrolimus in solid organ transplantation followed, and in April of 1994 it was approved for use by the Food and Drug Administration,

Effect of donor age and sex on the outcome of liver transplantation

We correlated donor and recipient factors with graft outcome in 436 adult patients who underwent 462 liver transplants. Donor variables analyzed were age, gender, ABO blood group, cause of death, length of stay in the intensive care unit, use of pressors or pitressin, need for cardiopulmonary resuscitation, terminal serum transaminases, and ischemia time. Recipient variables analyzed were age, gender, primary diagnosis, history of previous liver transplant, ABO blood group, cytotoxic antibody crossmatch, United Network for Organ Sharing (UNOS) status, and waiting time (except for the cross-match results, they were all known at the time of the operation). The endpoint of the analysis was graft failure, defined as patient death or retransplantation. Using multivariate analysis, graft failure was significantly associated with donor age, donor gender, previous liver transplantation, and UNOS 4 status of the recipient. The effect of donor age became evident only when they were older than 45 years. Livers from female donors yielded significantly poorer results, with 2-year graft survival of female to male 55% (95% CI, 45% to 67%); female to female, 64% (95% CI, 54% to 77%); male to male, 72% (95% CI, 66% to 78%); and male to female, 78% (95% CI, 70% to 88%). The only donors identified as questionable for liver procurement were old (> or = 60 years) women in whom the adverse age and gender factors were at least additive. However, rather than discard even these livers, in the face of an organ shortage crisis, their individualized use is suggested with case reporting in a special category.

Effect of donor age and sex on the outcome of liver transplantation

We correlated donor and recipient factors with graft outcome in 436 adult patients who underwent 462 liver transplants. Donor variables analyzed were age, gender, ABO blood group, cause of death, length of stay in the intensive care unit, use of pressors or pitressin, need for cardiopulmonary resuscitation, terminal serum transaminases, and ischemia time. Recipient variables analyzed were age, gender, primary diagnosis, history of previous liver transplant, ABO blood group, cytotoxic antibody crossmatch, United Network for Organ Sharing (UNOS) status, and waiting time (except for the cross-match results, they were all known at the time of the operation). The endpoint of the analysis was graft failure, defined as patient death or retransplantation. Using multivariate analysis, graft failure was significantly associated with donor age, donor gender, previous liver transplantation, and UNOS 4 status of the recipient. The effect of donor age became evident only when they were older than 45 years. Livers from female donors yielded significantly poorer results, with 2-year graft survival of female to male 55% (95% CI, 45% to 67%); female to female, 64% (95% CI, 54% to 77%); male to male, 72% (95% CI, 66% to 78%); and male to female, 78% (95% CI, 70% to 88%). The only donors identified as questionable for liver procurement were old (> or = 60 years) women in whom the adverse age and gender factors were at least additive. However, rather than discard even these livers, in the face of an organ shortage crisis, their individualized use is suggested with case reporting in a special category.

Long-term results after liver transplantation for primary hepatic epithelioid hemangioendothelioma

BACKGROUND

Hepatic epithelioid hemangioendothelioma (PHEHE) is a multifocal, low-grade malignant neoplasia characterized by its epithelial-like appearance and vascular endothelial histogenesis. The outcome of 16 patients treated with orthotopic liver transplantation (OLT) is the subject of this report.

METHODS

A retrospective study of 16 patients with HEHE (7 men, 9 women) with ages ranging from 24 to 58 years (mean 37 +/- 10.6 years). Follow-up intervals ranged from 1 to 15 years (median of 4.5 years).

RESULTS

Actual patient survival at 1, 3, and 5 years was 100, 87.5, and 71.3%, respectively. Disease-free survival at 1, 3, and 5 years was 81.3, 68.8, and 60.2%, respectively. The 90-day operative mortality was 0. Involvement of the hilar lymph nodes or vascular invasion did not affect survival. The 5-year survival of HEHE compares favorably with that of hepatocellular carcinoma at the same stage (stage 4A): 71.3 versus 9.8% (p = 0.001)

CONCLUSIONS

The long-term survival obtained in this series justifies OLT for these tumors even in the presence of limited extrahepatic disease.

Building clinical classifiers using incomplete observations--a neural network ensemble for hepatoma detection in patients with cirrhosis

One objective of liver transplant evaluation is to identify patients that harbor a hepatoma, but standard screening techniques are not sensitive enough. We trained neural network ensembles to predict the presence of hepatoma in patients with cirrhosis, based on information collected at the time of transplant evaluation. Network architecture and training were modified to handle missing observations. Three ensembles were trained: ensemble A using the subset with no missing observations (528 patients); ensemble B using the complete set, which included missing observations (853 patients); and ensemble C using the smaller subset, originally with complete data, but after a fixed number of observations were deleted (i.e., made "missing"). Ensemble performance on testing sets was very good. The areas under the ROC curves were 0.91, 0.89, and 0.90, for ensembles A, B, and C, respectively. Neural networks can successfully perform this classification task, and strategies can be developed that allow use of incomplete observations.

Liver transplantation without the use of blood products

OBJECTIVES

To examine the techniques and the outcome of liver transplantation with maximal conservation of blood products and to analyze the potential benefits or drawbacks of blood conservation and salvage techniques.

DESIGN

Case series survey.

SETTING

Tertiary care, major university teaching hospital.

PATIENTS AND METHODS

Four patients with religious objections to blood transfusions who were selected on the basis of restrictive criteria that would lower their risk for fatal hemorrhage, including coagulopathy, a thrombosed splanchnic venous system requiring extensive reconstruction, active bleeding and associated medical complications. All patients were pretreated with erythropoietin to increase production of red blood cells. All operations were performed at the same institution, with a 36-month follow-up.

INTERVENTIONS

Orthotopic liver transplantation that used blood salvage, plateletpheresis, and autotransfusion and the withholding of the use of human blood products with the exception of albumin.

MAIN OUTCOME MEASURES

Survival and postoperative complications, with the effectiveness of erythropoietin and plateletpheresis as secondary measures.

RESULTS

All patients are alive at 36 months after orthotopic liver transplantation. One patient, a minor (13 years of age), was transfused per a state court ruling. Erythropoietin increased the production of red blood cells as shown by a mean increase in hematocrit levels of 0.08. Platelet-pheresis allowed autologous, platelet-rich plasma to be available for use after allograft reperfusion. Three major complications were resolved or corrected without sequelae. Only one patient developed postoperative hemorrhage, which was corrected surgically. The mean charge for bloodless surgery was $174,000 for the three patients with United Network for Organ Sharing (UNOS) status 3 priority for transplantation. This result was statistically significant when these patients were compared with all the patients with UNOS status 3 priority during the same period who met the same restrictive guidelines (P < .05). Only 19 of 1009 orthotopic liver transplantations performed at our institution were similar according to the UNOS status and the fulfillment of the guidelines. The mean charge for these comparison patients was $327,000, 3.8% of which was related to transfusions.

CONCLUSIONS

Orthotopic liver transplantation without the use of blood products is possible. Blood conservation techniques do not increase morbidity or mortality and can result in fewer transfusion-related, in-hospital charges.

Early death or retransplantation in adults after orthotopic liver transplantation. Can outcome be predicted?

Early, reliable outcome prediction after a liver transplant would help improve organ use by minimizing unnecessary retransplantations. At the same time, early intervention in those cases destined to fail may ameliorate the high morbidity and mortality associated with retransplantation. The purpose of this study was to analyze several parameters that have been identified in the past as being associated with patient and graft outcome, and to try to develop a model that would allow us to make predictions based on data available in the early postoperative period. A total of 148 patients were followed in a prospective, observational study. Graft failure was defined as patient death or retransplantation within 3 months of surgery. Preoperative variables studied included patient demographics, need for life support, presence of ascites, serum bilirubin, serum albumin, prothrombin time, serum creatinine, and the results of the cytotoxic crossmatch. During the first 5 postoperative days, standard measurements included serum transaminases, serum bilirubin, ketone body ratio, prothrombin time, factor V, and serum lactate. Oxygen consumption was measured shortly after surgery, once the patients had rewarmed to 36 degrees C. There were 131 successful transplants (88.5%) and 17 failures (11.5%). Most of the variables studied were found to be associated with outcome (by univariate analysis) at different points in the early postoperative period. However, receiver operating characteristic curve analysis showed that the predictive ability of even the best parameter was not adequate to make decisions on individual patients. Multivariate analysis, using stepwise logistic regression, yielded a model with an overall accuracy of 92.7%. Again, receiver operating characteristic curve analysis suggested that this model did not achieve the discriminating power needed for routine clinical use. We are still not able to accurately predict outcome in the early posttransplant period. We must be very careful when evaluating parameters, or scoring systems, that are said to accomplish this. It is especially important in this era of cost containment, with its renewed pressures to guide therapy based on our perceived understanding of a patient's future clinical course.

Predicting outcomes after liver transplantation. A connectionist approach

OBJECTIVE

The authors sought to train an artificial neural network to predict early outcomes after orthotopic liver transplantation.

SUMMARY BACKGROUND DATA

Reliable prediction of outcomes early after liver transplantation would help improve organ use and could have an impact on patient survival, but remains an elusive goal. Traditional multivariate models have failed to attain the sensitivity and specificity required for practical clinical use. Alternate approaches that can help us model clinical phenomena must be explored. One such approach is the use of artificial neural networks, or connectionist models. These are computation systems that process information in parallel, using large numbers of simple units, and excel in tasks involving pattern recognition. They are capable of adaptive learning and self-organization, and exhibit a high degree of fault tolerance.

METHODS

Ten feed-forward, back-propagation neural networks were trained to predict graft outcomes, using data from 155 adult liver transplants. The data included information that was available by the second postoperative day. Ten separate training and testing data subsets were prepared, using random sampling, and the ability of the different networks to predict outcomes successfully was evaluated using receiver operating characteristic (ROC) curve analysis.

RESULTS

Four of the networks showed perfect discrimination, with an area under the ROC curve (Az) of 1.0. Two other networks also had excellent performance, with an Az of 0.95. The sensitivity and specificity of the combined networks was 60% and 100%, respectively, when using an output neuron activation of 0.6 as the cutoff point to decide class membership. Lowering the cutoff point to 0.14 increased the sensitivity to 77%, and lowered the specificity to 96%.

CONCLUSIONS

These results are encouraging, especially when compared to the performance of more traditional multivariate models on the same data set. The robustness of neural networks, when confronted with noisy data generated by nonlinear processes, and their freedom from a priori assumptions regarding the data, make them promising tools with which to develop predictive clinical models.

Xenotransplantation rejection is antibody-mediated in both sensitized and nonsensitized recipients

This study analyzes the mechanisms involved in xenotransplantation rejection between closely related species. Hamster hearts were transplanted heterotopically into both normal rats and rats previously sensitized by the transfusion of donor blood. Sequential ultrastructural and immunohistochemical analyses were performed on the grafts, spleens, and sera. The data obtained support the view that induced antibodies directed against the xenograft endothelium play a very important role in producing graft damage. Moreover, the demonstration of antibodies against myocyte determinants suggests that it is possible, in this particular model, that the antiendothelial antibodies are not the only ones involved in the injury process.

The latest advances in liver transplantation at the pittsburgh transplantation institute: evolution of FK506, liver-intestinal transplantation, clinical xenotransplantation, and the induction of graft acceptance

During the past 30 years orthotopic liver transplantation (OLTx) has become a highly successful form of therapy, and as of this writing it is being performed at more than 100 institutions in the U.S., and a similar number in Europe. This is testimony to the great advances achieved in this field since the 1960s and 1970s, when there were essentially only two places actively engaged in liver transplantation. Essential to its success have been the technical refinements introduced during the last three decades, which have allowed many surgeons around the world to be able to do the procedure safely. Liver transplantation is still considered as one of the most complex operations, and therefore the margin of error is small and attention to technical detail is crucial to a satisfactory outcome. This is magnified in importance since OLTx, unlike kidney, heart, pancreas and intestinal transplantation, lacks a back-up system, such as dialysis, ventricular assist device, insulin or total parenteral nutrition. Thus, the smallest mistake in the surgical management of the patient may prove fatal.

Liver xenotransplantation

During the past 30 years orthotopic liver transplantation has become a highly successful form of surgical treatments. The significant advances achieved in this field have led to an increased demand for organs and created a wide gap between organ availability and organ supply. A wider availability of organs for transplantation would allow an expansions rather than a contraction of the indications for transplantation, and, at the same time a relaxation of the patient selection criteria. All these facts clearly justify the renewed interest observed in the last decade in xenotransplantation. The original concept of xenografting, meaning the transplantation of cells, tissues, or organs between different species, is so ancient that it is easily recognizable in Greek and Roman mythology. The centaur Chiron, the teacher of Esculapius, and the Chimera are legendary examples of discordant xenogeneic creatures. However, it is only during this century that scientists have been able to bring this idea into the clinical arena. The early efforts were prompted by the shortage of humans organs at a time when there were few alternatives for treating end-stage organ failure.

Adult respiratory distress syndrome secondary to end-stage liver disease-successful outcome following liver transplantation

The adult respiratory distress syndrome (ARDS) complicating liver failure carries a 100% mortality. Two cases of ARDS that resolved following liver transplantation have been reported, one associated with acute allograft rejection, and the second due to sepsis. There is, however, a great reluctance to transplant these very-high-risk patients. We report the first series of patients with ARDS secondary to liver failure who successfully underwent OLTX. No patient had sepsis or pneumonia. Posttransplant mechanical ventilation was required for a median of 14 days (range 6-37 days). All patients in this series are alive and well, with a follow-up of 6-15 months. This demonstrates that ARDS associated with liver failure, an otherwise uniformly lethal complication, can respond dramatically to OLTX.

Orthotopic liver transplantation for life-threatening veno-occlusive disease of the liver after allogeneic bone marrow transplant

Veno-occlusive disease (VOD) of the liver is a serious and often lethal sequela to bone marrow transplantation. Although a history of prior hepatitis moderately increases the risk of VOD, reliable screening methods for identifying high risk patients are not available. New approaches to managing patients who develop serious VOD are needed. One approach may be the use of orthotopic liver transplantation in selected patients who are likely to die of the disease. In this report we describe a patient who underwent liver transplantation for life-threatening VOD following allogeneic transplantation for CML. Although this patient died early from interstitial pneumonitis, the orthotopic liver functioned well up to her death. Other reports describing successful liver transplants in patients with advanced VOD or graft-versus-host disease of the liver are discussed and the possible indications for liver transplantation for VOD after marrow transplantation are considered. Taken together, these reports suggest that orthotopic liver transplantation may be a feasible and potentially effective approach to managing select patients with life-threatening liver dysfunction after marrow transplantation.

The mechanism of the inhibitory action of neurotensin on pentagastrin-stimulated gastric secretion in dogs

Neurotensin is a potent inhibitor of pentagastrin-stimulated gastric acid secretion. This study was done to investigate the possible role of vagal innervation and of prostaglandins on this inhibitory effect. Five dogs with gastric cannulas were infused with pentagastrin (1 microgram per kilogram of body weight each hour) for 210 minutes. In the 60 to 150 minute period, neurotensin (5.5 micrograms per kilogram of body weight each hour) was infused. Neurotensin significantly decreased pentagastrin-stimulated gastric acid output, while the simultaneous administration of indomethacin (bolus of 1 milligram per kilogram of body weight plus infusion of 0.5 milligrams per kilogram each hour) abolished the effect of neurotensin. After truncal vagotomy, the inhibitory effect of neurotensin was again studied. Since the sensitivity of the stomach to pentagastrin decreased after vagotomy, the dose of pentagastrin was doubled (2 micrograms per kilogram of body weight each hour) in order to achieve acid stimulation comparable with the levels before vagotomy. After vagotomy, neurotensin inhibited the effect of pentagastrin in a manner similar to that shown before vagotomy. The inhibitory potency of neurotensin on gastrin-stimulated gastric secretion is independent of the vagus, but requires intact synthesis for prostaglandins.