Statistical grand rounds: design and analysis of studies with binary- event composite endpoints: guidelines for anesthesia research

Composite endpoints consisting of several binary events, such as distinct perioperative complications, are frequently chosen as the primary outcome in anesthesia studies (and in many other clinical specialties) because (1) no single outcome fully characterizes the disease or outcome of interest, and/or (2) individual outcomes are rare and statistical power would be inadequate for any single one. Interpreting a composite endpoint is challenging because components rarely meet the ideal criteria of having comparable clinical importance, frequency, and treatment effects. We suggest guidelines for forming composite endpoints and show advantages of newer versus conventional statistical methods for analyzing them. Components should be a parsimonious set of outcomes, which when taken together, well represent the disease of interest and are very plausibly related to the intervention. Adding components that are too narrow, redundant, or minimally influenced by the study intervention compromises interpretation of results and reduces power. We show that multivariate (i.e., multiple outcomes per patient) methods of analyzing a binary-event composite provide distinct advantages over standard methods such as any-versus-none, count of events, or evaluation of individual events. Multivariate methods can incorporate clinical importance weights, compensate for events occurring at varying frequencies, assess treatment effect heterogeneity, and are often more powerful than alternative statistical approaches. Methods are illustrated with an American College of Surgeons National Surgical Quality Improvement Program registry study that evaluated the effects of smoking on major perioperative outcomes, and with a clinical trial comparing the effects of crystalloids and colloids on major complications. Sample data files and SAS code are included for convenience.     

A stepwise logistic regression analysis of factors affecting morbidity and mortality after thoracic trauma: effect of epidural analgesia

Rib fractures and other chest wall injuries can lead to weak ventilation, atelectasis, and even death. Whereas such injuries in young patients are usually well tolerated, relatively minor chest wall trauma can be serious in elderly patients. Epidural analgesia, by improving pain control and ventilatory function, might improve morbidity and mortality rates compared to other forms of analgesia. Stepwise logistic regression was used to compare thoracic trauma patients more than 60 years of age treated with either epidural or parenteral (IV/IM) analgesia. In spite of more severe thoracic trauma in epidural patients as measured by the Abbreviated Injury Score for the chest (epidural = 3.3 +/- 0.1, IV/IM = 2.8 +/- 0.1; p less than 0.05) the use of epidural analgesia was an independent predictor of both decreased mortality (p = 0.0035) and a decreased incidence of pulmonary complications (p = 0.0088). Epidural analgesia has a positive effect on outcome in elderly trauma victims with chest wall injury and is useful in high-risk patients. Increased costs associated with epidural analgesia are minimal and are justified by improvements in outcome.     

Determinants of perioperative morbidity and mortality after pneumonectomy

A total of 197 consecutive patients undergoing pneumonectomy at the M.D. Anderson Cancer Center from 1982 to 1987 were reviewed. Sixty-five variables were analyzed for the predictive value for perioperative risk. The operative mortality rate was 7% (14/197). Patients having a right pneumonectomy (n = 95) had a higher operative mortality rate (12%) than patients having a left pneumonectomy (1%, p less than 0.05). The extent of resection correlated with the operative mortality rate (chest wall resection or extrapleural pneumonectomy, n = 39, 15%; versus simple or intrapericardial pneumonectomy, n = 158, 5%; p less than 0.05). Patients whose predicted postoperative pulmonary function, by spirometry and xenon 133 regional pulmonary function studies, was a forced expiratory volume in 1 second greater than 1.65 L, forced expiratory volume in 1 second greater than 58% of the preoperative value, forced vital capacity greater than 2.5 L, or forced vital capacity greater than 60% of the preoperative value had a lower operative mortality rate (p less than 0.05). Atrial arrhythmia was the most common postoperative complication (23%). Xenon 133 regional pulmonary function studies are useful in predicting the risks of pneumonectomy.     

Lung transplantation for pulmonary fibrosis: perioperative morbidity and mortality

Lung transplantation has become a valid therapeutic option for patients with pulmonary fibrosis in terminal stage, and the number of such interventions has increased exponentially in recent years. We undertook a retrospective study of 46 pulmonary fibrosis patients who received lung transplants from 1992 through 2002 with the aim of describing the most common intra- and early postoperative complications in the recovery unit. We also aimed to analyze the impact of each complication on mortality during the study period. The most frequent complications during surgery were reperfusion syndrome (47.5%), hemodynamic instability (41%), arrhythmias (23.9%), and pulmonary hypertension (15%), with exitus secondary to reperfusion syndrome. The most common postoperative complications were infection (56.5%), reimplantation response (45.7%), and kidney failure (19.6%). Overall mortality during the study period was 23.9%. The following complications were statistically significant in the univariate analysis of the relation with mortality: reperfusion syndrome (p=0.039), reimplantation response (p=0.039), kidney failure (p=0.013), rejection (p=0.016), and sepsis (p<0.001). The only complication that remained significant in the multivariate analysis was sepsis (p<0.001). In spite of the considerable progress made, intra- and postoperative complications continue to be a real threat for the transplanted lung patient. Sepsis was the strongest predictor of poor prognosis in the early recovery period.     

Pediatric anesthesia morbidity and mortality in the perioperative period

One of the most frequent questions asked of a pediatric anesthesiologist is "What are the risks of anesthesia for my child?" Unfortunately, few studies have examined the consequences of general anesthesia in children. We used data from a large pediatric anesthesia follow-up program at Winnipeg Children's Hospital (1982-1987) to determine rates of perioperative adverse events among children of different ages. A check-off form was completed by a pediatric anesthesiologist for each case (n = 29,220) and a designated follow-up reviewer examined all anesthesia forms and hospital charts to ascertain adverse effects for children less than 1 mo, 1-12 mo, 1-5 yr, 6-10 yr, and 11-16 yr of age in the intraoperative, recovery room, and postoperative periods. The majority of the children were healthy, and 70% had no preoperative medical conditions. Infants less than 1 mo old were more likely to be undergoing major cardiac or vascular procedures, whereas the older children had mainly orthopedic or otolaryngologic procedures. Infants less than 1 mo old had the highest rate of adverse events both intraoperatively and in the recovery room. The main problem in this age group was related to the respiratory and cardiovascular systems. In children over 5 yr of age, postoperative nausea and vomiting was very frequent, with about one-third of the children experiencing this problem. When all events were considered (both major and minor), there was a risk of an adverse event in 35% of the pediatric cases. This contrasts with 17% for adults. This morbidity survey helps to focus on areas of intervention and for further study.     

Reliability of pharmacodynamic analysis by logistic regression: mixed-effects modeling

BACKGROUND: Many pharmacologic studies record data as binary, yes-or-no, variables with analysis using logistic regression. In a previous study, it was shown that estimates of C50, the drug concentration associated with a 50% probability of drug effect, were unbiased, whereas estimates of gamma, the term describing the steepness of the concentration-effect relationship, were biased when sparse data were naively pooled for analysis. In this study, it was determined whether mixed-effects analysis improved the accuracy of parameter estimation. METHODS: Pharmacodynamic studies with binary, yes-or-no, responses were simulated and analyzed with NONMEM. The bias and coefficient of variation of C50 and gamma estimates were determined as a function of numbers of patients in the simulated study, the number of simulated data points per patient, and the "true" value of gamma. In addition, 100 sparse binary human data sets were generated from an evaluation of midazolam for postoperative sedation of adult patients undergoing cardiac surgery by random selection of a single data point (sedation score vs. midazolam plasma concentration) from each of the 30 patients in the study. C50 and gamma were estimated for each of these data sets by using NONMEM and were compared with the estimates from the complete data set of 656 observations. RESULTS: Estimates of C50 were unbiased, even for sparse data (one data point per patient) with coefficients of variation of 30-50%. Estimates of gamma were highly biased for sparse data for all values of gamma greater than 1, and the value of gamma was overestimated. Unbiased estimation of gamma required 10 data points per patient. The coefficient of variation of gamma estimates was greater than that of the C50 estimates. Clinical data for sedation with midazolam confirmed the simulation results, showing an overestimate of gamma with sparse data. CONCLUSION: Although accurate estimations of C50 from sparse binary data are possible, estimates of gamma are biased. Data with 10 or more observations per patient is necessary for accurate estimations of gamma.     

Predictors of perioperative morbidity and mortality in lung volume reduction surgery

Background. Selection criteria for lung volume reduction surgery are still being refined. We sought to determine whether preoperative features could be used to predict early morbidity or mortality.

Methods. We reviewed preoperative characteristics of the first 89 patients who underwent lung volume reduction surgery at the Alfred Hospital. Data included arterial blood gases, prednisolone use, pulmonary function tests, 6-minute walk test, and anesthetic time. Length of stay and reintubation for respiratory failure were used as markers of morbidity.

Results. Findings included Paco₂ of 43 ± 0.7 mm Hg, Pao₂ 70 ± 1.1 mm Hg, percent predicted values for forced expiratory volume in 1 second 29.6% ± 0.8%, TLCO% predicted 35.2 ± 1.4%, and 6-minute walk test of 315 ± 10.6 m (mean ± SEM). Mean length of stay was 19 ± 2 days, with 17 (19%) patients reintubated for respiratory failure. Mortality rate was 5.6% at 1 year post surgery, with no deaths in patients less than 65 years old. Multivariate analysis revealed that length of stay, reintubation and mortality were predicted by age and surgical time (p < 0.05), with no correlation with any other variables tested. Age greater than 70 years was associated with a significant risk of mortality (OR 9.0; p = 0.04).

Conclusions. Age greater than 70 years and anesthetic time greater than 210 minutes predict both perioperative morbidity and mortality.     

Revisional bariatric surgery: perioperative morbidity is determined by type of procedure

Background

Revisional bariatric procedures are on the rise. The higher complexity of these procedures has been reported to lead to increased risk of complications. The objective of our study was to compare the perioperative risk profile of revisional bariatric surgery with primary bariatric surgery in our experience.

Methods

A prospectively maintained database of all patients undergoing bariatric surgery by three fellowship-trained bariatric surgeons from June 2005 to January 2013 at a center of excellence was reviewed. Patient demographics, type of initial and revisional operation, number of prior gastric surgeries, indications for revision, postoperative morbidity and mortality, length of stay, 30-day readmissions, and reoperations were recorded. These outcomes were compared between revisional and primary procedures by the Mann–Whitney or Chi square tests.

Results

Of 1,556 patients undergoing bariatric surgery, 102 patients (6.5 %) underwent revisional procedures during the study period. Indications for revisions included inadequate weight loss in 67, failed fundoplications with recurrent gastroesophageal reflux disease in 29, and other in 6 cases. Revisional bariatric procedures belonged into four categories: band to sleeve gastrectomy (n = 23), band to Roux-en-Y gastric bypass (n = 25), fundoplication to bypass (n = 29), and other (n = 25). Revisional procedures were associated with higher rates of readmissions and overall morbidity but no differences in leak rates and mortality compared with primary procedures. Band revisions had similar length of stay with primary procedures and had fewer complications compared with other revisions. Patients undergoing fundoplication to bypass revisions were older, had a higher number of prior gastric procedures, and the highest morbidity (40 %) and reoperation (20 %) rates.

Conclusions

In experienced hands, many revisional bariatric procedures can be accomplished safely, with excellent perioperative outcomes that are similar to primary procedures. As the complexity of the revisional procedure and number of prior surgeries increases, however, so does the perioperative morbidity, with fundoplication revisions to gastric bypass representing the highest risk group.     

Influence of operation duration on perioperative morbidity in revision total hip arthroplasty

This retrospective study investigated the impact of patient and procedure-related parameters on the complication rate following revision total hip arthroplasty. Complications included vessel and nerve damage, periprosthetic femoral fracture, wound infection, wound bleeding, prosthesis dislocations, thromboembolism, cardiac and pulmonary complications, and death. The influence of operation duration, gender, revision status, ASA classification, and type of fixation of the primary implant on the perioperative morbidity was investigated in a sample of 60 revision procedures (cemented stems, cemented or cementless cups). Odds ratio [OR] and 95% confidence interval [CI] were estimated with multiple regression models. Perioperative morbidity was significantly correlated to operation duration (OR = 1.03; CI: 1.00-1.05), but not to age (OR = 1.01; CI: 0.93-1.09), gender (OR = 2.66; CI: 0.50-14.05), revision status (OR = 2.34; CI: 0.54-10.05), ASA classification (OR = 1.24; CI: 0.30-5.18), or type of fixation of the primary implant (OR = 2.49; CI: 0.47-13.17) Duration of the revision operation appeared as a predictive parameter for perioperative morbidity in revision total hip arthroplasty in our study group.     

Importance of dialysis adequacy in mortality and morbidity of chinese CAPD patients

BACKGROUND: In continuous ambulatory peritoneal dialysis (CAPD), the impact of dialysis adequacy on patient outcome is well established in Caucasian patients but is less clear in Asian patients. Recent evidence suggests that Asian dialysis patients enjoy better overall survival. We hypothesize that dialysis adequacy may be less important in determining outcome for this ethnic group. METHODS: We performed a single-center prospective observational study. From September 1995, we enrolled 150 existing and 120 new CAPD patients. They were followed for up to three years. We monitored dialysis adequacy and nutritional indices, including Kt/V, weekly creatinine clearance (CCr), residual glomerular filtration rate (GFR), normalized protein catabolic rate (NPCR), percentage of lean body mass (%LBM), and plasma albumin level. Clinical outcomes included mortality, technique failure, and duration of hospitalization. RESULTS: The duration of study follow-up was 22.1 +/- 12.3 months. In our study population, 136 were male. Seventy were diabetic (25.9%), and 212 were treated with 6 L exchanges per day (78.5%). The body weight was 59.3 +/- 9.4 kg. Baseline total Kt/V was 1.78 +/- 0.41, peritoneal Kt/V 1.48 +/- 0.36, and median residual GFR 0.98 mL/min (range 0 to 7.45). Two-year patient survival was 83.0%, and technique survival was 72.8%. Multivariate analysis showed that the duration of dialysis, diabetes, %LBM, index of dialysis adequacy (Kt/V or CCr), residual GFR, and requirement of a helper for CAPD exchanges were independent factors of patient survival; serum albumin, adequacy index (Kt/V or CCr), and requirement of a helper were independent factors of technique survival. Duration of dialysis, body weight, requirement of helper, cardiovascular disease, HBsAg carrier, serum albumin, and CCr had independent effects on hospitalization. The peritoneal component of Kt/V or CCr had no independent effect on any outcome parameter. When the prevalent and new CAPD cases were analyzed separately, Kt/V predicted survival only for new CAPD cases. CONCLUSIONS: Our results show that dialysis adequacy has significant impact on outcome of Asian CAPD patients. Although we have excellent medium-term patient and technique survival, this favorable outcome should not prevent health care workers from providing adequate dialysis to Asian patients. The reason of discrepancy in outcome between Asian and Caucasian dialysis patients requires further study.     

The effects of perioperative blood transfusion on morbidity and mortality after esophagectomy

The effect of blood transfusion on outcomes in esophageal surgery remains controversial. The contrasting conclusions drawn from a number of retrospective analyses with different methodologies create a landscape that is difficult to interpret. Because of the scope of esophageal resection, the need for blood transfusion cannot be eliminated. What recommendations then, if any, can be made for the practicing surgeon? First, surgeons and anesthesiologists need to reevaluate their transfusion thresholds. The age-old practice of keeping the hemoglobin above 10 g/dL has very little evidence-based support. A multicenter, randomized, controlled clinical trial in Canada demonstrated that a restrictive strategy of blood transfusion, in which patients were transfused only for a hemoglobin level of less than 7 g/dL, was at least as effective as and possibly was superior to a liberal transfusion strategy in critically ill patients. It has also been estimated that more than 25% of patients undergoing colorectal resections may receive at least one unit of unnecessary blood. Further, the immediate reduction in the hemoglobin concentration caused by the normovolemic hemodilution associated with surgery and crystalloid fluid replacement is not associated with any increased morbidity or mortality. If these data are examined in the context of the results of Langley and Tachibana indicating that a threshold amount of blood needs to be transfused to impact outcomes, it becomes even more important to limit transfusion to only the amount that is essential. Thus, surgeons and anesthesiologists should adopt a more stringent set of requirements for blood transfusion. Second, with the proven feasibility and reduction in infectious complications associated with autologous blood-donation programs, any patient who meets the criteria discussed here should be encouraged to participate in such a program. Although the effect of autologous blood on cancer outcomes remains unclear, the other advantages certainly make such a program worthy of consideration. This discussion leads to a final point, namely that patients should be encouraged, whenever possible, to participate in clinical trial research. The only way that the community of surgeons treating patients who have esophageal cancer can hope to address properly the question of how blood transfusion affects outcomes is with well-designed clinical trials. A large, multicenter, randomized trial (level I) would be ideal. Short of such a trial, inclusion criteria and study methodology should be discussed among various institutions to avoid the differences in studies that make direct comparisons of results among different investigators difficult and potentially meaningless. This measure would at least allow different level II to IV data to be compared directly with some validity.     

Prediction of mortality by logistic regression analysis in patients with postoperative enterocutaneous fistulae

The contrasting results of treatment of patients with postoperative enterocutaneous fistulae reflect the heterogeneity of the disease and depend on the patient's condition and the characteristics of the fistulae. For this reason, the use of a prognostic index, which enables such patients to be classified according to their risk of death, could be useful. In this study we propose a prognostic index based on a logistic regression analysis, obtained by using two (APACHE II score and serum albumin concentration) of the eight risk factors that have been retrospectively analysed in a series of 70 patients with postoperative enterocutaneous fistulae treated in our surgical department since 1981. The logistic regression equation indicates that patients with a probability of dying of less than 0.35 have a good prognosis, with a sensitivity of 90 per cent, a specificity of 90 per cent, a negative predictive value of 79 per cent, a positive predictive value of 96 per cent and an accuracy of 90 per cent. The predictive performance of the index has also been evaluated in a group of 17 patients studied prospectively, and this confirms the sensitivity and specificity of the model. This postoperative enterocutaneous fistulae index could be a helpful tool in clinical trials and surgical audit.     

The influence of obesity on perioperative morbidity and mortality in revision total hip arthroplasty

The significance of obesity as a risk factor for postoperative complications was determined in a consecutive series of 229 cases of revision total hip replacement. The body mass index (BMI) was used as an objective measure to classify the patients. The group-wise analysis of data included all medical and procedure-related complications, the number of fatal cases, operative time, requirement for analgesics, the number of transfusions and perioperative haemoglobin levels. The results of our study demonstrate a clear association between obesity and operative time, whereas no statistically significant relationships were observed between obesity and the other parameters. We conclude that obesity does not have any significant influence on perioperative morbidity and mortality but is clearly related to operation time and, therefore, to higher costs per operation. Received: 19 April 1999     

Impact of journal articles and grand rounds on practice: CT scanning in appendicitis

The purpose of this study was to determine whether a report in a high-impact journal published in January 1998 changed practice patterns and to further explore the impact of a review of the subject in a department of surgery grand rounds (January 2000). Charts from all patients undergoing appendectomy at our institution during three time periods (January to December 1997, January to December 1999, and January to June 2000) were reviewed. Rates of CT scanning, negative appendectomy, and perforated/ gangrenous appendicitis were compared for the three periods to determine the impact of the journal article and the subject review during grand rounds on practice patterns and outcomes. Charts from 230 (88%) of 262 patients who underwent appendectomy during the time periods were available for review. Age, percentage of male patients, temperature on admission, and white blood cell count did not differ among the groups. The rate of CT scanning increased significantly from 1997 to 1999 and again in 2000 (6.7%, 43%, and 70%, respectively; P < 0.001), whereas the proportion of perforated/gangrenous appendicitis decreased significantly from 33% in 1997 and 31% in 1999 to 13% in 2000 (P = 0.012). The use of CT scanning in appendicitis increased both after publication of a report in a high-impact journal and after review during grand rounds. A rate of CT scanning above 45 % appeared to affect outcomes as well. Presented at the Forty-Second Annual Meeting of The Society for Surgery of the Alimentary Tract, Atlanta, Georgia, May 20–23, 2001 (poster presentation).     

Statistical methods to predict morbidity and mortality: self assessment techniques for burn units

The medical records of 806 adult and paediatric burn patients were retrospectively reviewed. Patient data was summarized, coded and entered in a computer for subsequent analysis.This report describes the use of multiple regression analysis to produce equations useful for the prediction of the morbidity parameters: length of hospital stay, number of transfusions required and number of operative procedures. The multiple regression equations developed are useful as prediction tools, in patient medical audit and in assessing improvements in burn care.The classical technique of probit analysis for predicting the probability of mortality was used to develop the LA₅₀'s of burn injury. These LA₅₀'s were used as the basis for comparison of survival statistics between burned patients in this series and those registered in the National Burn Information Exchange (Feller, 1979). There is an apparent significantly improved survival in the children, young adults through 34 yrs and the older adults (60–74 yrs) in the current series of patients. This observation would support the adoption of the protocol of burn care used in treating these patients.The newer technique of discriminant analysis is also described. Discriminant analysis is a multifactorial method for discriminating between dichotomous outcomes (survivors and non-survivors). The technique of discriminant scoring proved to be 95.8 per cent accurate in predicting burn survival. Use of a burn severity scoring technique will also assist in recognition of the high risk patient.     

Continent urinary reconstruction versus ileal conduit: a contemporary single-institution comparison of perioperative morbidity and mortality

OBJECTIVES: To compare postoperative morbidity and mortality in a concurrent and contemporary series of patients who underwent radical cystectomy with ileal conduit versus orthotopic neobladder. METHODS: The data of 198 patients were reviewed, 117 with orthotopic reconstruction and 81 with ileal conduit during a 5-year time frame. Thirty-day morbidity, mortality, reoperative rates, and parameters associated with the surgical procedures were obtained from chart review. RESULTS: No perioperative or postoperative deaths occurred in either group. The median operative time for the ileal conduit was 201 minutes (range 140 to 373), and for the orthotopic neobladder, it was 270 minutes (range 230 to 425). The median blood loss was 389 and 474 mL, respectively. The median length of hospitalization was 8 days for the ileal conduit group and 7 days for the orthotopic neobladder group. Diversion-related complications recognized within 30 days that ultimately required a return to the operating room occurred in 3.4% of those with a neobladder and 1.2% of those with an ileal conduit. CONCLUSIONS: The orthotopic neobladder is a longer and technically more complex procedure than the ileal conduit procedure. However, no demonstrable difference in morbidity or perioperative complications were found between the two procedures in our review.