The Development of Aortic Insufficiency in Continuous-Flow Left Ventricular Assist Device Supported Patients

Transcription

1 The Development of Aortic Insufficiency in Continuous-Flow Left Ventricular Assist Device Supported Patients Ashim Aggarwal, MD, MRCP, Rashmi Raghuvir, MD, Paula Eryazici, MD, Gregory Macaluso, MD, Priya Sharma, BA, Christopher Blair, BS, CCRP, Antone J. Tatooles, MD, Pat S. Pappas, MD, and Geetha Bhat, PhD, MD Center for Heart Transplant and Assist Devices and Department of Cardiothoracic Surgery, Advocate Christ Medical Center, Oak Lawn, Illinois Background. Significant aortic insufficiency (AI) after left ventricular assist device (LVAD) placement affects device performance and end-organ perfusion. This study examined the development and progression of AI after implantation of continuous-flow LVAD. Methods. Seventy-nine patients undergoing Heart Mate II (Thoratec Corp, Pleasanton, CA) LVAD implantation for predominantly destination therapy (n 69 [87%]) were examined. Preoperative and postoperative echocardiograms for all patients were reviewed at the intervals of 0 to 3, 3 to 6, 6 to 12, 12 to 18, and 18 to 24 months. AI was graded on an interval scale of 0, none; 0.5, trivial; 1, mild; 1.5, mild to moderate; 2, moderate; 2.5, moderate to severe; and 3, severe. Development and progression of AI were analyzed. Results. The incidence of significant AI (mild or greater) was 52% (n 41). Median time to AI development was 187 days. The median duration of VAD support was 761 days. Mild AI developed in 41 patients (52%). No severe AI developed. In the Cox regression model (hazard ratio [95% confidence interval]), aortic valve closure (2.51 [1.06 to 5.89]; p 0.03), and age (1.04 [1.008 to 1.08]; p 0.01) were independent predictors of AI development. There was no difference in mortality rates in the two groups (p 0.40 by log-rank test). A mixed-model linear regression analysis showed a significant overall progression of AI over time ( standard error, ; p 0.006). Conclusions. AI develops over time in a significant number of Heart Mate II LVAD patients. AI is more common in patients with closed aortic valves and in the older age group. As more patients require long-term VAD support, the development of AI will need careful attention and monitoring. (Ann Thorac Surg 2013;95:493 9) 2013 by The Society of Thoracic Surgeons Heart failure currently affects 5.8 million Americans, with 670,000 new cases diagnosed annually and an average of 280,000 deaths per year [1]. The lifetime risk of developing heart failure has reached the epidemic proportion of 1 in 5 for both men and women by age 40 years [1]. When medical therapy fails to alleviate advanced end-stage heart failure, surgical alternatives, such as heart transplantation and mechanical circulatory support (MCS) devices, are viable options. Heart transplantation has been the preferred choice, but due to constraints, such as donor availability and strict eligibility criteria, this is a limited treatment option [2]. MCS devices for acute or chronic heart failure are used as a bridge to transplantation (BTT) or as long-term use as destination therapy (DT). The use of MCS has been growing due to enhanced survival and quality of life ratings. A potential obstacle that has arisen for successful long-term left ventricular assist device (LVAD) support is Accepted for publication Sept 7, Address correspondence to Dr Bhat, Advocate Christ Medical Center, 4400 W 95th St, Oak Lawn, IL 60453; geetha.bhat@advocatehealth. com. the native heart s ability to withstand the hemodynamic and ultrastructural fluctuations that are induced by prolonged MCS. One such unanticipated complication is the development of de novo aortic valve (AoV) lesions, which can lead to commissural fusion, stenosis, and aortic insufficiency (AI) [3, 4]. Significant AI can lead to ineffective LVAD output and end-organ malperfusion due to reduced effective forward flow. Studies have looked at the long-term incidence and the factors that possibly help us predict and prognosticate the future development of AI in patients on LVADs. However, these studies have mostly included patients with LVADs implanted as BTT [5, 6]. With limited donor availability and better LVAD technologies, more patients will be implanted with LVADs as DT. Longer durations of device support might lead to increased severity of AI with hemodynamic consequences that might affect long-term survival. This study examined the temporal trend of AI after Dr Tatooles discloses a financial relationship with Thoratec, Inc by The Society of Thoracic Surgeons /$36.00 Published by Elsevier Inc

2 494 AGGARWAL ET AL Ann Thorac Surg AI IN LVAD SUPPORTED PATIENTS 2013;95:493 9 Abbreviations and Acronyms AI aortic insufficiency AoV aortic valve BTT bridge to transplantation CI confidence interval DT destination therapy HM II Heart Mate II HR hazards ratio IABP intraaortic balloon pump LV left ventricle LVAD left ventricular assist device LVEDD left ventricular end diastolic diameter LVEF left ventricular ejection fraction LVESD left ventricular end systolic diameter MAP mean arterial pressure MCS mechanical circulatory support OR odds ratio RPM rotations per minute SD standard deviation SE standard error VAD ventricular assist device implantation of continuous-flow LVAD devices in patients implanted predominantly under the DT strategy to identify correlates of AI development and progression. Material and Methods This study was approved by the hospital Institutional Review Board and individual patient consent was waived. Patients We identified 154 patients who were implanted with Heart Mate II (HM II) LVADs (Thoratec Corp, Pleasanton, CA) between January 2005 and January The study excluded 75 patients: 48 with mild or greater AI before LVAD implant, 3 with previous AoV replacement, and 24 with a follow-up of less than 6 months. The final study population included 79 patients with continuous-flow HM II devices; of these, 74 (93.6%) had no AI and 5 (6.3%) had trivial AI on preoperative echocardiograms. Transthoracic echocardiograms from the 79 patients with an HM II LVAD implanted between January 2005 and March 2011 were retrospectively reviewed. Studies performed within 2 months preceding device placement were deemed baseline. Subsequent studies were categorized in the postoperative time intervals of 0 to 3, 3 to 6, 6 to 12, 12 to 18, and 18 to 24 months. AI Assessment Echocardiograms were performed according to American Society of Echocardiography guidelines and were reviewed by a single reader in a nonblinded manner. Three-beat image capture was used. AI was evaluated visually in the parasternal short-axis and long-axis views and was graded on an interval scale of absence of AI, 0; trivial, 0.5; trivial to mild, 0.75; mild, 1.0; mild to moderate, 1.5; moderate, 2.0; moderate to severe, 2.5; and severe, 3.0. The presence of AoV opening was evaluated visually and with M-mode imaging at each follow-up and was graded as full opening, intermittent opening (defined as 1 to 2 openings in 3 systoles), or full closure during 3 LV systoles. The AoV opening was timed with the onset of QRS complex signifying the onset of ventricular systole. Statistical Analysis Statistical analysis was performed using SPSS 19 software (SPSS Inc, Chicago, IL). Continuous variables are expressed as mean and standard deviation and were compared using the Student t test. The Wilcoxon rank sum test was used to determine differences in nonnormally distributed data. Categoric variables, expressed as number of patients and percentage, were compared using the 2 test or the Fisher exact test. Values of p less than 0.05 were considered significant. Univariate and Cox proportional hazards regression models were performed to identify risk factors for AI at any given time point. Kaplan-Meier survival curves were plotted for freedom from the development of AI and LVAD support. The effect of baseline clinical characteristics and baseline echocardiogram measurements on AI progression after LVAD implantation was evaluated using a mixed-effect linear regression model with restricted maximum likelihood estimates and compound symmetry covariance structure. In this model, time was treated as a continuous variable; an estimate of time variable interaction term was used to evaluate the change in AI severity over time after LVAD implantation for the presence or absence of a categoric variable or per unit measure of a continuous variable [5]. Results Entire Patient Population The analysis included 79 patients with HMII devices. Of these, 69 (87%) received the device for DT and 10 (13%) for BTT. The cohort was a mean age of years with a median duration of LVAD support of 761 days (range, 145 to 2434 days). Most patients were men (68 [85%]) and had ischemic cardiomyopathy (47 [59%]). The mean LV ejection fraction was Mild or greater AI developed in 41 of the 79 patients (52%) at a median of 187 days after LVAD implantation. Of these 41 patients, 5 (12%) progressed to mild to moderate AI and 4 (10%) progressed to moderate AI. No severe AI developed. AI vs No AI Table 1 lists the baseline characteristics comparing the patients with and without AI. The two groups were similar with respect to the demographic characteristics (sex, race, body mass index, and body surface area), except the age; patients with AI tended more often to be significantly older than those without AI ( vs

3 Ann Thorac Surg AGGARWAL ET AL 2013;95:493 9 AI IN LVAD SUPPORTED PATIENTS 495 Table 1. Baseline Characteristics Variables a Aortic Insufficiency Absent 38 (48) Present 41 (52) p Value Demographics Age, years Male sex 33 (86.6) 34 (82.9) 0.62 Race Caucasian 15 (39.5) 19 (46.3) African American 20 (52.6) 13 (31.7) 0.08 Body surface area, m Body mass index, kg/m MAP, b mm Hg Device strategy Destination therapy 31 (81.6) 38 (95.0) 0.06 Comorbidities Ischemic cardiomyopathy 19 (50.0) 27 (65.9) 0.15 Hypertension 26 (68.4) 25 (61.0) 0.48 Diabetes mellitus 16 (42.1) 17 (41.5) 0.95 Hyperlipidemia 15 (39.5) 15 (36.6) 0.79 IABP support 6 (15.8) 10 (24.4) 0.34 Echo variables LVEF Aortic root baseline, cm Aortic root change, cm LVESD, cm LVEDD, cm Aortic valve opening Closed 15 (39.5) 34 (82.9) Intermittent 3 (7.9) 3 (7.3) Open 20 (52.6) 4 (9.8) VAD support, months a Categoric variables are shown as number (%) and continuous variables as mean standard deviation. b MAP as measured within the first week postoperatively. IABP intraaortic balloon pump; LVEDD left ventricular end diastolic diameter; LVEF left ventricular ejection fraction; LVESD left ventricular end systolic diameter; MAP mean arterial pressure; VAD ventricular assist device years; p ). There was no significant difference in the comorbidities (hypertension, diabetes, hyperlipidemia) between the groups. The groups with and without AI did not differ in the duration of intraaortic balloon pump support ( vs days; p 0.69). Echocardiographic Variables On echocardiographic variables, the groups did not differ significantly in baseline aortic root diameter or change in aortic root size. There was a statistically significant difference in the AoV opening between the groups: patients with AI had predominantly closed AoVs (n 34 [68%]) compared to patients with no AI and predominantly opened AoVs (n 20 [48.8%]; p ). Post hoc analysis revealed a significant difference between closed and open groups (p ) and no difference amongst closed vs intermittent or open vs intermittent groups. Kaplan-Meier curves showed significant freedom from AI in the patients with open intermittent valve opening (p by log-rank test; Fig 1A). There were no significant differences in the LV end-systolic diameter or LV end-diastolic diameter in the groups. Patients with AI had a significantly greater duration of LVAD support ( vs months; p 0.01). Outcomes There were 23 deaths in the entire cohort. The causes of death were sudden death and intracranial bleed in 4 patients each; ischemic stroke, sepsis, gastrointestinal bleed, and multisystem organ failure in 2 patients each; and unknown in 7. Mortality was increased in the patients with AI (16 [39%] vs 7 [18%]; p 0.03). Logistic regression models were created to test the association of age and AI with death. Unadjusted, age and AI development were both associated with death (Table 1). However, the regression model showed AI was a significant (although only marginal) predictor of death (odds ratio, 3.14; 95% confidence interval [CI], to 9.87; p 0.05) after adjusting for age. The Kaplan-Meier survival curves

4 496 AGGARWAL ET AL Ann Thorac Surg AI IN LVAD SUPPORTED PATIENTS 2013;95:493 9 required in 11 patients (14%) for malfunctioning in 6 (54.5%), infection in 3 (27.7%), and thrombosis in 2 (18%). None of the patients had LVAD replacement as a direct consequence of AI. One patient (1.3%) underwent LVAD explantation after complete recovery of LV function. Predictors of AI Development Cox regression analysis was done using variables with p of less than 0.1 in the univariate analysis, which were age, race, body surface area, and AoV closure. Patients with open and intermittent opening were collapsed into one group for Cox regression analysis due to small numbers in the intermittent group and lack of significance in the univariate model. AoV closure (hazards ratio [HR], 2.51; 95% CI, 1.06 to 5.89; p 0.03) and age (HR, 1.04; 95% CI, to 1.08; p 0.01) were independent predictors of AI development at any given point in time. Progression of AI AI in our study was seen to progress over time ( standard error: ; p 0.006). Correlates of worsening AI over time are reported in Table 3. None of the clinical or echocardiographic variables predicted progression of AI in our study. Early vs Late AI Subgroup analysis was done to look for any possible differences accounting for development of early ( 180 days after LVAD implant, n 16) vs late ( 180 days after LVAD, n 25) AI. Increased baseline aortic root diameter ( vs cm; p 0.005) and female sex (6 [37.5%] vs 1 [4%]; p 0.009) favored early development of AI. However, a logistic regression model created using variables with p of less than 0.2 in univariate model showed baseline aortic root diameter was a marginal (p 0.05) independent predictor of early AI (OR, 11.6; 95% CI, 0.91 to ; p 0.05), whereas sex was not statistically significant (OR, 9.86; 95% CI, 0.89 to ; p 0.06). Fig 1. Kaplan-Meier curves demonstrate (A) freedom from aortic insufficiency (AI) in patients with open intermittent (blue line) and closed (green line) aortic valves (AoV) and (B) survival rates in patients with (, green line), and without (, blue line) AI after ventricular assist device (VAD) support. showed no difference of rate of survival by log-rank test (p 0.40; Fig 1B). Repeat hospitalizations for congestive heart failure were reviewed. There was no difference in patients with and without AI ( vs ; p 0.52). The mean speeds and the mean arterial pressures after LVAD at 3 and 6 months in outpatient settings were reviewed, and no difference was noted in the two groups (Table 2). The patients in AI group had increased LVAD speeds at 3 and 6 months after LVAD implantation, which could have possibly contributed to a greater frequency of AoV closure and thus increased predisposition to AI; however, this association is difficult to ascertain due to the retrospective nature of the study. LVAD replacement was Outcomes in Patients with Normal vs AoV Sclerosis There were 46 patients (58%) with AoV sclerosis and 33 (42%) with normal AoVs as detected on the transthoracic Table 2. Left Ventricular Assist Device Speed and Mean Arterial Pressure at 3 and 6 Months Variable Aortic Insufficiency Absent (Mean SD) Present (Mean SD) p Value At 3 months MAP, mm Hg VAD speed, 9,099 1,279 9,327 1, RPM At 6 months MAP, mm Hg VAD speed, RPM 9,134 1,296 9,233 1, MAP mean arterial pressure; RPM rotations per minute; SD standard deviation; VAD ventricular assist device.

5 Ann Thorac Surg AGGARWAL ET AL 2013;95:493 9 AI IN LVAD SUPPORTED PATIENTS 497 Table 3. Correlates of Worsening Aortic Insufficiency Characteristics AI Slope SE a p Value Age, years Male sex Body surface area, m Body mass index, kg/m Ischemic cardiomyopathy LVEDD, cm LVESD, cm AI SE b AoV opening (open intermittent) a Slope represents the change in AI over time for each referenced covariable. b Represents the change in AI severity for each covariable measured at each follow-up. AI aortic insufficiency; AoV aortic valve; LVEDD left ventricular end-diastolic diameter; LVESD left ventricular endsystolic diameter; SE standard error. echocardiogram before LVAD implant. The groups were compared to see if there was greater development of AI in patients with AoV sclerosis. AI developed in more patients with a sclerosed AoVs (27 [58.7%] vs 14 [42.4%]; p 0.15). There was no difference in mean time to AI ( days for sclerosed and days for normal valves; p 0.98). Patients with sclerosed AoVs were significantly older ( vs years; p 0.004). There was a no difference in aortic root diameter in sclerosed AoVs compared with normal AoVs (0.63 vs 0.60 cm; p 0.71). There was, however, no difference in mortality rates or repeat hospitalizations between the two groups. Comment Limitations in donor heart availability has led to the emergence in the use of LVADs in patients with endstage heart failure as a permanent alternative to heart transplantation or DT. A potential complication that has increasingly been recognized recently has been the development of AI. In patients with preexisting AI, the severity of insufficiency often progresses [3, 4, 7, 8]. Significant AI can lead to ineffective LVAD output and end-organ malperfusion due to reduced forward flow (recycling of regurgitant blood from the outflow graft into the LV inflow cannula) [5]. Cowger and colleagues [5] in their study of 78 patients with an 83% BTT population being supported with HM XVE and HM II LVADs demonstrated a nearly 64% incidence of significant AI by 18 months in patients implanted with the HM II LVAD. The predictors of progressive AI on postoperative echocardiograms included increasing aortic sinus diameter, an AoV that remained closed or only intermittently opened, and lower LV diastolic and systolic volumes. Pak and colleagues [6], in another study of 93 HM I and 73 HM II patients in a predominantly BTT population, reported AI development in 14.3% of HM II patients at median of 90 days (range, 7 to 364 days) after LVAD implantation. They demonstrated similar results, with more AI seen in patients with predominantly closed AoVs and aortic root dilatation, as assessed by a pathologist in explanted hearts of patients undergoing transplantation. The aortic root diameter, as measured by echocardiography in this study, however, was not significant. Both studies had enrolled and analyzed predominantly a BTT population. Our study, which enrolled a greater than 90% DT population, showed development of mild or greater AI in 52% patients at a median of 187 days. Severe AI did not develop in any patients. The predictors of AI development included a closed AoV (HR, 2.51; 95% CI, 1.06 to 5.89; p 0.03) and age (HR, 1.04; 95% CI, to 1.08; p 0.01) after LVAD implantation. In our study, neither the baseline aortic root diameter nor change in root diameter were significant predictors, as seen in the other studies. This could partly be explained by the technical limitations in precise detection of subcentimeter changes in aortic root dimensions by echocardiograms, as was evidenced in the study by Pak and colleagues [6]. However, AI developed relatively earlier in the patients with increased baseline aortic root diameters than in the patients with smaller aortic roots. Studies have suggested that alterations in fluid dynamics, histology, and geometry of the aortic apparatus contribute to the development of AI. High pressure and high velocity retrograde jets of regurgitant blood volumes contact the root side of closed AoVs and cause valvular damage and degeneration [8, 9]. AoV commissural fusion in LVAD patients is the major pathophysiologic process that has now been well established [6, 10, 11]. This leads to valve fibrosis and degeneration [8], thereby affecting leaflet coaptation and resultant AI. Thus, closed valves are prone to fusion and degeneration from lack of mobility in the setting of concomitant valve trauma mediated by high pressure and velocity blood flow [5]. The association of AI and age has been seen in many studies. The Framingham Heart Study was probably the first large population-based study to demonstrate a significant incidence of AI in an aging population [12]. Age was an independent predictor of AI (OR, 2.3; 95% CI, 2.0 to 2.7). Results from the Strong Heart Study by Lebowitz and colleagues [13] also demonstrated more frequent AI in participants aged older than 60 years compared with those younger than 60 years (14.4% vs 5.8%, p 0.001). Regression analysis showed that AI was independently related to older age (p 0.001). A study by Akasaka and colleagues [14] of 176 healthy volunteers also showed an amplified risk of AI with aging (r 0.81, p 0.001), with 89% of volunteers aged older than 80 years exhibiting AI. The effects of age on AI development potentially result from myxomatous degeneration of valves and their supporting structures [15, 16]. The other possible mechanisms postulated have been the effects of the relatively higher pressures and mechanical stress felt by the left side of the heart compared with the right side [14]. Thus, the rate of degradation as well other

6 498 AGGARWAL ET AL Ann Thorac Surg AI IN LVAD SUPPORTED PATIENTS 2013;95:493 9 injuries to the valves increases and results in AI. The implantation of a LVAD thus probably augments naturally felt stresses and pressures on the AoV, culminating in a significant clinical effect of age on development of AI. A potential strategy to reduce the development or progression of AI would be adjustment of LVAD speed under echocardiographic guidance to allow valve opening at the lowest possible speed without compromising LVAD flows or ventricular decompression. Another approach is afterload reduction by maintaining a tight control of the mean arterial pressure (70 to 80 mm Hg) to reduce aortic wall stress and valvular degeneration. In our study, the patients who were free of AI had relatively lower LVAD speeds than those with AI. This difference, however, did not attain statistical significance. The policy at our institution since this study has involved adjustment of LVAD speeds under echocardiographic guidance to allow intermittent opening of the AoV. These adjustments are made once the patient is stable and ready to be moved out of the acute care setting. No changes are made in the immediate postoperative period. The control of mean arterial pressure (as measured by arterial Doppler), however, did not seem to correlate with the development of AI. This observation most likely is due to a lack of consistency in blood pressure recordings by multiple health care providers. Because this investigation was an unblinded study of LVAD implantations at a single center, we cannot exclude bias and confounding in patient selection, echocardiography image acquisition and interpretation, and LVAD management that may have affected AI assessment. The small sample size and event rates could have limited the flexibility of multivariable modeling. However, the results of our study are in line with those of other studies [5, 6]. Future prospective, blinded studies are required to validate the results of our study. It might also be important to examine not only the opening or closure of the AoV but also the effect of the degree of leaflet excursion (partial opening vs complete opening) on the development and progression of AI. The effects of LVAD speed and control of mean arterial pressure also need to be studied and established in prospectively designed clinical trials. In conclusion, the development of AI during continuous-flow axial LVAD support is affected by the frequency and degree of AoV closure and increasing age. AI increases in patients with prolonged duration of LVAD support. As more patients are implanted with DT LVADs in the future, better management strategies aimed at prevention or retardation of AI need to be investigated and devised. This project was supported by the University of Illinois at Chicago (UIC) Center for Clinical and Translational Science (CCTS), Award Number UL1RR from the National Center for Research Resources. We are grateful to Weihua Gao at Design and Analysis Core for helping with statistical analysis. References 1. Writing Group Members, Lloyd-Jones D, Adams RJ, et al. Heart disease and stroke statistics 2010 update: a report from the American Heart Association. Circulation 2010;121: e Slaughter MS, Pagani FD, Rogers JG, et al. Clinical management of continuous-flow left ventricular assist devices in advanced heart failure. J Heart Lung Transplant 2010; 29:S Bryant AS, Holman WL, Nanda NC, et al. Native aortic valve insufficiency in patients with left ventricular assist devices. Ann Thorac Surg 2006;81:e Mudd JO, Cuda JD, Halushka M, Soderlund KA, Conte JV, Russell SD. Fusion of aortic valve commissures in patients supported by a continuous axial flow left ventricular assist device. J Heart Lung Transplant 2008;27: Cowger J, Pagani FD, Haft JW, Romano MA, Aaronson KD, Kolias TJ. The development of aortic insufficiency in left ventricular assist device-supported patients. Circ Heart Fail 2010;3: Pak SW, Uriel N, Takayama H, et al. Prevalence of de novo aortic insufficiency during long-term support with left ventricular assist devices. J Heart Lung Transplant 2010;29: Rao V, Slater JP, Edwards NM, Naka Y, Oz MC. Surgical management of valvular disease in patients requiring left ventricular assist device support. Ann Thorac Surg 2001;71: Samuels LE, Thomas MP, Holmes EC, et al. Insufficiency of the native aortic valve and left ventricular assist system inflow valve after support with an implantable left ventricular assist system: signs, symptoms, and concerns. J Thorac Cardiovasc Surg 2001;122: May-Newman K, Hillen B, Dembitsky W. Effect of left ventricular assist device outflow conduit anastomosis location on flow patterns in the native aorta. ASAIO J 2006;52: Rose AG, Park SJ, Bank AJ, Miller LW. Partial aortic valve fusion induced by left ventricular assist device. Ann Thorac Surg 2000;70: Connelly JH, Abrams J, Klima T, Vaughn WK, Frazier OH. Acquired commissural fusion of aortic valves in patients with left ventricular assist devices. J Heart Lung Transplant 2003;22: Singh JP, Evans JC, Levy D, et al. Prevalence and clinical determinants of mitral, tricuspid, and aortic regurgitation (the Framingham heart study). Am J Cardiol 1999;83: Lebowitz NE, Bella JN, Roman MJ, et al. Prevalence and correlates of aortic regurgitation in American Indians: the Strong Heart Study. J Am Coll Cardiol 2000;36: Akasaka T, Yoshikawa J, Yoshida K, et al. Age-related valvular regurgitation: a study by pulsed Doppler echocardiography. Circulation 1987;76: Pape LA. Pathogenesis and etiology of valvular heart disease. In: Dalen JE, Alpert JS, eds. Valvular heart disease. Boston: Little, Brown and Company; 1981: Pomerance A. Age-related cardiovascular changes and mechanically induced endocardial pathology. In: Silver MD, ed. Cardiovascular pathology. New York: Churchill Livingstone; 1983:87.