|Home | About | Journals | Submit | Contact Us | Français|
Endoscopic variceal ligation (EVL) and non-selective beta-blockers (beta-blockers) are both effective for primary prophylaxis for variceal hemorrhage, however the route of administration and side effects of these treatments are distinct. The objective of this study was to examine patient and physician preferences for primary prevention of variceal hemorrhage.
Untreated patients with newly diagnosed esophageal varices and practicing gastroenterologists were enrolled in the study. Patients with contraindications to either EVL or beta-blockers were excluded. Preferences for treatment were evaluated using an interactive computer task designed to elicit preferences based on subjects’ trade-offs related to route of administration, risk of fatigue, sexual dysfunction, dysphagia, shortness of breath and/or hypotension, procedure-related bleeding and perforation.
Fifty-three patients and 61 physicians were enrolled. Thirty-four (64%) patients and 35 (57%) physicians preferred EVL over beta-blockers. Patients’ preferences were most strongly influenced by the risks of shortness of breath or hypotension, fatigue, and procedure-related bleeding; whereas physicians placed greater importance on procedure-related bleeding, sexual dysfunction, and perforation. Forty-eight patients were prescribed beta-blockers; two were not given prophylaxis, three were lost to follow up.
Treatment preferences among both patients and physicians for primary prophylaxis of variceal hemorrhage vary significantly. Physicians in this study preferring EVL stated that they prescribe beta-blockers as first line therapy in order to remain compliant with guidelines. Physicians should discuss both EVL and beta-blockers with patients requiring primary prophylaxis for variceal hemorrhage. Future guidelines should incorporate individual patient preferences.
Screening for esophageal varices and use of prophylactic therapies to prevent a first variceal hemorrhage is an essential component of care for patients with cirrhosis (1). Two treatment options have been proven to be effective in preventing first variceal bleed in patients with cirrhosis and large varices: non-selective beta-blockers (beta-blockers) and endoscopic variceal ligation (EVL) (1, 2). Meta-analyses show that, compared to no treatment or placebo, beta-blockers reduce the risk of first variceal hemorrhage from approximately 30% to 14% over two years in patients with medium/large varices (3) and improve survival (4). Two recent meta-analyses of studies comparing beta-blockers and EVL show that EVL is marginally more effective than beta-blockers in preventing first variceal hemorrhage without any differences in mortality (5, 6). Expert consensus recommends beta-blockers as first line therapy and EVL for those with contraindications or intolerance to beta-blockers (7).
The principles of patient autonomy and informed consent prescribe that patients should be informed of all available treatment options and that their preferences should be incorporated into treatment decisions made in conjunction with their treating physician. This is particularly important when there is controversy as to the best treatment option, or when there are significant differences between treatments in terms of their likelihood, impact, or types of complications (8). To the best of our knowledge, patients’ preferences regarding treatment for primary prophylaxis for variceal hemorrhage have not been examined. Moreover, although guidelines have been published, clinicians’ values for the trade-offs involved in choosing EVL versus beta-blockers have not been described. Given this background, our objective was to assess patient and physician preferences for beta-blockers versus EVL in patients with cirrhosis who are candidates for primary prophylaxis of variceal hemorrhage.
Preferences were assessed using Adaptive Conjoint Analysis (ACA). ACA collects and analyzes preference data using an interactive computer program (Sawtooth Software ®). Conjoint analysis has a strong theoretical basis and obtains high levels of internal consistency (9–11). ACA uses an interactive format that engages participants’ attention and results in greater gains in knowledge compared to standard educational materials. In addition, ACA is administered by computer and thereby minimizes interviewer biases and facilitates data collection and management. Because ACA can be programmed to present the outcomes in random order, it eliminates ordering effects (11). The main advantage of ACA’s interactive design is that it allows a relatively large amount of information to be processed without resulting in information overload or respondent fatigue (9–12). This method, described in detail below, has been used to effectively describe patient preferences in patients for whom multiple competing treatment options with varying risk profiles and benefits exist (13–15).
The protocol for this study was approved by the Yale University School of Medicine Human Investigations Committee and the VA Connecticut Healthcare System’s Human Subjects Committee.
Patients with cirrhosis and newly diagnosed medium or large varices were identified by their treating physician from seven different participating hospital or endoscopy centers in Connecticut. Cirrhosis was diagnosed by the treating physician through liver biopsy and/or clinical presentation. As patients were enrolled from multiple sites, there was variability in the use of liver biopsy to diagnose cirrhosis. We accepted the diagnosis of cirrhosis by biopsy or by compatible clinical picture (laboratory and physical data) as determined by the patient’s treating physician, as all of these patients would be considered candidates for primary prophylaxis if esophageal varices were present. Patients were contacted by one of the investigators (AL) after endoscopy and invited to participate. Endoscopy was primarily performed to screen for esophageal varices in patients with cirrhosis. Patients with cirrhosis undergoing endoscopy for other purposes were included as long as there was no concern for acute gastrointestinal hemorrhage or any prior variceal hemorrhage. Patients were enrolled within one week of endoscopy so as not to delay initiating prophylactic therapy. Patients were excluded if they had prior variceal hemorrhage, previously received either EVL or beta-blockers, or had contraindications to either therapy such as significant pulmonary disease or hypotension. Patients were also excluded if they were non-English speaking, or if they had untreated or uncontrolled hepatic encephalopathy. After signing consent, patients received a brief standardized educational session regarding cirrhosis, esophageal varices and the risk of variceal hemorrhage, and the essential characteristics of prophylactic therapies with either beta-blockers or EVL (see Appendix A). Patients were told that both prophylactic therapies were equal in terms of efficacy. The educational session was conducted in person by one of the investigators (AL), at a place and time convenient to the patient (usually in a clinic room). The session was performed at the time of study enrollment, after informed consent and prior to the ACA.
Practicing gastroenterologists and hepatologists in academic and private practice settings in Connecticut were invited to participate and provided verbal consent. Physicians were also recruited at an international Hepatology meeting (AASLD 2006 Boston), where consecutive physicians attending a poster session were invited to participate. Physicians were excluded if they did not treat patients with liver disease. Fellows in gastroenterology were enrolled, providing they had at least one full year of fellowship training. Physicians who had assisted in patient recruitment were also included, providing that they had not been previously exposed to the ACA. These physicians did not take the ACA concurrently with their patients.
Characteristics included in the ACA task were risks of fatigue, sexual dysfunction, dysphagia, shortness of breath and/or hypotension (described as lightheadedness or having low blood pressure), procedure-related bleeding, perforation and the route of administration. Probabilities of each adverse event were based on randomized controlled trials and observational primary prophylaxis studies (16–19). The decision to couple shortness of breath with hypotension was made after testing the initial ACA models on practicing gastroenterologists and lay individuals. When shortness of breath and hypotension (both potentially “severe” side effects) were listed separately in the model this had an effect of biasing preference against beta-blockers, as there were more “severe” side-effects attributed to beta-blockers compared to EVL (procedure related bleeding). The severe side effects were therefore coupled in the model, allowing for only one “severe” side effect to be listed for each beta-blockers and EVL and thereby creating a more balanced model.
The ACA task first asked subjects to rate the route of treatment administration. Subjects were then asked to rank the relative importance of each characteristic on a 9-point scale ranging from Not as important to me to Extremely important to me. This information allows the software to compute an initial estimate of each subject’s utilities. In this context, “utility” is a number that represents the value that respondents associate with a particular characteristic, with higher utilities representing greater value. The software program then presents subjects with a series of trade-off questions (Figure 1). Each question presents a pair of options with a specific trade-off. ACA constructs pairs by examining all possible ways for levels of each characteristic to be combined. ACA is interactive, in that it uses the information obtained from each new paired comparison to update utility estimates and to select the next pair of options. Utility measures become more precise as subjects are asked to discriminate among competing risks and benefits in successive pairs. Final utilities are generated by regression analysis. Subsequently, these utilities are employed to derive individual subject’s preferences. Details regarding the models used to calculate ACA utilities and preferences are available at http://www.sawtoothsoftware.com/download/techpap/acatech.pdf.
After completing the task, subjects were shown a bar graph depicting the relative importance of each attribute (Figure 2). The relative importances reflect the extent to which each characteristic influences each subject’s treatment preference. The relative importances are calculated by dividing the range of utilities for each characteristic by the sum of ranges and multiplying by 100. The relative importances sum to 100.
After the ACA task, patients completed a written questionnaire. This included questions evaluating the acceptability of the task (i.e.: ease in understanding questions, ability of the task to help compare the pros and cons of treatment, amount of time required to complete the task), as well as a question evaluating the validity of the task: “The bar graph represents the way I feel about the specific treatment characteristics” answered on a five-point Likert scale. Self-reported demographic data, the presence of co-morbid medical or psychiatric illness, quality of life (20), level of physician trust (21), preference for medications over procedures (“For most health problems, I prefer to avoid taking prescription drugs” and “For most health problems I would rather take a prescription drug than have a procedure done”) (22) and risk perception (“What do you think your risk is of bleeding from varices is over the next 1–2 years?”) were also ascertained.
All data were collected in the presence of one of the investigators (AL). The presence of ascites, encephalopathy, and most recent serum laboratory tests were obtained from the medical record to assess the severity of liver disease by Child-Turcotte-Pugh score and Model End-Stage Liver Disease (MELD) score. Laboratory data and elements of the Child score were performed within 2 weeks of the endoscopy. We did not repeat lab work as part of study enrollment. Patients received prophylaxis from their treating physician. Final follow-up occurred four to six weeks after study enrollment to ask patients what type of prophylaxis they received.
Interviews were conducted at a time and place convenient to the physician, with one of the investigators present (AVL or AI). Physicians were presented with the following instructions: “Non-selective beta-blockers and EVL have similar efficacy in preventing first variceal hemorrhage. Which aspects of these therapies do you feel are most important when initiating primary prophylaxis in a patient with newly diagnosed esophageal varices (assuming they do not have contraindications to either treatment)?”
Physicians were administered the same ACA task as patients. Type and location of practice, physician gender, and approximate percentage of practice devoted to patients with liver disease was recorded. Percentage of time caring for patients with liver disease was used as a surrogate to determine level of expertise in hepatology. Because we found that physician preferences according to the ACA task were not consistently concordant with prescribing practice, we asked the last 30 physicians to discuss the discrepancy between their preferences predicted by ACA and their prescribing practices. Responses were transcribed at the time of interview.
The ACA data were downloaded into SAS (version 8.0) files for analysis. Descriptive analyses were performed to describe patient and physician characteristics and treatment preferences. In this setting, treatment preferences are presented as strength of preference for EVL and beta-blockers. Preferences are calculated by first summing the utilities of the levels corresponding to each option. The utilities are exponentiated and then rescaled so that they sum to 100. We performed bivariate analyses (using the Mann Whitney U test) and multivariable linear regression to ascertain associations between respondent characteristics and treatment preference.
This study was designed as a pilot project to explore the use of ACA in evaluating the question of primary prophylaxis for variceal hemorrhage. As our analysis was designed to be descriptive, sample size was not based on expected outcome. Our goal was to collect data from a group of patients that could be representative of a larger group of patients with cirrhosis.
Fifty-seven patients were eligible for enrollment during the study period and were invited to participate, four patients declined participation and 53 patients were actually enrolled. Patient characteristics are listed in Table 1. Among patients, treatment preference was stronger for EVL than beta-blockers, the mean strength of preference (±SD) for EVL was 61% (±30) versus a mean strength of preference of 39% (±30) for beta-blockers. In descending order of importance, patients ranked shortness of breath or low blood pressure, fatigue, bleeding, sexual problems, and route of administration as having the strongest impact on their preferences (Table 2).
In bivariate analysis (Table 3), younger age and higher education level were both associated with a stronger preference for EVL. Reporting a preference for medications over procedures and having a higher risk perception of variceal hemorrhage were positively associated with preference for beta-blockers. Site of enrollment, insurance status, or VA status did not impact preference. We also found no significant associations between CPT score (r = 0.1, p = 0.2), (MELD score (r = 0.17, p = 0.2), trust (r = −0.14, p = 0.3) or quality of life (r = −0.12, p = 0.4) and treatment preference. In a multivariable linear regression model (containing age, education, risk perception, and medications over procedures as predictor variables) education and preference for medications over procedures remained significant predictors of patient preference.
More than 90% of patients stated that the computer questions were easy to understand and helped them compare the two treatment options. When asked if the bar graph depicting the importances of the attributes represented the way they felt about the specific treatment characteristics, 48 (91%) patients stated “Agree or Strongly agree”, whereas four patients stated “Neutral” and one responded “Strongly disagree”.
Forty-eight patients received beta-blockers for primary prophylaxis, including two patients who received atenolol, a selective beta-blocker. Three patients were lost to follow-up. No patients received EVL, and two patients were not on any prophylactic therapy at one month after study enrollment.
61 physicians agreed to participate (Table 4). Physicians mean strength of preferences (±SD) was 57% (±30) for EVL and 43% (±30) for beta-blockers. There were no significant associations between physician characteristics and preference. Fellows did had a stronger preference for EVL compared to practicing physicians (77 versus 52, p=0.4), but this difference did not reach statistical significance, most probably due to the relatively small number of fellows studied. Percentage of time caring for patients with liver disease did not significantly impact preference.
Qualitative data were collected from 28 physicians. In asking physicians to explore discrepancies between their preference and practice, six of the 15 physicians whose preference did not correlate with their practice stated they prescribe beta-blockers to be compliant with current treatment guidelines. Other reasons for a discrepancy between preference and practice included cost, invasiveness, and allowing patients to make informed decisions.
Although treatment preferences were similar (Figure 3), patients and physicians differed with respect to the relative importance of specific treatment characteristics on their preferences (Table 2). The risk of shortness of breath or low blood pressure, route of administration, and dysphagia were ranked similarly by both patients and physicians. However, physicians were more concerned with the risk of procedure-related bleeding, sexual dysfunction, and perforation and less strongly influenced by the risk of fatigue compared to patients. When asked if the bar graph depicting the relatives importance of the attributes represented the way they felt about the specific treatment characteristics, 50 (82%) physicians stated “Agree or Strongly agree”, whereas nine (15%) stated “Neutral” and two (3%) responded “Disagree”.
To the best of our knowledge, this is the first study examining patient and physician preferences for prophylaxis of variceal hemorrhage in patients with cirrhosis. In direct contrast to current clinical practice and guidelines, we found that 64% of patients with cirrhosis requiring primary prophylaxis for variceal hemorrhage and 57% of prescribing physicians prefer EVL over beta-blockers. Of note, our results also demonstrated that treatment preferences vary significantly, indicating that this choice is value-based, and emphasizing the importance of incorporating individual patient preferences into the treatment planning process.
Shortness of breath or low blood pressure was perceived as the most important factor by both patients and physicians. However, patients were more influenced by the risk of fatigue and much less influenced by the risk of rare procedure-related complications compared to physicians. This finding is consistent with other studies demonstrating differences between physician and patient preferences. (23–26)
Physicians may be more strongly influenced by the rare, but serious risks of perforation and bleeding in part due to the availability heuristic. The availability heuristic or bias refers to the observation that the frequency of events that are easier to remember or imagine are falsely believed to be more frequent than events which are less easy to imagine or remember. This bias predicts that physicians who may have caused or heard about as a single catastrophic adverse event would be more strongly influenced by this risk than patients.
Despite recruiting patients from seven different centers from across Connecticut, none of the patients enrolled received EVL as primary prophylaxis for variceal hemorrhage. This pattern of practice is consistent with consensus guidelines (7), but may be questionable given the variability in patient preferences noted in this study.
The strengths of this study lie in the methods used to evaluate preferences as well as the successful recruitment of a substantial number of patients at the time of their clinical decision to start primary prophylaxis for variceal hemorrhage. ACA measures preferences based on how subjects evaluate specific risks and benefits. Subjects do not evaluate treatment alternatives directly. Values are computed based on how each respondent makes trade-offs between competing risks and benefits related to the treatment options under consideration. These values are then used to predict which option most closely suits each patient’s individual priorities.
There are also several important limitations that must be considered. We assumed that the efficacy of EVL and beta-blockers were equivalent. In reality, it may be that for an individual patient, EVL would be more effective due to the endoscopic characteristics of their varices, the severity of their liver disease, or adherence. On the other hand, additional beneficial effects of beta-blockers in preventing other complications of cirrhosis such as ascites were not presented, neither was the option of switching from beta-blockers to EVL proposed. Also, additional evidence supporting that bleeding from EVL ulcers can be fatal was not available at the time of the preparation of the ACA task and was therefore not included (27). Given the difficulty of recruiting patients with incident disease we were limited in terms of the total number of patients enrolled. In terms of data collection for physicians, data exploring the difference between preference and prescribing practice were only collected on approximately half the physicians, limiting our interpretation of this discordance. Lastly, cost was not considered in this study.
In summary, we found significant heterogeneity among physicians and patients regarding preferences for primary prophylaxis for variceal hemorrhage. These results suggest that physicians’ recommendations for primary prophylaxis of variceal hemorrhage should be based on explicit elicitation of individual patient preferences.
In addition to the patients donating their time for the completion of this study, the investigators wish to acknowledge and thank the referring physicians: Drs. Hong Gao, Caroline Loeser, Jason Rogart, Anish Sheth, Hannah Miller, Carlos Mena-Hurtada, Timothy Wong, Chris Daniak, Savita Srivastava, Jonathan White, Sarah Canavan, Hemchand Ramberan, Anil Nagar, Shivakumar Vignesh, Scott Swenson, Suchat Wongcharatrawee, Joan Cho, Deborah Proctor.
Dr. Fraenkel is supported by the K23 Award AR048826-01 A1 and the VA Health Services Research Department. Dr. Garcia-Tsao is supported by the K24 Award DK02727.
Essential characteristics of the two treatment options: