Endoscopic Ligation Compared with Sclerotherapy for Treatment of Esophageal Variceal Bleeding

Methods

Data Identification

Both of us separately did a computerized bibliographic search of research (MEDLINE, National Library of Medicine, Bethesda, Maryland) published from 1983 onward using the keywords varices and ligation; an additional search using the keywords varices and band was also done. We identified frequently cited references and used SCISEARCH (Scientific Citation Index on-line, Institute for Scientific Information, Philadelphia, Pennsylvania) to locate additional studies that cited these articles. We also scanned the reference lists of all articles obtained to identify additional research not found in the computerized bibliographic database searching.

Strategies used to obtain unpublished material included searching Federal Research in Progress (National Technical Information Service, Springfield, Virginia), NTIS (National Technical Information Service, Springfield, Virginia), Conference Papers Index (Cambridge Scientific Abstracts, Bethesda, Maryland), and BIOSIS Previews (Biosciences Information Service, Philadelphia, Pennsylvania). We also communicated with the manufacturer of the ligation device (Bard Interventional Products, Tewksbury, Massachusetts) and with investigators in the field of variceal hemorrhage, and we reviewed abstracts from national meetings (1988 to May 1994) of the American Gastroenterological Association, the American Society for Gastrointestinal Endoscopy, the American College of Gastroenterology, and the American Association for the Study of Liver Diseases.

Selecting Research

We independently reviewed the titles and abstracts of all studies and retrieved all articles that met the following inclusion criteria: 1) The study was a randomized comparison of endoscopic ligation and sclerotherapy; 2) the target population was composed of patients with esophageal variceal bleeding; and 3) outcome measures were rebleeding, mortality, complications, or treatment sessions to obliteration. We again applied these same criteria independently to the full text of articles that had passed the first eligibility screening to make a final selection of the studies to be included in the overview.

Methodologic Quality Assessment and Data Collection

The criteria used to assess the methodologic quality of the primary articles are listed in Table 1. Each study received a score based on specifications of the population, intervention, and outcome measures. Total methodologic quality scores were then used to rank studies. We independently did methodologic quality assessment and data abstraction, and disagreement was resolved by consensus. We recorded data in terms of number of patients rather than number of events, and solicited missing or ambiguous data from the authors of the studies. A copy of the completed methodologic quality assessment and data abstraction forms were sent to the corresponding author of each study, who was asked to correct erroneous assessments and provide missing information. We asked authors of abstracts to provide full information about methods and updated results.

Data Analysis

The κ statistic with quadratic weights was used to measure agreement between coders for research selection and validity assessment [15]; the degree of agreement rather than the presence or absence of agreement is captured by this method. The odds ratio (OR) was the measure of association used in this meta-analysis. The Breslow-Day method was used to test for homogeneity under the null hypothesis that the ORs were consistent across studies, and the corrected Mantel-Haenszel chi-square test (1 degree of freedom, two-tailed) was used to test whether the ORs differed systematically from a value of 1. The software used was OR2 × 2 x ka (Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario). To avoid problems of biased or unstable estimators when data were sparse, a modification of the Mantel-Haenszel technique (adding 0.5 to each cell when any cell in the category contained zero) was used to estimate the common OR [16, 17].

A priori, we developed four hypotheses to identify potential differences in treatment across studies. First, we proposed that study results might be related to study quality. Hence, a separate analysis was planned to compare the effect of treatment in studies with an overall methodologic quality score of at least 8 with the effect of treatment in studies with a score of less than 8. The cutoff point of 8 was chosen as the midpoint in the scoring system.

Second, we considered that patients with different causes of liver disease might derive different benefits from ligation therapy. Therefore, we planned a separate analysis for studies in which at least 75% of the patients had alcoholic cirrhosis and studies in which fewer than 75% of the patients had alcoholic liver disease.

Third, we reasoned that patients might have different baseline risks for rebleeding, complications, and death based on the severity of their liver disease. Therefore, we did a separate analysis of studies in which at least 40% of patients had Child C liver disease and studies in which fewer than 40% of patients had Child C liver disease.

Fourth, we questioned whether the results presented in fully published manuscripts would differ systematically from results presented only in abstract form. Hence, we also did a subgroup analysis for those studies published in full and those published only as abstracts.

Results

Study Selection

Computerized bibliographic searches identified 158 articles. Overall, 7 randomized trials [12-14, 18-21] involving 547 patients met the inclusion criteria (Table 2). Four were published in full in peer-reviewed journals [12-14, 20], and 3 were published in abstract form [18, 19, 21]. Agreement between reviewers for the selection of relevant articles was 100%. Data from the 7 individual studies for specific outcome measures are provided in Appendix Table 1. We excluded 2 studies [22, 23] that were reported to be randomized, controlled trials comparing ligation with sclerotherapy for the treatment of esophageal varices. Hashizume and colleagues [22] randomly assigned patients who did not present with bleeding to ligation or sclerotherapy at the initial session, and then all patients received sclerotherapy at 1-week intervals. Avgerinos and coworkers [23] did sclerotherapy on all patients at the time of initial hemorrhage and then assigned patients to receive either ligation or sclerotherapy at subsequent treatment sessions.

Methodologic Quality Assessment and Data Collection

Chance-corrected agreement for the methodologic quality assessment yielded κ values of 0.8 to 0.9. Six of the seven authors (including the authors of all three abstracts) provided corroborative, corrective, or missing information [12, 13, 18-21], which was used in the analysis. Selected design details for individual trials are outlined in Table 2. Methodologic quality scores ranged from 7 to 12 on a scale of 0 to 16. Health care workers caring for the patients were blinded to treatment in only one trial [18], and the assessment of rebleeding and other variables was unblinded in all trials.

The study population is described in Table 3. The mean age ranged from 46 to 56 years, and more than 99% of patients had cirrhosis [12, 13, 18-21]. The proportion of patients with alcoholic cirrhosis ranged from 23% to 100%, and the proportion with Child C liver disease ranged from 19% to 78% (Table 3). The mean follow-up period, when specified, ranged from 295 to 337 days [13, 14, 19]. In the five studies that reported having patients lost to follow-up, 7% of the ligation group and 8% of the sclerotherapy group were lost to follow-up [12, 14, 19-21].

Sodium tetradecyl sulfate (1% to 1.5% final dilution) was used as the primary sclerosing agent in five studies [12, 13, 18-20]; it was combined with 33% alcohol and saline in one study [18] and with 25% dextrose in another [13]. Ethanolamine (5%) and polidocanol (1%) were used in the other studies [14, 21]. Intravariceal injections were done in all studies. The same ligation equipment (Bard Interventional Products, Tewksbury, Massachusetts) was used in all trials. All treatment sessions were done at intervals of 1 to 3 weeks; four trials [12-14, 20] restricted their planned treatment intervals to approximately 1 week (5 to 10 days).

Criteria used to define rebleeding were similar across the studies: Specific definitions of rebleeding were provided in all trials, but only three trials required evidence of clinically important bleeding as defined by specific blood pressure or heart rate criteria, decreases in hemoglobin level, or need for transfusion [12-14]. The source of recurrent bleeding was documented in more than 90% of cases in four trials [13, 19-21] and in 60% to 89% of cases in the other three studies [12, 14, 18]. Definitions and protocols detailing methods used to evaluate patients for the outcomes of esophageal stricture, pulmonary infection, and bacterial peritonitis were not clearly stated.

Rebleeding and Mortality

Evidence of statistical heterogeneity was not noted for any of the analyses in our study (P > 0.25 by Breslow-Day test for all analyses). The results of the comparisons of ligation with sclerotherapy are shown in Table 4. We first examined the relative rates of hemostasis for the subset of patients with actively bleeding varices (Table 4). Five studies included a total of 106 patients with active bleeding and provided information on hemostasis [12, 13, 18, 19, 21]. The hemostasis rate for actively bleeding varices was similar for patients having had ligation and those having had sclerotherapy (OR, 1.14 [95% CI, 0.44 to 2.90]).

Only one of seven individual trials showed a significant decrease in rebleeding (Figure 1), but when data were pooled, rebleeding was significantly less common with ligation than with sclerotherapy (OR, 0.52 [CI, 0.37 to 0.74]). The rate of rebleeding due to varices was specified in five studies [13, 18-21], and the rate of rebleeding due to treatment-induced ulcers was reported in all seven studies. The rate of rebleeding due to varices was lower in the patients who received ligation (OR, 0.47 [CI, 0.29 to 0.78]), as was the rate of rebleeding due to treatment-induced ulcers (OR, 0.56 [0.28 to 1.15]; P = 0.16).

The ORs and CIs for mortality in the individual studies are shown in Figure 2. Overall, ligation was associated with lower mortality when compared with sclerotherapy (OR, 0.67 [CI, 0.46 to 0.98]). Mortality due to bleeding was reported in five studies [12, 13, 19-21], and the OR was 0.49 (CI, 0.24 to 0.996). Deaths not due to bleeding were also less frequent in the ligation group (OR, 0.61 [CI, 0.35 to 1.07]; P = 0.11).

Adverse Events

Because ligation therapy was developed to decrease complications associated with sclerotherapy, we also examined the complications associated with these two approaches (Table 4). Bleeding from treatment-induced ulcers was less common in patients treated with ligation, but the difference was not statistically significant. Esophageal strictures also occurred less frequently with ligation. Significant differences were not seen between treatments in the proportion of pulmonary infections [12-1418, 19, 21] or episodes of spontaneous bacterial peritonitis [12, 13, 18, 19, 21] that occurred (P > 0.20). Esophageal perforation was reported in 2 (0.7%) of 274 patients receiving ligation and 0 of 273 patients treated with sclerotherapy. Induction of variceal bleeding at the time of treatment, where specified [18, 19, 21], occurred in 1 (0.9%) of 108 patients receiving ligation and 0 of 107 patients treated with sclerotherapy. Complications leading to death were seen in 2 (1.0%) of 210 patients treated with ligation and 7 (3.3%) of 211 patients receiving sclerotherapy (OR = 0.47 [CI, 0.15 to 1.48]; P = 0.30) [12, 13, 18, 19, 21].

Variceal Obliteration

As shown in Table 4, variceal obliteration was achieved in similar proportions of patients treated with ligation and sclerotherapy; this proportion ranged from 27% to 90% (Table 5). However, fewer endoscopic treatment sessions were required to achieve variceal obliteration with ligation than with sclerotherapy in six of seven studies (Table 5): The differences were reported to be statistically significant in four trials (13, 14, 19, 20; Table 4, a P value of 0.056 was reported in one trial [12], and statistical analysis was not provided in another trial in which three fewer treatment sessions were required with ligation than with sclerotherapy [21]. Comparison of the number of endoscopic treatment sessions required to achieve variceal obliteration showed that 2.2 (CI, 0.9 to 3.5) fewer treatment sessions were needed with ligation than with sclerotherapy.

Subgroup Analyses

Separate subgroup analyses for rebleeding and mortality were done in relation to methodologic score, proportion of patients with alcoholic cirrhosis, proportion of patients with Child C liver disease, and publication status of the study (manuscript published in full or in abstract form). The results of these subgroup analyses are shown in Table 6. Estimates of efficacy were similar in the subgroups tested and had widely overlapping CIs, precluding statements about greater efficacy in specific subsets of patients or any difference in treatment effect associated with methodologic quality or publication status.

Discussion

The purpose of our meta-analysis was to compare the effects of endoscopic ligation with those of the standard therapy, endoscopic sclerotherapy, for the treatment of patients with esophageal variceal bleeding. Although a meta-analysis does not replace a large-scale, well-designed, randomized, controlled trial, individual studies may be limited by small sample sizes—especially for end points with relatively low incidences. By synthesizing all available data, the meta-analysis allows for a more precise estimate than that which can be obtained from the results of any individual study. In this systematic review, we have followed the accepted methodologic criteria [24, 25]: We used comprehensive data retrieval to minimize publication bias; defined explicit selection criteria for the identification of relevant studies; did thorough methodologic quality assessments of the primary studies; and made reproducible decisions about relevance, selection, and methodologic quality assessment.

The overall rebleeding rate seen with sclerotherapy in the seven studies reviewed was 47% (127 of 273 patients; see Appendix Table 1 [12-14, 18-21]. Although only one of the seven trials showed a significantly lower rebleeding rate with ligation than with sclerotherapy, all of the individual trials did report rebleeding rates that favored ligation (Figure 1). Synthesis of these similar results from seven trials including 273 patients showed a significant 50% reduction in rebleeding with ligation compared with sclerotherapy. Rebleeding caused by varices and rebleeding caused by treatment-induced ulcers were both reduced with ligation. Assuming a rebleeding rate of 50% in patients treated with sclerotherapy, as was found in the seven studies we reviewed, and using the 50% reduction in bleeding associated with ligation therapy, one would need to treat four patients with ligation instead of sclerotherapy in order to avert one rebleeding episode.

The mortality seen with sclerotherapy in the seven trials that we examined was 32% (88 of 273 patients; see Appendix Table 1 [12-14, 18-21]. Our overview shows that mortality was decreased by approximately one third with ligation therapy compared with sclerotherapy. Assuming a mortality rate of 30% in patients treated with sclerotherapy, as was found in the seven studies we reviewed, and using the 30% reduction in mortality associated with ligation, one would need to treat 10 patients with ligation instead of sclerotherapy to save one life.

Exsanguination is not the cause of death in most patients who present with variceal bleeding. In the studies we reviewed, the investigators attributed only 40% of deaths to bleeding. Furthermore, numerous randomized trials of surgical therapy compared with medical treatment or with sclerotherapy document that prevention of bleeding does not necessarily improve survival [8]. Therapies may have varying effects on survival due to different comorbidity and complication rates. Deaths due to bleeding and deaths due to other causes both occurred less frequently with ligation than with sclerotherapy, although the difference in mortality unrelated to bleeding was not significant (P = 0.11).

Endoscopic ligation requires placement of an opaque cylinder over the end of the endoscope. This decreases the endoscopic field of view and may allow pooling of blood. Thus, in a patient with active bleeding, visualization may be more impaired when treating with ligation than when treating with sclerotherapy. Although further evaluation in patients with active bleeding will be valuable, our overview shows that ligation and sclerotherapy achieved similar rates of initial hemostasis in the five trials reporting on patients whose varices were actively bleeding at the time of treatment. In the future, the introduction of clear plastic cylinders for ligation may allow for better visualization in patients with actively bleeding varices.

Evaluation of a new technology requires not only assessment of its efficacy but evaluation of its potential side effects and cost implications. Local complications were less common with ligation than with sclerotherapy. A reduction in esophageal strictures was the most consistent and marked finding in our analysis. Systemic complications, such as pulmonary infections and bacterial peritonitis, were not significantly different in the two treatment groups, although a trend toward decreased pulmonary infection and bacterial peritonitis in patients treated with variceal ligation was seen. However, the lack of blinding, of definitions for complications, and of specified protocols for identifying these complications limit the inferences that can be drawn from these data.

The proportions of patients with variceal obliteration were similar in the two treatment groups. However, the mean number of endoscopic treatment sessions required to achieve variceal obliteration was consistently lower in the ligation group than in the group that received sclerotherapy (Table 4). Six of the seven studies suggest that one to three fewer endoscopic treatment sessions will be required to achieve variceal obliteration with ligation. The major cost of either ligation or sclerotherapy relates to the performance of upper endoscopy; the costs of the equipment and material required for ligation or sclerotherapy are similar. In addition to cost differentials that may occur in relation to differences in efficacy (such as rebleeding) or complications, substantial cost savings is likely if one form of treatment requires fewer endoscopic treatment sessions. However, a comprehensive economic evaluation is warranted to rigorously evaluate all of the costs and consequences of these two endoscopic approaches in the management of patients with esophageal variceal bleeding.

In conclusion, on the basis of lower rates of rebleeding, mortality, and complications and the need for fewer endoscopic treatments, ligation should be considered the endoscopic treatment of choice for patients with esophageal variceal bleeding. Further large-scale studies with longer periods of follow-up will be helpful in more firmly establishing the utility of ligation therapy compared with sclerotherapy.