Methods

Top Methods Results Discussion Author & Article Info References

The study was conducted during the 4-year period between 1 August 1990 and 31 July 1994 at Grady Memorial Hospital, which is a 1000-bed city-county hospital primarily serving the indigent population of metropolitan Atlanta, Georgia. All patients with documented HIV infection who had esophagogastroduodenoscopy during the study period were prospectively identified. The study group was composed of consecutive patients with esophageal ulceration identified during the study period. The study was approved by the Emory University Human Investigations Committee, and written informed consent was obtained for all endoscopic examinations.

Endoscopic Techniques

Esophageal ulceration was defined as a visible break in the esophageal mucosa and a perceptible depth of at least 3 mm in the greatest dimension, as assessed using open biopsy forceps, with ulcer (granulation) tissue found on histopathologic examination. Several biopsy specimens (usually six or more) were obtained from identified ulcers, primarily from the ulcer base. A biopsy technique previously documented to obtain larger tissue samples [3] was used during the last 3 years of the study. In patients with more than one ulceration, biopsy specimens were obtained from the largest or most accessible lesion or lesions. If severe Candida esophagitis was concomitantly identified, the Candida plaques were usually removed to expose the underlying mucosa. All examinations were done using Olympus endoscopes, primarily video, and standard biopsy forceps (Olympus Corp., Lake Success, New York). In all but two patients, the procedure was done or supervised by one of the authors.

Histopathologic Evaluation

Biopsy specimens were immediately fixed in buffered formalin and submitted for pathologic examination. At least three to five levels from each specimen, with each glass slide containing two to four tissue sections, were stained with hematoxylin and eosin. Additional stains for infectious agents, including bacterial and acid-fast organisms, were done when standard staining was nondiagnostic and clinical suspicion was high. Fungal agents were easily visualized using routine hematoxylin and eosin staining. When identification of these agents required morphologic confirmation, special stains, including periodic acid-Schiff or Gomori methenamine silver stains, were done. Viral, fungal, and mycobacterial cultures were not done on biopsy specimens. All specimens with histopathologic features suggestive of but nondiagnostic for viral cytopathic effect, and specimens in which granulation tissue was identified in the absence of any viral cytopathic effect, were evaluated for herpes simplex virus and cytomegalovirus using in situ DNA hybridization (Pathogene DNA probe assay, Enzo Diagnostics, New York, New York) and the chromogen AEC (3-amino-9-ethyl carbazole; Dako Corp., Carpenteria, California). Positive control slides done for each biopsy specimen were examined. All histologic material was examined by the authors.

Criteria for Diagnosis

Viral esophagitis was defined as the presence of typical or atypical viral cytopathic effect with confirmation by in situ DNA hybridization as described above [4]. Because of the previously documented high sensitivity and specificity of the in situ DNA hybridization technique, viral cultures were not done on biopsy specimens [4]. Idiopathic esophageal ulceration was diagnosed when histologic examination showed granulation tissue with necrosis of the overlying epithelium and inflammation consistent with ulceration; when no evidence of viral cytopathic effect or another etiologic agent was identified using routine and special staining; when the results of in situ DNA hybridization for herpes simplex virus and cytomegalovirus were negative; and when clinical and endoscopic evidence for gastroesophageal reflux disease and drug-induced esophagitis were absent [5]. Drug-induced esophagitis was said to occur when a patient had been taking a drug recognized to cause ulcerative esophagitis in the absence of any other demonstrable cause and when the identified lesions healed endoscopically without specific therapy other than discontinuation of the offending medication. Gastroesophageal reflux disease was diagnosed when clinical, endoscopic, and pathologic findings were compatible with reflux disease and when reflux-associated therapy alone resulted in clinical and endoscopic remission. Patients were excluded if they had Candida esophagitis or esophageal ulcer of which Candida was considered the sole cause by clinical, endoscopic, and pathologic studies and by both clinical and endoscopic follow-up after antifungal therapy alone. Four such patients were identified during the study period.

Treatment Regimens

Cytomegalovirus esophagitis was treated with intravenous ganciclovir at a dose of 5 mg/kg body weight twice daily for 10 to 21 days. Foscarnet was administered at a dose of 60 mg/kg every 8 hours or at a dose of 90 mg/kg every 12 hours for 14 to 21 days. Herpes simplex virus esophagitis was treated with intravenous acyclovir (15 mg/kg per day) followed by oral acyclovir at a dose of 200 mg 5 times a day for a total of 14 to 21 days. Idiopathic esophageal ulceration was treated either with 40 mg of prednisone per day orally tapering 10 mg per week or with 40 mg of prednisone daily for 2 weeks. No specific therapy other than discontinuation of therapy with the offending medication was used for drug-induced esophagitis. Gastroesophageal reflux disease was treated with omeprazole, 20 to 40 mg daily. For patients in whom moderate to severe Candida infection was detected, therapy with fluconazole (100 daily) or ketoconazole (200 mg daily) was instituted. Maintenance therapy with ganciclovir, foscarnet, acyclovir, prednisone, or antifungal agents was not routinely administered after the initial course of therapy. Oral narcotic agents were routinely used for the symptomatic treatment of severe odynophagia.

After therapy, all patients were encouraged to have repeated endoscopic examinations to evaluate ulcer healing. Repeat biopsy specimens of persistent ulcerations or new endoscopic abnormalities were routinely obtained as described above. During follow-up, endoscopy was done for recurrent esophageal symptoms when possible or after failure of an empiric trial of oral systemic antifungal therapy. A clinical response was said to occur when esophageal symptoms were ameliorated by at least 50% as assessed on a scale of 1 to 10. An endoscopic response was defined as reepithelialization of at least 50% of the ulcer base. A complete response was defined clinically as complete amelioration of all esophageal symptoms and endoscopically as more than 90% reepithelialization of all esophageal ulceration. In most patients, endoscopic reexamination was done within 1 week of the completion of therapy.

Statistical Methods

Medians and ranges were used to characterize the sample for many of the non-normal continuous measures. Univariate associations were initially explored using contingency Table analyses, including the chi-square test and the Fisher exact test for categorical variables. The univariate associations with continuous measures were evaluated using nonparametric techniques, including the Pearson correlation coefficient and the Mann-Whitney test. To describe the relation between specific diagnoses and overall survival, Kaplan-Meier product-limit estimation was done. Similarly, the association between CD4 lymphocyte count and survival was explored. The statistical significance of the difference between survival curves was assessed using the Wilcoxon test statistic. To multivariately explore associations with survival, Cox proportional-hazards regression was used. A P value of less than 0.05 was considered statistically significant.

Results

Top Methods Results Discussion Author & Article Info References

During the 48-month study period, 303 patients with HIV infection had esophagogastroduodenoscopy at our center; 100 patients with esophageal ulcer were identified using this procedure. The clinical characteristics of these patients are shown in Table 1. Eighty-two patients were men; 70 patients were black. Most patients had longstanding HIV infection. However, esophageal disease was the index diagnosis of HIV infection or the acquired immunodeficiency virus syndrome (AIDS)-defining diagnosis in 38 patients. Previous AIDS-defining diagnoses included Pneumocystis carinii pneumonia in 45 patients, disseminated Mycobacterium avium complex infection in 16 patients, Kaposi sarcoma in 7 patients, Candida esophagitis in 5 patients, Mycobacterium tuberculosis infection in 5 patients, Toxoplasma meningitis in 1 patient, and cytomegalovirus pneumonitis in 1 patient. No patient had the HIV-associated seroconversion syndrome [6]. Twenty-seven patients were receiving zidovudine at the time of initial endoscopic examination [7]; 2 patients were receiving zalcitabine [8]. Ten patients were receiving long-term acyclovir and 48 were receiving Pneumocystis carinii prophylaxis. All patients had esophageal or esophagogastric symptoms. Odynophagia was present in all but 6 patients and was the most common indication for esophagogastroduodenoscopy. In 54 patients, an empiric trial of oral systemic antifungal therapy was given for esophageal symptoms but resulted in no amelioration. Esophageal symptoms other than odynophagia included spontaneous (nonswallowing) substernal chest pain in 61 patients, dysphagia in 23 patients, and substernal burning pain (heartburn) in 20 patients. At the time of endoscopy, 14 patients had an ulcer in the oropharynx; biopsy of these lesions was not routinely done. Oropharyngeal candidiasis was present at the time of endoscopy in 23 patients; in 17 of these patients [74%], Candida esophagitis was identified endoscopically. Twenty-seven patients had Candida esophagitis at the time of esophageal ulcer diagnosis. In these patients, either 1) Candida esophagitis was mild in severity or separate from the ulcers or 2) when coating the ulcers, fungal elements were only identified histopathologically in the superficial epithelium and not in ulcer tissue.

The median CD4 lymphocyte count at the time of diagnosis was 15 cells/mm³ (range, 1 to 994 cells/mm³). A significant difference in median CD4 lymphocyte count was found between patients with cytomegalovirus esophagitis (11 cells/mm³ [range, 1 to 120 cells/mm³]) and patients with idiopathic ulcer (19 cells/mm³ [range, 2 to 994 cells/mm³]) (P = 0.025).

Causes of Ulceration

The causes of esophageal ulceration are shown in Figure 1. Ulcers caused by cytomegalovirus esophagitis and idiopathic ulcers were the most common; they occurred in 50 and 41 patients, respectively. One patient had both cytomegalovirus esophagitis and idiopathic ulcer. In this patient, some but not all ulcers healed after prednisone therapy. Follow-up endoscopy with biopsy of the unhealed ulcers showed cytomegalovirus viral cytopathic effect; ganciclovir resulted in a complete symptomatic response and complete ulcer healing on subsequent endoscopy. One patient had been taking aspirin for 2 weeks at bedtime for headaches before the development of severe odynophagia. Endoscopy showed many small, shallow ulcerations throughout the esophagus. Histopathologic findings showed ulcer tissue with no specific cause. Aspirin therapy was discontinued and the symptoms resolved; endoscopic follow-up 3 weeks later showed an esophagus with normal appearance; clinical follow-up at 3 years showed no recurrent symptoms. In one patient who was excluded from the analysis, we found an esophageal ulcer at the time of endoscopy; however, we were unable to do biopsies secondary to poor cooperation during the procedure.

View larger version (13K): [in this window] [in a new window]   Figure 1. Causes of esophageal ulcer in 100 patients. CMV = cytomegalovirus; GERD = gastroesophageal reflux disease; HSV = herpes simplex virus; IEU = idiopathic esophageal ulcer.

Ten patients had more than one cause of ulcer diagnosed during long-term follow-up. Six patients with idiopathic ulcer were later diagnosed with cytomegalovirus esophagitis. Four patients with cytomegalovirus esophagitis had recurrent odynophagia and a diagnosis of idiopathic ulcer. In these patients, ganciclovir or foscarnet and prednisone resulted in clinical and endoscopic remission, respectively. One of these patients had cytomegalovirus esophagitis followed by idiopathic ulcer and, subsequently, herpes simplex virus esophagitis and recurrent idiopathic ulcer. One other patient had several distinct episodes of idiopathic ulcer and cytomegalovirus esophagitis. Candida esophagitis alone was documented endoscopically and histopathologically in 5 patients at the time of recurrent esophageal symptoms.

Response to Therapy

The results of therapy are shown in Table 2. A response to therapy was assessed in all treated patients. Ganciclovir or foscarnet was instituted in all but 8 patients with cytomegalovirus esophagitis alone; a clinical response was found in 29 patients (78%) and a complete response was documented in 16 patients (43%). Six of the 8 untreated patients died of other causes within 2 weeks of diagnosis, and esophageal symptoms spontaneously resolved in 2 patients. Foscarnet was administered to 6 of the 7 patients who did not respond to ganciclovir therapy; clinical and endoscopic response was seen in 5 of these 6 patients. One patient failed to respond to foscarnet therapy but had a complete clinical and endoscopic response to ganciclovir. Thirteen patients had a partial response to either ganciclovir or foscarnet; a complete clinical response was ultimately achieved with continued therapy in all but 3 of these patients.

All but five patients with idiopathic esophageal ulcer were treated with prednisone. A complete clinical response was found in all but two patients, one of whom subsequently responded to an increased dose of prednisone. One of the five untreated patients had a spontaneous clinical and endoscopic remission without any specific treatment. Of the other four, one died of other causes within 1 week, two refused treatment, and one failed follow-up. Sequential therapy with prednisone and ganciclovir resulted in clinical and endoscopic cure in the patient with concurrent idiopathic esophageal ulcer and cytomegalovirus esophagitis.

Of the four patients with both herpes simplex virus and cytomegalovirus esophagitis, therapy resulted in clinical and endoscopic remission, and one patient was treated successfully with acyclovir alone. The patient with esophagitis that was presumed to be aspirin induced responded to discontinuation of aspirin therapy alone. Omeprazole was uniformly effective for reflux esophagitis-associated ulceration. Relapse occurred in 27 of those patients with complete endoscopic and clinical response: Thirteen had idiopathic esophageal ulcer; 10 had cytomegalovirus esophagitis; 1 had reflux-associated ulceration; and 3 had relapse of both idiopathic ulcer and cytomegalovirus esophagitis.

Survival

Overall, survival from the time of diagnosis of esophageal ulceration was poor; median survival was 8.9 months (range, 2 days to > 42 months). No patients were lost to follow-up. Thirty-three patients were still alive (10 had cytomegalovirus esophagitis only, 15 had idiopathic esophageal ulcer only, 1 had herpes simplex virus esophagitis only, 4 had cytomegalovirus and herpes simplex virus esophagitis, 2 had reflux-associated ulceration, and 1 had drug-induced ulceration) at a mean follow-up after diagnosis of 13 months (range, 2 months to 42.5 months). Survival differences were related to CD4 lymphocyte count at diagnosis (P = 0.004). The median survival time for patients with CD4 lymphocyte counts of less than 15 cells/mm³ was 7.4 months; median survival time for patients with CD4 lymphocyte counts of more than 15 cells/mm³ was 12.4 months. Survival was longer for patients in whom esophageal ulcer was the index diagnosis of AIDS (P = 0.034); it did not appear to be related to previous AIDS-defining illnesses. There was a significant difference in median survival time between patients with cytomegalovirus esophagitis and patients with idiopathic esophageal ulcer (7.6 months compared with 13.1 months; P = 0.03) (Figure 2). Based on multivariate analysis, this difference in survival appeared to be explained by the CD4 lymphocyte count. Too few patients had other causes of ulcer to permit meaningful comparisons of survival. Of the 5 patients with herpes simplex virus esophagitis, 4 died at a median of 25 days after diagnosis (range, 13 days to 106 days); one patient was alive at 93 days. All of the patients with herpes simplex virus and cytomegalovirus esophagitis are currently alive at a median of 7.3 months from diagnosis (range, 4.5 months to 22.4 months). Patients who had no clinical or endoscopic response to therapy had no complication related to persistent ulceration, such as perforation or bleeding. No patient died as a direct result of esophageal disease, although ulcer-related bleeding in two patients with severe cirrhosis and hepatic encephalopathy may have contributed to death.

View larger version (12K): [in this window] [in a new window]   Figure 2. Kaplan-Meier survival curve for patients with cytomegalovirus (CMV) esophagitis and patients with idiopathic esophageal ulcer (IEU). A significant difference in median survival was found (P = 0.03; 7.6 months for patients with cytomegalovirus esophagitis compared with 13.1 months for patients with idiopathic esophageal ulcer). Survival was also related to mCD4 lymphocyte count.

Discussion

Top Methods Results Discussion Author & Article Info References

Of studies addressing the causes of esophageal ulceration in patients with HIV infection, ours is the largest to date in which 1) a definition of esophageal ulceration and a method of ulcer biopsy were defined; 2) state-of-the-art histopathologic techniques were used to exclude viral disease; 3) clinical and endoscopic follow-up after therapy was done to further validate the cause of ulcer and to determine the response to therapy; and 4) long-term follow-up was done to assess long-term outcome and survival. Our findings document the multiplicity of both opportunistic and nonopportunistic causes of esophageal ulcer in patients with HIV infection. The identified causes were all potentially treatable, and the high response rate to currently available therapies is encouraging.

Two previous studies [1, 2] prospectively evaluated the causes of esophageal symptoms in patients with HIV infection in whom esophageal ulcerations were reported. In a cohort of 154 patients, Connolly and colleagues [1] identified 48 patients with esophageal symptoms, all of whom had endoscopy. The most frequently identified cause of ulceration was Candida esophagitis, which occurred alone in 26 patients and with other possible causes in 12 patients. A "discrete ulcer" was found in each of 10 patients, in whom 12 causes of ulceration were ultimately identified. A specific definition of ulcer was not given. Four biopsy specimens were obtained from each ulcer, primarily from the edge. Cytomegalovirus and herpes simplex virus esophagitis were each the identified cause in 4 patients, idiopathic esophageal ulcer was seen in 3 patients, and 1 patient had reflux-associated ulceration. On follow-up, 1 patient had herpes simplex virus esophagitis followed by cytomegalovirus esophagitis and then ultimately by an idiopathic esophageal ulcer.

Bonacini and colleagues [2] identified 19 of 110 patients with esophageal symptoms in whom an erosion or ulcer was endoscopically identified. As in the study by Connolly and colleagues [1], approximately 25% of patients studied by Bonacini and colleagues had at least 2 causes identified, usually Candida infection plus another cause. Of the 19 patients with esophageal erosion or ulcer, 11 had a viral infection (7 had cytomegalovirus infection, 3 had herpes simplex virus infection, and 1 had both cytomegalovirus and herpes simplex virus infection). Six patients had no cause identified and were considered "undiagnosed"; these patients may have had idiopathic esophageal ulcer. We believe that some patients who had concomitant Candida esophagitis and positive histologic findings for cytomegalovirus or herpes simplex virus may have had esophageal ulcer, although this could not be endoscopically identified because of concomitant Candida plaques overlying the ulcer. A diagnosis of viral esophagitis was established either by identification of viral cytopathic effect or culture of biopsy specimens. Of the 33 patients with cytomegalovirus disease, only 14 (42%) had "positive" histologic findings [2]. Given the possibility of blood-borne contamination with cytomegalovirus due to the frequent occurrence of viremia in these patients [9-11], we rely, as do others [12, 13], on the histopathologic identification of cytomegalovirus to establish the diagnosis. In addition, only two or more biopsies were done, thus potentially limiting the number of patients in whom viral disease may have been a cause of ulceration but in whom only Candida esophagitis was diagnosed. The number of biopsies required to best exclude viral esophagitis is unknown, although more than three biopsies may be necessary [14]. In both studies [1, 2], in situ DNA hybridization was not done in patients with unexplained lesions to more thoroughly exclude viral esophagitis, and follow-up after therapy was not reported. We believe that the use of in situ DNA hybridization or other immunohistochemical techniques is helpful in diagnosing viral esophagitis when viral cytopathic effect is either atypical in appearance or infrequent in number.

A unique and important finding in our study is the documented high prevalence of idiopathic esophageal ulcer. Recognition of this ulcer is important because of the apparent efficacy of corticosteroid therapy [5, 15]. In addition, idiopathic ulcer may be confused with a "missed" viral lesion; this may result in repeated endoscopic examinations for repeated biopsies or in the administration of empiric antiviral therapy with its attendant expense and potential toxicity, thus delaying effective prednisone therapy. The high prevalence of idiopathic esophageal ulcer further emphasizes the importance of an extensive histopathologic evaluation to exclude viral disease: Such exclusion could eliminate the use of steroids with their potential side effects. Because of the apparent prevalence of idiopathic ulcer, its association with a low CD4 lymphocyte count, and the relatively poor prognosis associated with it, it should probably be considered an AIDS-defining illness.

In other immunocompromised hosts, herpes simplex virus esophagitis is a relatively common cause of esophageal ulceration, but it was relatively uncommon in our patients [16, 17]. Our findings are in contrast to those of Connolly and colleagues [1] and Bonacini and coworkers [2]. The differences found in prevalence may result from the number of patients evaluated, the definition of ulcer used, the number and technique of endoscopic biopsies, the histopathologic techniques used, and the absence of assessment of response to therapy. The small number of patients with herpes simplex virus esophagitis in our study cannot be totally ascribed to concomitant use of acyclovir, because only 10 patients were receiving this agent at the time of endoscopy. It is also unlikely that patients with idiopathic esophageal ulcer had undiagnosed herpes simplex virus esophagitis, given the numerous endoscopic biopsies done, the use of in situ hybridization for herpes simplex virus in biopsy specimens, and the clinical and endoscopic follow-up done after prednisone therapy.

As with any process that may have numerous causes, knowledge of the spectrum as well as the prevalence of these entities is critical, especially in patients with HIV infection, in whom the therapy for each of these disorders is diverse. In addition, many of these therapies are expensive and potentially toxic, which further emphasizes the importance of a specific diagnosis. Currently, management strategies incorporating empiric oral systemic therapy for Candida esophagitis are usually used for patients with HIV infection and new-onset esophageal symptoms. The low prevalence of herpes simplex virus esophagitis and reflux-associated esophageal ulcer in our study suggests that empiric therapy for these conditions is unwarranted in the persistently symptomatic patient (especially a patient with odynophagia after an empiric trial of antifungal therapy) unless other factors strongly suggest these processes. Although cytomegalovirus and idiopathic esophageal ulcer are common causes of ulcerative esophagitis, we would not recommend empiric ganciclovir or prednisone in patients who fail to respond to empiric antifungal therapy because of the potential toxicities of these agents.

The long-term survival of our patients was poor. In their study of 48 patients with esophageal disease of various causes, Connolly and colleagues [1] found an overall mean survival time of 5 months. The poor survival found in our study was not unexpected, however, because most of our patients had advanced HIV infection as documented by low CD4 lymphocyte counts (median, 15 cells/mm³) and previous AIDS-defining illnesses. Previous studies have documented reduced survival in patients with low CD4 lymphocyte counts, especially those of less than 40 cells/mm³ [18-21]. The median survival time after diagnosis in patients with Kaposi sarcoma [22] and Pneumocystis carinii pneumonia [23] is typically more than 1 year; in contrast, the survival for our patients, especially those with cytomegalovirus esophagitis, most closely paralleled that for patients with opportunistic diseases such as disseminated Mycobacterium avium complex infection [24-26], lymphoma [27-29], or severe cryptosporidial diarrhea [30], in whom a median survival time of 7 months or less is common. Survival also appeared to be longer for patients in whom esophageal ulceration was the AIDS-defining illness, and for those with idiopathic esophageal ulcer rather than cytomegalovirus esophagitis. The significant difference in survival between patients with cytomegalovirus esophagitis and those with idiopathic ulcer confirms the results of a previous study [31], which suggested that meaningful differences in survival may be seen in patients with slight differences in CD4 lymphocyte count (5 to 10 cells/mm³) even when this count is less than 60 cells/mm³. Given that esophageal disease per se was rarely the direct cause of death or a direct contributor to death, survival appears to be most closely related to the underlying stage of HIV infection rather than to the esophageal disease itself. Therefore, these disorders represent markers of severe immunodeficiency and shortened survival. Nevertheless, despite poor survival, evaluation and treatment of these patients was rewarding because a clinical response was always associated with a renewed ability to eat, discontinuation of therapy with pain medications, and a subjectively improved quality of life.

Esophageal ulcers in patients with HIV infection may result from numerous causes. Opportunistic causes are the most frequent, and specific therapy is available for each of these disorders, which results in a high response rate. We conclude that all symptomatic patients in whom an esophageal ulceration is identified endoscopically or radiographically should have several endoscopic biopsies of the lesion so that a firm diagnosis may be established and appropriate therapy promptly instituted. Quality of life may be improved with therapy, but long-term survival remains poor.

A portion of this work was previously presented in abstract form at the annual meeting of the American Gastroenterology Association, Boston, Massachusetts, 1993. Some of our patients have been previously reported on in studies evaluating the endoscopic appearances and treatment of esophagitis [5, 32-34].