ESOPHAGUS Taxol Effect, Colchicine Toxicity
Taxol, an antineoplastic agent with a novel mechanism of action, can be associated with striking mitotic arrest associated with epithelial necrosis and ulceration of the esophagus. As a class of drug, taxane chemotherapeutic agents are commonly used to treat malignancies of the esophagus, breast, and lung. Paclitaxel (Taxol®) chemotherapy has been associated with dramatic GI mucosal changes, accompanied by an increase in apoptosis1. Taxol induces these GI mucosal changes by binding to microtubules, thus promoting polymerization and inhibiting depolymerization. Electron microscopy has shown this central core of polymerized microtubules surrounded by dispersed chromatin, resulting in a “ring” structure during metaphase1. Taxanes have been shown to induce unique histologic changes within epithelium of the GI tract associated with cell necrosis in the GI tract1. As referenced by Hruban et al., in preclinical trials, taxol had been associated with gastritis and duodenal necrosis in mice and with colonic necrosis in dogs1. Reported human drug-related GI effects have included vomiting, diarrhea, mucositis, and neutropenic enterocolitis.
Since the mitotic arrest is associated with bundling of intermediate filaments secondary to accumulation of polymerized microtubules, the histologic correlate is the presence of arrested mitoses with ring forms. With taxol, the findings tend to be striking in the esophagus, whereas, in colchicine toxicity, the small bowel is more likely to be severely altered. The ring mitoses are accompanied by prominent apoptosis in all gastrointestinal tract sites and, regardless of the site, the alterations are found in the proliferative compartment. Thus, in esophageal squamous mucosa, they are encountered in the basal layer, whereas they are encountered in the gastric pits rather than the deeper glands. In the small bowel, the epithelial changes are found in the mid crypt but they are in the deep crypts in the colon. The surface cells are uninvolved in all sites. It was initially believed that finding this pattern of injury in a patient taking taxol indicated clinical toxicity. However, this is not necessarily the case. The medication is administered intravenously on an outpatient basis. If the patient happens to have a gastrointestinal tract biopsy or resection within 4 days of the administration of the medication, the histologic changes are encountered even in asymptomatic patients 2.
The histologic findings associated with colchicine toxicity are essentially the same as those encountered in patients taking taxol but are only seen in patients who have clinical toxicity. Since it is impossible to separate the effects of the two medications with certainty on histologic examination, it is our practice to contact the submitting clinician caring for the patient to correlate with medication history and determine the need for intervention (colchicine toxicity can require supportive care). That noted, colchicine toxicity is typically encountered in the antrum and small bowel.
Allergic/Eosinophilic Esophagitis Overall, eosinophilic esophagogastroenteritis (EG) is an uncommon benign inflammatory condition characterized by eosinophilic infiltration of the gastrointestinal (GI) tract. The diagnostic criteria are:
• gastrointestinal symptoms;
• eosinophilic infiltration of the gastrointestinal tract, usually with intraepithelial eosinophils;
• no evidence of parasitic infestation.
Most patients (in addition to their GI involvement) also have a history of allergy, asthma, drug sensitivities, peripheral eosinophilia, and increased IgE levels. Food and inhalant intolerance have been postulated as etiologic factors and some argue for allergy evaluation with skin prick testing in patients with eosinophilic esophagitis (EE). For example, Penfield et al. identified one or more food or inhalant allergens in 21/26 (81%) patients with EE who underwent skin testing 3. The value of such investigations in adults remains uncertain as there is no evidence to support a role for an elimination diet based on their results. In addition, a seasonal variation has been demonstrated with a predominance of newly diagnosed cases during the spring and summer months4-6. EG predominantly affects patients in their third to sixth decades. However, 15% to 20% of cases are seen in the pediatric age group and, in some cases, milk allergy may be demonstrated.
Any part of the GI tract (from the esophagus to the rectum) can be affected––the stomach and small bowel are commonly involved. Eosinophilic esophagogastroenteritis can show preferential involvement of the mucosa, muscularis propria, or serosa. Symptoms depend on the site and extent of eosinophilic infiltration. Mucosal disease can present as diarrhea, malabsorption, and protein-losing enteropathy. Submucosal disease presents as obstruction and abdominal pain, and patients can develop eosinophilic ascites with serosal involvement. Rarely do patients present with an acute abdominal emergency necessitating immediate laparotomy.
Patients with eosinophilic esophagitis (limited to esophagus) present with dysphagia, food impaction, and strictures. This finding is encountered in about 6% of adult patients undergoing upper endoscopy and there seems to be a male predominance. In children, there is a similar male predominance and probable whites are over-represented. Some of our gastroenterology colleagues provide useful information regarding the endoscopic appearance of the esophagus and pathologists should be familiar with the terminology in order to address the clinical concern. Linear furrows, esophageal rings, esophageal trachealization, and feline esophagus are some of the possible endoscopic scenarios. White mucosal specks, corresponding to foci of eosinophilic microabscesses, may also be seen. In some cases the mucosa appears grossly normal. The differential diagnosis is with reflux esophagitis. However, unlike in reflux esophagitis, the upper and mid esophagus are commonly affected with relative distal sparing. On histologic examination, superficial epithelial clusters of eosinophils that slough into the lumen are more common in eosinophilic esophagitis. Basal cell hyperplasia can be striking, even more so than in cases of GERD, as shown by Steiner et al. in a study where they found a more severe degree of basal cell hyperplasia in pediatric patients with EE than in those with GERD 7. Besides mucosal involvement, there may be eosinophilic infiltration present deep in the submucosa and muscularis propria, which could explain the fragility of the esophageal wall in these patients. Indeed, they are prone to endoscopic complications, namely esophageal tears or rupture, in some cases associated with full thickness esophageal inflammation and high eosinophil density 8-11. Spontaneous esophageal rupture has been reported in some cases 12-13. The eosinophilic infiltrate in EG can be patchy and multiple, localized, or diffuse. In 10% of cases, mucosal biopsies can be nondiagnostic, due to the patchy nature of the disease or mucosal sparing. Multiple and full thickness biopsies may be necessary to establish the diagnosis.
Some have proposed that an estimate of the number of eosinophils is a helpful feature to establish the diagnosis. In fact, a consensus agreement sponsored in part by the American Gastroenterological Association (AGA) recommends a minimum of 15 eosinophils/hpf to make a diagnosis of EE 14 although Dellon et al noted the poor reproducibility of such eosinophil counts 15. Because some clinicians feel it is necessary to provide concrete numbers, some pathologists are in the habit of counting and reporting eosinophils/hpf in biopsies from patients suspected of having EE. Taking this recommendation to heart, especially in isolation, can lead to misdiagnoses as it is well-known that significantly increased intraepithelial eosinophils can be seen in other settings such as severe GERD 16. On the other hand, work by Kephart et al. demonstrates significant extracellular deposition of eosinophil-derived neurotoxin (identified via immunofluorescence) in adult patients with EE, even in the face of relatively few intact eosinophils in some cases 17. This finding raises the possibility that, at least in some situations, counting may actually underestimate the degree of tissue involvement in EE. Our approach is as follows: if there are fewer than eight eosinophils per 40× field, reflux esophagitis is favored. Likewise, a count of >24 eosinophils per 40× field better supports eosinophilic esophagitis, especially when seen in a biopsy from the mid or proximal esophagus. Clinicopathologic correlation should be attempted for all cases, especially for those with intermediate “counts” (e.g. pH monitoring studies, stigmata of allergic disease, endoscopic impression). Since reporting number of eosinophils can be misleading, we would prefer not to report counts in our practice but we do offer estimates and we attempt to compare biopsies to prior ones, and estimate whether more or fewer eosinophils are present to help assess response to treatment. While we believe that distinguishing between reflux and eosinophilic esophagitis is a worthwhile endeavor (as patients with EE can show dramatic response to steroids), we also believe that making this diagnosis in a vacuum can result in erroneous diagnoses. A conversation with the gastroenterologist might be of benefit when the clinical and/or endoscopic settings are uncertain.
Reflux disease In the United States, gastroesophageal reflux disease 18 is the costliest gastrointestinal disorder and the most common physician diagnosis for gastrointestinal disorders in outpatient clinic visits 19. A 2002 estimate of the direct cost of GERD was over $9 billion 20. Not surprisingly, many patients undergo endoscopic biopsies in the course of evaluation for reflux symptoms. Findings in these biopsies can range anywhere from “classic” reflux changes (basal cell hyperplasia, elongation of vascular papillae, and intraepithelial eosinophils) to ulcers, intraepithelial lymphocytosis (which correlates poorly with pH studies), or “balloon” cells (which appear distended and have pale abundant pink cytoplasm).
The classic view on gastroesophageal reflux disease is that it is a reflection of acid reflux. However, we have become increasingly aware that the refluxed material, that is the contents of the stomach, consists of duodenal contents mixed with gastric secretions. These stomach contents probably contain a higher percentage of duodenal “juice” than in the past because modern patients often take proton pump inhibitors and have the benefit of acid suppression (which reduces the volume of gastric secretions and concentrates the secretions refluxed into the stomach from the duodenum). The question in this proton pump era is, “Are the effects of duodenal reflux more severe than those of acid reflux?” Since it is impossible to separate the effects of gastric acid from bile reflux in humans, which can be done in animals 21-22, this question can only be answered indirectly. At the time when Norman Barrett was writing about the condition that now bears his name 23-24, the complications of reflux were severe erosive esophagitis and debilitating stricture formation. But now that ulcers are readily healed by proton pump inhibitors, the objective is to detect early neoplasia because reflux has been shown as a preneoplastic condition 25.
On epidemiologic grounds, bile is a carcinogen. There are many examples of this phenomenon, but some are especially worth reiterating. For example, patients who have had a cholecystectomy (and who thus secrete constant low levels of bile rather than controlled pulses that accompany boluses of food) have a statistically significant increment in their risk of upper gastrointestinal tract carcinomas 26. Patients who are status post Billroth operations are prone to laryngeal carcinogenesis as well 27. But if the duodenum is diverted in patients who already have had ample injury to the esophagus (e.g., Barrett esophagus with damage to the lower esophageal sphincter), the risk of progression to dysplasia/carcinoma is reduced 28. In fact, regression of Barrett esophagus in obese patients following Roux-en-Y gastric bypass has been documented29-30. Similarly, diverting the duodenum in patients who already have low-grade dysplasia may forestall progression to invasive carcinoma 28.
In studying patterns of injury to the esophagus, or any part of the gastrointestinal tract, it is difficult to separate the cause of the damage when the GI tract has a limited repertoire of responses to a wide variety of insults. For example, we have seen patients with extensive atrophic gastritis (presuming little acid output, if any) whose esophageal biopsies display features identical to those of acid reflux. However, based on the available epidemiologic evidence, it seems likely that the initial injuries may be a result of acid but that carcinogenic bile reflux informs the cycles of repair. Another possibility is that esophageal mucosal injury stems from the combination of bile juices and acidic gastric contents. There is evidence that shows that bile acids combined with an acidic medium (but not either one alone) induces oxidative stress and DNA damage in ex-vivo Barrett tissue as well as in esophageal cell lines 31. For practical purposes, reflux esophagitis (gastroesophageal reflux disease or GERD) is the most common cause of esophagitis. Although most prevalent in adult Caucasian males, reflux esophagitis can occur in both men and women of all races, and even in infants and children.
GERD is caused by reflux of gastric contents into the esophagus, most prominently from reflux of gastric acid and pepsin. However, reflux of alkaline bile and pancreatic secretions (moving from the duodenum into the stomach and then into the esophagus) is increasingly recognized as a contributing factor to esophageal injury in GERD.
Predisposing factors include: decreased tone of the lower esophageal sphincter (LES) from alcohol, medications, hypothyroidism, pregnancy, scleroderma, etc.; interference with the function of a normal LES from nasogastric tubes; hiatal hernia; decreased clearance of refluxed material, as in achalasia; and delayed gastric emptying with accumulation of gastric secretions, as in diabetes. Obesity exacerbates all of the aforementioned and is associated with reflux disease. In many patients, no clear predisposing cause is identified.
Reflux esophagitis is an example of an “-itis” that often lacks a prominent component of inflammation. The pathology reflects injury to the squamous epithelium, followed by attempts of the epithelium to regenerate. Mild features of cellular injury include balloon cells (squamous cells with ballooned cytoplasm from accumulation of plasma proteins) and vascular lakes (dilated small blood vessels in the mucosa, not areas of hemorrhage, which often are seen endoscopically as erythema or redness). Severe injury can result in mucosal sloughing with erosions or ulcers. Regenerative changes include hyperplasia of the basal zone to >15% to 20% of the epithelial thickness. The upper limit of the basal cell layer can be defined as the level above which the nuclei are separated by a distance greater than the nuclear diameter32. There is elongation of the vascular papillae to greater than two thirds of the epithelial thickness. Inflammation is typically mild and includes scattered eosinophils. Less commonly, scattered neutrophils are present and can be more prominent in cases with more severe injury, including erosions and ulcers. Parakeratosis can be a component as well (but foci of parakeratosis should still be screened for fungal organisms). Fairly reproducible criteria that have been established by Fiocca et al as abnormal and associated with clinical reflux include:
Thickened basal layer (>15% or 5-6 layers)
Increased papillary length (>50% of the squamous thickness)
Dilated/widened intercellular spaces (which may appear as “bubbles” or “ladders”)
It is important to keep in mind that the severity and extent of the histologic changes that we see on a biopsy do not necessarily correlate well with the severity of the patient’s symptoms (heartburn or pyrosis).
Severe complications of GERD are unusual. Complications include development of an ulcer, bleeding from an ulcer, and a stricture formation resulting from scarring, due to deep injury. The complication of Barrett esophagus occurs in approximately 10% of patients with symptomatic reflux.
Dysplasia in Barrett Esophagus To help diagnose dysplasia in Barrett esophagus, criteria established at a consensus meeting follow 33. However, after noting these, attention will be drawn to cases that defy classification based on such criteria.
Grading Dysplasia in Barrett Esophagus: Algorithm
The following algorithm is based on four mucosal features in Barrett esophagus. The algorithm presupposes that the biopsy in question is taken from the esophagus containing compatible endoscopic features of Barrett esophagus, and that intestinal metaplasia is found.
1. SURFACE MATURATION compared to the underlying glands.
2. ARCHITECTURE of the glands.
3. CYTOLOGIC FEATURES.
4. INFLAMMATION and erosions/ulcers.
Each feature may vary, but all are combined to arrive at a diagnosis.
“Surface maturation” is assessed at low magnification and confirmed at high magnification. In nondysplastic Barrett esophagus, the proliferating nuclei in the most basal layers of the glands are larger, more hyperchromatic, and more stratified than those at the surface. In contrast, the surface nuclei are generally arranged in a monolayer with polarized basal nuclei. Characteristically, the glands in Barrett esophagus are mildly atypical, especially when viewed in comparison to adjacent nonmetaplastic gastric-type glands (either fundic or cardiac). Thus, an eye-catching feature when scanning a biopsy at low magnification is the tinctorial comparison between the deep portions of the biopsy and the surface. The possible patterns may be any of the following: (a) the glands have proportionally larger nuclei, (b) the glands and surface have similar nuclear size, or (c) the surface has proportionally larger nuclei.
The “architecture” of the tissue is also best assessed at low magnification. The glandular architecture of a biopsy is the relation between the glands and the lamina propria, and also encompasses the outline of the glands. Architectural abnormalities encompass both increased numbers of glands and changes in their shape. In the nondysplastic setting, the glands tend to be round with little budding, and are surrounded by abundant lamina propria. Crowding of normal-appearing glands is considered a mild architectural abnormality. Crowding of abnormal glands is a feature of dysplasia. Cribriform glands, cystic dilation, and necrotic luminal debris are considered severe architectural abnormalities.
“Cytologic features” are assessed at high magnification in zones selected as abnormal during the assessment of surface maturation and architecture. It must be recognized that some degree of nuclear enlargement and atypia is inherent in Barrett metaplasia in the absence of dysplasia, especially in the basal zone and in the columnar epithelium adjacent to squamous mucosa. In summary, cytologic atypia in Barrett esophagus can be due to (a) dysplasia, in which case it should be cytologically and architecturally unequivocal, (b) reactive changes, particularly associated with inflammation, and (c) inherent changes in the deeper glands of Barrett esophagus, in which case the changes are mild and mature towards the surface, or (d) any of the above and without certainty (a situation which would merit an interpretation of indefinite for dysplasia [IND]). Dysplastic cells are generally hyperchromatic. Once the cells have been interpreted as dysplastic, assigning a low- and high-grade category reflects a matter of degree along a morphologic continuum, a point emphasized in the 1988 criteria34. Also included in the assessment of cytologic features is the relationship of one nucleus to another, referred to as nuclear polarity. In “normal polarity,” the long axis of the nucleus remains perpendicular to the basement membrane, and the nuclei are aligned parallel to one another. “Loss of nuclear polarity” refers to the loss of this perpendicular orientation and a random, “jumbled” appearance of the nuclei, in relation to the basement membrane and one another.
“Inflammation and erosions/ulcers” are a potential component that add difficulty and are assessed at both scanning and high magnification. They can obscure a truly malignant lesion or impart worrisome cytologic alterations that are attributable to a reparative process.
Applying the Algorithm to Classify Dysplasia Barrett Esophagus: Negative for Dysplasia
In Barrett esophagus without dysplasia, the surface appears more mature than the underlying glands. That is, the nuclear-to-cytoplasmic ratio of surface cells is lower than that of the deeper glands. The architecture is normal, with abundant lamina propria between glands. The cytologic features are normal, noting that mitoses may be present in deeper glands, as well as nuclear stratification. The individual nuclei should have smooth nuclear membranes, and nucleoli, if present, should be small with smooth outlines. Nuclear polarity should be maintained in deep and superficial aspects of the biopsy. If inflammation is a component, reparative features may be present. In this setting, nuclear membranes should remain smooth, although the cells may display nuclear-to-cytoplasmic enlargement and nucleoli may become more prominent but retain smooth contours. The surface should show maturation compared to the deeper glands, but there may be some loss of surface mucin. A common pitfall in the diagnosis of Barrett esophagus is the presence of carryover intestinal mucosa from the duodenum since typically the duodenum is biopsied first with the same forceps used to biopsy the esophagus and tissue from the intestine can contaminate the container labeled as “esophagus”. These contaminants are typically small and seen detached from the esophageal mucosa. Some cases have a hypermucinous pattern reminiscent of hyperplastic polyps in the colon.
Barrett Esophagus: Indefinite for Dysplasia
Using the algorithm, the IND category includes cases that had deeper cytologic changes suggestive of dysplasia, but which showed surface maturation. But, other observers have used the indefinite category as a “waste can” 34. Cases indefinite for dysplasia could have normal architecture or some degree of glandular crowding. On cytologic evaluation, lesions could have hyperchromasia, nuclear membrane irregularities, and increased mitoses in the deeper aspects, and these matured to the surface. Loss of nuclear polarity was not a feature of IND. In the presence of inflammation, more striking architectural abnormalities were to be included in the indefinite category. This interpretation can also be applied when tangential embedding does not allow assessment between the glands and the surface. Some cases display peculiar hypermucinous features, and it is unclear whether they are neoplastic or reparative. We often offer an explanation for resorting to this category when it is assigned.
Barrett Esophagus: Low-Grade Dysplasia
In Barrett esophagus with low-grade dysplasia, the surface appears similar to the underlying glands at low magnification, or displays only slight maturation. The architecture may be mildly to markedly distorted with glandular crowding, although lamina propria should be identifiable between glands. The cytologic features are important and the changes should extend at least focally to the surface. This lack of surface maturation need not be diffusely present within all tissue fragments and, in fact, areas of abrupt transition between non-dysplastic and dysplastic epithelium are a clue that the changes are indeed neoplastic as opposed to reactive. Superficially located nuclei are irregular, hyperchromatic, mildly enlarged, and may show some degree of stratification and mucin loss. Mitotic figures may be seen at or close to the surface. More than mild nuclear enlargement is allowed if the other features support an interpretation of low-grade dysplasia. Loss of nuclear polarity and nucleoli are not features of low-grade dysplasia, though nuclear stratification, similar to that seen in colonic adenomas, is within the spectrum of low-grade dysplasia and may be present at the surface. Inflammation is typically minimal; cases with abundant inflammation and the other features of low-grade dysplasia are usually best classified in the IND category. If tangential embedding precludes evaluation of the surface, low-grade dysplasia can be diagnosed in the absence of inflammation if: (a) there are dysplastic features in the deep aspects and (b) the features of high-grade dysplasia (see below) are lacking.
Barrett Esophagus: High-Grade Dysplasia
As in low-grade dysplasia, in high-grade dysplasia, surface maturation is lacking. The architecture may show crowding of cytologically abnormal glands, or be markedly distorted with prominent glandular crowding and little intervening lamina propria. If the cytologic features are sufficiently dysplastic, lesser architectural distortion is acceptable. Most examples, similar to dysplasia in the uterine cervix, do not display prominent nucleoli, which tend to be present in either marked repair or when invasion has begun. Both of these latter situations are associated with ulcers. Nuclei may have either delicately clumped, dark heterochromatin and inconspicuous nucleoli, or prominent, irregular nuclei with irregularly clumped chromatin and irregular nucleoli. Markedly enlarged hyperchromatic cells are a feature of high-grade dysplasia and these may extend to the surface. Loss of nuclear polarity is seen in high-grade dysplasia. Mitoses are readily identifiable. Inflammation is typically minimal. There is some evidence to suggest that high-grade dysplasia, accompanied by an ulcer, is a worrisome feature for an associated, unsampled, invasive carcinoma and we suggest additional biopsies and/or endoscopic ultrasound when we see this pattern.
Certain features, when seen in association with high-grade dysplasia, should provoke concern for the possibility of an underlying, unsampled carcinoma. These include: 1) cribiform architecture, 2) dilated tubules containing necrotic debris, 3) associated ulceration, 4) neutrophils within dysplastic glands, and 5) extension of neoplastic cells into the overlying squamous epithelium in a Pagetoid pattern. Zhu and colleagues demonstrated that identification of 1 of these features in association with high-grade dysplasia is associated with carcinoma in a subsequent resection specimen in 39% of the cases. This figure increased to over 80% when 2 or more of these findings were present. In contrast, cases lacking any of these findings had no carcinoma in the resection specimen. Involvement of the overlying squamous epithelium with neoplastic glands in a Pagetoid pattern was associated with cancer in 100% of the cases35. Other than Pagetoid spread pattern, these features, in fact, overlap with features of intramucosal carcinoma. Likewise, identification of single Paget cells (intraepithelial glandular neoplastic cells) in a biopsy specimen is invariably associated with an underlying adenocarcinoma containing at least a focal, poorly differentiated component 36 Overall, we believe that sufficient cytologic alterations “trump” architectural pattern.
In the past, esophagectomy was typically offered as a treament in patients with high-grade dysplasia. Nevertheless, using modern techniques, endoscopic treatment has become the standard37-38. SEER data show that patients with-high grade dysplasia and early carcinomas have the same mortality whether managed endoscopically or surgically39. More current practice is to sample aggressively for occult invasive carcinoma and perform close follow-up or endoscopic intervention. Knowing local practices may also influence interpretation of such cases.
In addition to endoscopic mucosal resection (EMR), available endoscopic treatments include multipolar electrocoagulation, argon plasma coagulation, photodynamic therapy, radiofrequency ablation, and cryotherapy. There is most experience with photodynamic therapy but radiofrequency ablation is emerging as the preferred technique since it appears to have fewer complications than the others38, 40-41. . A source of trepidation in the past has been that these techniques would fail to ablate dysplastic mucosa underneath squamous mucosa (buried BE). Surely dysplastic mucosa can be encountered underneath squamous mucosa but is usually associated with surface dysplasia, at least in patients who have had photodynamic therapy42.