Tremendous expansion in access to and utilization of endoscopy for primary diagnosis or treatment monitoring of pediatric gastrointestinal (GI) diseases has occurred in the last decade. Although the overall impact of this trend in providing health care to children is undeniably positive, clinicians and pathologists paid little attention to the optimization of practice parameters with respect to the overall clinical utilization, impact of the number of biopsies on patients, or the financial costs associated with endoscopy and diagnostic pathology. No formal study exists to validate the current clinical practices with respect to the diagnostic value of the “routine” biopsy protocols. Remarkably, the tremendous progress made in defining consensus criteria and procedures for diagnosis and classification of pediatric inflammatory bowel disease (1,2) has not yielded specific recommendations for effective tissue sampling strategies.
Although pediatric endoscopy constitutes a small fraction of all GI endoscopies overall (for instance, 12% of all GI endoscopies with biopsy performed at the Massachusetts General Hospital [MGH] in 2007), pediatric GI biopsies often represent a disproportionate fraction of all GI mucosal biopsies submitted for pathological evaluation (for instance, 38% of all GI mucosal biopsies evaluated at MGH in 2007). In addition, as is commonly known to practitioners and formally documented in this report, the majority of pediatric GI biopsies represent normal tissues, which offer little or no new diagnostic information. Even with the most conservative estimates and the acknowledgment that practice parameters in 1 academic center may not represent individual practioner's practice parameters, this implies tens of thousands of nondiagnostic or noninformative pediatric GI biopsies per year. These observations are suggestive of overutilization of resources and may imply real health costs to pediatric patients associated with diagnostic procedures and real financial costs to the health care system associated with procedure time, specimen processing, and pathological evaluation.
To begin to explore this issue, we used a modeling approach to evaluate the application of left-right pooled biopsies (2-region protocol), common in adult GI practice in some institutions, including the MGH, or a more conservative 4-region protocol for nonfocal colonic biopsy in children. Nonfocal colonic biopsies include samples from an endoscopically normal colon, a colon with segmental disease (eg, proctitis), or a diffusely abnormal colon (eg, ulcerative pancolitis). These biopsies represent the bulk of the pediatric GI pathology workload and typically are performed in patients with established or suspected inflammatory diseases of the bowel. We emphasize that screening colonoscopies for dysplasia in the setting of long-standing inflammatory bowel disease and focal biopsies such as biopsies of mass lesions, polyps, and ulcers pose a different diagnostic challenge. We explicitly exclude discussion of such targeted biopsies and endoscopic excisions from this study.
PATIENTS AND METHODS
Patients and Dataset
We retrieved colonoscopy reports from the MGH pathology archives between 2005 and 2006 for pediatric GI patients in whom ≥1 biopsies were taken between the rectum and the cecum. We selected a de-identified series of 100 consecutive colonoscopies in patients with a suspected (new onset) or a known (follow-up) inflammatory disease of the intestines. We excluded 4 colonoscopies in which the only available specimens were endoscopically visible lesions such as polyps or ulcers, which fall into the category of focal biopsies or endoscopic resections. None of the endoscopies were performed for dysplasia screening in the setting of long-standing inflammatory bowel disease, in which multiple nonfocal biopsies are the well-established standard of care. The final dataset included 96 colonoscopies from 96 patients with a median age of 13 years (range 2–21 years) and a male-to-female ratio of 1.7:1. For these 96 patients, we included all “nonfocal” biopsies, defined as biopsies for which there was no designation of a discrete endoscopic lesion such as polyp, pseudopolyp, submucosal nodule, ulcer, or any identification other than the anatomical location. This led to the inclusion of a total of 770 specimens, representing an average of 8.0 specimens (range 4–13) per colonoscopy. Each specimen contained at least 1 and more commonly 2 fragments of tissue, the sum of which was considered 1 “biopsy” for the purpose of this analysis.
To determine how our institutional dataset, which represents a tertiary care center, compares to the general pediatric GI practice, we retrieved all of the pediatric GI pathology cases that were submitted for second opinion of inflammatory bowel disease to MGH between 2004 and 2011. A total of 83 referred cases had a complete colonoscopy as judged by the presence of at least 1 rectal and at least 1 ileal or cecal biopsy. The average number of nonfocal colon biopsy specimens was 6, with a range of 2 to 11. As such, the internal MGH practice appears biased toward a larger number of biopsies than the average pediatric GI practice. This difference highlights lack of commonly accepted practice standards and is of no consequence to our methods, results, or conclusions.
The standard of care in pathological evaluation is to incorporate all of the available diagnostic information (eg, clinical presentation, small intestinal disease, upper GI disease, history of inflammatory bowel disease, perianal fissures or skin tags, and abnormal imaging) in the final diagnostic interpretation. For purposes of our analysis, we focused narrowly on the diagnostic value of histopathology alone and ignored any and all other diagnostic information. This assumption intentionally (and artificially) maximizes the value of the colonic biopsies by implying that the other available data contribute no information of diagnostic value. As such, our results put an upper boundary on the diagnostic value of colonic biopsies.
Assignment of Diagnostic Value to Individual Biopsies
We assigned a graded diagnostic value to each biopsy based on its potential contribution to the overall pathological diagnosis or disease activity. We identified major pathological diagnoses such as chronic active colitis and granulomas that definitively define a disease process as class 1 diagnoses and assigned an arbitrary value of 100 to their first encounter. We assigned each subsequent encounter of the same diagnosis a value equal to 50% of the preceding incidence based on the concept that additional confirmatory biopsies increase confidence and add some incremental value, but not as much value as the prior instance. Thus, the first biopsy showing a granuloma receives a value of 100, the second a value of 50, the third a value of 25, and so on. Note that some important class 1 diagnoses such as rectal sparing and skip lesions require more than 1 biopsy by definition. For instance, 2 left-sided biopsies showing chronic active colitis plus 1 normal left-sided biopsy would be insufficient to diagnose rectal sparing if the biopsies are pooled as 1 specimen. In contrast, the same set of biopsies placed in 3 specimen containers and individually identified as rectum, sigmoid, and descending colon may lead to the identification of rectal sparing or a skip lesion. Although other class 1 diagnoses such as pseudomembranous colitis and dysplasia may occur in general practice, Table 1 lists only the diagnoses we encountered in the present study set.
Class 2 diagnoses included histopathological observations that did not specifically identify a disease process. Included in class 2 were pathological diagnoses such as “nonspecific inflammation” and “focal cryptitis.” We assigned the first instance of a class 2 diagnosis a value of 25 and each subsequent encounter of the same class 2 diagnosis received 50% of the value of the preceding diagnosis. As such, the first biopsy with “focal cryptitis” receives a value of 25, the second a value of 12.5, the third a value of 6.25, and so on. We also included repeated instances of normal biopsies in endoscopically normal colons in class 2. Although we assigned the first “normal” biopsy a value of 100 (ie, the maximum value assigned in class 1), because it confirmed the endoscopic impression of normal, we considered repeated confirmations of normal as diagnostically unnecessary and consequently assigned them to class 2 (ie, the second normal biopsy receives a score of 25).
Assignment of Biopsies to Anatomical Regions
The submitting physicians designated the region from which each sample was collected, including rectum, sigmoid, descending colon, splenic flexure, transverse colon, hepatic flexure, ascending colon, and cecum, with anywhere from 0 to 4 biopsies per region per colonoscopy. We coded pathological diagnoses for each patient in a 20-column table with positions labeled as rectum #1 to #4, sigmoid #1 to #3, descending #1 to #3, splenic flexure, transverse colon #1 and #2, hepatic flexure, ascending colon #1 and #2, and cecum #1 and #2. We assigned the samples to 2 alternative biopsy strategies: 2-region and 4-region. The 2-region protocol represents the left and right colon and it requires a minimum of 1 nonfocal biopsy per region. The 4-region protocol represents 4 anatomical landmarks, and it requires a minimum of 1 nonfocal biopsy per region. For the purposes of this study, we chose the 4 regions as rectum/distal sigmoid, proximal sigmoid/descending colon, mid-colon (transverse colon, hepatic and splenic flexures), and cecum/ascending colon. In each scenario, we assumed that the initial set of protocol biopsies (2 for the 2-region or 4 for the 4-region, depending on the strategy) represents the minimum standard of care and has a combined class 1 diagnostic value of 100, regardless of the histopathological findings. This assumption serves 2 purposes. First, it normalizes the starting diagnostic value of each protocol and enables comparison across the 2 strategies. Second, it conservatively maximizes the diagnostic value of any additional biopsies (above protocol) by assuming that the combined diagnostic value of 2 or 4 initial biopsies can never exceed 100, regardless of the pathological findings and any value they contain as the minimum standard of care.
To assign the actual patient biopsies into the 2- and 4-regions, we assigned random numbers between 1 and 20 to each biopsy (generated using the random sequence generator function in http://www.random.org). For each patient, we identified the 2 (or 4) standard-of-care biopsies as the biopsy in each anatomical region with the lowest assigned random number.
Following random selection of the initial 2- or 4-region protocol biopsies, which we performed for both strategies for all 96 patients, additional biopsies (above protocol) were added to the analysis set one at a time based on the next-lower-assigned random number, until all of the available biopsies in a colonoscopy were assigned. In each step, we calculated the incremental value of the added biopsy (above and beyond the protocol biopsies) based on the diagnostic value guidelines described above. The incremental value included diagnostic information from the initial protocol set, following the value assignment rules described above. As such, the incremental biopsy value would measure the diagnostic value of each additional biopsy in the context of the sum of the pathological data revealed in all of the previous biopsies. In addition, in each step, we calculated the average value of the colonic biopsies for the total biopsy count as the sum of the value of the protocol set (100) plus the incremental value of all of the additional biopsies divided by the total number of biopsies. The average biopsy value should remain at 100 as long as all of the biopsies above protocol provide new class 1 diagnoses, or temporarily increase if a new class 1 diagnosis is incidentally revealed at some point during the random sampling.
The 770 total biopsies included 495 (64%) histologically normal specimens and 275 (36%) with ≥1 pathological diagnoses. Figure 1 shows the average biopsy values for 2- and 4-region protocols, respectively. Under both biopsy strategies, the average value of the biopsies shows a downward trend, indicating that the average diagnostic value per biopsy decreases with the increasing number of biopsies. This suggests that successive samples generally tend to provide confirmatory rather than new diagnostic information.
For the 2-region protocol, we found that 1 additional random biopsy (above the 2 used for the 2 preselected regions) kept the average biopsy value unchanged in 11 cases (11.4%), which suggests added diagnostic value with 1 random biopsy above protocol only in a small fraction of cases. A closer look at these 11 cases revealed that 7 (64%) of these would have received a definitive diagnosis of Crohn disease based solely on the protocol set if other available diagnostic features such as terminal ileitis, which we arbitrarily excluded from consideration, had been taken into account. For the 4-region protocol, we found only 4 cases (4.2%) in which a single random biopsy above the 4 used for the 4-region protocol kept the average biopsy value unchanged. Of these 4 cases, 3 (75%) would have received a confirmed diagnosis of Crohn disease if other available clinical and pathological data, which we arbitrarily excluded from consideration, had been taken into account.
At any point during the analysis, the average biopsy value increased with 1 additional biopsy in 8 cases (8.3%) in the 2-region protocol and 4 cases (4.2%) in the 4-region protocol (highlighted by thick solid lines in Fig. 1). As before, the majority of these cases, including 5 of 8 in the 2-region protocol and 3 of 4 in the 4-region protocol, would have had a confirmed diagnosis of Crohn disease based on either clinical data and/or all of the available diagnostic features, including terminal ileitis and upper GI disease that are not captured in Figure 1. In all of these cases, regardless of a transient constancy or an increase in average biopsy value, the overall shape of the diagnostic value curves maintained a downward trend, consistent with the general trend in all of the data combined.
To better visualize the contribution of nonfocal biopsies to the overall diagnostic content of each colonoscopy, we show the average incremental value of the biopsies above protocol in Figure 2. The y-axis intercept of 100 represents the normalized diagnostic content of the 2- and 4-region protocols at baseline, which we show for emphasis. Additional biopsies above protocol were generally more informative in the 2-region protocol (solid bars) than in the 4-region protocol (open bars). In both scenarios, however, the average incremental value of the additional biopsies remained below 25, suggesting an average information content less than or equal to a class 2 diagnosis. The average incremental value was <10 at +4 and approached 0 at +8, suggesting essentially no diagnostic value when the total number of biopsies exceeds 6 to 8.
Figures 1 and 2 show the analysis of pooled biopsy samples into 2 or 4 specimen containers, consistent with the 2- and 4-region protocols, respectively. As such, we find no justification for physically separating nonfocal biopsy samples above and beyond 4 pooled specimen containers representing 4 main anatomical regions, such as rectosigmoid, left, mid-, and right colon.
Although widespread access to endoscopy has resulted in a tremendous expansion in its diagnostic and therapeutic impact, little attention has been paid to the efficiency and validity of biopsy strategies. A small number of previous studies have addressed the diagnostic value of colonic biopsies in endoscopically normal mucosa (3–8), and suggested pooled 2-region colonic biopsies to reduce histopathology workload (9). In spite of the large and rapidly increasing volume of pediatric GI biopsies, we are unable to identify specific biopsy recommendations based on value or cost of diagnostic information in pediatric GI practice. In the absence of evidence-based standards, sampling strategies vary widely by clinician and practice setting as documented in our analysis.
Although the adult GI practice has been focused traditionally on targeted biopsies to answer specific diagnostic questions, pediatric GI practice is largely dominated by nonfocal biopsies from all anatomic sites visualized at endoscopy. This approach has historically evolved to prevent children from undergoing additional procedures by aspiring to capture all of the possible pathologies at all of the possible anatomic sites. This strategy, however, ignores the wealth of knowledge in clinical medicine and disease pathophysiology that make identification of a completely unexpected disease process based on histology alone exceedingly unlikely. In addition, this approach results in overuse of supporting clinical services such as anesthesia and pathology, increasing the already large financial costs to the health care system with questionable overall value.
In an effort to begin a formal evaluation of the value of pediatric GI biopsies, we chose to address the simple and arguably obvious scenario of nonfocal colonic biopsies. To maximize the potential contribution of histopathology to the final diagnosis, we assumed that the final diagnosis (overall cumulative interpretation of the case) relies entirely on histopathological findings in colon biopsies, which ignores the rich supply of diagnostic information above and beyond histopathology in each and every case. Even under these conditions optimized to falsely elevate the value of colonic histopathology, we find little to no diagnostic value in obtaining numerous (>6) nonfocal biopsies, and no value in physically separating multiple biopsies from the same anatomical region into individual specimen containers.
Further exploration of our dataset reveals that the vast majority of colonoscopies in which additional nonfocal biopsies resulted in the identification of new diagnostic information were in patients with Crohn disease. Crohn disease is in fact a patchy and segmental disease, and it is not surprising that the type and nature of diagnostic information would vary from biopsy to biopsy, hence producing “new” diagnostic information using the methods used in our study. When these cases were revisited in their entirety, however, it became clear that although additional random biopsies occasionally provided “new” diagnostic information, the overarching diagnosis of Crohn disease was not affected or improved because of the new histopathological information. Specifically, the clinical findings such as perianal disease, abnormal imaging, and ileal pathology would have yielded the correct diagnosis regardless of the colonic histopathology. As such, colonic biopsies only contributed to the characterization of colonic disease activity and distribution rather than yielding fundamentally new diagnostic information.
The subjective assignment of diagnostic classes and values represents a limitation of our study. Although different values arguably could be used or different weights given to various encounters of different diagnostic entities, we believe that our conclusions would not significantly change with alternative value propositions that similarly assign decreasing value to repeated instances of confirmatory results. In addition, any alternative value proposition that would include additional clinical or pathological data, such as terminal ileal disease, abnormal imaging, and the like, would obviously reduce the diagnostic value of random colonic biopsies as discussed in the context of Crohn disease patients above. As such, one can consider the trends shown in Figures 1 and 2 being applicable to all comparable analyses, whereas the slopes may vary based on the model details.
Although we recognize that reducing the number of samples may lead to a real loss of valuable information in some cases, our results suggested a small number of instances for which an additional sample beyond the 2- and 4-region protocol added new information with diagnostic value. More important, with increasing availability of complementary diagnostic modalities such as high-resolution radiological imaging and advanced endoscopy techniques, practitioners, institutions, and professional societies must actively assess the costs and value of complementary diagnostic information provided by each of these complementary techniques, including the costs and value of histopathology. We recognize that these analyses must occur first at the level of large institutions and group practices as we have attempted here, but we hope that a convergence of these efforts will eventually result in adoption of national standards. As such, we note a second limitation of our study that only includes data from a single institution. We recognize that pediatric GI practices vary at the level of both the institution and individual practitioner, and our review of consult cases submitted to MGH from various outside sources confirms this variability. This variability is indeed a result of the lack of practice standards and a justification for why assessment and optimization of biopsy strategies are needed.
We suggest that nonfocal colonic biopsies can be safely limited to 4 pooled specimens representing 4 major anatomical segments of the large intestine without any diagnostic compromise. In addition, we find little value in collecting more than 6 to 8 adequate colonic biopsies overall. We suggest that individual pediatric GI clinicians and the institutions in which they practice explore the possibility of moving in this direction if their current practice is significantly different from this strategy. We emphasize that our study does not include targeted biopsies of focal lesions such as polyps and masses, the diagnostic workup of which must be considered separately. In addition, we emphasize that our analysis and conclusions are not relevant to screening for dysplasia in the setting of long-standing inflammatory bowel disease for which large numbers of negative or normal biopsies are of significant diagnostic and prognostic value. Follow-up studies are necessary to determine whether the value of information results obtained for other anatomical sites and scenarios, such as the value of terminal ileal biopsies in the adult inflammatory bowel disease (10,11), also applies to children with suspected inflammatory diseases of the GI tract.
We are grateful to Dr Alan Leichtner, Division of Gastroenterology, Children's Hospital Boston, for helpful discussions and critical reading of this manuscript.