Context.—Little has been reported on changes in pancreatic pathology practice after implementation of endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA).
Objectives.—We assessed the impact of EUS-FNA on cytologic diagnosis replacing histologic diagnosis for pancreatic disease and determined whether it fulfills Christensen criteria of a disruptive innovation effect.
Design.—Pattern of utilization during 20 years, diagnostic categories, and diagnostic accuracy of pancreatic cytology were compared before and after implementation of EUS-FNA. The disruptive effect of cytology relevant to biopsy was assessed by comparing the utilization trends and the accuracy of diagnosis over time.
Results.—The mean annual volume (standard deviation) of cytologic specimens increased from 24 (11) to 231 (10) after implementation of EUS-FNA, and that of histologic specimens increased from 97 (42) to 377 (148). The average percentage of annual cases managed by following cytology alone was 19% (10) before versus 51% (8) after implementation. The percentage managed by histology alone was 56% before versus 23% after implementation. Non–endoscopic ultrasound-guided fine-needle aspiration cytology decreased from 36% to 1%. Needle biopsies decreased from 7% to 1%, and other biopsy types from 29% to 9%. Unsatisfactory (7% versus 1%), atypical (16% versus 4%), and suspicious (16% versus 3%) diagnoses were significantly reduced. The accuracy of cytologic diagnosis significantly improved: the sensitivity (confidence interval) and specificity (confidence interval) for cancer diagnosis were 55% (38%–70%) and 78% (58%–89%) before versus 88% (84%–91%) and 96% (93%–98%) after implementation, respectively.
Conclusions.—Endoscopic ultrasound-guided fine-needle aspiration improved the accuracy of cytologic diagnosis, reduced the number of indeterminate diagnoses, and replaced the need for tissue biopsy. Given its cost and simplicity as compared with tissue biopsy, this trend represents a disruptive innovation effect.
In its recent Futurescape Conference Series, the College of American Pathologists introduced the concept of disruptive innovation as a means to ex ante (before the fact) predict the impact of new innovations and technologies on the future of pathology practice.1,2 This type of innovation has very specific features, and, unlike sustaining (incremental or radical) innovations, it produces a new product (or service) with secondary values(s) that appeals to new customers (or a fringe market), but not to established customers because the new product, compared with the incumbent product, underperforms regarding the main dimension of performance. Over time, however, the new product improves on the main performance measure and displaces the incumbent product from the market (utilization).3–8 Christensen reviewed hundreds of products in which disruptive innovation played a major role in their demise (for example mainframe computers being replaced by minicomputers, and the latter being replaced by desktops, etc) and he came up with a surprising and counterintuitive notion that companies are destined to fail if they focus on satisfying their best and most loyal customers without paying attention to fringe and low-end customers. The theory and its application, therefore, has generally but not universally been considered strategically important for the survival of new and particularly established products or firms.9–11 Anecdotally, this process is similar to the story of many defunct products and technologies in pathology practice, and has been suggested to play a major role in utilization of new technologies in pathology practice, but to our knowledge the theory has not been supported by empirical data and evidence in this field. In this study, we used data from our institution collected during 20 years and determined the impact of endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) on cytopathology and surgical pathology practice, framing the impact into the disruptive innovation theory.
MATERIALS AND METHODS
The study included all consecutive patients who underwent cytologic or histologic evaluation for pancreatic lesions at the University of Alabama at Birmingham Medical Center, Birmingham, Alabama, in the period from January 1990 to December 2010. Endoscopic ultrasound-guided fine-needle aspiration was introduced in our institution in July 2000. This medical center is the largest reference center for pancreatic cancer in the state and has an active National Institute of Health Pancreatic Specialized Program of Research Excellence grant. Included in the study were all patients with suspicious pancreatic lesions. Transplant and trauma cases were excluded. The most severe diagnosis was considered when there were duplicate specimens from the same lesion. Gastroenterology, surgery, or radiology units ordered the majority of cytology specimens, whereas surgery or radiology units ordered the majority of histology specimens.
Non–endoscopic ultrasound-guided fine-needle aspiration pancreatic samples included fluids, brushes, and smears. Fluids were obtained through endoscopic retrograde cholangiopancreatography or cholangiography and sent to the laboratory fresh, then concentrated and prepared as smears, cytospins, thin layer, and/or cell block. A combination of Diff-Quik and Papanicolaou stain was performed on these preparations. Endoscopic ultrasound-guided fine-needle aspiration smears were air dried before staining with Diff-Quik stains. In some cases, smears were fixed in 95% ethanol for Papanicolaou staining in the laboratory. Cell block preparations were attempted in each case. Flow cytometry and culture for microorganisms were occasionally ordered as ancillary tests. Rapid on-site interpretation was performed in each case of EUS-FNA. A final diagnosis was delivered after thorough microscopic review of all material. Percutaneous fine-needle aspiration was handled much the same as EUS-FNA, absent the ultrasound guidance and rapid interpretation. Needle biopsy, laparoscopic biopsy, and resection specimens were processed for routine gross and microscopic examination using hematoxylin–eosin. Eight cytopathologists before implementation and 5 cytopathologists after implementation reviewed and signed cytologic cases (4 overlapping both periods). On the other hand, 19 surgical pathologists reviewed and signed cases before implementation and 28 surgical pathologists after implementation of EUS-FNA; 80% of cases were signed by 8 surgical pathologists in each period, with 6 of these overlapping both periods.
To frame the impact of EUS-FNA on pathology practice, the physicians were considered as customers to the pathology practice with their respective departments and units as market segments. Because EUS-FNA is an improvement in cytologist techniques, we considered it a sustaining innovation to cytology, making cytology a disruptive innovation relative to histologic methods.3,7 Ten years before and 10 years after implementation of EUS-FNA, the following utilization parameters were collected and compared: the number of ordering physicians (customers) and their representative units (internal market segment), total and annual average of all and each type of pancreatic specimen, diagnostic categories, and accuracy of cytologic diagnoses, using histology as the gold standard. As appropriate, χ2 test or Student t test was used for comparison and statistical significance was considered at P < .05. Disruptiveness of cytology relative to biopsy was assessed by comparing the utilization trends and the accuracy of diagnosis over time. A substitution curve for cytology test orders relative to histology test orders was constructed as suggested by Christensen.5
During 20 years, a total of 2781 (242 before and 2539 after implementation) cytologic and 2338 (729 before and 1609 after implementation) histologic specimens were submitted for analysis. Fourteen physicians ordered the majority (80%) of cytologic tests before implementation of EUS-FNA, compared with 2 physicians ordering the same fraction after implementation (Pareto chart, Figure 1, A). On the other hand, 20 physicians ordered the majority (80%) of histologic tests, compared with 4 physicians ordering the same percentage after implementation (Figure 1, B). Virtually all the specimens came from 5 service units: gastroenterology, surgery, oncology, radiology, and internal medicine departments. After implementation, 92% of cytologic specimens came from gastroenterology compared with 59% before implementation (P < .001). After implementation, 94% of histologic specimens were submitted from surgery compared with 77% before implementation (P < .001). The fraction of all specimens submitted from gastroenterology increased from 16% before implementation (Figure 2, A) to 60% after implementation (Figure 2, B) (P < .001), whereas that submitted from surgery decreased from 64% before implementation (Figure 2, A), to 37% after implementation (Figure 2, B) (P < .001).
Viewed as annual trends, the total number of pancreatic specimens submitted to our laboratory increased gradually from merely 40 specimens in 1990 to 563 in 2010, with a dramatic increase after introduction of the EUS-FNA service (Figure 3). The mean annual volume (standard deviation) of cytologic specimens was 24 (12) before versus 231 (97) after implementation, 9 (6) non–EUS-FNA, 222 (102) EUS-FNA (P < .05), and that of histologic specimens was 73 (31) before versus 146 (59) after implementation (P < .05).
Unlike before implementation, more cytologic than histologic specimens were submitted after implementation of EUS-FNA (Figure 3). Gradually cytology replaced histology for the diagnosis and management decisions of pancreatic disease. Specifically, the average annual percentage (SD) of cases managed following cytology alone was 18% (10%) before versus 51% (8%) after implementation (P < .05) (Figure 4). In contrast, the percentage managed after histology alone was 56% (25%) before versus 23% (5%) after implementation (P < .05) (Figure 4). The substitution of cytology for histology was best illustrated in the Christensen substitution curve (Figure 5), which showed the log ratio of cytology percentage to histology percentage over time. Although the fitted line was not straight as suggested by Christensen, there was a trend upward, indicating continuous replacement of cytology for histology.
EUS-FNA replaced smears, brushings, and percutaneous FNAs, as well as tissue biopsy. Non-EUS FNA decreased from 36% to 1% of total cytologic pancreatic specimens (P < .05). Out of all pancreatic surgical pathology specimens, the total number of needle biopsies decreased from 7% to 1% (P < .05) and other biopsy types decreased from 29% to 9% (P < .05). Whipple procedures increased from 19% to 49% (P < .05), whereas partial resections remained essentially unchanged, 44% versus 41%.
As an example of improvement of accuracy, we assessed EUS-FNA diagnostic performance for ductal carcinoma of the pancreas using histology as a gold standard. After implementation, the unsatisfactory and indeterminate diagnoses decreased significantly: 7% before versus 1% after implementation (P < .05) for unsatisfactory, 16% before versus 4% after (P < .05) for atypical, and 16% before versus 3% after (P < .05) for suspicious diagnoses (Table 1). Additionally, more definitive cytologic diagnoses such as categorization of neoplastic and cystic diseases were observed after implementation (Table 1). The accuracy of cytologic diagnosis improved significantly; for ductal carcinoma, for example, the sensitivity (confidence interval) and specificity (confidence interval) were 54% (37%–70%) and 77% (58%–89%) before versus 88% (83%–91%) and 96% (93%–98%) after, respectively (Table 2). The diagnostic odds ratio (confidence interval) for a cytology positive for neoplasm before and after was 4.0 (1–13) and 200 (94–422) respectively (Table 2).
In this study, we demonstrate the impact of EUS-FNA on the practice of pancreatic pathology. The data presented here confirm the widely held view that EUS-FNA has practically replaced tissue biopsy and non–EUS-FNA cytology for the diagnosis of pancreatic lesions. Because EUS-FNA brought about a much-needed improvement (a sustaining innovation) in accuracy of cytologic diagnosis, it resulted in the displacement of histologic diagnosis in management of some pancreatic lesions. This process of displacement shows the characteristic 5 features of disruptive innovation as described by Christensen in his various writings3–7 (Figure 6). (1) At onset, a disruptive innovation underperforms regarding the main performance measure. As we have seen in this study, the sensitivity and specificity of cytologic diagnosis before EUS-FNA were very low compared with those of histology (the mainstream product). (2) It introduces secondary features that are attractive to some customers. Because of convenience, simplicity, cost, and prevention of unnecessary surgeries (secondary measure of performance), cytologic diagnosis in general will always be potentially disruptive to histologic diagnosis regardless of organ site. In addition, EUS-FNA fits Christensen's description of a disruptive technology: “technically straightforward, consisting of off-the-shelf components put together in a way that was often more simple than prior approaches.” 3 Endoscopic ultrasound-guided fine-needle aspiration consists of 3 technologies, each of which is disruptive in its own way: fiber-optic endoscopy was disruptive to rigid endoscopy, ultrasound to x-ray, and cytology to histology. It appeals to (3), a fringe market, but not to (4), the main users. In this study, cytology was ordered more by endoscopists than by other physicians because of convenience (while performing already planned procedures), simplicity, cost, and speed, particularly in patients who were not candidates for surgery. However, as seen in this study, the clinical decision would be compromised by the low accuracy when the specimens are smears, brushes, or aspirates, which were plagued by sampling issues and frequent lack of definitive diagnoses. Non–EUS-FNA preparations had very low sensitivity and specificity (Table 2). For these reasons, until recently the role of cytologic diagnosis in pancreatic cancer has been questioned, particularly among surgeons, listing causes such as poor performance, misinterpretation, management delay, complications, and dissemination of the disease.12 Finally, (5), over time, disruptive technology improves on the main performance measure to become attractive to the main users and displaces the mainstream product. In this study, we showed the remarkable improvement of sensitivity and specificity of cytologic diagnosis after the introduction of EUS-FNA. This is accompanied by an increase in utilization of cytology and obviates the need for tissue biopsy in a certain set of patients, thus making EUS-FNA/cytodiagnosis a disruptive innovation (Figure 6).
To our knowledge, this is the first example of an ex post (after the fact) prediction of disruptive innovation at work in anatomic pathology. Using empirical data, Christensen showed that disruptive innovation was behind the success of new and demise of incumbent products in hundreds of examples, including patient care and the steel, computer hardware, and service industries.3,7 Two examples he frequently cited as relevant to health care were the balloon angioplasty disrupting bypass surgery and the miniature glucose meter disrupting glucose assay in the clinical laboratory.5 One can think of a handful of other products in pathology practice that may be good examples of such disruptive technology.
In spite of its appeal, there are many critiques, shortcomings, and challenges facing this theory.9–11 Although Christensen has vigorously responded to this point and has published a book on “seeing what is next,” 4,13 still some academics, with strong arguments, doubt the ability of the theory to produce ex ante predictions.10 Proponents of the theory see in it a new way of strategic thinking especially useful in industries where technologic improvement is offered in excess of the consumer demand or need, such as in health care, aviation, and higher education. Detractors, on the other hand, question the generalizability of the theory, the inconsistency of the terms used, sampling methods, and the degree of disruption to firms or products in the market. Its inability to ex ante predict which products will become a “disruptive innovation in particular standouts.“9,10 A product's and the market's features may help in identifying the potentiality, but not the inevitability, of disruption. To this end, Christensen continuously asks readers to differentiate between disruptive and sustaining (radical or incremental) innovations that mean to improve products in the main dimension of performance to serve the same customers in the same market. Framed as such, a technology that automates biomarker assessment in tissue is not a disruptive innovation; neither is an imaging technology that helps cytotechnologists perform slide reading more efficiently. However, a low-cost human papilloma virus test is a disruptive innovation. To help better predict disruption, Christensen used the substitution curve in many disruptive products, such as disk drives, sportswear, and online education.6 In this study, the EUS-FNA substitution for biopsy curve showed some trends, but it did not strictly follow a straight line as suggested by Christensen. It would be interesting to see whether many of the suggested disruptive technologies discussed in the College of American Pathologists Futurescape Forum, such as imaging, telepathology, genome-wide sequencing, and informatics, would follow the features and model suggested above.
Several limitations exist within this study. First, pathology practice was treated as a market with internal customers who could be segmented into different groups depending on the unit of service from which the specimen is ordered. Although the patients are the final customers of pathologists, it has been acceptable to treat physicians as immediate customers. To this end, the internal market in the hospital that describes how different departments and internal customers carry transactions may not follow the dynamics, preference, and constraints of a general market. Second, we treated non–EUS-FNA cytology and EUS-FNA as the same product, with the latter adding value (sustaining improvement) to the former. Additionally, we did not inquire about the indications for EUS-FNA and or tissue biopsy in detail, nor did we stratify our patients into those who had resectable versus unresectable lesions. Because the trends were tracked during a 20-year period, other factors may explain the observed trends, including changes in guidelines, pathology or clinical staff, or referral base to our medical center. Finally, the preference of EUS-FNA over biopsy may change in the future, as has happened to breast FNA, which was replaced by biopsy because of the need for tissue adequate for biomarkers and receptor assessments.
In conclusion, EUS-FNA has quantitatively and qualitatively changed the scope of pancreatic pathology practice in such a way that EUS-FNA may be considered an example of a disruptive innovation. This adds to the already known impact of EUS-FNA or EUS alone on clinical decisions and outcomes of pancreatic disease.14,15 There is a need for more understanding of the mechanisms and processes through which new technologies, disruptive or sustaining, impact pathology practice.
From the Department of Pathology, Division of Anatomic Pathology (Dr Eltoum) and the Department of Pathology (Dr Alston and Ms Roberson), University of Alabama at Birmingham.
The authors have no relevant financial interest in the products or companies described in this article.
Presented in part at the 59th Annual Scientific Meeting of the American Society of Cytopathology; November 4–8, 2011; Baltimore, Maryland.