Upper GI biopsies for adenocarcinoma – how many biopsies should endoscopists take?

There is evidence that four or five gastric cancer biopsies are required for accurate HER2 interpretation. However, the number of biopsies that need to be taken to reach this number of viable cancer biopsies is without evidence. This study aimed to address this gap by assessing the number of biopsies required to gain at least four viable biopsies containing cancer.


Introduction
Human epidermal growth factor receptor 2 (HER2) testing is a requirement for a significant proportion of gastro-oesophageal adenocarcinomas, as response to trastumuzab is dependent on positive expression. 1 However, expression of HER2 may vary significantly within a tumour, and in a gastric resection specimen a threshold of 10% strong positive staining is regarded as positive, while in biopsies, a cluster of five positive cells showing strong membranous staining is regarded as positive. 2 Initial guidelines which were based on no specific evidence recommended that: 'an adequate number of viable biopsy specimens (ideally [6][7][8] are required'. 3 The recently published British Society of Gastroenterology (BSG) quality standards in upper gastrointestinal endoscopy 4 have also recommended a minimum of eight biopsies. Reported concordance between biopsy and resection cases varies markedly in the literature, at between 45 and 85%. [5][6][7][8] Two studies in particular, both published in 2015, have focused on how many biopsies are required to ensure maximum correlation between biopsies and the resected specimen. Ahn et al. 9 compared 702 paired biopsy and resection cases. They found a high degree of biopsy heterogeneity, but using receiver operating characteristic (ROC) analysis concluded that a minimum of four biopsies were required to minimise discrepancy. Tominaga et al. 10 used a different methodology by examining 84 gastric cancer resection cases and assessing six 'virtual biopsies' distributed evenly through the luminal compartment of the tumour. They concluded that five biopsies would be sufficient for complete correlation with the gastrectomy specimen. Interestingly, they also showed preferential expression in the luminal and lateral aspects of the tumour.
These studies provide vital evidence, but neither have considered the real-world scenario where many of the biopsies which are taken do not contain viable tumour. The objective of this audit study was to look at the actual number of targeted biopsies taken and determine how many of these fragments actually contain viable tumour (i.e. diagnostic). Our aim was to examine if the minimum standard for the recommended number of biopsies is being adhered to and, more importantly, determine the minimum number of biopsies required to obtain diagnostic material and reach the recommended requirement for HER2 testing.

Methods
This audit was registered with the NUH Trust online Clinical Audit system. One hundred and five consecutive cases of biopsy-proven upper gastrointestinal (GI) cancer during 2016 were retrieved using the Winpath laboratory information system. Cases were diagnostic biopsies with a proven diagnosis of upper GI cancer. We therefore did not include cases where the diagnosis of cancer had been missed, or not proved. The cases were then reviewed and the following recorded: 1. The total number of endoscopic biopsies taken. 2. The total number of endoscopic biopsies containing viable tumour. 3. The total number of endoscopic biopsies containing necrotic tumour. 4. The anatomical location, i.e. oesophageal or gastric. The total number of biopsies and percentage with viable tumour was calculated, which enabled determination of the number of biopsies required to be taken to establish a diagnosis of cancer. Further, the data were inspected to determine the number of biopsies required to yield four or five viable tumour fragments. Further, the number of biopsies required for viable tumour fragments was also calculated mathematically ( Figure 1).

Results
We assessed 105 diagnostic upper gastrointestinal biopsies. Thirty-five cases (33%) met the minimum national standards of at least eight biopsies. The relationship between the total number of biopsies and viable tumour fragments is shown in Table 1 viable tumour. On average, the chance of obtaining viable tumour with a single biopsy is therefore approximately 70%, and there is a 30% chance of missing tumour with each biopsy. The chance of missing tumour with multiple biopsies is therefore 0.3 n , etc. This translates to a 97% chance of obtaining at least one viable tumour biopsy with three biopsies. However, for optimal gastric HER 2 testing, evidence suggests that at least four or possibly five tumour biopsies are required. Seventy of 105 (67%) cases had at least four viable tumour biopsies, but only 47 of 105 (45%) had five or more.
Inspection of our data showed that if only four biopsies were taken, seven of 15 had four biopsies with tumour, seven of 14 with five biopsies, nine of 12 with six biopsies and 16 of 17 with seven biopsies (Table 2A). Taking further biopsies beyond seven did not appear to increase the likelihood of reaching four tumour viable biopsies. If, however, five viable tumour biopsies are required, at least 10 biopsies are required for a greater than 90% chance of five viable biopsies (Table 2B). Assuming a fixed probability for a viable tumour biopsy of 0.7, we have constructed a probability tree using the binomial distribution to calculate a 90% chance of obtaining at least four viable tumour biopsies. This amounted to eight biopsies being required for a 94% chance of obtaining at least four viable tumour biopsies. With seven biopsies, the probability of at least four viable tumour biopsies is less than 90%. This theoretical approach assumes a fixed probability for each biopsy and as a model system could be refined with the collection of more data.

Discussion
Trastuzumab (Herceptin) has become the standard of care for first-line chemotherapy in patients with metastatic gastro-oesophageal cancer for the small proportion of tumours which strongly express HER2. 11 Most of these patients will have had biopsies only and will not have undergone prior gastrectomy. Recent studies 9,10 suggest that four or five viable biopsies are required to reach optimal concordance for HER2 status between biopsy and gastrectomy specimen.
This audit study was undertaken to determine how frequently we met this standard, and how many biopsies were required to be taken to reach this number. BSG guidelines recommend taking eight tumour biopsies, but no evidence currently exists to support this. 4   This audit has shown that in our centre, which is likely to be representative of the United Kingdom in general, only 33% of all endoscopies met the required standard of eight biopsies. Seventy per cent of targeted biopsies contained diagnosable cancer which, all things being equal, would translate to three biopsies being required for a 97% chance of diagnostic success. Few previous studies have looked at this. In a prospective study, Choi et al. 12 showed very high diagnostic rates when three to four biopsies were taken in gastric and colorectal cancer, but a previous study in rectal cancer showed a less than 80% diagnostic rate in colorectal cancer with six biopsies, 13 while an even older upper GI study recommended seven biopsies for a greater than 98% chance of diagnosis. 14 We are not aware of any studies which have looked at the optimal number of biopsies to be taken to ensure that the target of four or five viable tumour biopsies are collected for HER2 testing. The evidence here suggests that a minimum of eight targeted biopsies should yield at least four viable tumour biopsies in more than 90% of cases, while five viable biopsies require at least 10 biopsies to be taken for a greater than 90% chance of success. Although there is some variation in the literature on the optimal number of viable biopsies needed, the current BSG upper GI quality standards guidelines recommendation of eight biopsies now has solid evidence to support it. The German guidelines similarly recommend eight biopsies be taken. 15 This study was limited by not including cases of intended tumour biopsy which did not yield a positive diagnosis at all and had to be repeated. However, it is likely that the same principle will apply to these cases, and that taking at least eight biopsies will avoid the requirement for repeat endoscopy other than in exceptional cases. It is likely that the biopsy yield for colorectal cancer will be even lower, as these tumours often have a large residual adenomatous component which may interfere with sampling the invasive component. It would certainly be worth conducting a similar study on lower GI cancer.
It is clear that there are multiple factors related to the chance of obtaining an adequate diagnostic biopsy. These include both tumour characteristics as well as endoscopist skill and experience. This study could not account for these factors, but gives a snapshot of actual practice in our institution, where endoscopies are performed by a very wide range of individuals. The evidence presented here suggests that taking at least eight biopsies will accomplish the task satisfactorily in the vast majority of cases. This study may also be of use to those planning the acquisition of fresh biopsy tissue for research, in particular the 100 000 Genomes Project and other large studies seeking to obtain high quality genetic material. As mentioned, the chance of obtaining viable tumour with one targeted biopsy is no better than 70%, so when harvesting upper GI tumour tissue for this purpose at least two and preferably three biopsies should be obtained.
In summary, the results of this audit show that we are currently reaching the national guideline target in only a third of cases, and that only 67% of our cases had sufficient tumour tissue to accurately grade HER2 staining at the four viable tumour biopsy threshold (and only 45% at the five biopsy mark). Furthermore, we have shown that taking at least eight biopsies would probably increase our yield of required tissue to more than 90%, supporting the national guidance.