Missed Diagnoses in Patients with Upper Gastrointestinal Cancers

• Introduction
• Patients and Methods
• Results
• Discussion
• References

Background and Study Aims: A few studies have been published on cancers missed at previous endoscopy, but detailed analyses of the causes for failure were lacking. The aims of our study were to determine the incidence of and causes for failure to detect oesophageal and gastric cancers after referral of patients to a surgical endoscopy unit.
Patients and Methods: Out of a consecutive series of 305 patients diagnosed with oesophageal and gastric cancers, we retrospectively identified patients who had undergone an endoscopy within 3 years before the diagnosis. The timing of previous endoscopies, indications for endoscopy, endoscopic findings and the number of biopsy specimens taken were recorded. Missed diagnoses were categorized as either definitely or possibly missed and the reasons for failure were documented.
Results: Of the 305 patients, 30 (9.8 %) had undergone a minimum of one endoscopy within the previous 3 years, 20 (67 %) of these within the previous 1 year. Sinister symptoms were present at the time of previous endoscopies in 75 % of patients with oesophageal cancer (n = 16) and in 57.2 % of patients with gastric cancer (n = 14). In 56 % of the patients with oesophageal cancers the initial diagnosis was oesophagitis or benign stricture; in 71.4 % of the patients with gastric cancers the initial diagnosis was gastritis, ulcer or ”suspicious lesion”. Among those patients with a definitely missed diagnosis (7.2 %), endoscopist errors accounted for the majority of failures (73 %) and the remainder were due to pathologist errors (27 %).
Conclusions: Missed cancers were a frequent finding in patients with oesophageal and gastric cancer who had undergone previous endoscopy, and errors by the endoscopists accounted for the majority of missed lesions. This study emphasizes the importance of identifying signs of early cancers and of having a low threshold for performing multiple biopsies of any suspicious-looking lesion.

Introduction

In Western countries, cancers of the oesophagus and stomach are usually at an advanced stage at the time of diagnosis. There has been considerable emphasis on early diagnosis of these cancers as this greatly improves the outlook, with 5-year survival rates of 80 - 95 % [1] [2] [3] [4]. However, despite the introduction of open-access endoscopy and the issuing of guidelines for general practitioners encouraging them to refer patients early, there does not appear to have been a dramatic improvement in the detection of early cancers in the UK [5].

This delay in diagnosis of upper gastrointestinal cancers may be explained by a number of factors:

• Patients often ignore their symptoms for some time before presenting to their general practitioner.

• There is a trend among general practitioners to commence patients on symptomatic treatment with proton pump inhibitors or H₂-blockers, which delays early referral.

• Patients may have to be seen in a hospital clinic before referral for endoscopy.

• Cancers may be missed at the time of endoscopy.

The current literature emphasizes the importance of early referral for endoscopy, but there are only a few studies from Western countries which have analysed the reasons for missed diagnosis at endoscopy [5] [6] [7]. In Japan, however, it is well recognised that the diagnosis of early cancer has greatly increased, following the identification of the various subtypes of early gastric cancer, and the routine use of serial gastric imaging techniques during endoscopy [8], and of chromoendoscopy [9] [10]. These two techniques are not widely practised in the UK.

The present study determined the number of patients with oesophageal and gastric cancers who had had a previous endoscopy up to 3 years before their diagnosis. The reasons for the failure to diagnose these cancers at the earlier endoscopies have been analysed.

Patients and Methods

A prospective database of all endoscopic examinations performed in the Surgical Endoscopy Unit was established on the introduction of an open-access endoscopy service at Dumfries and Galloway Royal Infirmary in January 1994. Information obtained from this database was cross-checked against a prospectively collected tumour registry of all upper gastrointestinal cancers which is maintained in the same hospital. This process identified patients with oesophageal and gastric cancers, throughout the entire region, who were diagnosed by endoscopy in the Unit between January 1994 and December 2001. The number of patients who had had at least one previous endoscopy within 3 years before their diagnosis was then determined and the reasons for failure to make a diagnosis at that earlier stage were analysed. Table [1] shows these reasons and the criteria according to which diagnoses were categorized as ‘definitely missed’ or possibly missed’. The number of previous endoscopic examinations within this period were documented for each patient, together with the mode of referral for each examination. Using the endoscopy database, the indications for endoscopy (such as benign or sinister symptoms), the findings and the number of biopsies taken were recorded for each endoscopy. A sampling error was defined as having occurred when fewer than four biopsy specimens were taken at the initial endoscopy from an abnormal area which subsequently corresponded to the site of the cancer. A pathological error was established if on review of the previous biopsy specimens, the pathologists described these as exhibiting features of either severe dysplasia that was suspicious of malignancy, or of intramucosal carcinoma.

The stage of the oesophageal and gastric cancers at the time of eventual diagnosis was established by means of a standard staging protocol and examination of the pathological specimen where possible, using the Union Internationale Contre le Cancer (UICC) TNM classification [11]. Statistical analysis was performed using the chi-squared test, with regard to the number of biopsy specimens taken at endoscopy.

Results

A total of 305 patients with oesophageal or gastric cancer were identified from the database, and among them 30 patients (9.8 %) were found to have had undergone at least one previous endoscopy within the 3-year period leading up to the final diagnosis. The mean age was 71 years (range 44 - 87) with no difference in the sex distribution (16 men, 14 women). The referral patterns for this group of 305 patients were as follows: open-access system, 53 %; outpatient referral, 40 %, and inpatient referral, 7 %.

A total of 40 endoscopic procedures were carried out in these 30 patients, with 24 patients (80 %) undergoing only one previous endoscopy prior to the final diagnosis. The remaining six patients underwent a total of 16 procedures (three patients had two procedures; two patients had three procedures, and one patient had four procedures). Among these patients, 67 % had their previous endoscopy within 1 year before the final diagnosis; 13 % within 1 - 2 years, and 20 % within 2 - 3 years before diagnosis.

There were 16 patients with oesophageal cancer, with involvement of the lower oesophagus and the oesophagogastric junction in ten patients, of the middle third in five patients and of the upper third in one patient (Table [2]). The remaining 14 patients had gastric cancers (Table [3]).

The indications for both the initial and the final endoscopy examinations are shown in Table [4]. Among the patients with oesophageal cancer, benign symptoms accounted for only 25 % of the indications at the initial endoscopy. However, sinister symptoms (dysphagia, weight loss, anorexia, anaemia, haematemesis or vomiting) were present in 75 % of patients even at the time of the initial endoscopy, and were present in nearly all of the patients at the time of their final endoscopy.

Among the patients with gastric cancer, dyspepsia or pain, without sinister symptoms, was the main feature in 35.7 % of patients at initial endoscopy but in only 14.3 % at final diagnosis. Sinister symptoms were present in 57.2 % and 71.4 % of patients at the initial and final endoscopy respectively. Two patients (14.3 %) had their cancers identified while they were receiving planned surveillance.

The findings at previous endoscopies as well as the site of the subsequently detected cancer, in patients with a final diagnosis of oesophageal cancer, are listed in Table [2]. A diagnosis of oesophagitis or stricture was made in nine patients (56.2 %), hiatus hernia in two patients (12.5 %) and gastric abnormalities in two patients (12.5 %). Endoscopy findings were reported as normal in three patients (18.8 %). One of the oesophagitis/stricture patients (no. 3) also had a hiatus hernia. Among the patients with gastric cancer (Table [3]), gastritis (generalized or localized inflammation) was seen in seven patients (50 %), prepyloric ulcer in one patient (7.2 %), suspicious lesions in two patients (14.2 %) and oesophageal abnormalities in three patients (21.4 %), with endoscopy findings reported as normal in one patient (7.2 %).

At previous endoscopies, 77 % of patients had fewer than four biopsy samples taken (inadequate biopsies), compared with 37 % at the final endoscopy (Figure [1]); adequate biopsies were taken in 23 % of patients at the initial endoscopy and in 63 % of patients at the final endoscopy. This difference was statistically significant, with P < 0.0001.

Overall, the missed diagnoses at previous endoscopies were categorized as definitely missed in 22 patients (7.2 %) and possibly missed in the remaining eight patients (2.7 %). The causes for the missed diagnoses and their categorizations are outlined in Tables [2], [3]. In the definitely missed diagnosis group, the majority of the cancers were missed due to endoscopist or clinician errors (73 %), pathologist errors accounting for the remaining 27 % (Table [5]). Among the missed diagnoses attributed to endoscopist or clinician error, sampling errors and unacceptable delays in follow-up accounted for 9 % each.

In the possibly missed diagnosis group, no abnormalities had been seen at previous endoscopies in five patients (62.5 %), biopsies were benign or inconclusive in two patients (25 %) and, in the remaining patient (12.5 %), inflammatory changes (gastritis) were seen but no biopsy was taken, probably because the lesion was thought to be insignificant. In this group, the previous endoscopy had been performed between 1 and 3 years before the final diagnosis and these were therefore all classified as possibly missed diagnoses.

The TNM stages of the oesophageal and gastric cancers are shown in Tables [2], [3] respectively. Of the patients with oesophageal cancer, only two patients had T2 disease, 12 patients had either T3 or T4 disease, and staging was not carried out for two patients. The picture was similar in the gastric cancer group, two patients having T1 disease and the remaining 12 patients having T3 or T4 disease.

According to UICC staging, the oesophageal cancers were distributed as follows: stage 0, stage 1, n = 0; stage IIA, n = 3 patients (18.7 %); stage IIB, n = 0; stage III, n = 9 patients (56.3 %); stage IV, n = 2 patients (12.5 %), and not staged, n = 2 patients (12.5 %). The gastric cancers were distributed as follows: stage 0, n = 0; stage IA, n = 2 patients (14.3 %); stage IB, n = 0; stage II, n = 2 patients (14.3 %); stage IIIA, n = 3 patients (21.4 %), stage IIIB, n = 2 patients (14.3 %), and stage IV, n = 5 patients (35.7 %).

Discussion

Only a few reports have been published regarding the incidence of false-negative or missed diagnoses of gastrointestinal cancers at previous endoscopies [5] [6] [7] [8] [12] [13] [14]. Hosokawa et al. [8], in their study, reported false-negative rates of 19 %. Of the 155 endoscopic examinations which were reported as false-negative, 111 procedures had been carried out within 3 years prior to the final diagnosis. The false-negative rates were higher in patients who had had a single previous endoscopy compared with patients who had undergone two or more previous examinations. Amin et al. [13] found that the diagnosis of gastric adenocarcinoma was missed at first endoscopy in 16 % of cases. Suvakovic et al. [5] reported that, among patients presenting with advanced disease within 3 - 4 years of a previous upper gastrointestinal investigation, the diagnosis of early gastric cancer was missed during the first examination in one in six patients (16.7 %). Our study has shown an overall definitely missed diagnosis rate of 7.2 %, and we feel that the diagnosis may have been missed at previous endoscopies in a further 2.7 %. We also noted that the majority of patients (80 %) had undergone a single previous endoscopy prior to the final diagnosis.

In our study, we considered a 3-year period prior to the final endoscopy as similar time periods had been used by previous authors [8] [15]. This time period was based on the tumour-doubling times, and Fujita [16] reported that the doubling time for gastric cancer originating on the mucosal surface of the stomach was approximately 2 - 3 years. Hosokawa et al. [8] had used this time-frame when determining the accuracy of gastroscopy in the diagnosis of gastric cancer, and so too had Rex et al. [15] in their study assessing the relative sensitivity of colonoscopy and barium enema in the detection of colorectal cancers.

In our study, 67 % of patients had their previous endoscopies within 1 year before the final diagnosis and, if we consider the tumour-doubling period to be as described above, these lesions should have been present at the previous examinations.

In our definitely missed diagnosis group, endoscopist errors accounted for the majority (73 %) of the missed lesions. Lesions were not seen by the endoscopist in 27 % of cases, and abnormalities in the vicinity of the subsequent cancer were seen but no biopsies taken in a further 14 %. It is likely that the endoscopist missed or did not recognize the early signs of the subsequent cancers, or the signs may have been masked by antisecretory therapy. The problems involved in recognising these early signs have been highlighted by Bramble et al. [17], and the value of additional techniques, such as chromoendoscopy, is well established [9] [10]. However, chromoendoscopy is not commonly used in the UK compared with countries such as Japan, where the prevalence of this disease is much higher.

Sampling errors accounted for 9 % of definitely missed diagnoses, and in a further 14 %, although an adequate number of biopsy samples were taken, the histological findings were inconclusive. In this group of patients, either the biopsy specimens may not have been entirely representative of the cancerous lesion or the numbers of biopsy samples were insufficient to enable diagnosis of the cancer. Previous studies have shown that adequate numbers of biopsy samples should be taken in order to improve diagnostic accuracy. Lal et al. [18] evaluated the optimal number of biopsy specimens which should be taken with regard to oesophageal cancers, and found that if two biopsy specimens were taken the diagnosis was achieved in 95.8 % of cases. This rate increased to 97.9 % if four specimens were taken and to 100 % for six specimens. Likewise in the case of gastric cancers, 97 % accuracy could be achieved when five biopsy specimens were taken [12]. For the purpose of our study, we considered that at least four biopsy specimens should be have been taken for adequate sampling. We found that inadequate biopsy sampling was more common at previous endoscopies than at the final diagnostic endoscopy (77 % vs. 37 %) and that this was statistically significant.

Pathologist errors were responsible for a definitely missed diagnosis in six patients (27 %). When they reviewed the biopsy specimens from previous examinations, the pathologists noted that cancers had been missed in four patients. At subsequent diagnosis, one of these patient had a malignant maltoma and the other three patients had gastric adenocarcinoma.

The staging of the cancers showed that the majority of oesophageal and gastric cancers were advanced at the time of diagnosis. This high percentage of advanced tumours in our patients reflects the pattern found in Western countries and, despite the introduction of an open-access endoscopy system (accounting for 53 % of referrals in our study), sadly, the detection rates for early cancers have not improved to any significant degree [5]. Although the view that early cancers can present with benign symptoms (such as dyspepsia) in the absence of sinister symptoms is controversial, it is worthy of comment that no fewer than 30 % of our patients presented in this way initially. By the time of they were finally diagnosed, however, only 7 % of the patients had no sinister symptoms.

In summary, we found that nearly 10 % of patients with upper gastrointestinal malignancy had had at least one previous endoscopy within the 3-year period leading up to their diagnosis, with an overall rate for definitely missed diagnoses of 7.2 %, and with errors by the endoscopist accounting for the majority of the missed diagnoses.

We believe that endoscopists could avoid misdiagnosis of oesophageal and gastric cancers by performing a careful examination, by developing their ability to recognize the early signs of cancer, by lowering the threshold for performing a biopsy, and by taking adequate numbers of biopsy specimens in order to reduce sampling errors. Additional techniques, such as chromoendoscopy might be useful for enhancing diagnostic accuracy.

References

• 1 Saragoni L, Gaudio M, Vio A. et al . Early gastric cancer in the province of Forli: follow-up of 337 patients in a high-risk region for gastric cancer.  Oncol Rep. 1998;  5 945-948
  PubMedSearch in Google Scholar
• 2 Muller J M, Erasmi H, Stelzner M. et al . Surgical therapy of oesophageal carcinoma.  Br J Surg. 1990;  77 845-857
  PubMedSearch in Google Scholar
• 3 Allum W H, Powell D J, McConkey C C. et al . Gastric cancer: a 25-year review.  Br J Surg. 1989;  76 535-540
  PubMedSearch in Google Scholar
• 4 Zhang Z, Wan J, Zhu C. et al . Direct gastroscopy for detecting gastric cancer in the elderly.  Chin Med J. 2002;  115 117-118
  PubMedSearch in Google Scholar
• 5 Suvakovic Z, Bramble M G, Jones R. et al . Improving the detection rate of early gastric cancer requires more than open-access gastroscopy: a 5-year study.  Gut. 1997;  41 308-313
  CrossrefPubMedSearch in Google Scholar
• 6 Dekker W, Tytgat G N. Diagnostic accuracy of fiberendoscopy in the detection of upper intestinal malignancy: a follow-up analysis.  Gastroenterology. 1977;  3 1415-1420
  PubMedSearch in Google Scholar
• 7 Llanos O, Guzman S, Duarte I. Accuracy of the first endoscopic procedure in the differential diagnosis of gastric lesions.  Ann Surg. 1982;  195 224-226
  PubMedSearch in Google Scholar
• 8 Hosokawa O, Tsuda E, Kidani K. et al . Diagnosis of gastric cancer up to 3 years after negative upper gastrointestinal endoscopy.  Endoscopy. 1998;  30 669-674
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Bhunchet E, Hatakawa H, Sakai Y. et al . Fluorescein electron endoscopy: a novel method to detect early gastric cancer not evident to routine endoscopy.  Gastrointest Endosc. 2002;  55 562-571
  CrossrefPubMedSearch in Google Scholar
• 10 Yokoyama A, Ohmori T, Makuuchi H. et al . Successful screening for early oesophageal cancer in alcoholics using endoscopy and mucosal iodine staining.  Cancer. 1995;  76 928-934
  PubMedSearch in Google Scholar
• 11 UICC .TNM classification of malignant tumours. 5th edn. Berlin; Springer-Verlag 1997
  Search in Google Scholar
• 12 Tatsuta M, Iishi H, Okuda S. et al . Prospective evaluation of diagnostic accuracy of gastrofibrescopic biopsy in diagnosis of gastric cancer.  Cancer. 1989;  63 1415-1420
  CrossrefPubMedSearch in Google Scholar
• 13 Amin A, Gilmour H, Graham L. et al . Gastric adenocarcinoma missed at endoscopy.  J R Coll Surg Edinb. 2002;  47 681-684
  PubMedSearch in Google Scholar
• 14 Gorski T F, Rosen L, Reither R. et al . Colorectal cancer after surveillance colonoscopy: false-negative examination or fast growth?.  Dis Colon Rectum. 1999;  42 877-880
  CrossrefPubMedSearch in Google Scholar
• 15 Rex D G, Rahmani E Y, Haseman J H. et al . Relative sensitivity of colonoscopy and barium enema for detection of colorectal cancer in clinical practice.  Gastroenterology. 1997;  112 17-23
  CrossrefPubMedSearch in Google Scholar
• 16 Fujita S. Biology of early gastric carcinoma.  Pathol Res Pract. 1978;  163 297-309
  CrossrefPubMedSearch in Google Scholar
• 17 Bramble M G, Suvakovic Z, Hungin A PS. Detection of upper gastrointestinal cancer in patients taking antisecretory therapy prior to gastroscopy.  Gut. 2000;  46 464-467
  CrossrefPubMedSearch in Google Scholar
• 18 Lal N, Bhasin D K, Malik A K. et al . Optimal number of biopsy specimens in the diagnosis of carcinoma of the oesophagus.  Gut. 1992;  33 724-726
  CrossrefPubMedSearch in Google Scholar

C. D. Auld, Consultant Surgeon

Department of General Surgery, Dumfries and Galloway Royal Infirmary

Bankend Road · Dumfries GD1 4AP · Scotland, United Kingdom

Fax: +44-1387-241088 ·

Email: c.auld@dgri.scot.nhs.uk

References

• 1 Saragoni L, Gaudio M, Vio A. et al . Early gastric cancer in the province of Forli: follow-up of 337 patients in a high-risk region for gastric cancer.  Oncol Rep. 1998;  5 945-948
  PubMedSearch in Google Scholar
• 2 Muller J M, Erasmi H, Stelzner M. et al . Surgical therapy of oesophageal carcinoma.  Br J Surg. 1990;  77 845-857
  PubMedSearch in Google Scholar
• 3 Allum W H, Powell D J, McConkey C C. et al . Gastric cancer: a 25-year review.  Br J Surg. 1989;  76 535-540
  PubMedSearch in Google Scholar
• 4 Zhang Z, Wan J, Zhu C. et al . Direct gastroscopy for detecting gastric cancer in the elderly.  Chin Med J. 2002;  115 117-118
  PubMedSearch in Google Scholar
• 5 Suvakovic Z, Bramble M G, Jones R. et al . Improving the detection rate of early gastric cancer requires more than open-access gastroscopy: a 5-year study.  Gut. 1997;  41 308-313
  CrossrefPubMedSearch in Google Scholar
• 6 Dekker W, Tytgat G N. Diagnostic accuracy of fiberendoscopy in the detection of upper intestinal malignancy: a follow-up analysis.  Gastroenterology. 1977;  3 1415-1420
  PubMedSearch in Google Scholar
• 7 Llanos O, Guzman S, Duarte I. Accuracy of the first endoscopic procedure in the differential diagnosis of gastric lesions.  Ann Surg. 1982;  195 224-226
  PubMedSearch in Google Scholar
• 8 Hosokawa O, Tsuda E, Kidani K. et al . Diagnosis of gastric cancer up to 3 years after negative upper gastrointestinal endoscopy.  Endoscopy. 1998;  30 669-674
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Bhunchet E, Hatakawa H, Sakai Y. et al . Fluorescein electron endoscopy: a novel method to detect early gastric cancer not evident to routine endoscopy.  Gastrointest Endosc. 2002;  55 562-571
  CrossrefPubMedSearch in Google Scholar
• 10 Yokoyama A, Ohmori T, Makuuchi H. et al . Successful screening for early oesophageal cancer in alcoholics using endoscopy and mucosal iodine staining.  Cancer. 1995;  76 928-934
  PubMedSearch in Google Scholar
• 11 UICC .TNM classification of malignant tumours. 5th edn. Berlin; Springer-Verlag 1997
  Search in Google Scholar
• 12 Tatsuta M, Iishi H, Okuda S. et al . Prospective evaluation of diagnostic accuracy of gastrofibrescopic biopsy in diagnosis of gastric cancer.  Cancer. 1989;  63 1415-1420
  CrossrefPubMedSearch in Google Scholar
• 13 Amin A, Gilmour H, Graham L. et al . Gastric adenocarcinoma missed at endoscopy.  J R Coll Surg Edinb. 2002;  47 681-684
  PubMedSearch in Google Scholar
• 14 Gorski T F, Rosen L, Reither R. et al . Colorectal cancer after surveillance colonoscopy: false-negative examination or fast growth?.  Dis Colon Rectum. 1999;  42 877-880
  CrossrefPubMedSearch in Google Scholar
• 15 Rex D G, Rahmani E Y, Haseman J H. et al . Relative sensitivity of colonoscopy and barium enema for detection of colorectal cancer in clinical practice.  Gastroenterology. 1997;  112 17-23
  CrossrefPubMedSearch in Google Scholar
• 16 Fujita S. Biology of early gastric carcinoma.  Pathol Res Pract. 1978;  163 297-309
  CrossrefPubMedSearch in Google Scholar
• 17 Bramble M G, Suvakovic Z, Hungin A PS. Detection of upper gastrointestinal cancer in patients taking antisecretory therapy prior to gastroscopy.  Gut. 2000;  46 464-467
  CrossrefPubMedSearch in Google Scholar
• 18 Lal N, Bhasin D K, Malik A K. et al . Optimal number of biopsy specimens in the diagnosis of carcinoma of the oesophagus.  Gut. 1992;  33 724-726
  CrossrefPubMedSearch in Google Scholar

C. D. Auld, Consultant Surgeon

Department of General Surgery, Dumfries and Galloway Royal Infirmary

Bankend Road · Dumfries GD1 4AP · Scotland, United Kingdom

Fax: +44-1387-241088 ·

Email: c.auld@dgri.scot.nhs.uk