Surg Endosc DOI 10.1007/s00464-014-3969-9
and Other Interventional Techniques
Preoperative endoscopy localization error rate in patients with colorectal cancer Fady Saleh • Thamer Al Abbasi • Michelle Cleghorn M. Carolina Jimenez • Timothy D. Jackson • Allan Okrainec • Fayez A. Quereshy
•
Received: 28 July 2014 / Accepted: 25 October 2014 Ó Springer Science+Business Media New York 2014
Abstract Background Preoperative repeat endoscopy in colorectal cancer (CRC) patients is considered by many to be an integral component of surgical planning. Little is known, however, about the utility of re-endoscopy. Methods A retrospective review of 342 consecutive patients undergoing elective surgical resection for CRC from January 2008 to December 2011 was performed. Patients were included if the initial endoscopist was different than the operating surgeon. A localization error was recorded if the final tumor location identified during surgery was in a different anatomical segment than that identified by endoscopy. The Chi-squared test was used to compare categorical variables. An error rate with a 95 % confidence interval was obtained using the exact binomial distribution. Results 298 patients were identified, 118 (39.6 %) of whom also underwent a preoperative re-endoscopy by the operating surgeon or partner. Nineteen patients had incorrect tumor localization at initial endoscopy, equivalent to a
This work was presented at the 2014 Canadian Society of Surgical Oncology Annual Scientific Meeting and the 14th World Congress of Endoscopic Surgery. F. Saleh T. A. Abbasi M. Cleghorn M. C. Jimenez T. D. Jackson A. Okrainec F. A. Quereshy (&) Division of General Surgery, Toronto Western Hospital, University Health Network, 399 Bathurst Street, Room 8MP-320, Toronto, ON M5T 2S8, Canada e-mail: [email protected] F. Saleh T. A. Abbasi T. D. Jackson A. Okrainec F. A. Quereshy Department of Surgery, University of Toronto, Toronto, ON, Canada
6.4 % error rate (95 % CI 3.88–9.78). In comparison, there were two localization errors on re-endoscopy, 1.69 % (95 % CI 0.21–6.00). Re-endoscopy was found to be protective against localization errors (P \ 0.05), correcting 10 of the 12 errors made at the initial endoscopy. The sensitivity of re-endoscopy as a diagnostic tool to detect errors was 83 % with a corresponding specificity of 100 %. The overall accuracy of re-endoscopy in preventing endoscopic localization errors was 92 % (95 % CI 81–100). Conclusions There is a small but important localization error rate in preoperative endoscopic evaluation of colorectal tumors. Re-endoscopy appears to be safe and may potentially identify and correct these errors and help with preoperative planning at the expense of delaying surgery. Further research is necessary to find ways to improve localization and identify which patients would benefit from re-endoscopy. Keywords error
Colorectal cancer Endoscopy Localization
In 2013, the American Cancer Society estimated approximately 140,000 cases of colorectal cancer (CRC), resulting in 50,830 total deaths [1]. CRC is the third most common cancer in both men and women in the United States. However, cancer incidence rates have been decreasing over the past two decades, partially attributable to increases in the use of CRC screening, and in particular colonoscopy, enabling the detection of polyps before they progress to cancer [2]. Colonoscopy remains the gold standard in diagnosis as well as preoperative tumor localization for malignant lesions of the colon and rectum. However, colonoscopic tumor localization has its limitations with a reported error rate of 11.3–21 % [3–5].
123
Surg Endosc
Given the benefits of laparoscopy and evidence demonstrating equivalent oncologic outcomes compared to open surgery [6–10], there has been a dramatic shift in practice from open to laparoscopic colorectal resections [11]. Preoperative localization has become essential to avoid surgical errors, conversion to open, and to reduce the need for intraoperative colonoscopy. Particularly as we enter the era of increasing proficiency in techniques to tackle larger lesions endoscopically, such as submucosal resection, the need to potentially perform an oncological surgical resection for a non-palpable lesion has brought to the forefront the importance of adequate localization [5]. Preoperative repeat endoscopy has subsequently become an integral part of surgical planning in an effort to reduce localization errors and was previously shown to be 40 % at our institution [12]. Certain interventions, such as colonoscopic tattooing, are considered to be standard practice with regard to preoperative tumor localization [13], failure to do so on the primary colonoscopy being a key indication for repeat endoscopy [12]. Repeat endoscopy may, however, be associated with certain incumbent drawbacks including increased cost, the potential for complications, patient discomfort, and delay to definitive treatment. Little is known with respect to its overall potential in reducing localization errors. The purpose of this study was to establish the initial preoperative colonoscopy localization error rate at a tertiary academic center and to evaluate preoperative reendoscopy as a diagnostic tool in terms of its ability to detect and prevent localization errors at initial endoscopy.
Methods A retrospective cohort study was performed on all patients with suspected CRC based on colonoscopy and who subsequently underwent surgical management at the University Health Network in Toronto, Canada between January 2008 and December 2011. Patients were excluded if they had emergency surgery, non-curative surgery (e.g., palliative), or a diagnosis of anal cancer. Patients were also excluded if the operating surgeon was the initial colonoscopist (42 patients in total), as all patients needed to be candidates for re-endoscopy prior to their elective operation. Additionally, at our tertiary institution, these patients were more likely to have had previous resections and were presenting with metachronous cancers, have atypical presentations such as partial obstructions or palpable masses, or have unusual diagnoses such as carcinoid tumors after appendectomy, making this group very different from typical screening or symptomatic patients. Finally, one additional patient, whose
123
tumor location was not obvious on the initial colonoscopy report, was excluded. The study protocol was approved by the institutional Research Ethics Board. Electronic patient records were used to collect data on baseline patient information and perioperative characteristics. All endoscopy reports, surgeon letters, and operative notes were used to obtain information about the details of the initial colonoscopy, re-endoscopy, and intraoperative findings. Specifically, tumor location, completeness of the procedure, specialty of the endoscopist, and where the procedure was performed (internal or external to the study institution) were recorded. Additionally, details on the endoscopist including years in practice and specialty training were obtained from the publically available registry of the College of Physicians and Surgeons of Ontario. Missing data were uncommon and are reported where relevant. Descriptive statistics were used to define the patient population and to establish our institutional re-endoscopy rate. Means were compared using the Student t test and categorical variables using the Pearson Chi-squared statistic. Median time to surgery was compared using the non-parametric Mann–Whitney test. Endoscopic error was defined as any change of one or more colorectal segments in location from either the initial colonoscopy or the re-endoscopy compared to intraoperative findings, which was considered the gold standard of tumor localization in this study. Colorectal segments were defined as ascending, hepatic flexure, transverse, splenic flexure, descending, sigmoid, and rectum as illustrated in Fig. 1. While other definitions for localization were possible such as distance from the anal verge, colonic segment was the most commonly reported localization method and rarely omitted, whereas distance from anal verge was seldom reported beyond the left side. Additionally, this study did not distinguish between the three levels of the rectum or rectal cancer distance from the anal verge, as this was an observational study and no gold standard for measurement was used to measure height of rectal lesions. 95 % confidence intervals for error rates were obtained from the exact binomial distribution. Univariate analyses were performed to identify predictors of endoscopic errors. Multivariable logistic regression using a step-wise elimination (cut off P \ 0.2) was performed to identify independent predictors of endoscopic errors. Sensitivity and specificity as well as the c-statistic (i.e., area under the receiver operating characteristic curve) were generated for preoperative re-endoscopy as an evaluative tool to identify localization errors of the initial colonoscopy. Statistical significance was set at P \ 0.05. All statistical analyses were performed using Stata/IC (version 12.1, Statacorp, College Station, Texas).
Surg Endosc Fig. 1 Tumor segmental location
Results Patient demographic and endoscopic characteristics are detailed in Table 1. A total of 298 patients met the inclusion/exclusion criteria. The mean age of patients was 64.7 years (SD 12.5), and 54.4 % of patients were female. The tumor location was most commonly right-sided (43.4 %), followed by the rectum (30.6 %), and left sided (25.9 %). A more detailed breakdown of tumor location is detailed in Fig. 1. The majority of patients had their initial colonoscopy performed within the study institution (55.2 %), and gastroenterologists performed 180 (63.4 %) of the initial colonoscopies. Table 1 Patient characteristics
demographics
and
primary
endoscopy
N
298
Mean age, years (±SD)
64.7 (12.5)
Female Site of primary endoscopy
162 (54.4)
Internal
158 (55.2)
External
128 (44.8)
Colonoscopist specialtya General surgeon
104 (36.6)
Gastroenterologist
180 (63.4)
Tumor locationb Right
129 (43.4)
Left
77 (25.9)
Rectum
91 (30.6)
Experience B10 years
61 (20.5)
[10 years
237 (79.5)
Previous colectomy
7 (2.4)
Complete initial colonoscopy
236 (85.5)
N (%) unless otherwise specified as mean (±SD, standard deviation). Missing data included: site: 12, specialty: 14, experience: 14 a b
Colonoscopist performing primary colonoscopy
Right includes ascending, hepatic flexure, and transverse colon; left includes splenic flexure, descending, and sigmoid colon
Table 2 describes the univariate preoperative predictors of localization errors based on the initial colonoscopy. Nineteen (6.4 %) localization errors were identified and confirmed intra-operatively. With the exception of tumor location, none of the variables measured were associated with endoscopic localization error. Notably, site of initial colonoscopy (internal versus external to the institution), OR 0.80 (95 % CI 0.27–2.35; P = 0.644), and specialty of the endoscopist (general surgeon versus gastroenterologist), OR 0.94 (95 % CI 0.28–2.87; P = 0.907), were not associated with localization errors. Using right-sided lesions as the baseline reference group, left-sided lesions had 4.7 times increased risk of a localization error (P = 0.012). Similarly, rectal lesions were also at an increased risk of localization error, but this was not statistically significant (OR 1.82, 95 % CI 0.47–6.96, P = 0.384). In the multivariable analysis, all variables with the exception of tumor location were eliminated from the model based on an a priori cutoff value and did not provide new information, and thus only the results of the univariate analysis are presented here. Figures 2 and 3 illustrate the endoscopic error rates. As previously mentioned, based on intraoperative confirmation of tumor location, the localization error rate of the initial endoscopy was 6.4 % (95 % CI 3.88–9.78). A total of 118 patients (39.6 %) underwent preoperative reendoscopy for surgical planning by the operating surgeon or surgical colleague. The localization error rate of the reendoscopy was 1.69 % (95 % CI 0.21–6.00) and was found to be protective against localization error compared to the initial colonoscopy, P = 0.049 (Fig. 3). While the majority of the 19 localization errors on initial endoscopy differed from the actual intraoperative location by one colorectal segment [11 patients (57.9 %)], 6 (31.6 %) differed by two segments, and 2 (10.5 %) differed by more than two segments. In addition, one patient had wrongsegment surgery, requiring a second resection because of incorrect localization reported as right-sided. More importantly, eleven (3.7 % of the total study population) of
123
Surg Endosc Table 2 Univariate predictors of localization errors Variables
No error
Error
OR (95 % CI)
P value
N (%)
279 (93.6)
19 (6.4)
Mean age (±SD)
64.8 (12.7)
64.1 (10.2)
1.00 (0.96–1.03)
0.810
Male
126 (92.7)
10 (7.4)
0.74 (0.26–2.10)
0.527
Female
153 (94.4)
9 (5.6)
Internal
149 (93.0)
9 (7.0)
0.80 (0.27–2.35)
0.639
External
119 (94.3)
9 (5.7) 0.94 (0.28–2.87)
0.907
Gender
Site of endo
Specialty Surgeon
98 (94.2)
6 (5.8)
GI Location
169 (93.9)
11 (6.1)
Right
125 (96.9)
4 (3.1)
Left
67 (87.0)
10 (13.0)
4.66 (1.41–15.44)
0.012
Rectum
86 (94.5)
5 (5.5)
1.82 (0.47––6.96)
0.384
0.96 (0.29–4.14)
0.948
1.00 (0.00–8.22)*
0.485
1.00
Experience B10 years
57 (93.4)
4 (6.6)
[10 years
222 (93.7)
15 (6.3)
Previous colectomy Yes No
7 (100.0) 272 (93.5)
0 (0.0) 19 (6.5)
OR odds ratio, CI confidence interval, SD standard deviation, GI gastroenterologist * Cornfield estimate Fig. 2 Tumor segmental localization error rates
Fig. 3 Overall localization error rates
123
the 19 errors resulted in a significant change in the planned operative approach: 5 errors resulted in a change from planned sigmoidectomy to low anterior resection, 1 from subtotal to extended right hemi-colectomy, 3 from right to left hemi-colectomy/sigmoidectomy, and 2 from left to right hemi-colectomy. This is likely an underestimate of change in operative procedure as changes from a (1) right to an extended right hemi-colectomy, (2) left to extended right hemi-colectomy for a peri-splenic flexure lesion, and (3) left hemi-colectomy to sigmoidectomy were not
Surg Endosc
included. Additionally, extent of rectal dissection for upper, middle, and lower rectal lesions was not considered in this study. Re-endoscopy was able to find and correct 10 of 12 errors that occurred in the 118 patients who were re-scoped preoperatively. As a diagnostic tool, the sensitivity of reendoscopy on finding localization errors of the initial endoscopy was 83 % with a specificity of 100 % (Table 3). The overall accuracy of re-endoscopy at preventing endoscopic localization errors was 92 % (95 % CI 81–100). Additionally, re-endoscopy identified two cancers previously not seen at initial colonoscopy. However, one of the costs to the patient of re-endoscopy is that it results in a significant delay of the elective operation. The median time to operation in patients undergoing re-endoscopy was 103 (interquartile range [IQR]: 55–154) compared to 41 days (IQR: 25–89), P \ 0.001.
Discussion In this retrospective cohort study of 298 consecutive CRC elective resections, we demonstrate a 6.4 % (95 % CI 3.88–9.78) localization error rate at initial colonoscopy. The localization error rate for the 118 (39.6 %) patients who underwent re-endoscopy was 1.69 % (95 % CI 0.21–6.00). Re-endoscopy was found to be protective against localization error when compared to the initial colonoscopy, P = 0.049. As a diagnostic tool, the sensitivity of re-endoscopy at finding localization errors of the initial endoscopy was 83 % with a specificity of 100 % and an overall accuracy at identifying (and thus preventing) endoscopic localization errors of 92 %. To our knowledge, this is the first study to establish the specificity and sensitivity of re-endoscopy, as well as its overall utility in preoperative CRC tumor localization. Colonoscopic localization errors may have significant deleterious effects on patient care both by potentially increasing the complexity of procedures when detected
Table 3 2 9 2 table for sensitivity and specificity of re-endoscopy to detect localization error at initial endoscopy, n = 118 Re-endoscopy positive for error Error at initial endoscopy* No error at initial endoscopy* Total
Re-endoscopy negative for error
Total
10
2
10
0
106
108
118
180
298
Sensitivity = 10/(10 ? 2) = 0.83; specificity = 106/(106 ? 0) = 1.00 * Determined by definitive intraoperative localization at time of surgery
intra-operatively, as well as the incumbent potential for wrong-segment surgery. In modern literature, colonoscopy has a reported localization error rate of 11.3–21 % [3–5] and an overall accuracy rate of 79.8–81 % [13, 14] even with the use of adjuncts such as tattoo localization and computed tomography [15]. Piscatelli et al. [5] found that their reported localization error rate of 21 % translated clinically to a different surgical procedure than was originally planned in 11 % of cases, and modification of the surgical approach in 4 % of cases, which is similar to the 3.7 % of patients in our study that had a change in their operative procedure. In an effort to avoid these misadventures, particularly in the laparoscopic era with limited tactile ability to palpate lesions, preoperative re-endoscopy has become an integral part of surgical planning and may contribute to further reducing localization errors, especially when adjuncts such as peri-lesional tattooing are employed [13]. There is, however, a paucity of evidence in the literature to support this frequent practice. We have previously established predictors of re-endoscopy in patients with CRC, namely an initial failure to tattoo the lesion, left-sided colonic neoplasms, and a planned laparoscopic resection [12]. With regards to predictors of localization error, we found tumor location to be the only significant predictor with left-sided lesions having 4.7 times increased risk of localization error (P = 0.012). No localization errors were found for right-sided lesions in this series, however. Piscatelli and colleagues have previously demonstrated that patients who had undergone a previous colorectal procedure were 4 times as likely to have an error made in their lesion localization. Our study did not corroborate this finding with only 7 patients (2.3 %) having previous colorectal surgery and no localization errors. Interestingly, they did not consider tumor location in their multivariate model, but did emphasize the particular clinical relevance of rectum localization, as these lesions were associated with the highest number of clinically significant (altered surgical procedure) errors in their series [5]. Despite the reported complication rate of colonoscopy, repeat endoscopic examination was not associated with any reported complications at our institution. It was, however, associated with a statistically significant delay in operative management (P \ 0.001). Current treatment guidelines suggest that all high-risk stage II and stage III patients with CRC receive adjuvant chemotherapy, ideally within an 8-week period postoperatively [16, 17]. Multiple studies have demonstrated that longer time to adjuvant chemotherapy is associated with worse survival among patients with resected CRC [18, 19]. While there is no clear evidence to suggest that delay to operative management from the time of initial diagnosis is clearly associated with a decrease in disease-free or overall survival in colon cancer patients [20–22], there may be an impact on survival in
123
Surg Endosc
rectal cancer with the delay to adjuvant chemotherapy. Delay in surgical management from the time of diagnosis may also contribute greatly to the anxiety and stress experienced by both patients and their family. This study is limited by its retrospective design and single institution experience with a particularly individualized referral pattern owing to regional practices. One difficulty encountered in this study was the lack of standardized reporting, considering the multitude of sites at which the initial colonoscopy was performed. Although we cannot attribute any errors of localization directly to failure in accurate reporting, we view this lack of standardization in reporting practices to be a potential source for error that is independent of colonoscopic technique. Additionally, while the error rate of the operating surgeon performing the repeat endoscopies was found to be statistically lower than that of the initial endoscopist, it is likely that this was aided by the prior knowledge of the presence of a lesion and adjunct studies such as computed tomography or magnetic resonance imaging. One might also speculate that the operating surgeon has a different vested interest in locating the tumor and therefore may spend more time ensuring correct localization, resulting in this improvement. This may introduce bias in the reported accuracy of re-endoscopy to identify localization errors by potentially over-estimating its efficacy. The accuracy of the operating surgeon as the initial endoscopist is a topic that requires additional study, but is beyond the scope of this paper in which there were very few patients screened by the operating surgeon. Although excluding these patients may have introduced a selection bias, they are not a comparable group to patients referred to our center by endoscopists in the community (including endoscopist–surgeons). Furthermore, it is unlikely that patients who had a repeat endoscopy were selected at random. While many surgeons routinely perform re-endoscopy for surgical planning (e.g., tattooing), re-endoscopy would have been expected in cases where there was concern about the completeness of the initial endoscopy or disagreement with subsequent imaging. In conclusion, localization errors appear to be common and can be detected and prevented using re-endoscopy. Reducing localization errors is particularly important with a growing trend toward laparoscopic resection of CRC and advances in endoscopic management of larger polyps and early stage cancers. While re-endoscopy appears to be a safe strategy to reduce these errors, it may result in a delay in surgical management, as shown in this study. Further studies are required to identify which patients would benefit most from preoperative re-endoscopy.
Disclosures Drs. Saleh, Al Abbasi, Jimenez, Jackson, Quereshy, and Cleghorn have no conflicts of interest or financial ties to disclose. Dr. Okrainec reports grants and personal fees from Covidien.
123
References 1. American Cancer Society (2013) Cancer facts and figures 2013. American Cancer Society, Atlanta 2. Winawer SJ, Zauber AG, Ho MN, O’Brien MJ, Gottlieb LS, Sternberg SS, Waye JD, Schapiro M, Bond JH, Panish JF, Ackroyd F, Shike M, Kurtz RC, Hornsby-Lewis L, Gerdes H, Stewart ET (1993) Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med 329:1977–1981 3. Vignati P, Welch JP, Cohen L (1994) Endoscopic localization of colon cancers. Surg Endosc 8:1085–1087 4. Cho YB, Lee WY, Yun HR, Lee WS, Yun SH, Chun HK (2007) Tumor localization for laparoscopic colorectal surgery. World J Surg 31:1491–1495 5. Piscatelli N, Hyman N, Osler T (2005) Localizing colorectal cancer by colonoscopy. Arch Surg 140:932–935 6. The Clinical Outcomes of Surgical Therapy Study Group (2004) A comparison of laparoscopically assisted and open colectomy for colon cancer. N Engl J Med 350:2050–2059 7. Guillou PJ, Quirke P, Thorpe H, Walker J, Jayne DG, Smith AM, Heath RM, Brown JM, MRC CLASICC trial group, (2005) Shortterm endpoints of conventional versus laparoscopic-assisted surgery in patients with colorectal cancer (MRC CLASICC trial): multicentre, randomised controlled trial. Lancet 365:1718–1726 8. Hewett PJ, Allardyce RA, Bagshaw PF, Frampton CM, Frizelle FA, Rieger NA, Smith JS, Solomon MJ, Stephens JH, Stevenson AR (2008) Short-term outcomes of the Australasian randomised clinical study comparing laparoscopic and conventional open surgical treatments for colon cancer: the ALCCaS trial. Ann Surg 248:728–738 9. Veldkamp R, Kuhry E, Hop WC, Jeekel J, Kazemier G, Bonjer HJ, Haglind E, Pahlman L, Cuesta MA, Msika S, Morino M, Lacy AM, COlon cancer Laparoscopic or Open Resection Study Group (COLOR) (2005) Laparoscopic surgery versus open surgery for colon cancer: short-term outcomes of a randomised trial. Lancet Oncol 6:477–484 10. Jayne DG, Thorpe HC, Copeland J, Quirke P, Brown JM, Guillou PJ (2010) Five-year follow-up of the Medical Research Council CLASICC trial of laparoscopically assisted versus open surgery for colorectal cancer. Br J Surg 97:1638–1645 11. Simunovic M, Baxter NN, Sutradhar R, Liu N, Cadeddu M, Urbach D (2013) Uptake and patient outcomes of laparoscopic colon and rectal cancer surgery in a publicly funded system and following financial incentives. Ann Surg Oncol 20:3740–3746 12. Al Abbasi T, Saleh F, Jackson TD, Okrainec A, Quereshy FA (2014) Preoperative re-endoscopy in colorectal cancer patients: an institutional experience and analysis of influencing factors. Surg Endosc. doi:10.1007/s00464-014-3549-z 13. Beretvas RI, Ponsky J (2001) Endoscopic marking: an adjunct to laparoscopic gastrointestinal surgery. Surg Endosc 15:1202–1203 14. Lam DT, Kwong KH, Lam CW, Leong HT, Kwok SP (1998) How useful is colonoscopy in locating colorectal lesions? Surg Endosc 12:839–841 15. Lee J, Voytovich A, Pennoyer W, Thurston K, Kozol RA (2010) Accuracy of colon tumor localization: computed tomography scanning as a complement to colonoscopy. World J Gastrointest Surg 2:22–25 16. National Institutes of Health (1990) Adjuvant Therapy for Patients with Colon and Rectum Cancer: NIH Consensus Statement, Vol. 8: National Institutes of Health, Bethesda (MD), pp 1-25 17. Figueredo A, Charette ML, Maroun J, Brouwers MC, Zuraw L (2004) Adjuvant therapy for stage II colon cancer: a systematic review from the Cancer Care Ontario Program in Evidence-based Care’s Gastrointestinal Cancer Disease Site Group. J Clin Oncol 22:3395–3407
Surg Endosc 18. Biagi JJ, Raphael MJ, Mackillop WJ, Kong W, King WD, Booth CM (2011) Association between time to adjuvant chemotherapy and survival in colorectal cancer: a systematic review and metaanalysis. JAMA 305:2335–2342 19. Bayraktar UD, Chen E, Bayraktar S, Sands LR, Marchetti F, Montero AJ, Rocha-Lima CM (2011) Does delay of adjuvant chemotherapy impact survival in patients with resected stage II and III colon adenocarcinoma? Cancer 117:2364–2370 20. Simunovic M, Rempel E, Theriault ME, Baxter NN, Virnig BA, Meropol NJ, Levine MN (2009) Influence of delays to
nonemergent colon cancer surgery on operative mortality, diseasespecific survival and overall survival. Can J Surg 52:E79–E86 21. Iversen LH, Antonsen S, Laurberg S, Lautrup MD (2009) Therapeutic delay reduces survival of rectal cancer but not of colonic cancer. Br J Surg 96:1183–1189 22. Yun YH, Kim YA, Min YH, Park S, Won YJ, Kim DY, Choi IJ, Kim YW, Park SJ, Kim JH, Lee DH, Yoon SJ, Jeong SY, Noh DY, Heo DS (2012) The influence of hospital volume and surgical treatment delay on long-term survival after cancer surgery. Ann Oncol 23:2731–2737
123
and Other Interventional Techniques
Preoperative endoscopy localization error rate in patients with colorectal cancer Fady Saleh • Thamer Al Abbasi • Michelle Cleghorn M. Carolina Jimenez • Timothy D. Jackson • Allan Okrainec • Fayez A. Quereshy
•
Received: 28 July 2014 / Accepted: 25 October 2014 Ó Springer Science+Business Media New York 2014
Abstract Background Preoperative repeat endoscopy in colorectal cancer (CRC) patients is considered by many to be an integral component of surgical planning. Little is known, however, about the utility of re-endoscopy. Methods A retrospective review of 342 consecutive patients undergoing elective surgical resection for CRC from January 2008 to December 2011 was performed. Patients were included if the initial endoscopist was different than the operating surgeon. A localization error was recorded if the final tumor location identified during surgery was in a different anatomical segment than that identified by endoscopy. The Chi-squared test was used to compare categorical variables. An error rate with a 95 % confidence interval was obtained using the exact binomial distribution. Results 298 patients were identified, 118 (39.6 %) of whom also underwent a preoperative re-endoscopy by the operating surgeon or partner. Nineteen patients had incorrect tumor localization at initial endoscopy, equivalent to a
This work was presented at the 2014 Canadian Society of Surgical Oncology Annual Scientific Meeting and the 14th World Congress of Endoscopic Surgery. F. Saleh T. A. Abbasi M. Cleghorn M. C. Jimenez T. D. Jackson A. Okrainec F. A. Quereshy (&) Division of General Surgery, Toronto Western Hospital, University Health Network, 399 Bathurst Street, Room 8MP-320, Toronto, ON M5T 2S8, Canada e-mail: [email protected] F. Saleh T. A. Abbasi T. D. Jackson A. Okrainec F. A. Quereshy Department of Surgery, University of Toronto, Toronto, ON, Canada
6.4 % error rate (95 % CI 3.88–9.78). In comparison, there were two localization errors on re-endoscopy, 1.69 % (95 % CI 0.21–6.00). Re-endoscopy was found to be protective against localization errors (P \ 0.05), correcting 10 of the 12 errors made at the initial endoscopy. The sensitivity of re-endoscopy as a diagnostic tool to detect errors was 83 % with a corresponding specificity of 100 %. The overall accuracy of re-endoscopy in preventing endoscopic localization errors was 92 % (95 % CI 81–100). Conclusions There is a small but important localization error rate in preoperative endoscopic evaluation of colorectal tumors. Re-endoscopy appears to be safe and may potentially identify and correct these errors and help with preoperative planning at the expense of delaying surgery. Further research is necessary to find ways to improve localization and identify which patients would benefit from re-endoscopy. Keywords error
Colorectal cancer Endoscopy Localization
In 2013, the American Cancer Society estimated approximately 140,000 cases of colorectal cancer (CRC), resulting in 50,830 total deaths [1]. CRC is the third most common cancer in both men and women in the United States. However, cancer incidence rates have been decreasing over the past two decades, partially attributable to increases in the use of CRC screening, and in particular colonoscopy, enabling the detection of polyps before they progress to cancer [2]. Colonoscopy remains the gold standard in diagnosis as well as preoperative tumor localization for malignant lesions of the colon and rectum. However, colonoscopic tumor localization has its limitations with a reported error rate of 11.3–21 % [3–5].
123
Surg Endosc
Given the benefits of laparoscopy and evidence demonstrating equivalent oncologic outcomes compared to open surgery [6–10], there has been a dramatic shift in practice from open to laparoscopic colorectal resections [11]. Preoperative localization has become essential to avoid surgical errors, conversion to open, and to reduce the need for intraoperative colonoscopy. Particularly as we enter the era of increasing proficiency in techniques to tackle larger lesions endoscopically, such as submucosal resection, the need to potentially perform an oncological surgical resection for a non-palpable lesion has brought to the forefront the importance of adequate localization [5]. Preoperative repeat endoscopy has subsequently become an integral part of surgical planning in an effort to reduce localization errors and was previously shown to be 40 % at our institution [12]. Certain interventions, such as colonoscopic tattooing, are considered to be standard practice with regard to preoperative tumor localization [13], failure to do so on the primary colonoscopy being a key indication for repeat endoscopy [12]. Repeat endoscopy may, however, be associated with certain incumbent drawbacks including increased cost, the potential for complications, patient discomfort, and delay to definitive treatment. Little is known with respect to its overall potential in reducing localization errors. The purpose of this study was to establish the initial preoperative colonoscopy localization error rate at a tertiary academic center and to evaluate preoperative reendoscopy as a diagnostic tool in terms of its ability to detect and prevent localization errors at initial endoscopy.
Methods A retrospective cohort study was performed on all patients with suspected CRC based on colonoscopy and who subsequently underwent surgical management at the University Health Network in Toronto, Canada between January 2008 and December 2011. Patients were excluded if they had emergency surgery, non-curative surgery (e.g., palliative), or a diagnosis of anal cancer. Patients were also excluded if the operating surgeon was the initial colonoscopist (42 patients in total), as all patients needed to be candidates for re-endoscopy prior to their elective operation. Additionally, at our tertiary institution, these patients were more likely to have had previous resections and were presenting with metachronous cancers, have atypical presentations such as partial obstructions or palpable masses, or have unusual diagnoses such as carcinoid tumors after appendectomy, making this group very different from typical screening or symptomatic patients. Finally, one additional patient, whose
123
tumor location was not obvious on the initial colonoscopy report, was excluded. The study protocol was approved by the institutional Research Ethics Board. Electronic patient records were used to collect data on baseline patient information and perioperative characteristics. All endoscopy reports, surgeon letters, and operative notes were used to obtain information about the details of the initial colonoscopy, re-endoscopy, and intraoperative findings. Specifically, tumor location, completeness of the procedure, specialty of the endoscopist, and where the procedure was performed (internal or external to the study institution) were recorded. Additionally, details on the endoscopist including years in practice and specialty training were obtained from the publically available registry of the College of Physicians and Surgeons of Ontario. Missing data were uncommon and are reported where relevant. Descriptive statistics were used to define the patient population and to establish our institutional re-endoscopy rate. Means were compared using the Student t test and categorical variables using the Pearson Chi-squared statistic. Median time to surgery was compared using the non-parametric Mann–Whitney test. Endoscopic error was defined as any change of one or more colorectal segments in location from either the initial colonoscopy or the re-endoscopy compared to intraoperative findings, which was considered the gold standard of tumor localization in this study. Colorectal segments were defined as ascending, hepatic flexure, transverse, splenic flexure, descending, sigmoid, and rectum as illustrated in Fig. 1. While other definitions for localization were possible such as distance from the anal verge, colonic segment was the most commonly reported localization method and rarely omitted, whereas distance from anal verge was seldom reported beyond the left side. Additionally, this study did not distinguish between the three levels of the rectum or rectal cancer distance from the anal verge, as this was an observational study and no gold standard for measurement was used to measure height of rectal lesions. 95 % confidence intervals for error rates were obtained from the exact binomial distribution. Univariate analyses were performed to identify predictors of endoscopic errors. Multivariable logistic regression using a step-wise elimination (cut off P \ 0.2) was performed to identify independent predictors of endoscopic errors. Sensitivity and specificity as well as the c-statistic (i.e., area under the receiver operating characteristic curve) were generated for preoperative re-endoscopy as an evaluative tool to identify localization errors of the initial colonoscopy. Statistical significance was set at P \ 0.05. All statistical analyses were performed using Stata/IC (version 12.1, Statacorp, College Station, Texas).
Surg Endosc Fig. 1 Tumor segmental location
Results Patient demographic and endoscopic characteristics are detailed in Table 1. A total of 298 patients met the inclusion/exclusion criteria. The mean age of patients was 64.7 years (SD 12.5), and 54.4 % of patients were female. The tumor location was most commonly right-sided (43.4 %), followed by the rectum (30.6 %), and left sided (25.9 %). A more detailed breakdown of tumor location is detailed in Fig. 1. The majority of patients had their initial colonoscopy performed within the study institution (55.2 %), and gastroenterologists performed 180 (63.4 %) of the initial colonoscopies. Table 1 Patient characteristics
demographics
and
primary
endoscopy
N
298
Mean age, years (±SD)
64.7 (12.5)
Female Site of primary endoscopy
162 (54.4)
Internal
158 (55.2)
External
128 (44.8)
Colonoscopist specialtya General surgeon
104 (36.6)
Gastroenterologist
180 (63.4)
Tumor locationb Right
129 (43.4)
Left
77 (25.9)
Rectum
91 (30.6)
Experience B10 years
61 (20.5)
[10 years
237 (79.5)
Previous colectomy
7 (2.4)
Complete initial colonoscopy
236 (85.5)
N (%) unless otherwise specified as mean (±SD, standard deviation). Missing data included: site: 12, specialty: 14, experience: 14 a b
Colonoscopist performing primary colonoscopy
Right includes ascending, hepatic flexure, and transverse colon; left includes splenic flexure, descending, and sigmoid colon
Table 2 describes the univariate preoperative predictors of localization errors based on the initial colonoscopy. Nineteen (6.4 %) localization errors were identified and confirmed intra-operatively. With the exception of tumor location, none of the variables measured were associated with endoscopic localization error. Notably, site of initial colonoscopy (internal versus external to the institution), OR 0.80 (95 % CI 0.27–2.35; P = 0.644), and specialty of the endoscopist (general surgeon versus gastroenterologist), OR 0.94 (95 % CI 0.28–2.87; P = 0.907), were not associated with localization errors. Using right-sided lesions as the baseline reference group, left-sided lesions had 4.7 times increased risk of a localization error (P = 0.012). Similarly, rectal lesions were also at an increased risk of localization error, but this was not statistically significant (OR 1.82, 95 % CI 0.47–6.96, P = 0.384). In the multivariable analysis, all variables with the exception of tumor location were eliminated from the model based on an a priori cutoff value and did not provide new information, and thus only the results of the univariate analysis are presented here. Figures 2 and 3 illustrate the endoscopic error rates. As previously mentioned, based on intraoperative confirmation of tumor location, the localization error rate of the initial endoscopy was 6.4 % (95 % CI 3.88–9.78). A total of 118 patients (39.6 %) underwent preoperative reendoscopy for surgical planning by the operating surgeon or surgical colleague. The localization error rate of the reendoscopy was 1.69 % (95 % CI 0.21–6.00) and was found to be protective against localization error compared to the initial colonoscopy, P = 0.049 (Fig. 3). While the majority of the 19 localization errors on initial endoscopy differed from the actual intraoperative location by one colorectal segment [11 patients (57.9 %)], 6 (31.6 %) differed by two segments, and 2 (10.5 %) differed by more than two segments. In addition, one patient had wrongsegment surgery, requiring a second resection because of incorrect localization reported as right-sided. More importantly, eleven (3.7 % of the total study population) of
123
Surg Endosc Table 2 Univariate predictors of localization errors Variables
No error
Error
OR (95 % CI)
P value
N (%)
279 (93.6)
19 (6.4)
Mean age (±SD)
64.8 (12.7)
64.1 (10.2)
1.00 (0.96–1.03)
0.810
Male
126 (92.7)
10 (7.4)
0.74 (0.26–2.10)
0.527
Female
153 (94.4)
9 (5.6)
Internal
149 (93.0)
9 (7.0)
0.80 (0.27–2.35)
0.639
External
119 (94.3)
9 (5.7) 0.94 (0.28–2.87)
0.907
Gender
Site of endo
Specialty Surgeon
98 (94.2)
6 (5.8)
GI Location
169 (93.9)
11 (6.1)
Right
125 (96.9)
4 (3.1)
Left
67 (87.0)
10 (13.0)
4.66 (1.41–15.44)
0.012
Rectum
86 (94.5)
5 (5.5)
1.82 (0.47––6.96)
0.384
0.96 (0.29–4.14)
0.948
1.00 (0.00–8.22)*
0.485
1.00
Experience B10 years
57 (93.4)
4 (6.6)
[10 years
222 (93.7)
15 (6.3)
Previous colectomy Yes No
7 (100.0) 272 (93.5)
0 (0.0) 19 (6.5)
OR odds ratio, CI confidence interval, SD standard deviation, GI gastroenterologist * Cornfield estimate Fig. 2 Tumor segmental localization error rates
Fig. 3 Overall localization error rates
123
the 19 errors resulted in a significant change in the planned operative approach: 5 errors resulted in a change from planned sigmoidectomy to low anterior resection, 1 from subtotal to extended right hemi-colectomy, 3 from right to left hemi-colectomy/sigmoidectomy, and 2 from left to right hemi-colectomy. This is likely an underestimate of change in operative procedure as changes from a (1) right to an extended right hemi-colectomy, (2) left to extended right hemi-colectomy for a peri-splenic flexure lesion, and (3) left hemi-colectomy to sigmoidectomy were not
Surg Endosc
included. Additionally, extent of rectal dissection for upper, middle, and lower rectal lesions was not considered in this study. Re-endoscopy was able to find and correct 10 of 12 errors that occurred in the 118 patients who were re-scoped preoperatively. As a diagnostic tool, the sensitivity of reendoscopy on finding localization errors of the initial endoscopy was 83 % with a specificity of 100 % (Table 3). The overall accuracy of re-endoscopy at preventing endoscopic localization errors was 92 % (95 % CI 81–100). Additionally, re-endoscopy identified two cancers previously not seen at initial colonoscopy. However, one of the costs to the patient of re-endoscopy is that it results in a significant delay of the elective operation. The median time to operation in patients undergoing re-endoscopy was 103 (interquartile range [IQR]: 55–154) compared to 41 days (IQR: 25–89), P \ 0.001.
Discussion In this retrospective cohort study of 298 consecutive CRC elective resections, we demonstrate a 6.4 % (95 % CI 3.88–9.78) localization error rate at initial colonoscopy. The localization error rate for the 118 (39.6 %) patients who underwent re-endoscopy was 1.69 % (95 % CI 0.21–6.00). Re-endoscopy was found to be protective against localization error when compared to the initial colonoscopy, P = 0.049. As a diagnostic tool, the sensitivity of re-endoscopy at finding localization errors of the initial endoscopy was 83 % with a specificity of 100 % and an overall accuracy at identifying (and thus preventing) endoscopic localization errors of 92 %. To our knowledge, this is the first study to establish the specificity and sensitivity of re-endoscopy, as well as its overall utility in preoperative CRC tumor localization. Colonoscopic localization errors may have significant deleterious effects on patient care both by potentially increasing the complexity of procedures when detected
Table 3 2 9 2 table for sensitivity and specificity of re-endoscopy to detect localization error at initial endoscopy, n = 118 Re-endoscopy positive for error Error at initial endoscopy* No error at initial endoscopy* Total
Re-endoscopy negative for error
Total
10
2
10
0
106
108
118
180
298
Sensitivity = 10/(10 ? 2) = 0.83; specificity = 106/(106 ? 0) = 1.00 * Determined by definitive intraoperative localization at time of surgery
intra-operatively, as well as the incumbent potential for wrong-segment surgery. In modern literature, colonoscopy has a reported localization error rate of 11.3–21 % [3–5] and an overall accuracy rate of 79.8–81 % [13, 14] even with the use of adjuncts such as tattoo localization and computed tomography [15]. Piscatelli et al. [5] found that their reported localization error rate of 21 % translated clinically to a different surgical procedure than was originally planned in 11 % of cases, and modification of the surgical approach in 4 % of cases, which is similar to the 3.7 % of patients in our study that had a change in their operative procedure. In an effort to avoid these misadventures, particularly in the laparoscopic era with limited tactile ability to palpate lesions, preoperative re-endoscopy has become an integral part of surgical planning and may contribute to further reducing localization errors, especially when adjuncts such as peri-lesional tattooing are employed [13]. There is, however, a paucity of evidence in the literature to support this frequent practice. We have previously established predictors of re-endoscopy in patients with CRC, namely an initial failure to tattoo the lesion, left-sided colonic neoplasms, and a planned laparoscopic resection [12]. With regards to predictors of localization error, we found tumor location to be the only significant predictor with left-sided lesions having 4.7 times increased risk of localization error (P = 0.012). No localization errors were found for right-sided lesions in this series, however. Piscatelli and colleagues have previously demonstrated that patients who had undergone a previous colorectal procedure were 4 times as likely to have an error made in their lesion localization. Our study did not corroborate this finding with only 7 patients (2.3 %) having previous colorectal surgery and no localization errors. Interestingly, they did not consider tumor location in their multivariate model, but did emphasize the particular clinical relevance of rectum localization, as these lesions were associated with the highest number of clinically significant (altered surgical procedure) errors in their series [5]. Despite the reported complication rate of colonoscopy, repeat endoscopic examination was not associated with any reported complications at our institution. It was, however, associated with a statistically significant delay in operative management (P \ 0.001). Current treatment guidelines suggest that all high-risk stage II and stage III patients with CRC receive adjuvant chemotherapy, ideally within an 8-week period postoperatively [16, 17]. Multiple studies have demonstrated that longer time to adjuvant chemotherapy is associated with worse survival among patients with resected CRC [18, 19]. While there is no clear evidence to suggest that delay to operative management from the time of initial diagnosis is clearly associated with a decrease in disease-free or overall survival in colon cancer patients [20–22], there may be an impact on survival in
123
Surg Endosc
rectal cancer with the delay to adjuvant chemotherapy. Delay in surgical management from the time of diagnosis may also contribute greatly to the anxiety and stress experienced by both patients and their family. This study is limited by its retrospective design and single institution experience with a particularly individualized referral pattern owing to regional practices. One difficulty encountered in this study was the lack of standardized reporting, considering the multitude of sites at which the initial colonoscopy was performed. Although we cannot attribute any errors of localization directly to failure in accurate reporting, we view this lack of standardization in reporting practices to be a potential source for error that is independent of colonoscopic technique. Additionally, while the error rate of the operating surgeon performing the repeat endoscopies was found to be statistically lower than that of the initial endoscopist, it is likely that this was aided by the prior knowledge of the presence of a lesion and adjunct studies such as computed tomography or magnetic resonance imaging. One might also speculate that the operating surgeon has a different vested interest in locating the tumor and therefore may spend more time ensuring correct localization, resulting in this improvement. This may introduce bias in the reported accuracy of re-endoscopy to identify localization errors by potentially over-estimating its efficacy. The accuracy of the operating surgeon as the initial endoscopist is a topic that requires additional study, but is beyond the scope of this paper in which there were very few patients screened by the operating surgeon. Although excluding these patients may have introduced a selection bias, they are not a comparable group to patients referred to our center by endoscopists in the community (including endoscopist–surgeons). Furthermore, it is unlikely that patients who had a repeat endoscopy were selected at random. While many surgeons routinely perform re-endoscopy for surgical planning (e.g., tattooing), re-endoscopy would have been expected in cases where there was concern about the completeness of the initial endoscopy or disagreement with subsequent imaging. In conclusion, localization errors appear to be common and can be detected and prevented using re-endoscopy. Reducing localization errors is particularly important with a growing trend toward laparoscopic resection of CRC and advances in endoscopic management of larger polyps and early stage cancers. While re-endoscopy appears to be a safe strategy to reduce these errors, it may result in a delay in surgical management, as shown in this study. Further studies are required to identify which patients would benefit most from preoperative re-endoscopy.
Disclosures Drs. Saleh, Al Abbasi, Jimenez, Jackson, Quereshy, and Cleghorn have no conflicts of interest or financial ties to disclose. Dr. Okrainec reports grants and personal fees from Covidien.
123
References 1. American Cancer Society (2013) Cancer facts and figures 2013. American Cancer Society, Atlanta 2. Winawer SJ, Zauber AG, Ho MN, O’Brien MJ, Gottlieb LS, Sternberg SS, Waye JD, Schapiro M, Bond JH, Panish JF, Ackroyd F, Shike M, Kurtz RC, Hornsby-Lewis L, Gerdes H, Stewart ET (1993) Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med 329:1977–1981 3. Vignati P, Welch JP, Cohen L (1994) Endoscopic localization of colon cancers. Surg Endosc 8:1085–1087 4. Cho YB, Lee WY, Yun HR, Lee WS, Yun SH, Chun HK (2007) Tumor localization for laparoscopic colorectal surgery. World J Surg 31:1491–1495 5. Piscatelli N, Hyman N, Osler T (2005) Localizing colorectal cancer by colonoscopy. Arch Surg 140:932–935 6. The Clinical Outcomes of Surgical Therapy Study Group (2004) A comparison of laparoscopically assisted and open colectomy for colon cancer. N Engl J Med 350:2050–2059 7. Guillou PJ, Quirke P, Thorpe H, Walker J, Jayne DG, Smith AM, Heath RM, Brown JM, MRC CLASICC trial group, (2005) Shortterm endpoints of conventional versus laparoscopic-assisted surgery in patients with colorectal cancer (MRC CLASICC trial): multicentre, randomised controlled trial. Lancet 365:1718–1726 8. Hewett PJ, Allardyce RA, Bagshaw PF, Frampton CM, Frizelle FA, Rieger NA, Smith JS, Solomon MJ, Stephens JH, Stevenson AR (2008) Short-term outcomes of the Australasian randomised clinical study comparing laparoscopic and conventional open surgical treatments for colon cancer: the ALCCaS trial. Ann Surg 248:728–738 9. Veldkamp R, Kuhry E, Hop WC, Jeekel J, Kazemier G, Bonjer HJ, Haglind E, Pahlman L, Cuesta MA, Msika S, Morino M, Lacy AM, COlon cancer Laparoscopic or Open Resection Study Group (COLOR) (2005) Laparoscopic surgery versus open surgery for colon cancer: short-term outcomes of a randomised trial. Lancet Oncol 6:477–484 10. Jayne DG, Thorpe HC, Copeland J, Quirke P, Brown JM, Guillou PJ (2010) Five-year follow-up of the Medical Research Council CLASICC trial of laparoscopically assisted versus open surgery for colorectal cancer. Br J Surg 97:1638–1645 11. Simunovic M, Baxter NN, Sutradhar R, Liu N, Cadeddu M, Urbach D (2013) Uptake and patient outcomes of laparoscopic colon and rectal cancer surgery in a publicly funded system and following financial incentives. Ann Surg Oncol 20:3740–3746 12. Al Abbasi T, Saleh F, Jackson TD, Okrainec A, Quereshy FA (2014) Preoperative re-endoscopy in colorectal cancer patients: an institutional experience and analysis of influencing factors. Surg Endosc. doi:10.1007/s00464-014-3549-z 13. Beretvas RI, Ponsky J (2001) Endoscopic marking: an adjunct to laparoscopic gastrointestinal surgery. Surg Endosc 15:1202–1203 14. Lam DT, Kwong KH, Lam CW, Leong HT, Kwok SP (1998) How useful is colonoscopy in locating colorectal lesions? Surg Endosc 12:839–841 15. Lee J, Voytovich A, Pennoyer W, Thurston K, Kozol RA (2010) Accuracy of colon tumor localization: computed tomography scanning as a complement to colonoscopy. World J Gastrointest Surg 2:22–25 16. National Institutes of Health (1990) Adjuvant Therapy for Patients with Colon and Rectum Cancer: NIH Consensus Statement, Vol. 8: National Institutes of Health, Bethesda (MD), pp 1-25 17. Figueredo A, Charette ML, Maroun J, Brouwers MC, Zuraw L (2004) Adjuvant therapy for stage II colon cancer: a systematic review from the Cancer Care Ontario Program in Evidence-based Care’s Gastrointestinal Cancer Disease Site Group. J Clin Oncol 22:3395–3407
Surg Endosc 18. Biagi JJ, Raphael MJ, Mackillop WJ, Kong W, King WD, Booth CM (2011) Association between time to adjuvant chemotherapy and survival in colorectal cancer: a systematic review and metaanalysis. JAMA 305:2335–2342 19. Bayraktar UD, Chen E, Bayraktar S, Sands LR, Marchetti F, Montero AJ, Rocha-Lima CM (2011) Does delay of adjuvant chemotherapy impact survival in patients with resected stage II and III colon adenocarcinoma? Cancer 117:2364–2370 20. Simunovic M, Rempel E, Theriault ME, Baxter NN, Virnig BA, Meropol NJ, Levine MN (2009) Influence of delays to
nonemergent colon cancer surgery on operative mortality, diseasespecific survival and overall survival. Can J Surg 52:E79–E86 21. Iversen LH, Antonsen S, Laurberg S, Lautrup MD (2009) Therapeutic delay reduces survival of rectal cancer but not of colonic cancer. Br J Surg 96:1183–1189 22. Yun YH, Kim YA, Min YH, Park S, Won YJ, Kim DY, Choi IJ, Kim YW, Park SJ, Kim JH, Lee DH, Yoon SJ, Jeong SY, Noh DY, Heo DS (2012) The influence of hospital volume and surgical treatment delay on long-term survival after cancer surgery. Ann Oncol 23:2731–2737
123
