Ann Surg Oncol DOI 10.1245/s10434-014-3800-9
ORIGINAL ARTICLE – COLORECTAL CANCER
Treatment Delay in Surgically-Treated Colon Cancer: Does It Affect Outcomes? Ramzi Amri, MSc, Liliana G. Bordeianou, MD, Patricia Sylla, MD, and David L. Berger, MD Division of General Surgery & Gastrointestinal Surgery, Massachusetts General Hospital and Harvard Medical School, Boston, MA
ABSTRACT Background. Treatment delay, or the time lapse between diagnosis and surgery, may have a detrimental effect on cancer outcomes. This study assesses the effect of treatment delay on cancer-related outcomes in a large, continuous series of surgically treated colon cancer patients. Methods. All surgical colon cancer cases at our center from 2004 through 2011 were reviewed. Patients who underwent preoperative chemotherapy, emergency admissions, palliative cases, and incidental and postoperative diagnoses were excluded. Treatment delay was correlated with outcomes in univariate and multivariate regression and proportional hazards models. Results. In 769 included patients, for every treatmentdelay quartile increase, odds of death decreased by an odds ratio (OR) of 0.78 (p = 0.001), and metastatic recurrence by OR 0.78 (p = 0.013). Shorter survival duration had a hazard ratio (HR) of 0.81 (p = 0.001) and shorter diseasefree survival HR 0.72 (p \ 0.001). Multivariate regression adjusting for baseline staging greatly reduces these ratios, and makes them non-significant. Similar patterns were shown in high-risk subsets, including stage III disease, ethnic minorities, patients with positive margins, and extramural vascular invasion. Conclusions. The inverse relation between treatment delay and survival and recurrence reflected adequate prioritization of advanced and high-risk cases and concurrently showed that, matched for stage and risk categories, treatment delay was not associated with worse cancer outcomes for patients with colon cancer. A reasonable delay between diagnosis and subsequent surgery is not detrimental to patient outcomes and permits more flexibility in scheduling and Ó Society of Surgical Oncology 2014 First Received: 22 February 2014 D. L. Berger, MD e-mail: [email protected]
justifies allowing time to complete proper preoperative evaluation and staging, improving the quality and safety of resection and treatment. Treatment delay is the time lapse between diagnosis and treatment. Since most malignancies have linear1 or even exponential2 growth models, onset of symptoms is often preceded by a long asymptomatic phase with increasing degeneration.3 The accelerating growth and degeneration is what theoretically makes diagnosis-to-treatment delays especially detrimental in cancer: Treatment delay is likely to take place during a critical phase with high tumor growth rates and marginally effective innate tumor defense mechanisms, increasing the risk of metastasis and narrowing treatment possibilities. Reports on the effect of treatment delay are mixed. Detrimental effects have been described in cancers of the breast,4 prostate,5 rectum, lung and pancreas,6 while other reports find no significant impact of treatment delay on outcomes of cancers of the stomach,6 bronchus,7 and pancreatic head.8 In colon cancer, the impact of treatment delay is controversial,6,9–11 perhaps owing to heterogeneities in tumor biology. The historical consensus for colon cancer is that most cases are relatively slow-growing. These cases result from the adenoma-adenocarcinoma sequence,12 which can take up to 15 years from onset to eventual progression to malignancy.13 However, an increasing number of exceptions with more aggressive behavior challenge this assumption. Prominently validated high-risk examples include cases presenting with tumorpositive lymph nodes, metastatic disease, extramural vascular invasion (EMVI),14 and comorbid inflammatory bowel disease (IBD).15 Patients with aggressive colon cancer forms may be harmed by the assumptions on slow progression if similar treatment delays are applied to them. This paper aims to assess the distribution of treatment delay and its effect on colon cancer outcomes in a large,
R. Amri et al.
continuous series of surgical colon cancer patients treated in a tertiary care center, both in its general population and in high-risk subpopulations. METHODS Patients A contiguous patient series encompassing all colon cancer patients operated on at Massachusetts General Hospital (MGH) from January 2004 through December 2011 was included in a data repository under an Institutional Review Board-approved protocol. We chose to include colon cancer only instead of colon and rectum malignancies as these diseases have considerable differences in treatment regimens and tumor biology.16 Since we investigated the link between treatment delay and surgical outcomes of colon cancer, our focus was on elective cases where the delay is malleable and the primary therapy is surgery. Therefore, of a total of 1,071 cases, 302 were not suitable for the purpose of this research, including urgent admissions requiring immediate treatment (n = 102), patients with postoperative diagnosis (no preoperative confirmation, or incidental discovery; n = 82), referrals without an accurate diagnosis date (n = 70), patients who underwent neoadjuvant chemotherapy (n = 31), and palliative cases (n = 17). Of the 769 patients eligible at baseline, 28 patients without follow-up had to be excluded from further analysis. The remaining 741 patients were electively admitted and underwent surgery for colonic adenocarcinoma with curative intent. Demographics, surgical admission, pathology, and follow-up data were gathered from patient records and the MGH cancer registry. Survival information was also acquired from the Social Security Death Index. Definitions and Objectives Treatment delay was defined as the time interval between diagnosis and surgery. The diagnosis date was considered the date when colon cancer was clinically established. In virtually all cases, this was by a colonoscopy with a (preliminary) pathology indicating adenocarcinoma. Metastatic presentation was defined as any case presenting with metastatic disease, or diagnosed with metastatic disease within 30 days of the operation, allowing for perioperative staging and biopsies to fall into this category. Cases with metastatic disease established later than 30 days postoperatively, without suspicion during admission, are considered metastatic recurrences. Follow-up-related outcomes included overall survival (OS), duration of survival, local and distant recurrence rates, and disease-free interval duration.
Patients potentially at risk for more serious disease were also assessed as subgroups separately from the overall population. These included patients with pre-existing polyposis, pre-existing colon cancer, IBD, symptomatic diagnosis, ethnic minority patients,17 as well as pathological characteristics including American Joint Committee on Cancer (AJCC) stages III and IV, positive resection margins (proximal, distal, and/or radial), and EMVI. Analysis and Statistical Methods All statistical analysis was performed using the SPSS statistical suite (IBM Corp. Released 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.). A p value of below 0.05 (two-tailed) was considered significant. Pearson’s Chi square test was used to compare nominal variables between groups, while a one-way analysis of variance (ANOVA) was used for continuous characteristics. The distribution of treatment delay duration was reviewed statistically (mean, median, standard deviation, and interquartile range [IQR]) and visually over a twodimensional frequency plot, dichotomizing the included population into those diagnosed during admission and those diagnosed before admission. This analysis also reviewed delay distributions in excluded patients with preoperative diagnosis. Links between treatment delay and presentation characteristics were assessed using a binary logistic regression for dichotomous outcomes and a Cox proportional hazards model for time-related continuous outcomes. In order to yield results that are intuitive to interpret, treatment delay was divided over quartiles. This means that the resulting odds ratios (ORs) or hazard ratios (HRs) are the estimate for two subsequent quartiles. Finally, to ensure any significant findings were validated, the links between treatment delay and long-term outcomes were also corrected for covariates that might act as confounders using multivariate models. These covariates included age, comorbidity burden (expressed by the Charlson comorbidity score,18 excluding the contribution of colon cancer to the score), pathological staging, and, lastly, duration of follow-up. Covariates per model were determined by clinical significance and model fit (R2). RESULTS Perioperative Characteristics The distribution of treatment delay is shown on Fig. 1, along with the inclusion flowchart; the figure shows that treatment delay has a log-normal distribution, with a mean delay of 31.5 days (SD 40.1). Table 1 shows detailed baseline
Treatment Delay in Surgically-Treated Colon Cancer
Treatment Delay Categories
120
Inclusion diagram
Emergency diagnosis Palliative case Neoadjuvant chemotherapy Included (In–stay Diagnosis) Included (Preadmission Diagnosis)
Enrollment
Surgically treated colon cancer patients 2004-2011 (n=1071)
Excluded (n=302) • • • • •
Frequency
100
80
Diagnosis date unknown Postoperative dx/confirmation Emergency surgery Palliative surgery Neoadjuvant chemotherapy
(n=70) (n=82) (n=102) (n=17) (n=31)
Included (n=769)
Follow-Up Status st on December 1 2012
60 Alive (n=549)
Deceased (n=220)
Active follow-up
Cause of death
At MGH (n=491) Affiliated/corresponding (n=30)
Colon cancer-related Other/multiple causes
Lost to follow-up:
40
(n=109) (n=101)
Moved outside of region (n=20) Moved outside of country (n=5) Voluntary FU withdrawal (n=3)
20 Survival Analysis Analyzed (n=521) No follow-up (n=28)
0 0
1
10
100
Analyzed (n=741)
Analyzed (n=220)
1000
Treatment delay (days) FIG. 1 Treatment delay distribution and inclusion. The left pane shows all preoperatively diagnosed cases of colon cancer plotted over the x-axis following their respective treatment delay durations. Treatment delay has been scaled logarithmically to allow display of
outliers. This also demonstrates the lognormal distribution of our population. The right pane states the flowchart for inclusions, detailing causes for exclusion. This also details survival and followup status, which are a last cause for further exclusions. dx, diagnosis
characteristics: no unexpected differences were encountered between eligible and excluded patients at baseline. For the 741 patients included for final analysis, baseline characteristics distributed over quartiles are shown in Table 2. Median treatment delays for each quartile were 8, 19, 29, and 55 days, respectively, for quartiles 1 through 4. Demographics were not significantly different between treatmentdelay quartiles, except for fluctuations in median age. In terms of presentation characteristics, there was an increasing trend in comorbidity burden (p = 0.001), and screening diagnoses (p \ 0.001) with longer treatment delay. This came along with a decrease in the number of symptomatic diagnoses (p \ 0.001) and metastatic disease established within 30 days of admission (p = 0.001). Accordingly, in surgical pathology, lower delay quartiles had substantially higher rates of EMVI (p \ 0.001) and stage III disease (p = 0.010), and lower rates of stage I (p \ 0.001) and stage II (p = 0.021) disease. Surgical characteristics were largely similar over treatment-delay quartile intervals, with comparable rates of laparoscopic surgery, conversion, margin positivity, and lymph node yields, except a non-significant higher multivisceral resection rate in the first two quartiles (12.1 and 15.3 % vs. 8.7 and 8.6 %).
Univariate Analysis of Treatment Delay with Outcomes Univariate relationships between long-term outcomes and treatment delay including follow-up, survival, and disease-free survival, as well as quartile differences, are displayed in Fig. 2. In the analyzed population, over a median follow-up of 142 weeks (range 0–430; IQR 170 weeks), the median survival was 189 weeks (range 0–462; IQR 190). Median disease-free survival was 103 weeks (range 0–430; IQR 176). Interestingly, there was a tendency for death and metastatic disease to occur less often in the longer treatmentdelay quartiles. This difference was statistically significant for rates of metastatic recurrence (p = 0.031), OS (p = 0.005), and cancer-related death (p \ 0.001). In univariate regression, this led to lower odds of death and recurrence to occur in the higher quartiles of treatment delay. For every two consecutive treatment-delay quartiles, this included an OR of 0.78 for death (p = 0.001), an OR of 0.78 of metastatic recurrence (p = 0.013), an HR of 0.81 (p = 0.001) for shorter survival duration, and an HR of 0.72 (p \ 0.001) for shorter disease-free survival were identified using Cox regression.
R. Amri et al. TABLE 1 Baseline characteristics and comparison of included and excluded subsets Includeda
Urgent admission
Postoperative diagnosis
Diagnosis date unknown
Neoadjuvant therapy
Palliative surgery
N (%)
769 (71.8)
102 (9.5)
82 (7.7)
70 (6.5)
31 (2.9)
17 (1.6)
–
Age (mean, SD)
67 (±13.8)
68.2 (±14)
65.9 (±14.4)
65.8 (±13.1)
57.6 (±11.8)
63 (±17.8)
0.005
Gender, male (%)
50.8
48.0
51.2
57.1
54.8
52.9
0.89
Ethnicity, non-White (%)
9.8
12.7
6.1
8.6
22.6
23.5
0.051
Symptomatic presentation (%)
62.3
90.2
48.8
45.7
80.6
94.1
\0.001
Metastatic presentation (%)
22.1
43.1
18.3
15.7
80.6
100
\0.001
History of polyps (%) History of colon cancer (%)
14.1 2.7
7.8 1
13.4 3.7
7.1 2.9
6.5 0
11.8 0
0.23 0.73
History of IBD (%)
4
2
3.7
1.4
3.2
0
0.73
a
p value
Includes 28 patients excluded from further analysis due to lack of follow-up (cf. Fig. 1)
TABLE 2 Baseline and surgical characteristics by treatment-delay quartile Overall
Q1
Q2
Q3
Q4
p value
N
741
182
189
184
185
–
Median delay (days)
23
8
19
29
55
–
Delay range
0–798
0–13
14–23
24–37
38–798
–
Gender, male (%)
51.4
52.2
46.0
52.2
55.1
0.35
Age (median, IQR)
67 (21)
68 (24)
65 (24)
67.5 (19)
69 (19)
0.032
Ethnicity, non-White (%)
9.6
9.3
7.9
8.2
13.0
0.32
Charlson score (mean ± SD)
0.76 – 1.2
0.69 – 1.1
0.55 – 1
0.77 – 1.2
1.04 – 1.4
0.001
Symptomatic disease (%)
61.8
78.0
61.4
62.5
45.4
\0.001
Screening diagnosis (%)
29.5
13.7
25.9
33.2
44.9
\0.001
30-Day metastasis (%)
10.1
16.5
11.6
8.2
4.3
0.001
Surgery and pathology Multivisceral resection (%)
11.2
12.1
15.3
8.7
8.6
0.12
Laparoscopic (%)
29.5
24.2
30.2
33.7
29.7
0.25
Converted (%)
13.8
12.1
15.6
14.6
12.1
0.96
I
26.2
14.8
20.1
26.6
43.2
\0.001
II
30.8
35.7
34.4
31.0
22.2
0.021
III
33.6
40.7
36.5
32.1
24.9
0.010
IV
0.434
AJCC stage (%)
5.7
6.6
6.3
6.5
3.2
High-grade tumor (%)
20.2
21.6
24.6
18.9
14.9
0.16
EMVI (%)
28.6
37.4
32.2
27.2
17.8
\0.001
Positive margin (%)
9.3
13.2
9.0
7.6
7.6
0.21
Lymph node yield (M)
18
20
19
17
17
0.11
Sections highlighted in bold denote statistically significant differences (p \ 0.05) EMVI extramural vascular invasion
Specific assessments performed for high-risk subgroups largely reiterated this pattern in point estimates, with statistically significant findings in patients with stage III disease (n = 249), patients with positive EMVI (n = 213), positive margins (n = 69), ethnic minority patients (n = 71), and symptomatic patients (n = 457). In all cases, one or more of the negative
long-term outcomes were at relative lower odds of occurring in the longer treatment-delay quartiles. Patients with stage IV disease, or a history of colon cancer or IBD had non-significant treatment delayrelated increases in the odds for the aforementioned negative outcomes. Details of these findings are displayed in Table 3.
100
100
80
80
Proportion alive (%)
Metastasis–free proportion (%)
Treatment Delay in Surgically-Treated Colon Cancer
60
40
60
40
20
20
0
0 0
1
2
3
4
5
6
7
8
0
1
2
3
Disease–free Survival (years)
Follow-up (Median, weeks) Stage I Stage II Stage III Stage IV
4
5
6
Overall
Q1
Q2
Q3
Q4
P
142
133
142
155
135
0.24
146.5 150.5 134.5 124
109 133 138 120
111 151 135 109
181 150 152 95.5
150 162 101 124
0.12 0.87 0.26 0.76
103
88.5
100
123
103
0.13
134.5 124.5 83
109 109 69.5
111 118 74
170 129 113
134 132 77
0.28 0.84 0.35
189
163.5
172
210.5
207
0.031
229 197 163 124
181 175 156 120
169 195 162 109
274 195 199 95.5
229.5 215 146.5 124
0.019 0.87 0.23 0.74
Metastatic recurrence Cancer-related death
12.7% 14.7%
14.8% 22.5%
16.4% 16.9%
12.5% 12.5%
7.0% 7.0%
0.037
ORIGINAL ARTICLE – COLORECTAL CANCER
Treatment Delay in Surgically-Treated Colon Cancer: Does It Affect Outcomes? Ramzi Amri, MSc, Liliana G. Bordeianou, MD, Patricia Sylla, MD, and David L. Berger, MD Division of General Surgery & Gastrointestinal Surgery, Massachusetts General Hospital and Harvard Medical School, Boston, MA
ABSTRACT Background. Treatment delay, or the time lapse between diagnosis and surgery, may have a detrimental effect on cancer outcomes. This study assesses the effect of treatment delay on cancer-related outcomes in a large, continuous series of surgically treated colon cancer patients. Methods. All surgical colon cancer cases at our center from 2004 through 2011 were reviewed. Patients who underwent preoperative chemotherapy, emergency admissions, palliative cases, and incidental and postoperative diagnoses were excluded. Treatment delay was correlated with outcomes in univariate and multivariate regression and proportional hazards models. Results. In 769 included patients, for every treatmentdelay quartile increase, odds of death decreased by an odds ratio (OR) of 0.78 (p = 0.001), and metastatic recurrence by OR 0.78 (p = 0.013). Shorter survival duration had a hazard ratio (HR) of 0.81 (p = 0.001) and shorter diseasefree survival HR 0.72 (p \ 0.001). Multivariate regression adjusting for baseline staging greatly reduces these ratios, and makes them non-significant. Similar patterns were shown in high-risk subsets, including stage III disease, ethnic minorities, patients with positive margins, and extramural vascular invasion. Conclusions. The inverse relation between treatment delay and survival and recurrence reflected adequate prioritization of advanced and high-risk cases and concurrently showed that, matched for stage and risk categories, treatment delay was not associated with worse cancer outcomes for patients with colon cancer. A reasonable delay between diagnosis and subsequent surgery is not detrimental to patient outcomes and permits more flexibility in scheduling and Ó Society of Surgical Oncology 2014 First Received: 22 February 2014 D. L. Berger, MD e-mail: [email protected]
justifies allowing time to complete proper preoperative evaluation and staging, improving the quality and safety of resection and treatment. Treatment delay is the time lapse between diagnosis and treatment. Since most malignancies have linear1 or even exponential2 growth models, onset of symptoms is often preceded by a long asymptomatic phase with increasing degeneration.3 The accelerating growth and degeneration is what theoretically makes diagnosis-to-treatment delays especially detrimental in cancer: Treatment delay is likely to take place during a critical phase with high tumor growth rates and marginally effective innate tumor defense mechanisms, increasing the risk of metastasis and narrowing treatment possibilities. Reports on the effect of treatment delay are mixed. Detrimental effects have been described in cancers of the breast,4 prostate,5 rectum, lung and pancreas,6 while other reports find no significant impact of treatment delay on outcomes of cancers of the stomach,6 bronchus,7 and pancreatic head.8 In colon cancer, the impact of treatment delay is controversial,6,9–11 perhaps owing to heterogeneities in tumor biology. The historical consensus for colon cancer is that most cases are relatively slow-growing. These cases result from the adenoma-adenocarcinoma sequence,12 which can take up to 15 years from onset to eventual progression to malignancy.13 However, an increasing number of exceptions with more aggressive behavior challenge this assumption. Prominently validated high-risk examples include cases presenting with tumorpositive lymph nodes, metastatic disease, extramural vascular invasion (EMVI),14 and comorbid inflammatory bowel disease (IBD).15 Patients with aggressive colon cancer forms may be harmed by the assumptions on slow progression if similar treatment delays are applied to them. This paper aims to assess the distribution of treatment delay and its effect on colon cancer outcomes in a large,
R. Amri et al.
continuous series of surgical colon cancer patients treated in a tertiary care center, both in its general population and in high-risk subpopulations. METHODS Patients A contiguous patient series encompassing all colon cancer patients operated on at Massachusetts General Hospital (MGH) from January 2004 through December 2011 was included in a data repository under an Institutional Review Board-approved protocol. We chose to include colon cancer only instead of colon and rectum malignancies as these diseases have considerable differences in treatment regimens and tumor biology.16 Since we investigated the link between treatment delay and surgical outcomes of colon cancer, our focus was on elective cases where the delay is malleable and the primary therapy is surgery. Therefore, of a total of 1,071 cases, 302 were not suitable for the purpose of this research, including urgent admissions requiring immediate treatment (n = 102), patients with postoperative diagnosis (no preoperative confirmation, or incidental discovery; n = 82), referrals without an accurate diagnosis date (n = 70), patients who underwent neoadjuvant chemotherapy (n = 31), and palliative cases (n = 17). Of the 769 patients eligible at baseline, 28 patients without follow-up had to be excluded from further analysis. The remaining 741 patients were electively admitted and underwent surgery for colonic adenocarcinoma with curative intent. Demographics, surgical admission, pathology, and follow-up data were gathered from patient records and the MGH cancer registry. Survival information was also acquired from the Social Security Death Index. Definitions and Objectives Treatment delay was defined as the time interval between diagnosis and surgery. The diagnosis date was considered the date when colon cancer was clinically established. In virtually all cases, this was by a colonoscopy with a (preliminary) pathology indicating adenocarcinoma. Metastatic presentation was defined as any case presenting with metastatic disease, or diagnosed with metastatic disease within 30 days of the operation, allowing for perioperative staging and biopsies to fall into this category. Cases with metastatic disease established later than 30 days postoperatively, without suspicion during admission, are considered metastatic recurrences. Follow-up-related outcomes included overall survival (OS), duration of survival, local and distant recurrence rates, and disease-free interval duration.
Patients potentially at risk for more serious disease were also assessed as subgroups separately from the overall population. These included patients with pre-existing polyposis, pre-existing colon cancer, IBD, symptomatic diagnosis, ethnic minority patients,17 as well as pathological characteristics including American Joint Committee on Cancer (AJCC) stages III and IV, positive resection margins (proximal, distal, and/or radial), and EMVI. Analysis and Statistical Methods All statistical analysis was performed using the SPSS statistical suite (IBM Corp. Released 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.). A p value of below 0.05 (two-tailed) was considered significant. Pearson’s Chi square test was used to compare nominal variables between groups, while a one-way analysis of variance (ANOVA) was used for continuous characteristics. The distribution of treatment delay duration was reviewed statistically (mean, median, standard deviation, and interquartile range [IQR]) and visually over a twodimensional frequency plot, dichotomizing the included population into those diagnosed during admission and those diagnosed before admission. This analysis also reviewed delay distributions in excluded patients with preoperative diagnosis. Links between treatment delay and presentation characteristics were assessed using a binary logistic regression for dichotomous outcomes and a Cox proportional hazards model for time-related continuous outcomes. In order to yield results that are intuitive to interpret, treatment delay was divided over quartiles. This means that the resulting odds ratios (ORs) or hazard ratios (HRs) are the estimate for two subsequent quartiles. Finally, to ensure any significant findings were validated, the links between treatment delay and long-term outcomes were also corrected for covariates that might act as confounders using multivariate models. These covariates included age, comorbidity burden (expressed by the Charlson comorbidity score,18 excluding the contribution of colon cancer to the score), pathological staging, and, lastly, duration of follow-up. Covariates per model were determined by clinical significance and model fit (R2). RESULTS Perioperative Characteristics The distribution of treatment delay is shown on Fig. 1, along with the inclusion flowchart; the figure shows that treatment delay has a log-normal distribution, with a mean delay of 31.5 days (SD 40.1). Table 1 shows detailed baseline
Treatment Delay in Surgically-Treated Colon Cancer
Treatment Delay Categories
120
Inclusion diagram
Emergency diagnosis Palliative case Neoadjuvant chemotherapy Included (In–stay Diagnosis) Included (Preadmission Diagnosis)
Enrollment
Surgically treated colon cancer patients 2004-2011 (n=1071)
Excluded (n=302) • • • • •
Frequency
100
80
Diagnosis date unknown Postoperative dx/confirmation Emergency surgery Palliative surgery Neoadjuvant chemotherapy
(n=70) (n=82) (n=102) (n=17) (n=31)
Included (n=769)
Follow-Up Status st on December 1 2012
60 Alive (n=549)
Deceased (n=220)
Active follow-up
Cause of death
At MGH (n=491) Affiliated/corresponding (n=30)
Colon cancer-related Other/multiple causes
Lost to follow-up:
40
(n=109) (n=101)
Moved outside of region (n=20) Moved outside of country (n=5) Voluntary FU withdrawal (n=3)
20 Survival Analysis Analyzed (n=521) No follow-up (n=28)
0 0
1
10
100
Analyzed (n=741)
Analyzed (n=220)
1000
Treatment delay (days) FIG. 1 Treatment delay distribution and inclusion. The left pane shows all preoperatively diagnosed cases of colon cancer plotted over the x-axis following their respective treatment delay durations. Treatment delay has been scaled logarithmically to allow display of
outliers. This also demonstrates the lognormal distribution of our population. The right pane states the flowchart for inclusions, detailing causes for exclusion. This also details survival and followup status, which are a last cause for further exclusions. dx, diagnosis
characteristics: no unexpected differences were encountered between eligible and excluded patients at baseline. For the 741 patients included for final analysis, baseline characteristics distributed over quartiles are shown in Table 2. Median treatment delays for each quartile were 8, 19, 29, and 55 days, respectively, for quartiles 1 through 4. Demographics were not significantly different between treatmentdelay quartiles, except for fluctuations in median age. In terms of presentation characteristics, there was an increasing trend in comorbidity burden (p = 0.001), and screening diagnoses (p \ 0.001) with longer treatment delay. This came along with a decrease in the number of symptomatic diagnoses (p \ 0.001) and metastatic disease established within 30 days of admission (p = 0.001). Accordingly, in surgical pathology, lower delay quartiles had substantially higher rates of EMVI (p \ 0.001) and stage III disease (p = 0.010), and lower rates of stage I (p \ 0.001) and stage II (p = 0.021) disease. Surgical characteristics were largely similar over treatment-delay quartile intervals, with comparable rates of laparoscopic surgery, conversion, margin positivity, and lymph node yields, except a non-significant higher multivisceral resection rate in the first two quartiles (12.1 and 15.3 % vs. 8.7 and 8.6 %).
Univariate Analysis of Treatment Delay with Outcomes Univariate relationships between long-term outcomes and treatment delay including follow-up, survival, and disease-free survival, as well as quartile differences, are displayed in Fig. 2. In the analyzed population, over a median follow-up of 142 weeks (range 0–430; IQR 170 weeks), the median survival was 189 weeks (range 0–462; IQR 190). Median disease-free survival was 103 weeks (range 0–430; IQR 176). Interestingly, there was a tendency for death and metastatic disease to occur less often in the longer treatmentdelay quartiles. This difference was statistically significant for rates of metastatic recurrence (p = 0.031), OS (p = 0.005), and cancer-related death (p \ 0.001). In univariate regression, this led to lower odds of death and recurrence to occur in the higher quartiles of treatment delay. For every two consecutive treatment-delay quartiles, this included an OR of 0.78 for death (p = 0.001), an OR of 0.78 of metastatic recurrence (p = 0.013), an HR of 0.81 (p = 0.001) for shorter survival duration, and an HR of 0.72 (p \ 0.001) for shorter disease-free survival were identified using Cox regression.
R. Amri et al. TABLE 1 Baseline characteristics and comparison of included and excluded subsets Includeda
Urgent admission
Postoperative diagnosis
Diagnosis date unknown
Neoadjuvant therapy
Palliative surgery
N (%)
769 (71.8)
102 (9.5)
82 (7.7)
70 (6.5)
31 (2.9)
17 (1.6)
–
Age (mean, SD)
67 (±13.8)
68.2 (±14)
65.9 (±14.4)
65.8 (±13.1)
57.6 (±11.8)
63 (±17.8)
0.005
Gender, male (%)
50.8
48.0
51.2
57.1
54.8
52.9
0.89
Ethnicity, non-White (%)
9.8
12.7
6.1
8.6
22.6
23.5
0.051
Symptomatic presentation (%)
62.3
90.2
48.8
45.7
80.6
94.1
\0.001
Metastatic presentation (%)
22.1
43.1
18.3
15.7
80.6
100
\0.001
History of polyps (%) History of colon cancer (%)
14.1 2.7
7.8 1
13.4 3.7
7.1 2.9
6.5 0
11.8 0
0.23 0.73
History of IBD (%)
4
2
3.7
1.4
3.2
0
0.73
a
p value
Includes 28 patients excluded from further analysis due to lack of follow-up (cf. Fig. 1)
TABLE 2 Baseline and surgical characteristics by treatment-delay quartile Overall
Q1
Q2
Q3
Q4
p value
N
741
182
189
184
185
–
Median delay (days)
23
8
19
29
55
–
Delay range
0–798
0–13
14–23
24–37
38–798
–
Gender, male (%)
51.4
52.2
46.0
52.2
55.1
0.35
Age (median, IQR)
67 (21)
68 (24)
65 (24)
67.5 (19)
69 (19)
0.032
Ethnicity, non-White (%)
9.6
9.3
7.9
8.2
13.0
0.32
Charlson score (mean ± SD)
0.76 – 1.2
0.69 – 1.1
0.55 – 1
0.77 – 1.2
1.04 – 1.4
0.001
Symptomatic disease (%)
61.8
78.0
61.4
62.5
45.4
\0.001
Screening diagnosis (%)
29.5
13.7
25.9
33.2
44.9
\0.001
30-Day metastasis (%)
10.1
16.5
11.6
8.2
4.3
0.001
Surgery and pathology Multivisceral resection (%)
11.2
12.1
15.3
8.7
8.6
0.12
Laparoscopic (%)
29.5
24.2
30.2
33.7
29.7
0.25
Converted (%)
13.8
12.1
15.6
14.6
12.1
0.96
I
26.2
14.8
20.1
26.6
43.2
\0.001
II
30.8
35.7
34.4
31.0
22.2
0.021
III
33.6
40.7
36.5
32.1
24.9
0.010
IV
0.434
AJCC stage (%)
5.7
6.6
6.3
6.5
3.2
High-grade tumor (%)
20.2
21.6
24.6
18.9
14.9
0.16
EMVI (%)
28.6
37.4
32.2
27.2
17.8
\0.001
Positive margin (%)
9.3
13.2
9.0
7.6
7.6
0.21
Lymph node yield (M)
18
20
19
17
17
0.11
Sections highlighted in bold denote statistically significant differences (p \ 0.05) EMVI extramural vascular invasion
Specific assessments performed for high-risk subgroups largely reiterated this pattern in point estimates, with statistically significant findings in patients with stage III disease (n = 249), patients with positive EMVI (n = 213), positive margins (n = 69), ethnic minority patients (n = 71), and symptomatic patients (n = 457). In all cases, one or more of the negative
long-term outcomes were at relative lower odds of occurring in the longer treatment-delay quartiles. Patients with stage IV disease, or a history of colon cancer or IBD had non-significant treatment delayrelated increases in the odds for the aforementioned negative outcomes. Details of these findings are displayed in Table 3.
100
100
80
80
Proportion alive (%)
Metastasis–free proportion (%)
Treatment Delay in Surgically-Treated Colon Cancer
60
40
60
40
20
20
0
0 0
1
2
3
4
5
6
7
8
0
1
2
3
Disease–free Survival (years)
Follow-up (Median, weeks) Stage I Stage II Stage III Stage IV
4
5
6
Overall
Q1
Q2
Q3
Q4
P
142
133
142
155
135
0.24
146.5 150.5 134.5 124
109 133 138 120
111 151 135 109
181 150 152 95.5
150 162 101 124
0.12 0.87 0.26 0.76
103
88.5
100
123
103
0.13
134.5 124.5 83
109 109 69.5
111 118 74
170 129 113
134 132 77
0.28 0.84 0.35
189
163.5
172
210.5
207
0.031
229 197 163 124
181 175 156 120
169 195 162 109
274 195 199 95.5
229.5 215 146.5 124
0.019 0.87 0.23 0.74
Metastatic recurrence Cancer-related death
12.7% 14.7%
14.8% 22.5%
16.4% 16.9%
12.5% 12.5%
7.0% 7.0%
0.037
