Cancer stage describes the severity of an individual’s cancer on the basis of the extent of the primary tumor and its spread through the body. Clinicians in oncology consider that stage at diagnosis is essential for management. Researchers and epidemiologists also consider that the cancer stage is essential for scientific exchange on calendar periods or geographic areas as it provides a common standardized language to communicate on incidence, treatment, or outcomes. Staging relies on physical examinations, imaging tests, pathology reports, and surgical reports. The TNM classification, on the basis of the extent of the tumor (T), the extent of spread to the lymph nodes (N), and the presence of metastasis (M), is the most important standardized tool used in oncology. Yet, in terms of the presence of metastasis, there is no consensual definition of synchronous, that is, at the ‘time of diagnosis’, or metachronous, that is, delayed or tumor progression. The cutoff to distinguish between the two in the international scientific literature may vary between 1 and 12 months between the diagnosis of the primary and the diagnosis of the metastasis (Neo et al., 2011; Elferink et al., 2015; Goey et al., 2016; Angelsen et al., 2017). Although this imprecise definition has no clinical impact, as treatment is delivered according to the patient’s clinical status, it may markedly affect the results of epidemiological studies and clinical trials. The proportion of synchronous metastatic cancers will increase mechanically with the increase in the delay defining the cutoff. This may introduce artifactual differences in the comparison of the proportion of cancers according to stages in epidemiological studies. In clinical trials, the absence of a precise definition of this cutoff could introduce a bias in the selection of patients to be included and in the measurement of disease-free survival, a usual endpoint in such trials. With more than 40 000 new cases each year in France, among whom 50% of patients will die within the 5 years following diagnosis, colorectal cancer is a major public health problem. Mass-screening programs, improvements in adjuvant and palliative chemotherapies, or the development of targeted treatments, for example, need to be evaluated over time, and such evaluations require precise staging to help determine the best strategy of care for individual patients.
The objective of this study was therefore to determine how variations in the cutoff for the definition of synchronous metastatic colorectal cancer can influence stage-specific incidence and survival of stage IV patients.
Patients and methods
Two French population-based cancer registries, which include all digestive tract cancers, cover the resident populations of the well-defined administrative areas of Côte d’Or and Saône-et-Loire (Burgundy, 1 053 000 inhabitants). The two registries involved have worked together for many years and use identical standardized data collection, recording, and validation procedures. Information was collected from the multiple concerned sources: public and private pathology laboratories, university hospital (including the Cancer Centre), general hospitals, private surgeons, gastroenterologists, radiotherapists and medical oncologists, the hospital administrative database, the Regional Health Services database, and death certificates. Because death certificates are somewhat unreliable, they were only used to identify missing cases. Quality checks are carried out using local controls and the DepEdit software provided by the International Agency for Research on Cancer. The quality and completeness of the registries is certified every 4 years through an audit carried out by the National Institute for Health and Medical Research, the French National Cancer Institute, and the French Public Health Institute.
Information on vital status was collected primarily through public services at the birthplace or requests to the National Register for the Identification of Physical Persons. The medical files of the general practitioner or gastroenterologist were used if the birthplace was unknown.
Between 2007 and 2013, 4636 patients presented with colorectal adenocarcinoma and were registered. Demographic, clinical, and tumor-related characteristics were collected routinely. Tumor extension at diagnosis was classified according to the TNM classification. Resected patients whose pathology reports were not available in the registry were excluded (stage unknown n=17). In total, 4619 colorectal adenocarcinomas were included in the analyses. Tumors not eligible for resection and without synchronous metastasis were classified as unresected regional tumors (n=178). The cancer site was classified according to the International Classification of Diseases for Oncology, third revision (C19 and C20). The date of diagnosis of the first metastasis (mostly the date of the positive morphology examination) was collected precisely through a special survey conducted during the 18 months following the diagnosis of the primary cancer. Metastases diagnosed during the first month before the date of incidence, that is, mostly the date of the endoscopy, were considered present at diagnosis. Among the 1037 metastases diagnosed (Table 1) within the first month, 974 were diagnosed during the morphology work-up examination and 63 peroperatively during the surgical resection. The time ‘T’ in months was defined as the time between the primary colorectal cancer diagnosis and the diagnosis of the metastasis. With ‘i’, the cutoff, varying from 1 to 12 months, the patients diagnosed with metastasis were assigned iteratively to the ‘synchronous group’ when T≤i and to the ‘metachronous group’ when T>i. For the highest cutoff (i=12 months), 1254 metastases were considered synchronous (Table 1).
To estimate survival in the metachronous group, we took into account metastases diagnosed between 12 and 18 months after the diagnosis of the primary colorectal cancer. A total of 1347 patients were diagnosed with metastases within the first 18 months and included in the survival analysis.
Percentage distribution and incidence rates by stages of the colorectal cancer were described by varying the cutoff between diagnosis of the primary and diagnosis of the metastasis from 1 to 12 months. Age-standardized incidence was calculated by the direct method using the world standard population. Incidence by stage was calculated for each cutoff. Incidence rates were compared using the direct comparison method (Esteve et al., 1994).
Net survival in the synchronous and metachronous groups was calculated from the date of the metastasis diagnosis. Net survival is defined as survival that would be observed if cancer was the only possible cause of death. Net survival estimated using the excess mortality rate takes into account competing risks of death from other causes without requiring the cause of death. The observed mortality rate in the population of patients is the sum of the expected mortality rate in the general population and the excess mortality rate (mortality due to cancer) in the population of patients. Expected and excess mortality rates are defined on the basis of the same demographic characteristics. The expected mortality rates in the population were obtained from life tables provided by the National Institute of Statistics and Economic Studies. Net survival was estimated using the method proposed by MP Perme (Perme et al., 2012) and the prognostic value of each cutoff was tested using a flexible parametric net survival model developed by Nelson et al. (2007). Survival for each group defined for each cutoff was compared using the likelihood ratio test. Relative differences in net survival were calculated as the ratio of the difference in net survival between the metachronous and synchronous groups divided by survival in the synchronous group at 1 and 3 years. Prognostic receiver operating characteristic (ROC) curves methodology, defined as 1−S(t)high-risk as a function of 1−S(t)low-risk for each cutoff i (Combescure et al., 2014), was used. Assuming that survival in patients with synchronous metastasis would be lower than that in patients with metachronous metastasis (Ghiringhelli et al., 2014), the synchronous group was defined as ‘high risk’ and the metachronous group was defined as ‘low risk’. Prognostic ROC curves were obtained from survival estimated by the flexible parametric survival model. As survival curves did not reach 0, the ROC curves were incomplete. Areas under the curve were obtained by extrapolating beyond the last follow-up time under the ‘noninformative’ assumption (Combescure et al., 2014).
Analyses were carried out using Stata 14 (Stata Corp., College Station, Texas, USA) and RStudio for the prognostic ROC curve.
A total of 4619 patients (male/female sex ratio: 1.2:1) were considered. The mean age at diagnosis was 71.46 years (SD: 12.68). In all, 67% of the tumors were located in the colon and 33% were located in the rectosigmoid junction or the rectal ampulla. Overall, 1254 patients presented with metastasis within 12 months of the primary diagnosis of colorectal cancer (Table 1); of these, 82% (1037/1254) presented within the first month.
Table 1 presents the distribution of stages according to the 12 cutoff times between the primary and the metastasis diagnosis. The proportion of stage IV cases increased from 22.5% for the smallest cutoff (0–1 month) to 27.2% for the largest cutoff (0–12 months), corresponding to a relative variation in the distribution of +20.9%. Accordingly, the proportion of stage III (−11.8%), of unresected regional tumors (−9.8%), and to a lesser extent of stage II (−4.8%) decreased, whereas that of stage I did not vary. The age-standardized incidence of stage IV increased significantly from 6.0/100 000 when considering metastasis diagnosed within the first month to 7.1/100 000 when including all metastases diagnosed until 12 months after the diagnosis of colorectal cancer (P<0.001) (Table 2). The corresponding incidence of stage III patients decreased from 5.8/100 000 to 5.1/100 000 (P=0.036). The incidence of stage I and of unresected regional tumors did not vary significantly according to the time intervals.
The comparison of net survival between the groups of synchronous and metachronous metastatic patients according to the 12 cutoff times is presented in Table 3 and Fig. 1. Net survival was nearly 53% 1 year after the diagnosis of synchronous metastasis and 21% at 3 years and did not vary with increasing cutoff times. In contrast, net survival for metachronous metastases increased as the cutoff increased from 1 to 12 months. At 3 years, it was around 25% up to i=4 months, and then increased to 31% for i=12. Net survival was significantly higher for the metachronous than for the synchronous group from i=5 months onwards. Relative variations showed that, for i varying from 5 to 12 months, the 1-year net survival for the metachronous group was over 10% higher than that for the synchronous group. Similarly, at 3 years, the relative variation was over 30%. These relative variations between the two groups increased steadily with increasing cutoff times.
The area under the ROC curve for survival varied according to the cutoff times between 0.52 (i=3 months) and 0.57 (i=12 months).
This study shows that selection bias may occur when observational studies dealing with stage at diagnosis are compared in the absence of a precise and a standardized definition of synchronous metastasis for colorectal cancer. When the interval for the definition of metachronous metastasis was increased from 1 to 12 months, we found a relative variation in the proportion of cancers of +21% for stage IV, −12% for stage III, and −5% for stage II. Similarly, the relative difference between net survival in patients with metachronous or synchronous metastases ranged between 16 and 53%.
Improvements in statistics and methodology in epidemiology in recent decades have led to an increasing number of international comparisons on the frequency, prognosis, or patterns of care in colorectal cancer. Understanding the reasons for inter-country or intra-country differences should help improve cancer control, prevention, and treatment strategies. For this purpose, population-based cancer registries are the best tool to provide unbiased and representative epidemiological indicators. As they include all cases diagnosed among the inhabitants of a chosen geographic area, they enable public health comparisons by eliminating the unavoidable selection bias related to hospital series or clinical trials. Stage at diagnosis is the most important prognostic factor for colorectal cancer. Previous epidemiologic studies from around the world have established that differences in the distribution of stage at diagnosis partly explained international differences in survival. All of the authors questioned the robustness of their results and addressed the limits of data quality in terms of the stage classification, the thoroughness of clinical pre staging investigations, or the completeness of data. The identification and control of these potential limits and the use of validated statistical standardization methods enable robust international comparisons between epidemiological indicators. Nevertheless, we showed that the absence of precise definitions in terms of the timing of the occurrence of distant metastasis may modify the selection of cases and may partly bias the results. This phenomenon indicates the paradox attributed to Will Rogers (Sormani, 2009). Differences in the chosen cutoff effectively lead to metastatic cancers being categorized as nonmetastatic (or vice versa). As these cases have on average a poorer (or better) prognosis than those of the category in which they were classified, estimated survival in both groups is modified. As a result, observed geographical differences in prognosis in both categories may be partly artifactual.
When the cutoff for the definition of metachronous metastasis was increased from 1 to 12 months, we found a relative variation in the proportion of cancers of +21% for stage IV, −12% for stage III, and −5% for stage II. In the large international CONCORD study, Tumors in Dukes’ stage A or B tumors were of similar frequency in Europe and in the USA. In contrast, Dukes’ D tumors were twice as common in Europe (21%) as in the USA (10%) and patients diagnosed at an advanced stage (i.e. metastatic cases plus unresected cases) were more common in Europe (29%) than in the USA (20%) (Allemani et al., 2013). The stage at diagnosis varied more widely between the participating European countries than between the participating US states. One cannot thus exclude the possibility that these differences might be explained by greater inter-registry variations in cutoffs for the definitions of synchronous events among European countries than among participating SEER areas.
Stage is also an important factor to disentangle the reasons for worldwide differences in colorectal cancer survival. International differences in 1 or 3-year net survival have been described to be wider previously for patients with advanced stage than for those with stages I or II (Maringe et al., 2013). The explanations usually provided are variations in diagnostic delay or in the thoroughness of staging procedures. We underlined that the wide variety of cutoffs to define a metastasis as synchronous in population-based studies may partly contribute toward variations in survival.
Thus, these potential variations have to be taken into account when tumor stage distribution is compared across different populations. As the number of diagnoses of metastases beyond the first month after colorectal cancer diagnosis is low compared with that of metastases diagnosed in the first month, variations in the cutoff should not strongly affect geographical comparisons of the incidence of colorectal cancer by stage. In our population, we found a variation of +1.11/100 000 of incident stage IV cases depending on whether metastases were diagnosed during the first month or during the 12 months following the diagnosis of colorectal cancer.
Our study has several limitations. The number of metastases occurring in each monthly interval was small and numbers were not balanced for survival analyses, thus limiting the power of the study. We could not analyze colon and rectum locations separately despite a higher recurrence in rectal cancer than in colon cancer (Bouvier et al., 2015; Cottet et al., 2015).
Our study suggests that an objective definition of a relevant cutoff to distinguish between synchronous and metachronous metastasis is required for epidemiological and clinical scientific exchanges. Survival in patients with metachronous metastasis was significantly better than that in patients with synchronous metastasis when the cutoff between synchronous and metachronous was over 4 months after the primary diagnosis. Nevertheless, this result has to be confirmed through other studies before recommending a threshold of 5 months to define a synchronous metastasis in staging classifications.
The authors thank Stéphanie Normand for her helpful assistance in the validation of the database. They also thank the French Public Health Institute (SpF) and the National Cancer Institute (INCa) for financial support.
Conflicts of interest
There are no conflicts of interest.
Allemani C, Rachet B, Weir HK, Richardson LC, Lepage C, Faivre J, et al (2013). Colorectal cancer
survival in the USA and Europe: a CONCORD high-resolution study. BMJ Open 3:e003055.
Angelsen JH, Horn A, Sorbye H, Eide GE, Loes IM, Viste A (2017). Population-based study on resection rates and survival in patients with colorectal liver metastasis
in Norway. Br J Surg 104:580–589.
Bouvier AM, Launoy G, Bouvier V, Rollot F, Manfredi S, Faivre J, et al (2015). Incidence and patterns of late recurrences in colon cancer patients. Int J Cancer 137:2133–2138.
Combescure C, Perneger TV, Weber DC, Daures JP, Foucher Y (2014). Prognostic ROC curves: a method for representing the overall discriminative capacity of binary markers with right-censored time-to-event endpoints. Epidemiology 25:103–109.
Cottet V, Bouvier V, Rollot F, Jooste V, Bedenne L, Faivre J, et al (2015). Incidence and patterns of late recurrences in rectal cancer patients. Ann Surg Oncol 22:520–527.
Elferink MA, De Jong KP, Klaase JM, Siemerink EJ, De Wilt JH. Statistical methods in cancer research. Descriptive epidemiology (volume IV). IARC Scientific Publication No. 128. Lyon; International Agency for Research on Cancer 1994. p. 302.
Esteve J, Benhamou E, Raymond L (1994). International Agency For Research on Cancer (WHO). Statistical methods in cancer research. descriptive epidemiology. IARC Scientific Publications no. 128.IV.
Ghiringhelli F, Hennequin A, Drouillard A, Lepage C, Faivre J, Bouvier AM (2014). Epidemiology and prognosis of synchronous and metachronous colon cancer metastases: a French population-based study. Dig Liver Dis 46:854–858.
Goey KK, T Lam-Boer J, de Wilt JH, Punt CJ, van Oijen MG, Koopman M (2016). Significant increase of synchronous disease in first-line metastatic colorectal cancer
trials: Results of a systematic review. Eur J Cancer 69:166–177.
Maringe C, Walters S, Rachet B, Butler J, Fields T, Finan P, et al (2013). Stage at diagnosis and colorectal cancer
survival in six high-income countries: a population-based study of patients diagnosed during 2000–2007. Acta Oncol 52:919–932.
Nelson CP, Lambert PC, Squire IB, Jones DR (2007). Flexible parametric models for relative survival, with application in coronary heart disease. Stat Med 26:5486–5498.
Neo EL, Beeke C, Price T, Maddern G, Karapetis C, Luke C, et al (2011). South Australian clinical registry for metastatic colorectal cancer
. ANZ J Surg 81:352–357.
Perme MP, Stare J, Esteve J (2012). On estimation in relative survival. Biometrics 68:113–120.
Sormani MP (2009). The Will Rogers phenomenon: the effect of different diagnostic criteria. J Neurol Sci 287 (Suppl 1):S46–S49.
Keywords:Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
bias; cancer registry; colorectal cancer; metastasis; staging