Declining Risk of Colorectal Cancer in Ulcerative Colitis
Declining Risk of Colorectal Cancer in Ulcerative Colitis
In the present study, that included all articles reporting incidence of CRC in UC patients (independently of their design), we have found that the overall risk of CRC in 181 923 UC patients was 1.69 per 1000 py, 0.91/1000 py in the first decade after the diagnosis of UC, 4.07/1000 py in the second and 4.55/1000 py in the third decade. These updated figures are lower than those reported by Eaden et al. in their meta-analysis, 3/1000 py overall, and 2%, 8% and 18% for each decade after diagnosis. These rates are considerably higher than the incidence of CRC reported in the general population. The incidence of colorectal cancer in general population ranges from 0.4/1000 py in Australia, New Zealand, USA and Western Europe to 0.1/1000 py in Africa.
We would like to emphasise that, in our systematic review, 72% of the total of the studies (58 of 81 studies) and 87% of the studies published after 1990 (46 of 53 studies) reported an incidence rate lower than 3/1000 py, the overall incidence rate that Eaden et al. reported in their meta-analysis. However, studies performed after the publication of that meta-analysis have reported lower incidence rates of CRC in UC patients. It is important to note the striking downward trend in the incidence rate over the last few decades: the incidence rates decreased from 4.29/1000 py in studies published in the 1950s to 1.21/1000 py in studies published in the last decade. Several hypotheses may be put forward to explain this trend, such as the adherence to endoscopic surveillance programmes, higher rates of colectomy in patients with dysplasia or a more aggressive control of the inflammatory activity. In this respect, 5-aminosalicylates have been suggested to have a chemoprophylactic impact on CRC associated with UC; although some studies have failed to prove their preventive effect, others have shown that these agents seem to interrupt the progression from healthy mucosa to dysplasia and carcinoma. Recent studies have reported that thiopurines would reduce the risk of CRC in UC patients, also by controlling the inflammation. Although data on the impact of anti-TNF agents in the development of CRC are still lacking, it is conceivable that an optimal control of mucosal inflammation could decrease this risk.
In the present study, several sub-analyses have been performed to identify patients with a higher risk of developing CRC that could be taken into account for surveillance programmes. To this respect, in agreement with previous data, the risk of CRC in UC patients was increased with larger extensions of the disease. We also observed that this risk was increased with longer durations of the UC, although it was much lower than that previously reported. Thus, incidence rates of CRC at 10, 20 and 30 years after diagnosis of UC in our study and in Eaden's study were 0.91 vs. 2, 4.07 vs. 7 and 4.55 vs. 12, all measured in cases per 1000 py.
Based on the results from Eaden's study, considering that the risk of CRC increases with the time of evolution of the UC, several guidelines have established that the first screening colonoscopy should be offered 8 to 10 years after the diagnosis in patients with extensive colitis, and 15 years after the onset of symptoms in patients with left-sided colitis. After that, surveillance colonoscopies should be repeated every 1–2 years. New ECCO guidelines published in 2013 recommend a screening colonoscopy 8 years after the onset of colitic symptoms. Ongoing surveillance should be repeated every 1, 3 or 5 years depending on the risk factors. Nevertheless, a Cochrane meta-analysis did not find clear evidence that endoscopic surveillance prolonged survival in patients with extensive colitis, although it showed that CRCs tend to be detected at an earlier stage. New endoscopic techniques, such as chromoscopy, have been shown to be much more efficient for the detection of dysplasia than the performance of random biopsies; their incorporation into endoscopic surveillance might optimise the screening time points.
Regarding the risk of CRC in paediatric UC population, although studies are scarce, they have also reported higher incidence rates than the overall rate for adults with UC, thus supporting previous findings, namely, that early onset of UC could confer a higher risk of developing CRC.
Considering only population-based studies – that is, excluding studies from referral centres – the incidence rate of CRC is even lower. Our sub-analysis of population-based studies revealed that the incidence rate was 1.24 per 1000 py vs. 1.9 per 1000 py for studies with different designs (tertiary or referral hospital-based cohorts, private clinic series, and surgical series). Of note, only 5 of the 19 population-based studies had been included in the meta-analysis by Eaden et al., as the other 14 were published after 2000.
To have an overview of the complete history of the development of CRC in UC patients, a sub-analysis of the studies in which the surveillance had started at the diagnosis of the disease was performed, finding that the incidence rate was even lower than the overall incidence rate, and closer to the general population without UC.
Our analysis is subject to a series of limitations. Our analysis covers a large number of studies in which high heterogeneity of design, size, length of follow-up and results exists, leading to pooled statistical heterogeneity. We have to assume that, even trying to reduce heterogeneity, this could not be alleviated even performing sub-analyses. The analysis is also limited by the small sample size of some cohorts, extracted from studies carried out by surgical teams or studies based on surveillance programmes in referral centres with active follow-up. These limitations could lead to higher incidence rates of CRC in small cohorts. Consequently, our results may be overestimated due to the inclusion of a considerable number of studies performed in referral centres (vs. population-based studies), mainly when a rare event like CRC is evaluated. Another limitation of our paper may be the differential incidences according to the geographical area. This may partially explain the significant heterogeneity of the results. However, we believe that despite the limitations described, our results are sufficiently robust to be considered valid and provide valuable updated information about the current risk of CRC in UC.
In conclusion, our meta-analysis estimates an overall incidence rate of CRC of only 1.58 per 1000 py in UC patients; this incidence rate is considerably lower than previously estimated. As described in precedent literature, the extension and the duration of the disease increase the risk of CRC in this population. On the other hand, the incidence of CRC in UC patients seems to have decreased during the last few decades, and this might be explained by a tighter control of the inflammation, higher colectomy rates, the use of drugs with chemopreventive effects and a better adherence to endoscopic surveillance programmes in high-risk patients. All these factors should be taken into account when critically reviewing current CRC surveillance programmes in UC patients.
Discussion
In the present study, that included all articles reporting incidence of CRC in UC patients (independently of their design), we have found that the overall risk of CRC in 181 923 UC patients was 1.69 per 1000 py, 0.91/1000 py in the first decade after the diagnosis of UC, 4.07/1000 py in the second and 4.55/1000 py in the third decade. These updated figures are lower than those reported by Eaden et al. in their meta-analysis, 3/1000 py overall, and 2%, 8% and 18% for each decade after diagnosis. These rates are considerably higher than the incidence of CRC reported in the general population. The incidence of colorectal cancer in general population ranges from 0.4/1000 py in Australia, New Zealand, USA and Western Europe to 0.1/1000 py in Africa.
We would like to emphasise that, in our systematic review, 72% of the total of the studies (58 of 81 studies) and 87% of the studies published after 1990 (46 of 53 studies) reported an incidence rate lower than 3/1000 py, the overall incidence rate that Eaden et al. reported in their meta-analysis. However, studies performed after the publication of that meta-analysis have reported lower incidence rates of CRC in UC patients. It is important to note the striking downward trend in the incidence rate over the last few decades: the incidence rates decreased from 4.29/1000 py in studies published in the 1950s to 1.21/1000 py in studies published in the last decade. Several hypotheses may be put forward to explain this trend, such as the adherence to endoscopic surveillance programmes, higher rates of colectomy in patients with dysplasia or a more aggressive control of the inflammatory activity. In this respect, 5-aminosalicylates have been suggested to have a chemoprophylactic impact on CRC associated with UC; although some studies have failed to prove their preventive effect, others have shown that these agents seem to interrupt the progression from healthy mucosa to dysplasia and carcinoma. Recent studies have reported that thiopurines would reduce the risk of CRC in UC patients, also by controlling the inflammation. Although data on the impact of anti-TNF agents in the development of CRC are still lacking, it is conceivable that an optimal control of mucosal inflammation could decrease this risk.
In the present study, several sub-analyses have been performed to identify patients with a higher risk of developing CRC that could be taken into account for surveillance programmes. To this respect, in agreement with previous data, the risk of CRC in UC patients was increased with larger extensions of the disease. We also observed that this risk was increased with longer durations of the UC, although it was much lower than that previously reported. Thus, incidence rates of CRC at 10, 20 and 30 years after diagnosis of UC in our study and in Eaden's study were 0.91 vs. 2, 4.07 vs. 7 and 4.55 vs. 12, all measured in cases per 1000 py.
Based on the results from Eaden's study, considering that the risk of CRC increases with the time of evolution of the UC, several guidelines have established that the first screening colonoscopy should be offered 8 to 10 years after the diagnosis in patients with extensive colitis, and 15 years after the onset of symptoms in patients with left-sided colitis. After that, surveillance colonoscopies should be repeated every 1–2 years. New ECCO guidelines published in 2013 recommend a screening colonoscopy 8 years after the onset of colitic symptoms. Ongoing surveillance should be repeated every 1, 3 or 5 years depending on the risk factors. Nevertheless, a Cochrane meta-analysis did not find clear evidence that endoscopic surveillance prolonged survival in patients with extensive colitis, although it showed that CRCs tend to be detected at an earlier stage. New endoscopic techniques, such as chromoscopy, have been shown to be much more efficient for the detection of dysplasia than the performance of random biopsies; their incorporation into endoscopic surveillance might optimise the screening time points.
Regarding the risk of CRC in paediatric UC population, although studies are scarce, they have also reported higher incidence rates than the overall rate for adults with UC, thus supporting previous findings, namely, that early onset of UC could confer a higher risk of developing CRC.
Considering only population-based studies – that is, excluding studies from referral centres – the incidence rate of CRC is even lower. Our sub-analysis of population-based studies revealed that the incidence rate was 1.24 per 1000 py vs. 1.9 per 1000 py for studies with different designs (tertiary or referral hospital-based cohorts, private clinic series, and surgical series). Of note, only 5 of the 19 population-based studies had been included in the meta-analysis by Eaden et al., as the other 14 were published after 2000.
To have an overview of the complete history of the development of CRC in UC patients, a sub-analysis of the studies in which the surveillance had started at the diagnosis of the disease was performed, finding that the incidence rate was even lower than the overall incidence rate, and closer to the general population without UC.
Our analysis is subject to a series of limitations. Our analysis covers a large number of studies in which high heterogeneity of design, size, length of follow-up and results exists, leading to pooled statistical heterogeneity. We have to assume that, even trying to reduce heterogeneity, this could not be alleviated even performing sub-analyses. The analysis is also limited by the small sample size of some cohorts, extracted from studies carried out by surgical teams or studies based on surveillance programmes in referral centres with active follow-up. These limitations could lead to higher incidence rates of CRC in small cohorts. Consequently, our results may be overestimated due to the inclusion of a considerable number of studies performed in referral centres (vs. population-based studies), mainly when a rare event like CRC is evaluated. Another limitation of our paper may be the differential incidences according to the geographical area. This may partially explain the significant heterogeneity of the results. However, we believe that despite the limitations described, our results are sufficiently robust to be considered valid and provide valuable updated information about the current risk of CRC in UC.
In conclusion, our meta-analysis estimates an overall incidence rate of CRC of only 1.58 per 1000 py in UC patients; this incidence rate is considerably lower than previously estimated. As described in precedent literature, the extension and the duration of the disease increase the risk of CRC in this population. On the other hand, the incidence of CRC in UC patients seems to have decreased during the last few decades, and this might be explained by a tighter control of the inflammation, higher colectomy rates, the use of drugs with chemopreventive effects and a better adherence to endoscopic surveillance programmes in high-risk patients. All these factors should be taken into account when critically reviewing current CRC surveillance programmes in UC patients.
