Original article 原创文章
Rates and characteristics of postcolonoscopy colorectal cancer in the Swedish IBD population: what are the differences from a non-IBD population?
瑞典 IBD 人群结肠镜检查后结直肠癌的发病率和特征:与非 IBD 人群有什么区别?
  1. Jessica Stjärngrim1,
  2. Anders Ekbom1,
  3. Ulf Hammar2,
  4. Rolf Hultcrantz1,
  5. Anna M Forsberg1

  6. 杰西卡·斯特亚恩格林1
  7. 安德斯·埃克博姆1
  8. 乌尔夫·哈马尔2
  9. 罗尔夫·赫尔特克兰茨1
  10. 安娜·福斯贝里1
  1. Correspondence to Dr Anna M Forsberg, Department of Medicine, Solna (MedS), Karolinska Institutet, Stockholm S 17176, Sweden; anna.forsberg@ki.se
    通信Anna M Forsberg 博士,瑞典斯德哥尔摩 S 17176 卡罗林斯卡医学院 Solna (MedS) 医学系;anna.forsberg@ki.se

Abstract 抽象

Objective The rate of postcolonoscopy colorectal cancer (PCCRC) is considered a key quality indicator of colonoscopy; little is known about PCCRC in IBD.
目的结肠镜检查后结直肠癌 (PCCRC) 的发病率被认为是结肠镜检查的关键质量指标;对 IBD 中的 PCCRC 知之甚少。

Design A population-based cohort study of colonoscopies in Sweden from 2001 to 2010 was conducted. Individuals with a colorectal cancer (CRC) detected within 36 months after a colonoscopy were identified and stratified on UC, Crohn’s disease (CD) or non-IBD. The CRCs were classified as detected CRCs (dCRC) (0–6 months) or as PCCRCs (6–36 months). PCCRC rates were calculated by the number of false negative/(the number of true positive+the number of false negative) colonoscopies. Poisson regression analysis was employed to examine the association between PCCRC and IBD (CD and UC) diagnosis, age, gender, location, time period and comorbidities.
设计对 2001 年至 2010 年瑞典的结肠镜检查进行了基于人群的队列研究。在结肠镜检查后 36 个月内检测到的结直肠癌 (CRC) 个体被识别出来,并根据 UC、克罗恩病 (CD) 或非 IBD 进行分层。CRC 分为检测到的 CRC (dCRC) (0-6 个月) 或 PCCRC (6-36 个月)。PCCRC 率通过假阴性数/(真阳性数 + 假阴性数)结肠镜检查次数计算。采用泊松回归分析检查 PCCRC 和 IBD (CD 和 UC) 诊断、年龄、性别、位置、时间段和合并症之间的关联。

Results We identified 348 232 colonoscopies in 270 918 individuals. Of these, 27 123 were performed on 14 597 individuals with CD, and 51 572 were performed on 26 513 individuals with UC. There were 13 317 CRCs in the non-IBD group, 133 in the CD group and 281 in the UC group. The PCCRC rate in the CD group was 28.3% and 41.0% in the UC group. The RR for a PCCRC was 3.82 (95% CI 2.94 to 4.96) in CD and 5.89 (95% CI 5.10 to 6.80) in UC, compared with non-IBD. The highest risk was observed among rectal cancer location in CD and in younger individuals with UC.
结果我们在 270 918 例个体中确定了 348 232 例结肠镜检查。其中,对 14 597 例 CD 患者进行了 27 123 例,对 26 513 例 UC 患者进行了 51 572 例。非 IBD 组有 13 317 个 CRC,CD 组有 133 个,UC 组有 281 个。CD 组的 PCCRC 率分别为 28.3% 和 UC 组的 41.0%。与非 IBD 相比,CD 和 UC 的 PCCRC RR 分别为 3.82 (95% CI [2.94, 4.96])和 5.89 (95% CI [5.10, 6.80])。在 CD 的直肠癌位置和患有 UC 的年轻个体中观察到的风险最高。

Conclusion The high rates of PCCRC in young patients with UC and for rectal cancer location in CD might affect future performance of IBD surveillance.
结论年轻 UC 患者的 PCCRC 发生率高,以及 CD 中直肠癌位置的高发生率可能会影响 IBD 监测的未来表现。

  • cancer prevention
  • inflammatory bowel disease
  • colonoscopy
  • colorectal cancer
  • cancer epidemiology

Statistics from Altmetric.com
来自 Altmetric.com 的统计数据

Request Permissions 请求权限

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
如果您希望重复使用本文的任何或全部内容,请使用下面的链接,该链接将带您进入版权结算中心的 RightsLink 服务。您将能够获得快速的价格和即时许可,以多种不同的方式重复使用内容。

Significance of this study
本研究的意义

What is already known on this subject?
关于这个主题已经知道什么?

  • The rate of postcolonoscopy colorectal cancer (PCCRC) is considered as a key quality indicator of colonoscopy, but little is known about PCCRC among individuals with IBD.
    结肠镜检查后结直肠癌 (PCCRC) 的发病率被认为是结肠镜检查的关键质量指标,但对 IBD 患者的 PCCRC 知之甚少。

  • Known risk factors in the non-IBD population are older age, right-sided location of tumour, history of polypectomy, diverticular disease and family history of colorectal cancer.
    非 IBD 人群的已知危险因素是年龄较大、肿瘤右侧位置、息肉切除术史、憩室病和结直肠癌家族史。

  • Studies of subgroups of IBD indicate that the risks of PCCRC are higher than in the non-IBD population, but no studies have included an entire IBD population.
    对 IBD 亚组的研究表明,PCCRC 的风险高于非 IBD 人群,但没有研究包括整个 IBD 人群。

What are the new findings?
有哪些新发现?

  • This study of the adult Swedish IBD population, using a previously suggested benchmark method, shows a substantially increased risk of PCCRC in IBD compared with the non-IBD population.
    这项针对瑞典成年 IBD 人群的研究,使用先前建议的基准方法,显示与非 IBD 人群相比,IBD 患 PCCRC 的风险显着增加。

  • There seems to be a difference in the profile of risk factors in the IBD population, compared with the non-IBD population.
    与非 IBD 人群相比,IBD 人群的危险因素概况似乎存在差异。

  • In the UC group, the younger age groups seem to be at highest risk, and among individuals in the Crohn’s disease (CD) group, rectal cancer location, increases the risk substantially.
    在 UC 组中,较年轻的年龄组似乎风险最高,而在克罗恩病 (CD) 组的个体中,直肠癌位置大大增加了风险。

How might it impact on clinical practice in the foreseeable future?
在可预见的未来,它会对临床实践产生什么影响?

  • By increasing the awareness of the risk of PCCRC among patients with IBD, especially when it comes to younger patients with UC and in the rectum among those with CD, when designing surveillance programmes, this has to be taken into account.
    通过提高 IBD 患者对 PCCRC 风险的认识,尤其是当涉及到年轻的 UC 患者和 CD 患者的直肠时,在设计监测计划时,必须考虑到这一点。

Background 背景

Colonoscopy has a central role in identifying and preventing colorectal cancer (CRC). The rate of postcolonoscopy colorectal cancer (PCCRC)—that is, a CRC diagnosed 6–36 months after a colonoscopy negative for CRC—has been recognised as a key quality indicator of colonoscopy.1 2 This is because the majority of PCCRCs are considered to be the result of missed or incompletely resected lesions.3 Known risk factors for PCCRC are examination by a non-gastroenterologist, right-sided location, diverticular disease, history of polypectomy, older age and a family history of CRC.4
结肠镜检查在识别和预防结直肠癌 (CRC) 方面起着核心作用。结肠镜检查后结直肠癌 (PCCRC) 的发生率——即在结肠镜检查后 6-36 个月诊断为 CRC——已被公认为结肠镜检查的关键质量指标。1 2这是因为大多数 PCCRCs 被认为是漏诊或未完全切除病灶的结果。3 已知的 PCCRC 危险因素是非胃肠病学家的检查、右侧位置、憩室病、息肉切除术史、高龄和 CRC 家族史。4

Individuals with IBD, that is, Crohn’s disease (CD) or UC, have been excluded in most studies of PCCRC. The natural history of CRC in IBD is considered to be different than that in sporadic cancer.5 Individuals with IBD have an increased risk of CRC, and CRC represents 10%–15% of all IBD deaths.6–10 For those with a substantially increased risk, regular surveillance colonoscopies are recommended to find premalignant lesions at an early stage, starting 8–10 years after disease onset.11 12 Furthermore, recommendations on IBD surveillance now recommend a shift from white-light endoscopy with random biopsies to high-definition chromoendoscopy with directed biopsies to improve the dysplasia detection rate.13 14
在大多数 PCCRC 研究中,患有 IBD,即克罗恩病 (CD) 或 UC 的个体已被排除在外。IBD 中 CRC 的自然病程被认为与散发性癌症不同。5 IBD 患者患 CRC 的风险增加,CRC 占所有 IBD 死亡人数的 10%-15%。6–10对于那些风险显着增加的人,建议定期监测结肠镜检查,以便在发病后 8-10 年开始及早发现癌前病变。11 12此外,关于 IBD 监测的建议现在建议从白光内窥镜检查随机活检转变为高清色素内窥镜检查和定向活检,以提高异型增生检出率。13 14

PCCRC rates in the normal population vary in different studies between 2.6% and 9.0%, partly depending on the definition of PCCRC and which exclusion criteria has been applied on the dataset.1 15 A new method has been proposed by Morris et al using colonoscopies instead of CRCs as a denominator, gaining a more standardised method that can be followed prospectively.1

Little is known about the PCCRC rates among individuals with IBD—a group that undergoes many colonoscopies. Studies on elderly American patients with IBD and those previously hospitalised for their IBD in England showed an increased risk of a CRC to be a PCCRC.1 16 A recently published Dutch study exhibited a high proportion (45.0%) of PCCRCs among IBD patients with CRC.17 To our knowledge, there has been no national studies on PCCRC in an entire adult IBD population.

The aim of this study was to evaluate the rates of PCCRC in the adult Swedish IBD population and to investigate whether there is an increased rate of PCCRC in an IBD population compared with PCCRC in a non-IBD population. We also aimed to evaluate the pattern of associated risk factors for PCCRC and to compare the pattern in the IBD population with that of the population without IBD.

Method 方法

The setting for this cohort study was the entire Swedish population (9.5 million inhabitants at the end of the study).
这项队列研究的背景是整个瑞典人口(研究结束时为 950 万居民)。

Registers 寄存 器

The National Patient Register has collected and published information regarding inpatients at public hospitals since the 1960s. It was renamed the National Hospital Discharge register in 1987 when it began to include all inpatient care in Sweden. Since 2001, the National Outpatient Register has contained information on outpatient visits, including day surgeries from both private and public caregivers but not primary healthcare.18 Surgical procedures, including colonoscopies, have been recorded according to the Swedish version of the NOMESCO Classification of Surgical Procedures since 1997. The coverage is not complete, but it is substantial due to the importance of reporting for reimbursement.
自 1960 年代以来,国家患者登记册 (National Patient Register) 收集并公布了有关公立医院住院患者的信息。它于 1987 年更名为国家医院出院登记册,当时它开始包括瑞典的所有住院护理。自 2001 年以来,国家门诊登记册 (National Outpatient Register) 一直包含门诊就诊信息,包括私人和公共护理人员的日间手术,但不包括初级医疗保健。18 自 1997 年以来,包括结肠镜检查在内的外科手术已根据瑞典版的 NOMESCO 外科手术分类记录下来。覆盖范围并不完整,但由于报告报销的重要性,覆盖范围很广。

The Swedish Cancer Register gathers information regarding all incidents of malignant disorders in Sweden. It was established in 1958 and is estimated to have around 96% coverage of all malignancies.19
瑞典癌症登记处收集有关瑞典所有恶性疾病事件的信息。它成立于 1958 年,估计覆盖了所有恶性肿瘤的 96% 左右。19

In this study, we employed the registers from 1987 to 2013 to identify information on patients’ comorbidities as well as diagnoses of CRC and of IBD using the International Classification of Diseases (ICD)-O3/10 codes from 1997 and ICD-9 for the years 1987–1996.
在这项研究中,我们使用 1987 年至 2013 年的登记册,使用 1997 年的国际疾病分类 (ICD)-O3/10 代码和 1987-1996 年的 ICD-9 来识别患者的合并症以及 CRC 和 IBD 的诊断信息。

Study population, PCCRC definition and calculation of PCCRC rate
研究人群、PCCRC 定义和 PCCRC 率的计算

We used the National Hospital Discharge Register and the National Outpatient Register to identify and include all colonoscopies performed during the years 2001–2010 in Sweden on individuals 18 years and older, using the operation codes for colonoscopy and colonoscopy with biopsy (UJF32 and UJF35). Colonoscopy was denoted as an ‘IBD-colonoscopy’ if the individual was diagnosed with UC or CD at the time of or 6 months after the colonoscopy to ensure that all newly diagnosed IBD cases were included. In cases in which an individual was diagnosed with both UC and CD, the latest diagnosis was chosen. By using personal identification numbers, the individuals that had undergone colonoscopies were individually linked to the Swedish Cancer Register. All individuals without a previous CRC with a CRC diagnosed within 0–36 months after a colonoscopy were selected (2001–2013), regardless of whether the CRC was subsequently detected by a colonoscopy. All colonoscopies had a follow-up time of at least 36 months. CRCs without a colonoscopy examination 0–36 months prior to the diagnosis were not included in the dataset. Stage T0 and Tis were excluded. The CRC was defined as a PCCRC if it was detected within 6–36 months of a colonoscopy in which no cancer was detected. The colonoscopy was reported as a false-negative examination. The CRC was defined as ‘detected’ (dCRC) if registered within 6 months of a colonoscopy, and this was assigned as a true-positive examination.
我们使用国家医院出院登记册和国家门诊登记册来确定并纳入 2001-2010 年在瑞典对 18 岁及以上个体进行的所有结肠镜检查,使用结肠镜检查和结肠镜检查活检的操作代码(UJF32 和 UJF35)。如果个体在结肠镜检查时或结肠镜检查后 6 个月被诊断患有 UC 或 CD,则结肠镜检查称为“IBD 结肠镜检查”,以确保包括所有新诊断的 IBD 病例。在个体同时被诊断患有 UC 和 CD 的情况下,选择最新的诊断。通过使用个人身份证号码,接受结肠镜检查的个体被单独链接到瑞典癌症登记册。选择所有既往无 CRC 并在结肠镜检查后 0-36 个月内诊断出 CRC 的个体 (2001-2013),无论 CRC 随后是否通过结肠镜检查检测到。所有结肠镜检查的随访时间至少为 36 个月。诊断前 0-36 个月未进行结肠镜检查的 CRC 未包含在数据集中。排除 T期 0 和 Tis。如果在结肠镜检查后 6-36 个月内检测到 PCCRC,则 CRC 被定义为 PCCRC,其中未检测到癌症。结肠镜检查报告为假阴性检查。如果在结肠镜检查后 6 个月内登记,CRC 被定义为“检测到”(dCRC),这被指定为真阳性检查。

The PCCRC rate was calculated by using the number of colonoscopies as the denominator. The denominator was the sum of the false-negative and the true-positive colonoscopies. The numerator was the number of false-negative colonoscopies. For individuals who underwent multiple colonoscopies, only the first colonoscopy that detected the CRC and the first colonoscopy that failed to detect the CRC were included in the calculation. In these cases, the CRCs were considered both as a dCRC and as a PCCRC and hence contributed twice in the denominator. Colonoscopies performed after a CRC diagnosis were censored. Hence, the PCCRC rate was expressed as: the number of false negative divided by (the number of true positive+the number of false negative) colonoscopies, described in detail by Morris et al.1 Due to different exclusion criteria and methods, some results of this study differ somewhat compared with a study of the Swedish non-IBD population in the same colonoscopy cohort.20
PCCRC 率是以结肠镜检查次数为分母计算的。分母是假阴性和真阳性结肠镜检查的总和。分子是假阴性结肠镜检查的数量。对于接受多次结肠镜检查的个体,仅包括第一次检测到 CRC 的结肠镜检查和第一次未能检测到 CRC 的结肠镜检查。在这些情况下,CRC 既被视为 dCRC 又被视为 PCCRC,因此在分母中贡献了两倍。在 CRC 诊断后进行的结肠镜检查被删失。因此,PCCRC 率表示为:假阴性数除以(真阳性数 + 假阴性数)结肠镜检查,由 Morris 等人详细描述。1 由于排除标准和方法不同,本研究的一些结果与同一结肠镜检查队列中瑞典非 IBD 人群的研究略有不同。20

We also recorded the proportion of individuals who underwent a colonoscopy within 2 years from the first as an indicator of multiple colonoscopies. This could only be assessed for colonoscopies performed during 2003–2010.
我们还记录了从第一次开始 2 年内接受结肠镜检查的个体比例,作为多次结肠镜检查的指标。这只能对 2003-2010 年期间进行的结肠镜检查进行评估。

Covariates 协变量

The covariates considered in the analysis were: age; sex; location of the cancer; stage of the cancer; whether the colonoscopy was performed at a university hospital; comorbidities such as diabetes mellitus, ischaemic heart disease (IHD), primary sclerosing cholangitis (PSC), diverticulosis and former colorectal polyp in an earlier colonoscopy; and if polypectomy was recorded during the colonoscopy. Two time periods were compared: 2001–2005 and 2006–2010. Age was analysed with six age groups; the largest age group 60–70 years was set as a reference.
分析中考虑的协变量是:年龄;性;癌症的位置;癌症的分期;结肠镜检查是否在大学医院进行;早期结肠镜检查中的合并症如糖尿病、缺血性心脏病 (IHD)、原发性硬化性胆管炎 (PSC)、憩室病和前结直肠息肉;以及结肠镜检查期间是否记录了息肉切除术。比较了两个时间段:2001-2005 年和 2006-2010 年。对 6 个年龄组进行了年龄分析;最大的年龄组 60-70 岁被设定为参考。

The ICD-10 codes were employed to define the anatomical location of the tumour. The location of the cancer in colorectum was recorded and categorised as right sided, left sided (proximal vs distal to the splenic flexure) or undefined. A distinction between rectal, colonic and undefined was also made. A rectal location was established when having the ICD-10 code C20. A CRC was classified as ‘undefined’ if information about the site was missing, unspecified or in case of multiple CRCs in different segments.
ICD-10 代码用于定义肿瘤的解剖位置。记录癌症在结直肠中的位置,并将其分类为右侧、左侧(脾曲近端与远端)或未定义。还区分了直肠、结肠和 undefined。当 ICD-10 代码为 C20 时,建立了直肠位置。如果有关站点的信息缺失、未指定或不同部分有多个 CRC,则 CRC 被归类为“未定义”。

Statistical analysis 统计分析

The Student’s t-test and χ2 test were used for continuous (age) and categorical variables, respectively. Occurrence of PCCRC was defined as the outcome in all further analyses. Relative risks (RRs) were calculated using multiple Poisson regression with estimation of cluster robust sandwich SEs, using the individual as the unit of analysis. Poisson regression is mainly used for count data, but it can also be used to calculate RRs if SEs are adjusted appropriately.21 The analyses were stratified by study group (UC, CD and non-IBD). A Poisson regression analysis on all data with the study group as the covariate was also performed. Other covariates in the analysis are mentioned in the section on covariates. IHD was not included in the CD analyses since there were too few to obtain reliable results. In addition, a generalised Chow test was used to test for interaction between the covariates age group, sex, former polyp diagnosis and location of cancer of CD and non-IBD, between UC and non-IBD and between CD and UC, respectively.22
学生 t 检验和 χ2 检验分别用于连续 (年龄) 和分类变量。PCCRC 的发生率被定义为所有进一步分析的结果。使用多元泊松回归计算相对风险 (RRs),并使用个体作为分析单位估计集群稳健夹心 SEs。泊松回归主要用于计数数据,但如果 SE 经过适当调整,它也可以用于计算 RR。21 分析按研究组 (UC、CD 和非 IBD) 分层。还对以研究组为协变量的所有数据进行了泊松回归分析。分析中的其他协变量在协变量一节中提到。IHD 不包括在 CD 分析中,因为太少而无法获得可靠的结果。此外,采用广义 Chow 检验分别检测协变量年龄组、性别、既往息肉诊断以及 CD 和非 IBD 癌症位置、UC 和非 IBD 之间以及 CD 和 UC 之间的相互作用。22

Sensitivity analysis 敏感性分析

We excluded cancer stage from our main analyses due to the large amount of missing data, but we also fitted models in which we adjusted for stage. We also tried models in which we excluded cases with PSC (since PSC could not be included in the multivariate models).
由于大量缺失数据,我们将癌症分期排除在主要分析之外,但我们也拟合了针对分期进行调整的模型。我们还尝试了排除 PSC 病例的模型(因为 PSC 不能包含在多变量模型中)。

As a sensitivity analysis of the variable age, it was modelled in the multivariate analyses using cubic splines with four knots.23 To further test the significance of the variable age for the risk of PCCRC, we performed a Wald test in the multivariate analyses on the variable age, both as a categorised variable in one model and as a continuous one with cubic splines in another one. All analyses were conducted using Stata V.13.1.
作为可变年龄的敏感性分析,它在多变量分析中使用具有四个结的三次样条进行建模。23 为了进一步检验可变年龄对 PCCRC 风险的意义,我们在可变年龄的多变量分析中进行了 Wald 检验,在一个模型中作为分类变量,在另一个模型中作为具有立方样条的连续变量。所有分析均使用 Stata V.13.1 进行。

Results 结果

Overall, 348 232 colonoscopies in 270 918 individuals were identified during the years 2001–2010. Out of these, 27 123 (8%) colonoscopies were performed on 14 597 (5%) individuals with CD, and 51 572 (15%) colonoscopies were performed on 26 513 (8%) individuals with UC (table 1). There were 41 110 individuals with CD or UC who underwent at least one colonoscopy during the time interval, which is equivalent to 67% of the IBD population in Sweden 2010.24 Of the colonoscopies during this time period, 23% were performed on individuals with IBD and 77% on the non-IBD population. Age records in the cohort and in the subgroups are presented in table 1.
总体而言,在 2001-2010 年期间,在 270 918 人中确定了 348 232 例结肠镜检查。其中,对 14 597 例 (5%) CD 患者进行了 27 123 例 (8%) 结肠镜检查,对 26 513 例 (8%) UC 患者进行了 51 572 例 (15%) 结肠镜检查(表 1)。有 41 110 名 CD 或 UC 患者在这段时间内至少接受了一次结肠镜检查,相当于 2010 年瑞典 IBD 人群的 67%。24 在此期间的结肠镜检查中,23% 是对 IBD 患者进行的,77% 是对非 IBD 人群进行的。表 1 列出了队列和子组中的年龄记录。

Table 1 表 1

Descriptive data of the cohort and of the cancers diagnosed within 36 months of a colonoscopy during the years 2001–2010
2001-2010 年期间队列和结肠镜检查后 36 个月内诊断出的癌症的描述性数据

In total, 13 731 cases of CRC in the interval of 0–36 months were identified during 2001–2013, out of which 133 were among individuals with CD and 281 were among individuals with UC. In the non-IBD group, there were 13 317 CRCs identified. The IBD group, particularly the men in the IBD group, was observed to be younger than the group without IBD. Among men in the CD group, those with PCCRCs were significantly younger than those with dCRCs (p=0.030). The same phenomenon was observed in the UC group (p=0.001). In all women and in men in the non-IBD group, no significant difference in age was seen between dCRC and PCCRC (table 1). In all groups, approximately half of the PCCRCs were diagnosed as stage T3/T4: 19/42 (45.2%) in CD, 71/141 (50.3%) in UC and 435/858 (50.7%) in the non-IBD group.
2001-2013 年期间,共确定了 13 731 例间隔 0-36 个月的 CRC,其中 133 例为 CD 患者,281 例为 UC 患者。在非 IBD 组中,共鉴定出 13 317 例 CRC。观察到 IBD 组,尤其是 IBD 组中的男性比没有 IBD 的组年轻。在 CD 组的男性中,PCCRCs 患者显著年轻于 dCRC 患者 (p=0.030)。在 UC 组中观察到相同的现象 (p=0.001)。在所有女性和非 IBD 组男性中,dCRC 和 PCCRC 之间的年龄差异无统计学意义(表 1)。在所有组中,大约一半的 PCCRCs 被诊断为 T3/T4 期: CD 为 19/42 (45.2%),UC 为 71/141 (50.3%),非 IBD 组为 435/858 (50.7%)。

Table 2 shows the descriptive data of the PCCRCs, and the PCCRC rates with the colonoscopies as the denominator. In the IBD groups, 25% of individuals underwent multiple colonoscopies. The corresponding figure in the non-IBD group was 8%. The PCCRC rate among the group with CD was 28.3%, and it was 41.0% in the UC group, which was nearly five and seven times higher than the non-IBD group, where the PCCRC rate was 6.3%. Among those with IBD and concomitant PSC, nearly all CRCs were a PCCRC. Therefore, we also calculated the PCCRC rates excluding this group, which did not influence the results substantially (data not shown).

Table 2

Descriptive data of the PCCRCs, and the PCCRC rates with the colonoscopies as the denominator

In the CD group, 22/39 (56%) of the PCCRCs were left sided, and three were undefined. In the UC group, 60/127 (47%) of the PCCRCs were left sided and 14 were undefined. In the UC group, we noticed a nearly linear negative correlation (r=−0.96, p<0.001) between age and rate of PCCRC: the younger the individual, the higher the rate of PCCRC. The distribution of the time-point in the interval of 6–36 months is to be different in UC compared with CD and non-IBD with less PCCRCs early in the interval. The proportion of all PCCRCs diagnosed every 6 months in the period 6–36 months after the colonoscopy is shown in figure 1.

Figure 1

The absolute numbers of postcolonoscopy colorectal cancers (PCCRCs) diagnosed every 6 months during the period 6–36 months after the colonoscopy. The percentage is the proportion of PCCRC in each group—CD, UC and non-IBD, respectively. The pattern differs slightly between CD, UC and non-IBD, so that one-third is diagnosed the first 6 months in the CD and non-IBD groups but less than one-fifth in the UC group. CD, Crohn’s disease.

The results of the univariate analysis of the RR of developing a PCCRC within 36 months of a colonoscopy are presented in table 3. In UC, a significantly increased risk of PCCRC was noted in the two youngest age groups, whereas age did not seem to influence the risk of PCCRC in CD and in non-IBD.

Table 3

Univariate analyses for the relative risk of developing a PCCRC within 36 months of a colonoscopy

In the multivariate analysis, the RR for an IBD-CRC to be a PCCRC compared with a CRC in an individual without IBD was significantly increased: 3.82 (95% CI 2.92 to 4.96, p<0.001) in the CD group and 5.89 (95% CI 5.10 to 6.80, p<0.001) in the UC group (table 4).

Table 4

Multivariate analyses for the relative risk of developing a PCCRC within 36 months of a colonoscopy

Among individuals with CD, there was an increased risk of a CRC to be a PCCRC if it was left sided and even more so if located in the rectum. However, when adjusting for the other covariates, only the increased risk if located in the rectum remained significant. The RR for PCCRC in the UC group was significantly higher in the younger age groups in the univariate analysis; in the multivariate analysis, this only remained significant among individuals with UC aged 30–40 years. In the IBD groups, unlike the non-IBD group, the finding of a polyp in a previous examination was not significantly associated with a higher risk of PCCRC.

The significantly increased risk of PCCRC in right-sided location of the tumour, among women and in individuals with diverticulosis and IHD, was not observed among those with IBD, though it was observed in the non-IBD group.

Whether the examination was performed in a university hospital or not did not seem to make a difference in any of the groups. A significantly lower RR for PCCRCs was noted over time in the non-IBD and UC group, but no significant improvement could be observed in the CD group.

Sensitivity analyses

In the multivariate analyses, a Wald test for the variable age, both expressed as six age categories or as cubic splines with four knots, showed insignificant results (p>0.05) for CD and for non-IBD. However, in UC, age was a significant risk factor for PCCRC both for age expressed as categories as well as cubic splines.

Stage was excluded from the main analyses, but we fitted models in which we adjusted for stage. Models were fitted in which cases with PSC were excluded. These results did not differ substantially in any of these cases compared with those models that are presented in tables 3 and 4.

Interaction analyses

In interaction analyses for UC compared with non-IBD, there were three variables that were significantly different risk factors for PCCRC: male gender (p=0.010), age 30–40 years (p=0.024) and former polyp diagnosis (p=0.001). A Wald test for overall comparison between the models showed a significant difference between UC and non-IBD (χ2=482.4, p<0.001).

For CD compared with non-IBD, no individual variable had a significant difference in risk, but the Wald test showed a significant overall difference between the two groups (χ2=69.4, p<0.001). There were no significant differences between UC and CD for any individual variable, nor for the overall model (χ2=18.2, p=0.077).

Discussion

This register-based population cohort study of the rates and characteristics of PCCRC in the Swedish IBD population shows a substantially increased risk of PCCRC compared with the non-IBD population. The characteristics of PCCRCs in the IBD population seem to differ from the non-IBD population. Some key differences that we noted were the tendency of an increased risk in the youngest age groups and that age as a risk factor for PCCRC seems to be of importance mainly in the UC group.

Furthermore, an increased risk was observed for rectal cancer location in individuals with CD.

The increased risk among women in the non-IBD group was not observed among the IBD groups, nor the increased risk in the diagnosis of diverticular disease or history of previous benign colorectal polyp.

High rates of PCCRC in IBD

A recent single-centre study showed previously unknown high-grade dysplasia (HGD)/CRC in 29% of the colectomy specimens in patients with IBD.25 Some of the PCCRCs in our study could be explained by a finding of an HGD in the index colonoscopy, and in cases of slow handling, the diagnosis of a CRC from the colectomy specimen will be noted more than 6 months after the index examination. However, at hospitals that have appropriate handling, the CRC diagnosis may be recorded within 6 months and therefore counted as a dCRC. The perioperative finding of CRC in strictures in the CD group, which is a known risk factor for a finding of a CRC, could contribute to the high PCCRC rates. An explanation could also be that, as opposed to international recommendations, the surveillance colonoscopies are not performed in endoscopic remission, which makes it more probable to miss important lesions.11 Another explanation could be that IBD-CRCs are considered more fast-growing than non-IBD-CRCs and that the index colonoscopy therefore was true negative.26 The high PCCRC rate of T3/T4 tumours does, however, indicate that increased rates of missed lesions at the index colonoscopy are an issue. There is a consensus that the PCCRC rates should be lowered.27 However, it is difficult to speculate on what would be an acceptable PCCRC rate in the Swedish IBD population. Nevertheless, we have reason to believe that the high PCCRC rates at least partly indicate that we are still not making use of the full potential of colonoscopies in patients with IBD and that there is room for improving surveillance strategies and colonoscopy quality.

Time to PCCRC diagnosis

In CD and non-IBD, half of the PCCRCs were diagnosed 6–18 months after the colonoscopy. One explanation could be that an early re-examination is performed due to persisting symptoms. For UC, the PCCRCs are more evenly distributed in the interval. That might be taken as an indication that PCCRCs in UC have a different genesis compared with the other two groups. This could be an issue for further studies.

Comparison with other studies on PCCRC in IBD

One important matter when comparing studies on PCCRC rates is that different calculation methods may be employed. The key issue is what the denominator contains.1 Our PCCRC rates would have been higher, if we had employed CRC in the denominator. Therefore, the consensus paper on PCCRC is of great importance in order to standardise the methods, making comparisons possible.27 It is also important to have uniform exclusion criteria.27

In an American study of elderly patients with IBD, the PCCRC rate was lower: 15%.16 In that study, CRC was used as a denominator, while we used colonoscopies; and different exclusion criteria were employed compared with ours making the results difficult to compare. Though, their rate seems to be substantially lower than ours and that may have several explanations, such as different quality of IBD surveillance or different disease severeness in the cohorts. In an English study with the same design as ours by Morris et al,1 an increased risk of PCCRC among individuals who had previously been hospitalised for IBD was noted. Their PCCRC rates were 32% for CD and 36% for UC, respectively. This study population represents the most severely ill part of the IBD population, which might make the results difficult to compare; yet, one would expect that severely ill patients had a higher risk of PCCRC compared with those with a milder disease. A recently published Dutch study by Wintjens et al 17 exhibited a PCCRC rate of 45%, which is considerably higher than the American and English studies. Their rate is somewhat higher than ours, but they have employed a longer time interval for PCCRC. Different calculation methods for the PCCRC rate reduces the conditions for comparison of the figures.

Age as a risk factor for PCCRC

Our results concerning the impact of age for the risk of PCCRC were partially surprising, and they seem to be contrary to some other studies.1 16 The results, that is, younger age was noticed to be a risk factor for PCCRC in UC and no significant risk factor in the rest of the population, were tested with the same outcome, both by sensitivity analyses as well as interaction analyses. Our results may be partially supported by a study by Lutkins et al, which found that 22%–28% of all CRCs in IBD patients develop early, before the recommended starting point of surveillance.28 In a study by Shi et al, it was observed that in a population-based setting, more elderly onset patients (age >60 years) had flat dysplasia with shorter disease duration, which could indirectly support the notion that the PCCRC risk would be higher in older age groups. However, they did not report a higher standardised incidence ratio among the elderly onset patients compared with those of non-elderly onset disease.29

Rectal cancer in CD and risk of PCCRC

The observation that rectal location of the cancer in the CD group had a greatly increased risk of being a PCCRC was not expected. This could perhaps partly be explained by technical and visualisation difficulties when performing colonoscopy in patients with CD with rectal manifestation. Complications in this area, such as fistulation and strictures, might also contribute to the high PCCRC rate. The way we perform a rectal examination in patients with CD may need additional care, such as changing to a slimmer endoscope or only approving examinations performed in endoscopic remission.

Previous diagnosis of colorectal polyps and risk of PCCRC

The presence of a polyp at a previous colonoscopy in an individual with UC did not affect the risk of future PCCRC, while in the non-IBD population, a three times increased risk was observed. One explanation could be, that CRC in UC does not follow the classical adenoma–CRC pathway,30 which is consistent with recent studies showing that adenomas are more seldom found in patients with UC.31 32

Time trends

In a large and long-running surveillance programme in IBD, initial work showed that 16 out of 30 CRC were interval CRC.33 More recent work from the same centre showed that despite decreased colectomy rate for dysplasia, interval cancers decreased over four decades and that surveillance appears to now detect cancer at an earlier stage as well as reduces the risk of developing interval cancer.34 The surveillance of IBD with total colonoscopy started in a larger scale in the early 1990s in Sweden.35 The registers can only deliver complete colonoscopy data from 2001. That means that we can only discuss whether the surveillance is becoming better, not whether surveillance in itself lowers the PCCRC rate compared with the earliest time period. What we have noticed is that the risk of PCCRC in UC is lower in the later time period (2006–2010) compared with the earlier (2001–2005), but we do not know with safety that this could be attributed to better surveillance strategies.

Strengths of the study

To our knowledge, no previous national study on PCCRCs in an entire adult IBD population exists. The study is large, encompassing almost 350 000 colonoscopies. We also have reason to believe that registers in Sweden are of good quality and validity.36

Weaknesses of the study

A register-based study such as this one has its weaknesses. We lack information on the extent of colitis, disease duration and characteristics of the mucosal lesions identified at the index examination. Information on the indication for the colonoscopy and adherence to potential surveillance programmes is also missing. Other weaknesses include the lack of information on the quality of the index colonoscopy, such as completeness of the procedure and the quality of bowel preparation. One important drawback is that in the IBD population, we do not have information on previous findings of dysplasia without polyps, particularly as low-grade dysplasia is considered a risk factor for development of advanced neoplasia.34 Furthermore, the coverage on reported colonoscopies is not complete, but we have no reason to believe that this loss of information would be differential. Although this study has a very large study population, the absolute number of CRCs in the IBD population is fairly low, and some of insignificant results in the IBD subgroups could be due to lack of power.

Conclusion

The PCCRC rates in the IBD groups seem to be high. The high rates in the more advanced stages could indicate that these PCCRCs truly represent missed lesions at the index or surveillance colonoscopy, highlighting the need for finding better surveillance strategies.

The high rates of PCCRC in young patients with UC and for rectal cancers in CD might affect the way we employ surveillance in the IBD group in the future.

References

Footnotes

  • Contributors Planning the project: all authors. Data collection: AMF and UH. Statistical analysis: AMF, UH and JS. Analysis of the results: all authors. Drafting the manuscript: JS, AMF and UH. Revision of the manuscript: AMF. Critical reading and approving the manuscript: all authors. All the authors have approved the revised manuscript.

  • Funding Financial support was provided through the regional agreement on medical training and clinical research (ALF) between Stockholm County Council and Karolinska Institutet by the Swedish Society of Medicine and by the Swedish Cancer Society.

  • Competing interests None declared.

  • Ethics approval The study was approved by the local ethical committee at Karolinska Institutet (Dnr: 2015/690-31/2).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Patient consent for publication Not required.