Introduction

For slightly more than a decade, primary screening for colorectal cancer (CRC) by colonoscopy has been recommended and offered to the general population at average risk for CRC in the United States (US) and Germany. In both countries, screening colonoscopy is generally offered on an “opportunistic” basis, i.e. it depends on individuals to request screening or on their health advisors to recommend screening, and individuals are not invited from a centralized register [1].

Post-polypectomy surveillance using colonoscopy is an integral part of the CRC screening process and required after approximately 20-30% of screening colonoscopies. It has been shown to be an effective tool for prevention of CRC if a good compliance is assured [2,3]. Current guidelines in the US and Germany distinguish between low-risk- and high-risk-adenoma situations in which the recommended surveillance interval is 3 years and 5 (to 10) years, respectively [4,5]. A shorter follow-up is needed in case of incomplete removal of adenomas or, independent of screening, due to signs and symptoms of gastrointestinal diseases.

So far, only limited evidence exists on the actual utilization of surveillance colonoscopy in community practice. Studies conducted in the US reported overuse of surveillance colonoscopy particularly among low-risk individuals but also underuse among individuals with advanced findings at screening [6-9]. Some evidence also suggests that physicians tend to recommend colonoscopic surveillance more frequently than necessary according to guidelines [10-13]. Furthermore, the communication between endoscopists, primary care physicians and patients about results of screening colonoscopies and consequences for surveillance may often be suboptimal [11,14]. Very little evidence on the utilization of surveillance colonoscopy in community practice is available from the German healthcare system despite the introduction of colonoscopy as primary screening offer in 2002. Two recent studies suggested frequent utilization of additional colonoscopies along with substantial adenoma yield in the first 3 years after screening colonoscopy [15], and also substantial inter-physician variation in follow-up habits after screening colonoscopy [16].

In this study, we aimed to assess patient adherence to physician recommendations for surveillance in a follow-up period of up to 4 years after screening colonoscopy in the German opportunistic CRC screening program.

Methods

A follow-up study of screening colonoscopy participants in Saarland, Germany, was conducted using statutory health insurance (SHI) claims and routine screening colonoscopy documentation data.

Results

A total of 20,058 individuals with screening colonoscopies performed between 2007 and 2009 were included in the study. Figure 1 shows a flow diagram of the identification of the study population. Characteristics of the study population are given in Table 1. Slightly more women (54%) than men were included and the mean age was 65 years. An early surveillance colonoscopy after 3, 6 or 12 months was recommended for 3% and surveillance colonoscopy after 36 months was recommended for 18% of the study population. In the majority of screening colonoscopies, 70%, no surveillance (i.e., another screening colonoscopy after 10 years) was recommended.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Flowchart.

https://doi.org/10.1371/journal.pone.0082676.g001

The cumulative utilization of follow-up colonoscopy according to different surveillance recommendations up to 4 years after the screening colonoscopy is depicted in Figure 2. Utilization started to increase markedly shortly before surveillance was due. The increase was sharper with shorter intervals. Among individuals for whom surveillance after 3, 6 and 12 months was recommended, 72%, 62% and 44%, respectively, had a follow-up colonoscopy within the subsequent 2 years after screening, as shown in Table 2. Of those with recommended surveillance after 36 months, 36% had a follow-up colonoscopy until 1 year after the due date. In this subgroup, the cumulative utilization of follow-up colonoscopy up to 4 years varied slightly depending on whether low-risk adenomas (35%) or high-risk adenomas (43%) had been detected (sensitivity analysis not shown in Tables or Figures). However, substantial variation was observed in the age-groups 55-64 years (42%), 65-74 years (35%) and ≥75 years (17%) (sensitivity analysis not shown in Tables or Figures). Of those without surveillance recommendation, 7% had an additional colonoscopy within 3 years and 12% had an additional colonoscopy within 4 years after screening colonoscopy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Colonoscopy utilization after screening colonoscopy according to physician recommendations for surveillance.

https://doi.org/10.1371/journal.pone.0082676.g002

The associations between non-adherence to physician recommendations for surveillance colonoscopy (assuming margins that tolerate a delay of 33%) and a set of potential predictors are shown in Table 4. Older age and length of surveillance interval were most strongly associated with non-adherence. In the age group ≥75 years, the odds of non-adherence was increased more than 3-fold compared to age group 55-65 years. Non-adherence was increased more than 2-fold in case of a surveillance interval of 36 months, as opposed to a short-term interval of 3 months. Furthermore, non-adherence was more frequent if no polypectomy had been performed at screening and if the screening colonoscopy was negative for adenomas.

Discussion

In this study, screening colonoscopy participants in the German “opportunistic” CRC screening program were followed-up for utilization of additional colonoscopies for up to 4 years using statutory health insurance claims data. Colonoscopy utilization started to increase shortly before recommended surveillance colonoscopies were due, but the cumulative percentages of colonoscopy utilization were still low a year after the recommended surveillance dates. For the different recommended intervals and definitions of adherence commonly only less than 50% of participants could be considered adherent to surveillance recommendations. Thus, surveillance was often either not utilized or only with considerable delay.

Non-adherence was more pronounced with increasing length of the surveillance interval. An extension of the surveillance interval in the low-risk adenoma situation from 3 to 5 years in 2008 could partially explain the low levels of adherence among participants with screen-detected low-risk adenomas [19]. However, as a subgroup analysis showed, adherence was also low among participants with high-risk adenomas. Furthermore, non-adherence was more pronounced with high age (especially ≥75 years). This observation may partly be explained by increasing age-related comorbidities (along with increased risk of adverse effects) and immobility in the elderly population which could hamper colonoscopy utilization. Concerning the stopping of surveillance, the 'European guidelines for quality assurance in colorectal cancer screening and diagnosis' state that the decision for or against a surveillance examination "should depend not only on adenoma characteristics, but also on the patient’s age and wishes, and the presence of significant co-morbidity" [20]. Furthermore, negative screening colonoscopy and screening colonoscopy without polypectomy were also positively associated with non-adherence to surveillance recommendations, but these subgroups were small and it seems likely that other reasons than adenoma surveillance could have led to the recommendation of a follow-up, e.g. repeat colonoscopy due to inadequate bowel preparation.

Significant waiting times for colonoscopy might in theory cause patients to be falsely classified as non-adherent to surveillance. However, with respect to the German healthcare system, it would be exceptional if a surveillance colonoscopy could not be performed within a short time (presumably less than a month) after contacting a gastroenterologist’s office. Problems to schedule a colonoscopy are therefore unlikely to have a major impact on the results.

Most of the published evidence on surveillance colonoscopy utilization stems from the US healthcare system. It suggests that rates of follow-up colonoscopies may be higher in the US than in the present German study, although comparability is limited due to different outcome measures [7,8,21,22]. Based on data from the late 1990s, the fraction of colonoscopies done for adenoma surveillance in the US was reported to be 24%, whereas it was only 11% in Bavaria, Germany, in 2006 [23,24]. A recent study from the German healthcare system suggested utilization rates of additional colonoscopies in the same magnitude as observed in the present study, but did not assess adherence to recommendations [15]. It could only track patients if they returned to the same gastroenterologist who performed screeening, whereas here any further colonoscopies irrespective of the provider could be taken into account. Another study, based on the same cohort as the former study, indicated that follow-up habits might vary substantially between physicians [16].

Alerting systems and reminders have been shown to effectively increase adherence to surveillance [25,26]. Given the observed low adherence, it appears sensible to consider their widespread implementation. Furthermore, health services research studies on physician-patient communication concerning screening results and surveillance requirements might help to better explain low levels of adherence to surveillance recommendations and to identify further risk factors for non-adherence, which could be targets for interventions to improve adherence to surveillance.

Apart from surveillance, participation in screening colonoscopy in the German “opportunistic” CRC screening program is also known to be low [18,27]. European guidelines have recently advocated implementation of “organized” CRC screening to ensure high coverage, and initiatives have been started in Germany to prepare personal invitations to CRC screening on a nationwide basis, i.e. to introduce a central element of “organized” screening [28,29]. The present study supports suggestions that colonoscopy screening should be regarded as a “package” that includes surveillance [30]. In forthcoming reforms of the CRC screening system in Germany, which are likely move towards more “organized” forms of screening, it may be of importance to also consider invitations and reminders with respect to surveillance in order to assure a high quality of the CRC screening program overall.

This analysis focused on individual recommendations made by the physicians who performed screening colonoscopy. An assessment of the appropriateness of these recommendations according to guidelines was deemed beyond the scope of this article. Studies conducted in the US healthcare system have suggested that physicians often deviate from guidelines by recommending earlier or unnecessary colonoscopic surveillance, e.g. in case of hyperplastic (i.e. non-adenomatous) polyps [12,31]. In this study, only a very small fraction of participants without adenoma findings had a recommendation for a surveillance colonoscopy suggesting that inappropriate recommendations for surveillance may have been rare in practice.

Several limitations are important to be considered for the interpretation of the present results. First, this study is based on routine data which were not primarily collected for scientific purposes. Several patient-level factors were unavailable that would have been of interest because they possibly impact on adherence to surveillance recommendations, e.g. education, smoking, and obesity. Also, physician-level factors such as experience, colonoscopy volume or specialization were not available but might influence adherence to surveillance colonoscopy. Second, due to a lack of similar studies and no available expert-consensus, the margins chosen to define adherence are subject to a certain degree of arbitrariness. However, even when fairly liberal adherence margins were applied in sensitivity analyses the overall observation of low adherence levels remained unchanged. Third, only surveillance recommendations up to 3 years were considered in the present study. There were very few recommendations for 5 years (classified under ‘other interval’ in Table 1) because at the time when most of the included screening colonoscopies were conducted, the recommended interval was still 3 years both in the high-risk and low-risk adenoma situation [32]. Fourth, possible losses to follow-up due to unobserved mortality might have resulted in an underestimation of surveillance rates. Using age- and sex-specific life-tables of the Saarland population, we estimated that 7% of the study population who had a recommendation for colonoscopic surveillance in 2007 can be expected to have died in the subsequent 4 years. Thus, the potential underestimation of the cumulative utilization in this time-period caused by mortality may be at most 7 percentage points (probably lower because people with severe comorbidity are less likely to have undergone screening colonoscopy). Fifth, it is unknown whether additional colonoscopies were actually performed for surveillance. They may partly also have been done primarily for signs and symptoms of gastrointestinal diseases. The codes used for health insurance claims do not distinguish between diagnostic and surveillance colonoscopies. According to the findings among participants with no surveillance recommendation, up to 12% may have had another colonoscopy within 4 years for diagnostic reasons. Finally, when predictors of non-adherence of surveillance colonoscopy were assessed in logistic regression models, only a small proportion of the evidence was based on individuals with a short-term surveillance recommendation of 3 or 6 months. In additional stratified analyses by recommended interval, the identified predictors of non-adherence (age-group, polypectomy, adenoma findings) did not reach statistical significance for these short-term intervals (results not shown). Therefore, confirmation of the predictors for short-term intervals in larger samples is required.

The results observed in this study reflect community practice in Saarland, Germany, over the time-period from 2007 to 2011. Although this is a regional assessement, due to identical incentive schemes for physicians the results may, in principle, be generalizable to the SHI system in the whole of Germany. Variation is likely to exist between physician practices. Physicians partially invite or remind their patients of the due surveillance colonoscopy, but the extent of this practice is unclear.

In summary, this study suggests frequent non-adherence to physician recommendations for surveillance colonoscopy in community practice, particularly with longer surveillance intervals. Increased efforts to improve adherence, including introduction of more elements of an organized CRC screening program (e.g. widespread use of alerting systems and reminders with invitations to surveillance colonoscopy), seem necessary to assure a high-quality CRC screening process. The upcoming changes to the CRC screening system in Germany in order increase screening uptake and to implement Eurpean guidelines on CRC screening may provide an opportunity to also increase the utilization of surveillance colonoscopy.

Acknowledgments

The authors are grateful to the regional Association of Statutory Health Insurance Physicians in Saarland, Germany, for provision of the data used for this study.