Guidelines recommend that CRC survivors receive regular surveillance via colonoscopy[2–7, 10, 11, 25, 32]. Previous studies found that the rate of surveillance colonoscopy among CRC survivors is low; however, these studies used older data sets which limit generalizability[12–15, 17, 19, 21, 33]. Using rich medical record and interview data from a recent multi-regional population-based cohort of CRC survivors, we extended prior research on surveillance colonoscopy among CRC survivors. This study using medical record abstraction included survivors with a range of insurance types, ages, and ethnic backgrounds, strengthening inferences about external validity based on study findings.
We found that only 49% of CRC survivors received surveillance colonoscopy. This rate is somewhat lower than previously-reported rates at 18 months (52-61%)[12–15]. The prior studies' slightly higher estimates may simply be due to additional analysis time or inclusion of other colon examinations (such as sigmoidoscopy and barium enema)[12–15]. However, given that colonoscopies were 70-97% [12–14] of all colon examinations, the use of other examinations does not fully account for the higher utilization rate in earlier studies. Cooper and Payes suggested that survivors may be delaying their first colonoscopy until after 1 year. It is possible that since 2002, the average time of receipt of the first postoperative colonoscopy has become further delayed beyond 14 months. Notably, sensitivity analyses suggested that are findings are robust with respect to length of post-surgical follow-up duration (13 months, 13.5 months, 14.5 months, and greater than 15 months after diagnosis). Moreover, excluding patients with both rectal and colon cancer (versus only one of the two) yielded identical results for both the direction and statistical significance of effect sizes.
There is large variation in surveillance colonoscopy among subpopulations. The greatest variation was found between study sites, where we found rates ranging from 37% to 57%. This finding is similar to the geographic variation reported using Surveillance Epidemiology and End Results (SEER) data and in a Canadian registry study[18, 19, 21, 33]. This regional variation may be a function of differences in practice patterns or health care delivery systems. Our results show that the utilization of surveillance colonoscopy is lowest in managed care organizations. Because the sites of care are either geography-based, as in the Los Angeles and Alabama sites, or based around a health care system, as in the VA and the managed care sites, it is difficult to disentangle these different possible influences on colonoscopy use in the present study.
Clinical factors were strongly associated with colonoscopy use. Rectal cancer survivors were less likely to than those with colon cancer to receive colonoscopy, which is consistent with previous reports[14, 16, 17, 20]. Because colonoscopies comprised 96-99% of all procedures performed, the higher use of sigmoidoscopy among rectal patients is unlikely to account for this finding[12, 14, 17]. Instead, less frequent colonoscopy use among rectal cancer survivors may reflect a selection bias. Rectal cancer survivors are more likely than colon cancer survivors to get neoadjuvant therapy (chemotherapy or radiation), but our criteria of surgery within one month excluded many survivors who received neoadjuvant therapy. Indeed, due to censoring, the proportion of participants who received neoadjuvant therapy fell from 9% to 1% among colon cancer survivors and from 30% to 0% among rectal cancer survivors. Because patients who receive neoadjuvant therapy may be more likely to have a timely colonoscopy, it is also possible that those who were most likely to adhere to colonoscopy guidelines were excluded from the analysis.
Survivors with more severe comorbidities were less likely to receive a colonoscopy. Previous studies using SEER data among Medicare fee-for-service populations have shown similar findings[16, 18, 20], although two studies[14, 15] of managed care populations showed no effect of comorbidities on colonoscopy use. Patients with more severe comorbidities may have shorter life expectancies, leading physicians to be less likely to offer, and patients less likely to accept, surveillance colonoscopy. As well, survivors with more severe comorbidities may be too frail to undergo colonoscopies.
Seeing a primary care physician in the first year after diagnosis is associated with a higher likelihood of receiving a colonoscopy. Similarly, those individuals who received adjuvant chemotherapy were more likely to receive a colonoscopy. Two recent studies found that CRC survivors who were seen by both an oncologist and a primary care physician received the highest proportion of recommended ongoing cancer preventive care (such as cervical cancer screening and mammography) compared to survivors who saw either type of physician (or had seen neither)[34, 35]. Primary care visits and chemotherapy use may be indicators of access to high-quality care and having a medical home, thereby increasing referral for timely surveillance colonoscopies. If primary care providers are responsible for most referrals, improving coordination of care as patients move from acute cancer care to ongoing care may increase adherence to colonoscopy guidelines. The Institute of Medicine has prioritized facilitating this transition, with the goal of improving ongoing care for cancer survivors.
Alternatively, prevention-oriented survivors may be more likely to visit a primary care physician, to seek out adjuvant chemotherapy to reduce the risk of recurrence, and, similarly, to request a colonoscopy to reduce their risk of cancer. Future longitudinal studies should address survivors' attitudes toward colonoscopy and CRC prevention and whether these attitudes affect colonoscopy use.
Sociodemographic factors were not related to colonoscopy use. Although previous studies have found associations between sociodemographic factors and colorectal surveillance, results have conflicted across studies[12–18, 20, 21]. Taken together with the current study, this suggests that sociodemographic factors are inconsistent predictors of colonoscopy use. Instead, the pattern of results appears to be that survivors are more likely to undergo surveillance colonoscopy if they have the best prognosis (those with mildest or no comorbidities, earlier stage, and use of adjuvant chemotherapy) or have a primary care physician.
This study has limitations. First, inclusion criteria of surgery within 1 month after diagnosis could limit the findings. However, sensitivity analyses that permitted longer lags between diagnosis and surgery, as well as a sensitivity analysis that did not limit timing of the surgery, showed similar results. It is still possible that those with later surgeries and those with early post-surgical colonoscopies had different patterns of colonoscopy use than their counterparts. Second, this study sample was largely insured, albeit to different extents and with different insurance types. Despite high levels of insurance, however, differences in surveillance colonoscopy use were observed in this study. Third, we were unable to assess the clinical impact of nonadherence in our sample. Fourth, as we could not assess reasons for colonoscopy use (i.e., diagnostic versus surveillance), some colonoscopies may have been diagnostic instead of for routine surveillance. This is unlikely, though, because Cooper et al looked at indications for colonoscopy use among cancer survivors and found that 95% of colonoscopies performed were routine. Fifth, the study assessed colonoscopy use by 14 months after surgery. Although significant variations in colonoscopy use were apparent even in this limited time frame, a longer analysis period may have accommodated delays in scheduling colonoscopies and revealed additional colonoscopies that would still be adherent to clinical practice guidelines in effect during the study period. A longer analysis period also would have included colonoscopies that were delayed due to long wait times. Finally, survivors who were missing data on surgery and staging were excluded from the analytic sample, and this may have created a selection bias. Although we do not know whether those with missing surgery data did not have colorectal cancer surgery or whether their surgery data are missing, the colonoscopy surveillance guidelines are targeted only to patients who do have surgery. A subset of patients lacking surgery data presumably did not have surgery and are appropriately excluded from the analytic sample, limiting the bias in our estimates. Patients who lacked staging information had a 41% colonoscopy rate overall. The lower rate of colonoscopies compared to the rate in the analytic sample (49%) may simply reflect the presence of patients with advanced disease in the unstaged sample; patients with advanced disease are not appropriate candidates for surveillance colonoscopy.