Results
From a total of 2,212 unique studies identified using the pre-specified systematic search strategy, 12 observational studies (10 population-based studies and 2 multicenter studies) were included in the quantitative synthesis. Six RCTs, including one pooled multi-cohort analysis of patients enrolled in eight randomized chemoprevention trials, thirteen studies with the absence of information on all CRCs (resulting in inability to estimate prevalence of interval CRCs vis-à-vis the total number of CRC cases), and one prospective study with lack of adequate follow-up were excluded. There were two studies performed in the same Surveillance, Epidemiology, and End Results (SEER)-Medicare cohort with completely overlapping time periods of which only one was included; sensitivity analysis was performed using the other study, which was also used for risk factor assessment in case data were not available from the primary study. Figure 1 shows the schematic diagram of study selection.
(Enlarge Image)
Figure 1.
Flowchart summarizing study identification and selection. CRC, colorectal cancer.
Characteristics and Quality of Included Studies
Table 1 describes the characteristics of the included studies. Overall, these 12 studies reported on 139,813 CRC cases, of which 7,912 were identified as interval CRCs diagnosed between 1988 and 2011. Ten studies were categorized as population-based studies; one of these studies, performed using the SEER-Medicare database, was nationally representative but not indicative of all patients in the population. Two studies were conducted in Ontario with only slight temporal overlap (1997–2002 and 2000–2005). Nishihara and colleagues performed a prospective cohort study following patients from two cohorts (Nurses' Health Study and Health Professionals Follow-up study) from Massachusetts over 20 years. Haseman and colleagues identified all interval CRCs identified across 20 hospitals in Indiana between 1988 and 1993. Nine of the included studies were retrospective in nature; three studies were prospective with adequate follow-up. Brenner and colleagues reported the results of their prospective population-based case–control study as two distinct cohorts—patients with negative index colonoscopy (i.e., no polyps identified at initial colonoscopy) and patients with polypectomy at index colonoscopy. Seven studies were performed in North America (three in Canada and four in the United States, four studies in the Europe (two in Germany, one in Denmark, and one in the Netherlands, and one study in New Zealand. All studies were deemed to be of high quality. Individual quality items for each of the included studies are provided in the Supplementary Table S1 http://www.nature.com/ajg/journal/v109/n9/suppinfo/ajg2014171s1.html.
Prevalence of Interval CRC
Overall. The prevalence of interval CRC ranged from 1.8 to 9.0% in the included studies. On pooled analysis, the prevalence of interval CRCs was 3.7% (95% CI=2.8–4.9%; Figure 2). This corresponds to 1 in 27 (95% CI=20–36) CRCs being interval CRCs.
(Enlarge Image)
Figure 2.
Pooled prevalence of interval colorectal cancers (CRCs), as compared with all CRCs. Numbers represented as proportion of interval CRCs per all CRCs. Overall, this transforms into a pooled prevalence of interval CRCs of 3.7% (95% confidence interval (CI)=2.8–4.9%).
Site Specific. In nine studies, 53,847 proximal CRCs were identified, of which 4,615 were classified as interval CRCs. On pooled analysis, the prevalence of proximal interval CRCs was 6.5%. (95% CI=4.9–8.6%; Figure 3). This corresponds to 1 in 15 (95% CI=12–20) proximal CRCs being interval CRCs. In the same nine studies, 77,922 distal CRCs were identified, of which 2,726 were classified as interval CRCs. On pooled analysis, the prevalence of distal interval CRCs was 2.9% (95% CI=2.0–4.2%; Figure 3). This corresponds to 1 in 34 (95% CI=24–50) distal CRCs being interval CRCs. As compared with detected CRCs, interval CRCs were 2.4 times more likely to be proximal CRCs than distal CRCs (OR=2.36; 95% CI=2.11–2.65).
(Enlarge Image)
Figure 3.
Site-specific pooled prevalence of interval colorectal cancers (CRCs; proximal vs. distal). Numbers represented as proportion of interval CRCs per all CRCs. Overall, this transforms into a pooled prevalence of proximal interval CRCs of 6.5% (95% confidence interval (CI)=4.9–8.6%) and of distal interval CRCs of 2.9% (2.0–4.2%).
Subgroup Analysis. Considerable heterogeneity was observed in the overall as well as site-specific analysis (I for prevalence of overall, proximal, and distal interval CRCs: 99%, 98%, and 98%, respectively). On subgroup analysis, the pooled prevalence of overall and proximal interval CRCs was significantly higher in population-based studies, as compared with multicenter studies, partially accounting for the observed heterogeneity (Pinteraction<0.01; Table 2). Studies performed in the Europe observed a lower prevalence of proximal and distal interval CRCs as compared with North America (Pinteraction<0.01).
Sensitivity Analysis. When we used an alternative definition of interval CRCs as those diagnosed within 6–60 months of previous colonoscopy, the pooled prevalence of interval CRCs was 4.3% (95% CI=2.6–6.9%; four studies). Sensitivity analysis, replacing one study performed using the SEER-Medicare database with another, did not significantly change the prevalence estimate (prevalence of interval CRC, 4.1%; 95% CI=3.2–4.3). On excluding the negative colonoscopy cohort study by Brenner et al. (no polyps identified in index colonoscopy), the pooled prevalence of interval CRCs in population-based studies was unchanged (prevalence=4.4%; 95% CI=3.9–5.2%); likewise, on excluding the study performed using the SEER-Medicare database (older individuals), the pooled prevalence of interval CRCs in population-based studies was 4.4% (95% CI=3.0–6.4%). To assess whether any one study had a dominant effect on the pooled prevalence of interval CRC, each study was excluded and its effect on the main summary estimate and I test for heterogeneity was evaluated; no study markedly affected the overall prevalence of interval CRC or degree of heterogeneity. As substantial heterogeneity was observed in the overall analysis, evaluation for publication bias using funnel plot asymmetry was not appropriate.
Time-trend Analysis. Based on eight studies from which data allowed estimation of interval CRC risk at different time points, there was a declining trend in the prevalence of interval CRCs from 4.8% (95% CI=2.9–7.8) in 1990s to 4.2% (95% CI=2.3–7.4) between 2000 and 2005, and 3.7% (95% CI=2.6–5.2) beyond 2005; however, this trend was not statistically significant (P=0.68).
Quality of Evidence. Using the GRADE approach for assessing the quality of evidence, our summary estimate on prevalence of interval CRC in this systematic review based on observational studies and with the high degree of heterogeneity among the studies, was classified as very low-quality evidence.
Risk factors for interval CRCs
Clinical Factors.Table 3 reports the baseline clinical characteristics comparing interval CRCs to detected CRCs. Seven studies (out of eleven studies for which data were available) observed that patients diagnosed with interval CRCs were slightly older than patients with detected CRCs. Using adjusted OR reported in individual studies, older patients (>65–70 years) had a higher risk of interval CRCs than younger individuals (<65–70 years; adjusted OR=1.15, 95% CI=1.02–1.30, I=20%, four studies; unable to estimate unadjusted OR). Females were no more likely than males to develop interval CRCs (using males as reference: adjusted OR=1.06, 95% CI=0.93–1.20, I=71%, six studies; unadjusted OR=1.00, 95% CI=0.87–1.15, I=80%, nine studies), although some studies individually reported a higher prevalence of interval CRCs in females as compared with males. On pooled analysis, patients with interval CRCs were 1.6 times more likely to have a family history of CRC (variably defined as history of CRC in any first-degree relative or only those occurring at age <50 years) as compared with detected CRCs (adjusted OR=1.64, 95% CI=1.40–1.90, I=0%, two studies; unadjusted OR=1.87, 95% CI=1.41–2.47, I=37%, four studies). Patients with interval CRC were 4.3 times more likely to have been diagnosed with diverticular disease as compared with patients with detected CRCs (adjusted OR=4.25, 95% CI=2.58–7.00, I=96%, four studies; unadjusted OR=4.55, 95% CI=3.04–6.82, I=96%, five studies); site-specific prevalence of interval CRCs in patients with diverticular disease was not available. Likewise, four studies (out of five studies which reported on comorbidities) reported a higher prevalence of comorbidities in patients who developed interval CRC, as compared with patients with detected CRC; on meta-analysis, the adjusted OR for risk of interval CRC in patients with multiple comorbidities using the Charlson comorbidity index was 2.00 (95% CI=1.77–2.27, I=26%, four studies; unable to estimate unadjusted OR). Smoking was not found to be a risk factor for interval CRCs, based on results from two studies. Three (out of four studies which reported this outcome) studies observed that patients who developed interval CRCs, as compared with detected CRCs, were more likely to have undergone a polypectomy on their index colonoscopy. On pooling these data, we observed that as compared with patients with detected CRCs, patients with interval CRC were 1.6 times more likely to have had a polypectomy at their index colonoscopy, although this difference was not statistically significant (adjusted OR=1.57, 95% CI=0.97–2.56, I=94%, four studies; unadjusted OR=2.48, 95% CI=0.95–6.45, I=99%, four studies).
Endoscopy-related Factors.Table 4 reports the endoscopy-related characteristics comparing interval CRCs to detected CRCs. In studies from Canada, surgeons were performing more colonoscopies than gastroenterologists; in contrast, in studies from the United States, gastroenterologists were the most common group of endoscopists.
As compared with patients with detected CRCs, patients with interval CRCs were more likely to have had their index colonoscopy by a non-gastroenterologist (particularly by an internist or family practitioner) than by a gastroenterologist (index colonoscopy by internist or family practitioner: adjusted OR=1.53, 95% CI=1.32–1.77, I=43%, four studies; index colonoscopy by surgeon: adjusted OR=1.15, 95% CI=1.03–1.28, I=25%, four studies; unadjusted OR for index colonoscopy by non-gastroenterologist, 1.19, 95% CI=0.96–1.49, I=88%, six studies; unable to estimate unadjusted OR by specialty; Table 4). One study reported that colonoscopies performed in academic centers had lower rates of interval CRC, although this was not apparent in two other studies; the pooled unadjusted OR for interval CRC in academic medical centers as compared with the non-academic setting was 1.09 (95% CI=0.75–1.57, I=82%; unable to estimate adjusted OR). Using polypectomy rate and procedure completion rates as surrogate measures of quality of endoscopists from claims-based studies, endoscopists with higher rates of interval CRCs had lower polypectomy rates (two studies) and slightly lower procedure completion rate (one study). On meta-analysis comparing endoscopists with the highest quartile of polypectomy rate with the endoscopists with lowest quartile of polypectomy rate, the risk of interval CRC was significantly lower in the former (adjusted OR=0.70, 95% CI=0.63–0.77, I=0%, two studies; unable to estimate unadjusted OR). The studies evaluating effect of procedure volumes on risk of interval CRCs used different cutoffs to define high and low procedure volumes. There was no clear evidence whether high or low procedure volumes affected the prevalence of interval CRCs.
Biology-related Factors. Nishihara and colleagues observed that interval CRCs were more likely to have CpG island methylator phenotype (CIMP; OR=2.19, 95% CI=1.14–4.21) and microsatellite instability (MSI; OR=2.10, 95% CI=1.10–4.02), as compared with non-interval CRCs; somatic BRAF, KRAS, and PIK3CA mutations were not significantly associated with interval CRCs. Other single-center studies, not included in the quantitative synthesis, also observed that interval CRCs were more likely to be CIMP positive and MSI high, as compared with detected CRCs, and less likely to be KRAS mutated.
Outcomes of Interval CRCs
Stage and Grade of CRCs.Table 5 reports the outcomes of patients with interval and detected CRCs. As compared with detected CRCs, interval CRCs were less likely to be diagnosed at an advanced stage (stage III or IV; OR=0.79, 95% CI=0.67–0.94, I=70%, eight studies). There was no significant difference in the grade distribution of interval and detected CRCs; interval CRCs were just as likely as detected CRCs to be well- to moderately differentiated (G1/G2: OR=0.87, 95% CI=0.71–1.06, I=50%, five studies).
Survival of Interval CRCs. In a population-based cohort from Utah, Samadder et al. observed a better prognosis associated with interval CRCs as compared with detected CRCs, with a 37% lower risk of mortality (hazard ratio (HR)=0.63, 95% CI=0.49–0.81). This survival benefit was seen both for early-stage cancer (stage 1: HR=0.77, 95% CI=0.52–1.15) as well as advanced stage cancer (stage 3: HR=0.50, 95% CI=0.29–0.88; stage 4: HR=0.48, 95% CI=0.29–0.80). However, Singh et al. did not find a survival benefit associated with diagnosis of interval CRCs (interval CRC vs. detected CRC: HR=0.99, 95% CI=0.84–1.17) in their population-based cohort in Manitoba, Canada. Likewise, in the Danish nationwide study, Erichsen et al. did not observe any difference in 1-year (mortality rate ratio=0.92, 95% CI=0.82–1.0) and 5-year survival (mortality rate ratio=1.0, 95% CI=0.88–1.20) of interval and detected CRCs. Single-center studies also did not find differences in 5-year survival of patients with interval or detected CRCs.