Cancer Causes & Control

, Volume 25, Issue 10, pp 1329–1336

Colonoscopy reduced distal colorectal cancer risk and excess cancer risk associated with family history

Authors

  • Sophie Morois
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
  • Vanessa Cottet
    • Registre Bourguignon des Cancers DigestifsUniversité de Bourgogne
    • INSERM U866Université de Bourgogne
    • CHU DIJONUniversité de Bourgogne
  • Antoine Racine
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
  • Françoise Clavel-Chapelon
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
  • Franck Carbonnel
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
    • Service d’HépatogastroentérologieCentre Hospitalier de Bicêtre
  • Nadia Bastide
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
    • INSERM U1018, Team 9Institut Gustave Roussy, Espace Maurice Tubiana
    • UMRS 1018Université Paris Sud 11
Original paper

DOI: 10.1007/s10552-014-0438-7

Cite this article as:
Morois, S., Cottet, V., Racine, A. et al. Cancer Causes Control (2014) 25: 1329. doi:10.1007/s10552-014-0438-7

Abstract

Purpose

Colonoscopy efficacy at preventing proximal colorectal cancer (CRC) is questioned, and little is known about efficacy in high-risk versus medium-risk populations. We investigated the relationship between colonoscopy screening, family history of colorectal cancer (FHCC), and CRC risk by site.

Methods

Among 92,078 women of the E3N prospective cohort, 692 CRCs have been diagnosed after a median follow-up of 15.4 years. Cox proportional hazard models estimated adjusted hazards ratios according to subsites of cancer and FHCC.

Results

A personal history of colonoscopy (PHC; n = 37,470) was associated with decreased rectal and distal colon cancer risks (hazard ratio (HR) = 0.57; 95 % Confidence Interval (CI) = 0.42–0.78 and HR = 0.37; 95 % CI = 0.26–0.52, respectively), but not proximal colon cancer risk (HR = 0.87; 95 % CI = 0.64–1.18). In women with no prior colonoscopy, those with FHCC had a 80 % higher CRC risk than those without FHCC. In women with previous colonoscopy, CRC risk was similar in women with and without FHCC (p for interaction = 0.04).

Conclusions

Results showed colonoscopy ability to prevent distal cancers, but not proximal cancers in women. Colonoscopy screening also reduced the excess risk of women with FHCC to that of women with no FHCC.

Keywords

Colorectal cancerColonoscopyFamily historyProspective studyScreeningEfficacy

Introduction

Colorectal cancer (CRC) stands among the three leading causes of cancer-related mortality in many industrialized countries, including France [1]. The overall 5-year survival rate is around 55 %, and only 20 % for cancers diagnosed at stage 4 [2]. Randomized controlled trials have demonstrated screening sigmoidoscopy to reduce incidence and mortality of distal CRC [3]. Colonoscopy allows examination of the whole colorectum and is thus considered in France as the optimal screening examination in both high- and average-risk patients. However, some observational studies have casted doubt on its efficacy regarding proximal tumors [46]; other studies described reductions in incidence and mortality for both distal and proximal cancers after a colonoscopy, but with stronger risk reductions for distal than for proximal cancers [79]. The optimal screening strategy is thus still debated in view of the lower cost of sigmoidoscopy [9]. Endoscopic screening could enable to diagnose cancer at an early stage, at least distal colon cancer, but few studies simultaneously considered stage and tumor site [10].

First-degree relatives of an index CRC case are considered at a higher risk than the general population. French academic societies currently recommend that subjects with one first-degree relative with a CRC before the age of 60, or two relatives whatever the age the cancer arose, should undergo a screening colonoscopy after the age of 45, or five years before the cancer arose in the relatives. However, few studies evaluated the efficacy of colonoscopy at preventing CRC in subjects with a family history of colorectal cancer (FHCC) [1113]. In France, systematic cancer screening program by fecal occult blood test (FOBT) in the general population aged 50–74 was launched in 2003 in 21 administrative areas and extended to the whole territory in 2009; full colonoscopy is the reference examination after positive FOBT. Thus, the efficacy of colonoscopy at preventing proximal, distal, and rectal cancers in different settings (gender, FHCC) or at enabling early cancer diagnosis is crucial for an effective screening program.

We investigated (1) colonoscopy efficacy at preventing proximal and distal colorectal tumors in women and (2) whether colonoscopy efficacy varies according to FHCC, in a large prospective cohort of French middle-aged women.

Materials and methods

The E3N cohort study

The E3N prospective cohort was initiated in 1990 to study risk factors for the most frequent cancers in women [14] and includes 98,995 French women aged 40–65 at baseline and covered by the MGEN, the national teachers’ health insurance plan. All women signed an informed consent, in compliance with the rules of the CNIL (French National Commission for Data Protection and Individual Freedom) from which approval was obtained.

Self-administered questionnaires were completed every 24–36 months and provided data on lifestyle, family and personal history of disease, and occurrence of medical events such as colonoscopy or diagnosis of CRC or polyp since the last follow-up questionnaire (questionnaires available in French on http://www.e3n.fr).

Colonoscopy screening and FHCC

Information on colonoscopy screening was self-reported at each questionnaire. In a validation study on a random sample of 480 women who declared at least one colonoscopy, history of colonoscopy was confirmed in 466, thus a positive predictive value of 97 %; a German study in a similar setting reported a positive predictive value of 91 % [15]. We reviewed all colonoscopy reports in a random sample of 143 women; colonoscopy reached the cecum in 117 (82 %), or various sites of the proximal colon but not the cecum in 14; in one woman, it was impossible to explore beyond the sigmoid, and the section attained was not reported in 11; in 11 of the 26 incomplete procedures, a double contrast barium enema or another colonoscopy was performed within two years.

FHCC among first-degree relatives (parents, siblings, and children) was self-reported at the 1990 and 2000 questionnaires.

Cases

Cancer occurrence was self-reported, and a few additional cases were identified from death files (n = 20). Pathology reports were obtained for 97 % of incident cases. We did not include cases for which pathology reports had not been obtained (n = 30). We only considered invasive adenocarcinoma cases. Proximal colon included the cecum, ascending colon, hepatic flexure, and transverse colon; distal colon included the splenic flexure, descending colon, and sigmoid colon; rectum included the rectosigmoid junction and rectum; anal cancers were excluded.

Populations and follow-up

Follow-up started at the date they responded to the 1990 questionnaire. Participants contributed person-years of follow-up until the date of cancer diagnosis, the date of the last completed questionnaire, or June 2008 (date at which the 9th questionnaire was mailed), whichever occurred first.

Among the 98,995 cohort women, we excluded those with no follow-up (n = 2,085), those diagnosed with a cancer other than a basal cell carcinoma before baseline (n = 4,832), and cases with no pathology report (n = 30). Thus, 92,048 women were available for the study.

Statistical analyses

Cox proportional hazards models with age as timescale were used to estimate hazard ratios (HR) and 95 % Confidence Intervals (CI) for CRC.

We considered exposure to colonoscopy as time dependent, since questions were asked at each follow-up questionnaire. Women contributed person-years to the “no-colonoscopy” group from baseline (reply to the 1990 questionnaire) until the date they first replied “yes” to the colonoscopy question, used as a proxy for the exact date of colonoscopy. Then, they contributed person-years to the “colonoscopy” group from this date until the end of follow-up. Colonoscopies reported in the same questionnaire as the cancer were not considered since they could have been performed before the cancer was diagnosed, for cancer diagnosis, or later for short-term surveillance.

We similarly considered a time-dependent variable for the cumulative number of declared colonoscopies (0, no colonoscopy/1, one report/2, colonoscopy reports in two or more questionnaires).

FHCC was considered as ever/never, including reports of at least one first-degree relative with CRC in the 1990 questionnaire, then updating the variable with reports at the 2000 questionnaire.

We adjusted models on potential confounders: physical activity, smoking status (time dependent), FHCC (time dependent), educational level, and body mass index (BMI; time dependent). Data missing for BMI (n = 2,078, i.e., 2.3 %) and smoking status (n = 1,068, i.e., 1.2 %) were replaced by the modal value. In sensitivity analyses, we investigated potential confounding by alcohol-free energy, alcohol, calcium, vitamin D, folate, and fiber intakes in women with available dietary data (73.5 % of the initial study population, n = 469 cancer cases), and by MHT (menopausal hormone therapy) use.

HRs by subsites were estimated using a competing risk method where cases from other sites were censored at the date of diagnosis. We used the Wald χ2 test to compare HRs between sites (colon vs. rectum, proximal vs. distal colon), and between stages at diagnosis (stage 1/2 vs. 3/4). We stratified analyses on year 2002 (which provided two datasets of similar number of cases) to investigate the potential impact of higher colonoscopy quality as well as of CRC mass screening in France, launched in 21 administrative areas in 2003 and nationwide in 2009. We tested for potential interactions between PHC and FHCC, abnormal bowel movements, and BMI.

All tests were two-sided with p value set at 0.05. All analyses were performed using statistical analysis systems (SAS) software, version 9.2 (SAS Institute, Inc, Cary, North Carolina).

Results

During 1,421,800 person-years (median follow-up = 15.4 years, SD = 4.5), incident CRC was diagnosed in 662 women: 422 colon cancers (189 proximal colon and 233 distal colon), 236 rectal cancers, and four of unknown site. Characteristics of participants according to PHC and FHCC at the end of follow-up are listed in Table 1. Women with a PHC had more often a FHCC than those with no PHC. Mean ages at diagnosis according to tumor site and stage at diagnosis are presented in Table 2. Mean age at diagnosis was similar in women with and without a FHCC (62.8 and 63.1 years, respectively), but was higher in women with than without a PHC (65.1 and 62.4, respectively).
Table 1

Selected characteristics of participants, overall and according to family history of colorectal cancer, or personal history of colonoscopy at the end of follow-up: the E3N study

 

Total

Family history of colorectal cancer

Personal history of colonoscopy

No

Yes

No

Yes

Number

92,048

81,206

10,842

54,589

37,459

Number of cases

662

579

83

498

164

Age at baseline (years), mean (SD)

49.3 (6.6)

49.2 (6.6)

50.1 (6.6)

48.8 (6.6)

49.9 (6.6)

Age at diagnosis (years), mean (SD)

64.7 (7.8)

64.5 (7.8)

66.0 (7.4)

63.7 (8.0)

66.2 (7.2)

Baseline characteristicsa

Total physical activity (METs-h/week), n (%)

 ≤26

22,444 (24.4)

19,811 (24.4)

2,633 (24.3)

12,939 (23.7)

9,505 (25.4)

 (26–36)

23,598 (25.6)

20,818 (25.6)

2,780 (25.6)

14,008 (25.7)

9,590 (25.6)

 (36–52)

22,827 (24.8)

20,148 (24.8)

2,679 (24.7)

13,713 (25.1)

9,114 (24.3)

 >52

23,179 (25.2)

20,429 (25.2)

2,750 (25.4)

13,929 (25.5)

9,250 (24.7)

Body mass index (kg/m2), n (%)

 <18.5

3,972 (4.3)

3,566 (4.4)

406 (3.7)

2,353 (4.3)

1,619 (4.3)

 (18.5–22)

40,925 (44.5)

36,156 (44.5)

4,769 (44.0)

23,964 (43.9)

16,961 (45.3)

 (22–25)

31,489 (34.2)

27,784 (34.2)

3,705 (34.2)

18,700 (34.2)

12,789 (34.1)

 (25–30)

12,921 (14.0)

11,290 (13.9)

1,631 (15.0)

7,844 (14.4)

5,077 (13.6)

 ≥30

2,741 (3.0)

2,410 (3.0)

331 (3.1)

17,328 (3.2)

1,013 (2.7)

Smoking status, n (%)

 Never smoker

50,428 (54.8)

44,400 (54.7)

6,028 (55.6)

29,661 (54.3)

20,767 (55.4)

 Past smoker

28,317 (30.8)

24,964 (30.7)

3,353 (30.4)

16,603 (30.4)

11,714 (31.3)

 Current smoker

13,303 (14.4)

11,842 (14.6)

1,461 (13.5)

8,325 (15.3)

4,978 (13.3)

Number of years schooling, n (%)

 <12 years

12,223 (13.3)

10,720 (13.2)

1,503 (13.9)

7,166 (13.1)

5,057 (13.5)

 12–14 years

48,003 (52.1)

42,305 (52.1)

5,698 (52.5)

28,230 (51.7)

19,773 (52.8)

 >14 years

31,822 (35.6)

28,181 (35.7)

3,641 (33.6)

19,193 (35.2)

12,629 (33.7)

Characteristics at the end of follow-up

Personal history of colonoscopy

 No

54,589 (59.3)

51,192 (63.0)

3,397 (31.3)

 Yes

37,459 (40.7)

30,014 (37.0)

7,445 (68.7)

Family history of colorectal cancer

 No

81,206 (88.2)

51,192 (93.8)

30,014 (80.1)

 Yes

10,842 (11.8)

3,397 (6.2)

7,445 (19.9)

MET Metabolic equivalent task

aMissing values: see methods section

Table 2

Ages at diagnosis for cancer cases

 

Colorectal

Proximal colon

Distal colon

Rectum

N cases

Age at diagnosis (mean, SD)

N cases

Age at diagnosis (mean, SD)

N cases

Age at diagnosis (mean, SD)

N cases

Age at diagnosis (mean, SD)

Overall

662

63.1 (7.7)

189

65.0 (7.7)

233

63.0 (7.0)

236

61.7 (7.8)

Stage 1 or 2

331

63.4 (7.7)

89

65.8 (7.6)

113

63.1 (7.2)

125

61.9 (7.9)

Stage 3 or 4

321

62.9 (7.6)

96

64.3 (7.6)

119

62.9 (7.0)

106

61.5 (7.9)

SD Standard deviation

PHC was associated with decreased CRC risk (HR = 0.56; 95 % CI = 0.47–0.68), similarly so for colon and rectal cancers (HR = 0.57; 95 % CI = 0.45–0.71 and HR = 0.57; 95 % CI = 0.42–0.78, respectively; phomogeneity = 0.95; Table 3). The inverse association between colonoscopy and cancer risk was observed in the distal, but not the proximal colon (HR = 0.37; 95 % CI = 0.26–0.52 and HR = 0.87; 95 % CI = 0.64–1.18, respectively, phomogeneity < 0.01). All associations were stronger when women reported a colonoscopy in more than one questionnaire (for CRC: HR = 0.43; 95 % CI = 0.33–0.55; p for linear trend <0.01). Associations with colonoscopy were similar for cancers diagnosed at stage 3/4 (HR = 0.52; 95 % CI = 0.40–0.68) and for those diagnosed at stage 1/2 (HR = 0.59; 95 % CI = 0.46–0.77).
Table 3

Association between colorectal cancer risk and personal history of colonoscopy. E3N cohort, France (n = 92,048)

 

Colorectal cancer

Colon cancers

Rectal cancers

Proximal colon cancers

Distal colon cancers

Person-years

Cases (N)

HRa

95 % CI

HRb

95 % CI

Cases (N)

HRb

95 % CI

Cases (N)

HRb

95 % CI

Cases (N)

HRb

95 % CI

Cases (N)

HRb

95 % CI

Personal history of colonoscopy

 No

996,257

498

1.00

Reference

1.00

Reference

314

1.00

Reference

180

1.00

Reference

124

1.00

Reference

190

1.00

Reference

 Yes

425,543

164

0.58

0.49–0.70

0.56

0.47–0.68

108

0.57

0.45–0.71

56

0.57

0.42–0.78

65

0.87

0.64–1.18

43

0.37

0.26–0.52

 phomogeneity

      

Colon versus rectum: 0.95

Proximal versus distal: <0.01

 No

996,257

498

1.00

Reference

1.00

Reference

314

1.00

Reference

180

1.00

Reference

124

1.00

Reference

190

1.00

Reference

 1 report

234,605

99

0.72

0.58–0.89

0.70

0.57–0.88

59

0.65

0.50–0.87

40

0.81

0.58–1.15

36

1.01

0.70–1.47

23

0.42

0.27–0.65

 >1 report

190,938

65

0.45

0.35–0.58

0.43

0.33–0.55

49

0.48

0.35–0.66

16

0.32

0.19–0.54

29

0.72

0.48–1.10

20

0.33

0.20–0.52

 ptrend

  

<0.01

 

<0.01

  

<0.01

  

<0.01

  

0.18

  

<0.01

 

 phomogeneity

      

Colon versus rectum: 0.51

Proximal versus distal: <0.01

Diagnosis at stage 1 or 2

Personal history of colonoscopy

 No

996,257

244

1.00

Reference

1.00

Reference

145

1.00

Reference

95

1.00

Reference

55

1.00

Reference

90

1.00

Reference

 Yes

425,543

87

0.62

0.49–0.80

0.59

0.46–0.77

57

0.64

0.46–0.87

30

0.56

0.37–0.85

34

0.99

0.64–1.54

23

0.42

0.26–0.67

 phomogeneity

      

Colon versus rectum: 0.62

Proximal versus distal: 0.01

Diagnosis at stage 3 or 4

Personal history of colonoscopy

 No

996,257

248

1.00

Reference

1.00

Reference

167

1.00

Reference

81

1.00

Reference

68

1.00

Reference

99

1.00

Reference

 Yes

425,543

73

0.52

0.40–0.68

0.52

0.40–0.67

48

0.48

0.35–0.67

25

0.60

0.37–0.94

28

0.70

0.45–1.10

20

0.33

0.21–0.55

 phomogeneity

Versus stage 1/2: 0.45

Versus stage 1/2: 0.23

Versus stage 1/2: 0.84

Versus stage 1/2: 0.28

Versus stage 1/2: 0.53

      

Colon versus rectum: 0.47

Proximal versus distal: 0.03

aUnadjusted

bAdjusted on physical activity, smoking status, colorectal cancer in first-degree relatives, educational level, and body mass index

When considering only PHC at least one year before the end of follow-up, results were not substantially modified (HR for CRC = 0.51; 95 % CI = 0.42–0.61). Associations between PHC and cancer risk were similar before and after year 2002 (for CRC: phomogeneity before vs after 2002 = 0.88, data not tabulated).

Additional adjustment on dietary factors (N = 469 cancer cases) or on MHT use did not substantially modify the results. There was no interaction between PHC and abnormal bowel movements or BMI (all pinteraction > 0.10).

Overall, FHCC was not associated with cancer risk (HR = 1.13; 95 % CI = 0.90–1.43) whatever the stage (stage 1/2, HR = 1.33; 95 % CI = 0.98–1.81; stage 3/4, HR = 0.94; 95 % CI = 0.66–1.34) or the site (not tabulated). However, FHCC was associated with an increased CRC risk in women with no previous colonoscopy (HR = 1.84, 95 % CI = 1.38–1.46), but not in those with previous colonoscopy (HR = 1.07, 95 % CI = 0.72–1.56, pinteraction = 0.04; Table 4). Results were similar for all sites. In a model that included all combinations of PHC and FHCC, with no colonoscopy and no FHCC as the reference, HRs for FHCC and no colonoscopy, colonoscopy and no FHCC, and colonoscopy and FHCC were, respectively, 1.61 (1.20–2.14); 0.61 (0.51–0.75); and 0.60 (0.410.86). Thus, the absolute annual incidence rate in women who did not perform a colonoscopy was 57.8/100,000 women with no FHCC and 93.1/100,000 women with FHCC; the incidence rates were 35.3/100,000 in those with previous colonoscopy without FHCC and 34.7/100,000 women in those with previous colonoscopy and FHCC.
Table 4

Association between colorectal cancers risk and family history of colorectal cancer according to personal history of colonoscopy. E3N cohort, France (n = 92,048)

Family history of colorectal cancer

Women with no personal history of colonoscopy

Women with a personal history of colonoscopy

Women with versus without personal history of colonoscopy

Person-years

Cases (N)

HRa

95 % CI

Person-years

Cases (N)

HRa

95 % CI

phomogeneity

Colorectal cancers

 No

771,515

446

1.00

Reference

513,870

133

1.00

Reference

0.04

 Yes

39,232

52

1.84

1.38–2.46

97,183

31

1.07

0.72–1.56

 

Colon cancers

 No

771,515

282

1.00

Reference

513,870

86

1.00

Reference

0.22

 Yes

39,232

32

1.77

1.22–2.55

97,183

22

1.16

0.72–1.85

 

Rectal cancers

 No

771,515

161

1.00

Reference

513,870

47

1.00

Reference

0.10

 Yes

39,232

19

1.92

1.19–3.10

97,183

9

0.90

0.44–1.83

 

Proximal colon cancers

 No

771,515

113

1.00

Reference

513,870

53

1.00

Reference

0.47

 Yes

39,232

11

1.46

0.79–2.72

97,183

12

1.02

0.55–1.91

 

Distal colon cancers

 No

771,515

169

1.00

Reference

513,870

33

1.00

Reference

0.46

 Yes

39,232

21

1.98

1.26–3.12

97,183

10

1.38

0.68–2.79

 

aAdjusted on physical activity, smoking status, colorectal cancer in first-degree relatives, educational level, and body mass index

Discussion

In this large prospective cohort of French middle-aged women, colonoscopy was significantly associated with reduced distal but not proximal colon cancer risk, with a stronger association when reported at least twice during follow-up. FHCC was associated with an increased risk of CRC only in women with no screening colonoscopy; women with previous colonoscopy had a similar cancer risk whether or not they had FHCC. To the best of our knowledge, only two prospective studies previously investigated the relationship between endoscopic screening and CRC risk, and most endoscopies were restricted to the sigmoid [10, 16].

The inverse association between colonoscopy screening and CRC risk agrees with previous studies on CRC incidence [4, 6, 7, 10, 1618] or mortality [5, 16, 19]. However, recent studies suggested that colonoscopy might prevent only distal cancers [4, 5, 20, 21], which led to challenge the cost-benefit balance of colonoscopy as opposed to sigmoidoscopy [22]. In France, colonoscopy has been the reference examination for several decades and is fully reimbursed. Although we also observed reduced distal but not proximal cancer after colonoscopy screening, our results suggest that repeated colonoscopy could also prevent some proximal cancers and that stage at diagnosis of these tumors could be less advanced in people with colonoscopy screening. This has been little investigated so far, and with limited power [10].

Several hypotheses have been proposed to explain the observed differences in colonoscopy efficacy between the distal and proximal colon. These include lesser quality of colon cleansing in the proximal than in the distal colon, and the importance of trained and experienced endoscopists; indeed, the proximal colon may be difficult to reach and explore properly, especially in women [23]. In addition, proximal adenomas are often flat, and difficult to identify and remove [24], even by an experienced endoscopist. However, analyses stratified on year 2002, which could account for progress in endoscopic techniques, led to similar results in the two periods. New techniques such as virtual chromoendoscopy are being developed to improve detection of flat lesions [25]. In addition, tumors of the proximal colon more often develop from serrated adenomas which often harbor-specific features, such as microsatellite instability, BRAF mutations, or high gene methylation as compared with distal tumors. It has been suggested that these features could be associated with fast tumor progression [26]. Therefore, the adenoma–carcinoma process could be faster in proximal than in distal tumors. Indeed, in our study, mean age at diagnosis of late-stage proximal cancers was younger than that of stage 1/2 cancers, which suggests rapid progression of some aggressive tumors that could escape colonoscopic prevention. The development of new biomarkers “with potential to improve screening effectiveness and user-friendliness” [27] is thus of particular interest.

Regarding distal tumors, colonoscopy efficacy was higher for distal colon than for rectal cancers. Only two studies separately considered distal colon and rectal tumors, with either similar associations for distal colon and rectal cancers [7] or stronger associations for rectal than for distal colon cancers [4]. A shorter mean sojourn time in the rectum than in the distal colon has been reported [28], consistently with our findings of a decrease in rectal cancer risk after at least two colonoscopy reports. Thus, the long-term risk reduction conferred by a single colonoscopy [29] may mostly concern distal colon cancers, while prevention of rectal cancers may request regular endoscopic examinations.

Our results confirm the increased risk of CRCs associated with FHCC in first-degree relatives, although only in women with no colonoscopy screening [25]. Those with previous colonoscopy had a similar risk to those with no FHCC, as already reported [7, 30]. These results support current French recommendations for people with FHCC, i.e., that they must undergo colonoscopy screening from the age of 45 or five years before the age CRC occurred in the relative. Adherence to screening colonoscopy in people with FHCC should improve, since even in our educated and health-conscious population over 30 % had not undergone a colonoscopy by the end of follow-up while they were all over age 45 at that time.

Strengths of our study include its prospective design, large population, long follow-up, regularly updated data on screening practices including colonoscopies, and adjustment for potential confounders. Pathology reports were obtained in all CRC cases included in our analyses (97 % of self-reported CRCs). Colonoscopies are always performed by trained gastroenterologists in France, mostly under full sedation, which has been described to provide better results than when performed by other health professionals [18]. Although our study population was composed of educated volunteers, it is of interest that the site distribution was quite similar to that reported for the 1991–2005 period in a French digestive cancer registry [31], i.e., proportions of 31.3, 33.2, and 35.5 % for the proximal and distal colon, and the rectum, respectively, which compares to figures of 28.7, 35.4, and 35.9, respectively, in our study.

Limitations of our study include it being restricted to women; however, investigating the efficacy of colonoscopy for proximal cancers in women is particularly relevant since proximal cancers are more frequent in women than in men [32], and tumors may be technically more difficult to diagnose in women because of pelvic adhesions [23]. Still, results similar to ours have been obtained in male populations [16]. Another limitation is self-report of colonoscopies. However, in the validation study, there was a 97 % positive predictive value, similar to that reported in a German study in a similar setting [15]. We did not validate non-report of colonoscopy, but the above-cited German study reported a 100 % negative predictive value [15], and in our population of educated women, mostly teachers, who were regularly asked to report all screening examinations, forgetting such a demanding and time-consuming examination is unlikely. Moreover, we lacked information on quality and completeness of colonoscopy. Dates of colonoscopies were approximate since they could have happened any time between two questionnaires (i.e., every 2–3 years). This probably resulted in a mean one to one and a half year delay between the true date and the date it was reported. Moreover, when considering number of colonoscopies, we considered only number of questionnaires where a colonoscopy had been reported and may have missed some short-term repeated colonoscopies (e.g., short-term verification after removal of polyps). However, for surveillance purposes, only middle-term colonoscopies are really of importance. Finally, while we did not collect the reasons for polyp-free colonoscopies, we collected colonoscopy reports for validating polyps; in a random sample of those, major reasons for colonoscopies were FHCC or adenomas, recent changes in bowel movements and abdominal pain. Since the mass screening program was launched in France at the end of our study follow-up, the impact and potential bias of colonoscopies for a positive FOBT were likely to be very low at most. Moreover, our study population consisted in women of a higher mean educational level than the general French population, with higher compliance to prevention strategies. Thus, it is likely that the proportion of colonoscopies was higher in our population than in the overall French population, emphasizing the needs for screening education especially in high-risk subjects such as those with FHCC. FHCC in first-degree relatives was also self-reported with potential under or over reporting. However, a case–control study with a face-to-face interview reported FHCC in 11.2 % of general population controls [7], very close to our 12 % figure at the end of follow-up. We lacked information on age at diagnosis of the index cases, and only six CRC cases had more than one family member with a history of CRC, which prevented analyses according to age at diagnosis or number of family index cases. Finally, we lacked information on aspirin and nonsteroidal anti-inflammatory drugs, which have been associated with a lower risk of colorectal tumors [33], and might modulate the association between colonoscopy, FHCC and CRC risk. However, in the Nurses’ Health Study, aspirin offered no additional benefit compared with colonoscopy alone [16].

In conclusion, this prospective study adds to the evidence for the efficacy of colonoscopy at preventing distal cancers and to a lesser extent proximal colon cancers in women; it also demonstrated that colonoscopy screening can reduce the excess risk of women with FHCC to that of women with no FHCC. Further studies that separately investigate the three sites of CRC should thus be encouraged. Our findings are of importance to public health professionals and should encourage improvements of techniques as well as setting up new means of preventing proximal colon cancer. Efforts should also be made to improve recommendation compliance for subjects at high or average risk of CRC.

Acknowledgments

The E3N cohort is being carried out with the financial support of the “Mutuelle Générale de l’Education Nationale” (MGEN); European Community; French League against Cancer (LNCC); Gustave Roussy Institute (IGR); French Institute of Health and Medical Research (Inserm). Sophie Morois was supported by a doctoral grant from the French ministry of research. The authors are indebted to all participants for providing the data used in the E3N Study and to practitioners for providing pathology reports. They are grateful to M. Fangon, L. Hoang, R. Chait and M. Niravong for their technical assistance.

Conflict of interest

None.

Copyright information

© Springer International Publishing Switzerland 2014