Antonio Federici, Paolo Giorgi Rossi, Piero Borgia, Francesco Bartolozzi, Sara Farchi and Gabriella Gausticchi

83 12 ORIGINAL ARTICLE The immunochemical faecal occult blood test leads to higher compliance than the guaiac for colorectal cancer screening programmes: a cluster randomized controlled trial Antonio Federici, Paolo Giorgi Rossi, Piero Borgia, Francesco Bartolozzi, Sara Farchi and Gabriella Gausticchi 1... J Med Screen 2005;12:83 88 See end of article for authors affiliations... Correspondence to: Paolo Giorgi Rossi, Agency for Public Health, Lazio Region, Via di S. Costanza 53, 00198 ; giorgirossi@asplazio.it Accepted for publication 26 January 2005... Introduction: We conducted a cluster-randomized trial aimed at assessing the effect of the type of faecal occult blood, guaiac or immunochemical test on screening compliance. Methods: We sampled 130 general practitioners (GPs) who consented to participate in the trial. We randomly allocated half of them to the guaiac (Hemo-Fec) and half to the immunochemical test (OC- Hemodia). We sampled 2/10 of the GPs 50 75-year-old patients (n ¼ 7332) and randomly divided this population into half. One half was invited to be screened at the GP s office and the other to the nearest gastroenterology ward. The principal outcome was the percentage of returned tests. Results: The immunochemical test had a compliance of 35.8% and the guaiac of 30.4% (relative risk [RR] 1.20; 95% confidence interval [CI] 1.02 1.44). The difference was mostly due to a higher probability of returning the sample: 93.8% and 88.6% for immunochemical and guaiac, respectively (RR 1.06; 95% CI 1.02 1.10). The guaiac test had a higher prevalence of positives (10.3% versus 6.3%, RR 0.603; 95% CI 0.433 0.837). There was a higher variability in the results obtained with the guaiac test compared with the immunochemical (F[1, 12] ¼ 16.25; P ¼ 0.0017). Conclusions: Compliance is more likely with the immunochemical than the guaiac test, independent of the provider. Guaiac tests show a higher variability of the results among centres. The successful implementation of a screening programme requires a period of standardization of the test reading in order to avoid unexpected work overload for colonscopy services. Colorectal cancer (CRC) is the most frequently diagnosed cancer, and the second biggest killer cancer in Europe. 1 About 75% of CRC cases occur in people without known risk factors and the incidence becomes relevant in terms of public health from age 50. 2 CRC screening is an effective measure in reducing cancer mortality. 3 5 The European Union recommends annual screening by Faecal Occult Blood Test (FOBT) 6 for people over 50 years of age. There are two types of FOBT available for primary screening: the guaiac and the immunochemical tests. The first has been used for large population trials to look for evidence of the efficacy of screening programmes in reducing CRC mortality. 3 5 Several recent papers have reported that the immunochemical test is at least as specific and sensitive as the guaiac, and have concluded that the evidence obtained in favour of testing with guaiac can be extended to the immunochemical test without new trials. 7 9 Furthermore, the immunochemical test is specific for human haemoglobin and does not require any dietary restrictions. Although it is more expensive, it is interpreted automatically and requires only one evacuation, instead of the three required by the guaiac test. Due to these advantages, some authors have included the possibility of using immunochemical test for screening, 10 while other authors have not. 11 A screening programme, in order to be effective, needs to reach a large part of the target population (i.e. it must have high compliance). Literature about the determinants of noncompliance provides few commonly accepted conclusions, mostly because the analysed determinants are contextspecific. 11 16 Only a few works have included the test characteristics in the determinants analysed, but this information is one of the most extensively generalizable in the study of compliance. 17 20 The Lazio region conducted a series of pilot studies to maximize the compliance and to plan a CRC screening programme with an evidence-based organization. 21 The experimental phase surveyed general practitioner s (GP s) screening knowledge and practice, and then asked them to participate in a randomized controlled trial. The aim of this trial was to evaluate the effect of the type of test, guaiac versus immunochemical, on compliance. POPULATION AND METHODS The Lazio region has 5.3 million inhabitants and includes the city of Rome; the CRC screening target population (i.e. people aged 50 74) is 1.5 million (Box 1). The design of the trial has been described in detail elsewhere; 21 briefly we report the study characteristics. We selected 13 s to accurately represent gastroenterology units (five university s, two large research s, six local s) and geographic areas (seven in metropolitan Rome, two in the outskirts of Rome, four in www.jmedscreen.com Journal of Medical Screening 2005 Volume 12 Number 2

84 Federici et al. Box 1 Sample size and power of the study With the hypothesis of absence of interaction, the study size was calculated to obtain a power of 90% (with alpha 0.05) to detect an RR of 1.44 for the type of test, with an expected response rate in the lowest group of 18% (data from a previous pilot study), an intracluster correlation of 0.1 and an average cluster size of 55. As a consequence, the minimum detectable RR for the provider factor was 1.17. The resulting study size was 130 clusters and 7150 subjects. 24 The power reduction due to the interaction and due to the consequent stratified analysis allowed us to detect a RR of 1.5 for the type of test and of 1.25 for the provider. small cities and rural areas of the province). We included in the survey all GPs with an office in the 13 districts (out of 50 in Lazio) that have participating s. The letter for study participation was sent in June 2002. The conditions for GP participation were: more than 100 patients in the target population (age 50 74); a PC in the office; willingness to give the list with names and updated addresses of the practice population to the centre coordinating the study. The randomized controlled trial was planned with a fourarmed factorial design: two test providers (the GP and the ) and two types of tests (guaiac Hemo-Fec [Roche Diagnostic, Mannheim, Germany] and immunochemical OC-Hemodia [Eiken, Tokio, Japan, distributed by Alpha Wasserman, Milan, Italy]). The trial outcome was the percentage of compliance: the number of people who picked up and returned the sample container for the test over the total number of invited people. In each of the 13 districts we sampled 10 eligible GPs. The GPs were randomly assigned to the immunochemical or to the guaiac test (5/5). We sampled 2/10 of the target population of each GP; about 1/10 of the population was randomized to the GP arm and 1/10 to the arm (Figure 1). We decided to assign 1/10 of the target population to be invited over a one-month period in order to simulate the GP s routine workload for a yearly screening. We analysed the lists of randomized patients to ensure that cohabitants were assigned to the same arm: the second member on the list with the same telephone number if assigned to a different arm was rejected and substituted. The coordinating centre contacted the two randomized populations by mail: the letter sent to the GP arm group was signed by the GP and directed the patient to the GP s office for testing instructions. Another letter was sent to the arm group, also signed by the GP, that directed the patient to the gastroenterology centre of the for screening. An economic incentive of about h1000 was paid to the participating GPs, plus h10 for each compliant patient. Each patient received the following test instructions: for the guaiac test (Hemo-Fec) three different evacuations are needed after abstaining from meat and anticoagulant drugs for three days; the immunochemical test (OC-Hemodia) requires a single evacuation and has no dietary restrictions except the abstention from anticoagulant use. The guaiac test has a paper sampler and container and is interpreted with visual colorimetric methods; the immunochemical test has a plastic sampler and container and is interpreted with an automatic optical sensor (OC-Sensor). All tests were analysed at the gastroenterology centre of the district, where we installed the optical reading machines for the immunochemical test. Patients who were positive for the FOBT were referred immediately, by the GP or by the personnel, for a colonoscopy at the gastroenterology centre. All participants in the study were informed about the aims, methods, and use of the tests. Both s and GPs were equipped with EpiData 2.1 software for data entry. Statistical methods The analysis of the effects of provider and type of test had a factorial design. To determine if a stratified analysis was necessary, the interaction between provider and type of test was estimated with a log-likelihood ratio test between the logistic regression models (Stata 7 command: logistic; 22 this analysis was performed not taking into account the effects of practice clustering for increased power). To quantify the effect of the test, we adopted a logistic regression model taking into account the 13 districts and Eligible GPs 1194 Participation to trial 902 292 Sampling 130 Sampling of target population 2/10 = 3604 65 GP randomization Guaiaco vs. Immuno. 64 1 withdrawal SSampling of target population 2/10 = 3716 Guaiaco 1797 Individual randomization GP vs. Guaiac MMG 1807 Immuno Individual randomization GP vs. Immuno MMG Figure 1 Flow chart of the trial randomization. Only one GP was excluded after randomization. Given the definition of the outcome (i.e. proportion people returning the FOBT on the total sampled population), there is no exclusion after randomization at the individual level. Journal of Medical Screening 2005 Volume 12 Number 2 www.jmedscreen.com

Type of FOBT and screening compliance 85 the 130 GPs used in the randomization. We also included the variable provider in order to adjust the balancing among arms (Stata 7 command: xi:svylogit). 22 The same models were used for positive outcomes and inadequate samples. All odds ratios given by logistic regressions were converted into relative risks (RRs) using the formula proposed by Zhang and Yu. 23 The association between prevalence of positives from the guaiac and immuno by centre was tested performing a linear regression using the logarithm of the proportions (Stata 7 command: regress). RESULTS Compliance to the study and randomization Out of the 1194 GPs invited, only 292 expressed interest in participating in the study (24.5% range among districts, 12.3 39%). We randomly sampled 130 compliant GPs, 10 from each district. These GPs were randomized to the two test types, with five GPs in one arm and five in the other for each district. One GP, in the guaiac arm, fell ill during the study and it was impossible to substitute or to re-contact him. The sampling of 2/10 of the practice target population meant 3604 patients were assigned to the guaiac arm and 3716 to the immunochemical (Figure 1). Table 1 shows that there are no significant differences by gender, age, residence and assigned provider. The GPs involved in the trial did not differ from the eligible GP population in the 13 health districts; the available variables for the target population do not show any significant differences compared with the total target population of the 13 districts. Figure 1 shows the number of subjects in each of the four study arms. The effect of the type of test Given the factorial design, we had to check for the presence of interaction before choosing the strategy for the analysis. The interaction detected was small: RR for the effect of the immunochemical versus the guaiac test 1.22 (95% confidence interval [CI] 0.94 1.58) in the arm, and 1.18 (95% CI 1.03 1.37) in the GPs arm (Wald s test: F[1, 115] ¼ 0.32; P ¼ 0.53). In the absence of interaction, we proceeded analysing the two factors as independent. The type of test influences compliance: the immunochemical test had a compliance of 35.8% and the guaiac had a compliance of 30.4%. RR, adjusted for provider, was 1.20 (95% CI 1.02 1.44), the intra-cluster correlation was 0.056 and the design effect 4.1. The difference was completely due to a higher probability of returning the test and not due to picking it up: 93.8% for immunochemical and 88.6% for guaiac of returned tests (RR 1.06, 95% CI 1.02 1.10; Figure 2), the intracluster correlation was 0.052 and the design effect 3.9. The GPs had higher compliance rate than s: 50.2% versus 16.2%, RR 3.40 (95% CI 3.13 3.70). Table1 Results of the randomization Percent of target population Guaiac Immunochemical Statistical test P n ¼ 7320 3604 3716 Gender 46.9% men 45.8% men w 2 (1) ¼ 0.90 0.34 Age (years) 50 59 43.8% 42.7% 60 69 39.0% 41.3% w 2 (2) ¼ 4.51 0.1 70+ 17.2% 16.0% w 2 (linear trend) ¼ 0.003 0.96 Residence Rome 53.3% 48.5% Other 46.7% 51.5% t ¼ 0.28 0.78 Provider Hospital 49.9% 50.0% GPs 50.1% 50.0% w 2 (1) ¼ 0.01 0.92 Target population assigned to the tests by gender, age, residence and test provider assigned Guaiac Immunochemical Hospital 1797 GP 1807 Hospital GP 1498 870 1508 786 299 (16.6%) 937 (51.9%) 350 (18.8%) 1072 (57.7%) 34 107 20 68 265 831 332 1009 (14.7%) (46.0%) (17.9%) (54.3%) 30.4% Compliance 35.8% RR of return given the invitation immuno vs Guaiaco (adj. for provider) RR of return given the pick up immuno vs Guaiaco (adj. for provider) = 1.20 (95%CI 1.02-1.44) = 1.06 (95%CI 1.02-1.10) Figure 2 Summary results of the trial. The flow chart represents the steps of invitation, picking up and returning of the test in the four arms of the trial. The RRs are calculated with a logistic model adjusting for provider (the odds ratios have been converted using the formula given by Zhang J and Yu KF). 23 www.jmedscreen.com Journal of Medical Screening 2005 Volume 12 Number 2

86 Federici et al. Secondary outcomes Difference in mean prevalence The guaiac test had a higher prevalence of positives than the immunochemical test (10.3% versus 6.3%, RR 0.603; 95% CI 0.433 0.837). The higher prevalence was mostly due to the results from three centres that reported a prevalence of positive guaiac tests of 32%, 27% and 26%, as shown in Figure 3; the mean prevalence in the other ten centres was 5.4%. There was no difference in mean prevalence between the two providers (10.2% versus 11.5% in and GPs, respectively, for the guaiac and 5.5% versus 6.6% for the immunochemical). Variability among centres Figure 3 shows the higher variability of the results obtained with the guaiac test than with the immunochemical (F[1, 12] ¼ 16.25; P ¼ 0.0017). The unexpected high values observed in centres A, E and I were either due to unspecific interpretation of the colorimetric test, common to s and GPs, or due to incorrect dietary recommendations, specific to each provider. Table 2 shows the prevalence of positives with guaiac test by provider: for districts A and E there was no difference between provider, for the district I there was higher prevalence in the GPs arm. In general, there was strong homogeneity among GPs, taking into account the effect of the district (random effect model, Prevalence of positive tests (%) 0.35 0.3 0.25 0.2 0.15 0.1 0.05 immuno guaiac 0 A B C D E F G H I L M N O Centre Figure 3 Prevalence of positive tests (on the y-axis) by centre (on the x-axis), and type of test. Diamonds for immunochemical and squares for guaiac tests. likelihood ratio test of rho ¼ 0: w 2 (1) ¼ 0.00, P>w 2 ¼ 0.9978). There was no association between the prevalence observed with the guaiac and the immunochemical tests by district: the regression of natural logarithm of the two proportions gives a non-significant negative coefficient ( 1.317464; 95% CI 2.897206 to 2.622769; F[1, 11] ¼ 3.37; P ¼ 0.0934). The proportion of inadequate samples The guaiac had a higher proportion of inadequate samples than the immunochemical test: 2.1% versus 1.1%. This difference that is not significant taking into account the variability among GPs and districts (RR 1.91; 95% CI 0.80 4.71) Positive predictive value Secondary diagnostic level (colonoscopy) had a compliance of 70.1% with no significant difference between centres. The number of cancers and high-grade adenomas detected (17 versus 15) by the two tests were similar, but the positive predictive value was slightly higher for the immunochemical test (29% versus 20%), although not significantly (P ¼ 0.2; Table 3). DISCUSSION We observed a 20% higher compliance when the immunochemical test rather than the guaiac test was proposed. The effect did not depend on the provider, giving a similar ratio Table 3 Results of referral rate and results of the colonoscopies performed, by type of test Immunochemical test Guaiac test n % n % Tests performed 1301 1064 Referred to colonoscopy 82 6.3 110 10.3 Performed colonoscopies 58 70.7 76 69.1 Cancer and high-grade 17 29.3 15 19.7 adenoma Low-grade adenoma 9 15.5 9 11.8 Negative 32 55.2 52 68.4 Table 2 Prevalence of guaiac positive tests by provider and centre Hospital GPs Random effect model Centre Test Positive % Test Positive % Z P A 28 10 35.7 84 26 31.0 0.467 0.641 B 10 0 0.0 47 2 4.3 n.a. C 6 0 0.0 55 4 7.3 n.a. D 34 2 5.9 98 9 9.2 1.025 0.306 E 18 5 27.8 40 10 25.0 0.223 0.823 F 35 0 0.0 74 1 1.4 n.a. G 18 1 5.6 37 3 8.1 0.321 0.748 H 22 2 9.1 62 5 8.1 0.289 0.772 I 19 1 5.3 36 14 38.9 2.252 0.024 L 33 1 3.0 87 2 2.3 0.228 0.819 M 16 6 37.5 56 2 3.6 3143 0.002 N 12 2 16.7 42 1 2.4 1.651 0.099 O 9 0 0.0 64 1 1.6 n.a. Total 260 30 11.5 782 80 10.2 0.43 0.667 The model tests the probability that the prevalence of positives will not differ between the two providers, taking into account the effect of each GP. n.a., not applicable: the test is not reliable when there are no positives in one of the two populations Journal of Medical Screening 2005 Volume 12 Number 2 www.jmedscreen.com

Type of FOBT and screening compliance 87 even at very different levels of baseline compliance. The difference was entirely due to a lower level of test returns. This finding is consistent with the plausibility of the hypothesis: the guaiac test discourages compliance because of the required dietary restrictions, the necessity of collecting three evacuations and the disgust in handling the faeces in the paper sampler. 16 There is solid evidence that screening with FOBT reduces CRC mortality; most government agencies recommend it as the best solution or as a good alternative to colonoscopy and flexosigmoidoscopy. 2 In Europe, current legislation recommends annual FOBT testing as CRC screening strategy in 50 74-year-old people without any specific risk factors. 6 The proof of efficacy, in an experimental setting, must be confirmed by the assessment of effectiveness in a public health setting. 25 A major concern for the reduction in effectiveness of screening programmes is low levels of compliance. 25,26 The characteristics of the available tests for CRC screening, in the opinion of several authors, can reduce compliance, and the choice of the primary screening test must take into consideration the population compliance to dietary restrictions. 10 Our study design allowed us to demonstrate that the type of FOBT determines compliance, and consequently this information should be taken into account in the planning of screening programmes. One of the characteristics required for a screening test is acceptability in the population; 27 in this trial we compared the acceptability of two tests in terms of individual preference. 28 Our results are different from those of Ko and collaborators, 29 obtained in an observational study. This could be related to contextual differences; in fact, a behavioural factor like compliance is strongly influenced by the culture of the community as well as study design. In an observational study rather than a randomized trial, a 20% difference could be masked by stronger determinants such as the GP or patient characteristics. Other factors may be more effective in increasing compliance, such as GP involvement, which was tested simultaneously in this trial (RR 3.4), but usually these are factors that cannot be easily applied to other situations (i.e. they are peculiar to health service or cultural characteristics). We considered the measurable test characteristics relevant for the implementation of a screening programme as secondary outcomes: the prevalence of positive tests, the variability among centres, the prevalence of inadequate samples, and the positive predictive value. The only secondary outcome that was significantly different between FOBTs was the higher prevalence of positive tests using guaiac. Generally, the immunochemical test performed better in our study, identifying approximately the same number of high-grade adenomas and carcinomas (i.e. higher positive predictive value) and giving a lower rate of inadequate samples. Even though these results were not statistically significant, they are strongly suggestive of easier applicability of the immunochemical test in a screening programme. 30,31 The major concern emerging from the results of the guaiac test is the high variability of the prevalence of positive tests between districts, not correlated with the results of the immunochemical test. This variability is higher than previously observed in large population trials, 3 5 although the phenomenon is well known: if we look at the review by Young, 10 the prevalence of positives for the immunochemical test ranges from 3.2% 7.5%, while in the same populations the guaiac test ranges from 2.0% 20.8%. Our setting is very close to the normal conditions in which public health services operate. We tried to understand the reasons for this variance and, given our study design, it may be a problem linked to the provider or to the interpretation of the test. In one of the three outlying districts, the abnormal prevalence of positives was present only for the GP provider. Since the interpretation was centralized, the problem must be related to the provider himself (i.e. incorrect dietary recommendations or incorrect storage of the samples). In the other two centres, the high prevalence was found for both providers, GPs and, and could be related to the interpretation of the tests 30 or to storage problems at the gastroenterology unit. In our setting, the GPs showed scarce interest in the trial only 24.5% agreed to participate. This could have been a problem if we had intended to involve the GPs actively in the organization of the screening programme. The low degree of participation of the GPs contacted suggests that we worked with a group that does not represent the whole GP population, which could lead to questions regarding the external validity of the compliance obtained. Nevertheless, the effect of the test on compliance does not depend on the provider: the immunochemical test increases compliance by 20% both in situations with overall compliance of 17% and in situations with overall compliance of 50% (i.e. the and the GP arm, respectively). The overall compliance obtained probably overestimates the compliance in the general population, but the effect of the type of test is independent from the level of compliance. A second problem is the short observation time. We tried to reproduce the average workload per time unit (month). Using this design we probably magnify the effect of the initial set-up problems. We compromised between the information needed and the resources available. Furthermore, this study design does not permit compliance to subsequent rounds of screening to be observed. An initial set-up problem that we probably overestimated is the variability of the prevalence of positive results, although test interpretation was centralized in specialized centres. A training phase for accurate interpretation of the guaiac FOBT must be planned in the implementation of CRCS programmes; without it, predicting the number of colonoscopies needed will be impossible. CONCLUSIONS The immunochemical test guarantees a higher compliance to CRC screening than guaiac, and this difference is independent of the test provider. The guaiac test showed a higher variability between centres of the prevalence of positive tests and, although this result could be due to an initial set-up problem, this finding must be taken into account in the planning of CRC screening programmes. Further cost benefit research is needed to evaluate whether the advantages of the immunochemical test in terms of compliance and reliability outweigh its higher costs.... Authors affiliations Dr Antonio Federici, Agency for Public Health, Lazio Region, Dr Paolo Giorgi Rossi, Agency for Public Health, Lazio Region, Dr Piero Borgia, Agency for Public Health, Lazio Region, Rome, Italy Dr Francesco Bartolozzi, Campus Biomedico, University Hospital, www.jmedscreen.com Journal of Medical Screening 2005 Volume 12 Number 2

88 Federici et al. Sara Farchi, Agency for Public Health, Lazio Region, Dr Gabriella Gausticchi, Agency for Public Health, Lazio Region, REFERENCES 1 The European Group for Colorectal Cancer Screening. Recommendation to include colorectal cancer screening in public health policy. J Med Screen 1999;6:80 1 2 American Gastroenterological Association. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112:594 642 3 Hardcastle JD, Chamberlain JO, Robinson MHE, et al. Randomised controlled trial of faecal occult blood screening for colorectal cancer. Lancet 1996;348:1472 7 4 Kronborg O, Fenger C, Olsen J, et al. Randomised study of screening for colorectal cancer with faecal occult blood test. Lancet 1996;348: 1467 71 5 Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for rectal occult blood. N Engl J Med 1993;328:1365 71 6 European Union Council. Recommendation of 2 December 2003 on Cancer Screening, 2003/878/EC 7 Allison JE, Tekawa IS, Ransom LJ, et al. A comparison of fecal occultblood tests for colorectal cancer screening. N Engl J Med 1996;334: 155 9 8 Greenberg PD, Bertario L, Gnauck R, et al. A prospective multicenter evaluation of new fecal occult blood tests in patients undergoing colonoscopy. Am J Gastroenterol 2000;95:1331 8 9 Miyoshi H, Oka M, Sugi K, et al. Accuracy of detection of colorectal neoplasia using an immunochemical occult blood test in symptomatic referred patients: comparison of retrospective and prospective studies. Intern Med 2000;39:701 6 10 Young GP, St John DJ, Winawer SJ, et al. Choice of fecal occult blood tests for colorectal cancer screening: recommendations based on performance characteristics in population studies. A WHO (World Health Organisation) and OMED (World Organization dor Digestive Endoscopy) Report. Am J Gastroenterol 2002;97:2499 507 11 Immunochemical versus guaiac fecal occult blood tests. Technol Eval Cent Asses Program Exec Summ 2004;19:1 3 12 Blalock SJ, DeVellis BM, Sandler RS. Participation in fecal occult blood screening: a critical review. Prev Med 1987;16:9 18 13 Box V, Nichols S, Lallemand RC, et al. Haemoccult compliance rates and reasons for non-compliance. Public Health 1984;98:16 25 14 Farrands PA, Hardcastle JD, Chamberlain J, et al. Factors affecting compliance with screening for colorectal cancer. Community Med 1984;6:12 9 15 Jepson R, Clegg A, Forbes C, et al. The determinants of screening uptake and interventions for increasing uptake: a systematic review. Health Technol Assess 2000;4:1 33 16 Lindholm E, Berglund B, Haglind E, et al. Factors associated with participation in screening for colorectal cancer with faecal occult blood testing. Scand J Gastroenterol 1995;30:171 6 17 Cole SR, Young GP. Effect of dietary restriction on participation in faecal occult blood test screening for colorectal cancer. Med J Aust 2001;175:195 8 18 Cole SR, Young GP, Esterman A, et al. A randomised trial of the impact of new faecal haemoglobin test technologies on population participation in screening for colorectal cancer. J Med Screen 2003;10:117 22 19 King J, Fairbrother G, Thompson C, et al. Colorectal cancer screening: optimal compliance with postal faecal occult blood test. Aust N Z J Surg 1992;62:714 9 20 Pignone M, Campbell MK, Carr C, et al. Meta-analysis of dietary restriction during fecal occult blood testing. Eff Clin Pract 2001;4:150 6 21 Rossi GP, Federici A, Bartolozzi F, et al. Trying to improve the compliance to colorectal cancer screening: a complex study design for a complex planning question. Control Clin Trials doi: 10.1016/j.cct.2005.01.005 22 Stata Corporation. Stata Statistical Software: Release 7.0. College Station, TX: Stata Corporation, 2001 23 Zhang J, Yu KF. What s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690 1 24 Hsieh FY. Sample size formulae for intervention studies with the cluster as unit of randomization. Stat Med 1988;7:1195 201 (Erratum in Stat Med 1997;16:1300) 25 Steele RJC, Parker R, Patnick J, et al. A demonstration pilot trial for colorectal cancer screening in the United Kingdom: a new concept in the introduction of healthcare strategies. J Med Screen 2001;8:197 203 26 Vernon SW. Participation in colorectal cancer screening: a review. J Natl Cancer Inst 1997;89:1406 22 27 Morrison AS. Screening in Chronic Disease. 2nd edn. New York, NY: Oxford Press, 1992 28 Hynam KA, Hart AR, Gay SP, et al. Screening for colorectal cancer: reasons for refusal of faecal occult blood testing in a general practice in England. J Epid Comm Health 1995;49:84 6 29 Ko CW, Dominitz JA, Nguyen TD. Fecal occult blood testing in a general medical clinic: comparison between guaiac-based and immunochemicalbased tests. Am J Med 2003;115:111 4 30 Castiglione G, Sala P, Ciatto S, et al. Comparative analysis of results of guaiac and immunochemical tests for faecal occult blood in colorectal cancer screening in two oncological institutions. Eur J Cancer Prev 1994;3:399 405 31 Nakama H, Zang B, Abdul Fattah ASM, et al. Does stool collection method affect outcomes in immunochemical fecal occult blood testing? Dis Colon Rectum 2001;44:871 5 Journal of Medical Screening 2005 Volume 12 Number 2 www.jmedscreen.com