American Journal of Gastroenterology ISSN 0002-9270C© 2008 by Am. Coll. of Gastroenterology doi: 10.1111/j.1572-0241.2008.01875.xPublished by Blackwell Publishing

CLINICAL REVIEWS

Cochrane Systematic Review of Colorectal CancerScreening Using the Fecal Occult Blood Test (Hemoccult):An Update

Paul Hewitson, B.A. (Hons), M.MSc,1 Paul Glasziou, M.B.B.S., Ph.D., F.A.F.P.H.M., F.R.A.C.G.P.,2

Eila Watson, B.Sc., Ph.D.,3 Bernie Towler, M.B.B.S. M.PH.,4 and Les Irwig, M.B.B.Ch., Ph.D., F.F.P.H.M.51Department of Primary Health Care, 2Centre for Evidence Based Medicine, Department of Primary HealthCare, University of Oxford, Oxford, United Kingdom; 3School of Health and Social Care, Oxford BrookesUniversity, Oxford, United Kingdom; 4Department of Health and Aging Services, Macarthur, Australia; and5Screening and Test Evaluation Program, School of Public Health, University of Sydney, Sydney, Australia

BACKGROUND: Reducing mortality from colorectal cancer (CRC) may be achieved by the introduction ofAND AIMS: population-based screening programs. The aim of the systematic review was to update previous

research to determine whether screening for CRC using the fecal occult blood test (FOBT) reducesCRC mortality and to consider the benefits, harms, and potential consequences of screening.

METHODS: We searched eight electronic databases (Cochrane Library, MEDLINE, EMBASE, CINAHL, PsychINFO,AMED, SIGLE, and HMIC). We identified nine articles describing four randomized controlled trials(RCTs) involving over 320,000 participants with follow-up ranging from 8 to 18 yr. The primaryanalyses used intention to screen and a secondary analysis adjusted for nonattendance. Wecalculated the relative risks and risk differences for each trial, and then overall, using fixed andrandom effects models.

RESULTS: Combined results from the four eligible RCTs indicated that screening had a 16% reduction in therelative risk (RR) of CRC mortality (RR 0.84, 95% confidence interval [CI] 0.78–0.90). There was a15% RR reduction (RR 0.85, 95% CI 0.78–0.92) in CRC mortality for studies that used biennialscreening. When adjusted for screening attendance in the individual studies, there was a 25% RRreduction (RR 0.75, 95% CI 0.66–0.84) for those attending at least one round of screening using theFOBT. There was no difference in all-cause mortality (RR 1.00, 95% CI 0.99–1.02) or all-causemortality excluding CRC (RR 1.01, 95% CI 1.00–1.03).

CONCLUSIONS: The present review includes seven new publications and unpublished data concerning CRC screeningusing FOBT. This review confirms previous research demonstrating that FOBT screening reduces therisk of CRC mortality. The results also indicate that there is no difference in all-cause mortalitybetween the screened and nonscreened populations.

(Am J Gastroenterol 2008;103:1541–1549)

INTRODUCTION

Colorectal cancer (CRC) represents one of the most signif-icant and preventable causes of cancer death today. CRC isthe third most commonly diagnosed cancer and the secondhighest cause of cancer mortality in the United States and theUnited Kingdom (1, 2). The prospect of decreasing mortalityand increasing detection of early-stage CRC has stimulatedcalls for the introduction of mass screening programs. Indeed,population screening for CRC has been strongly advocatedby a number of prominent organizations for a number of years(3–6). However, despite these endorsements, participationrates for CRC screening remain low (7, 8), and there remains

a continuing debate regarding the most effective screeningmodality for people who are at average risk of developingCRC (9–11). One routine CRC screening modality that hasbeen the most intensively and extensively investigated is thefecal occult blood test (FOBT) (12).

It is almost a decade since the Cochrane review of screen-ing for CRC using FOBT was first published (13). A meta-analysis of the included studies in this review found that therewas a 16% reduction in CRC mortality, and further, a 23%reduction in CRC mortality when adjusting for screening at-tendance in the individual studies. However, criticism hasbeen directed at the lengthy delay in updating the findings ofthis review with the recently published evidence, and further,

1541

1542 Hewitson et al.

the failure of this review to include an analysis of all-causemortality between the screened and control populations (14,15). This is especially concerning, given the recent publica-tions suggesting that there was a significant increase in all-cause mortality for the screened population when controllingfor CRC mortality and that FOBT screening may be poten-tially more harmful than beneficial (16, 17). Therefore, themain objective of the present Cochrane review was to iden-tify and evaluate any new available evidence that may revisethe results of the previous review (13). The aims were to de-termine whether screening for CRC using FOBT (guaiac orimmunochemical) reduces CRC mortality and to consider thebenefits, harms, and potential consequences of screening.

METHODS

Updated Literature SearchThe review reports an update of the previously publishedCochrane Collaboration systematic review (13). To identifyappropriate studies published, we conducted a search usingCochrane Library (2005 Version 1), MEDLINE, EMBASE,CINAHL, PsychINFO, AMED, SIGLE, and HMIC electronicdatabases (searches performed using WebSPIRS—full strate-gies used are published elsewhere (18) and are available onrequest). There were no restrictions on the language of thearticles. All searches were conducted from 1989 to February2005 (an updated search was conducted in February 2006).

The studies identified in the searches published before1996 were compared to the existing reference list for the pre-vious Cochrane review (13). The references of all retrievedrelevant studies were searched for additional trials. We alsowrote to the principal investigators of four of the trials (Funen,Goteborg, Minnesota, and Nottingham) to clarify aspects ofthe methods and results and asked for any unpublished datain the areas of quality and trial outcomes. Three of the fourinvestigators replied.

Study SelectionAll randomized controlled trials (RCTs) of repeated screen-ing (e.g. either annually or biennially) for CRC using FOBT(guaiac or immunochemical) were considered for inclusion inthe review. Further investigation following a positive FOBTcould be colonoscopy or sigmoidoscopy and double-contrastbarium enema (DCBE), including the removal of colorectalneoplasms (cancers or adenomas) found at diagnostic inves-tigation. The guaiac test slides may or may not have been re-hydrated. Trial participants must be “volunteers” who agreeto take part in the trial, or individuals or households identifiedfrom general practitioner records or population registers.

Main OutcomesThe primary outcome was CRC mortality. Other process mea-sures assessed included: the sensitivity of the FOBT (guaiacor immunochemical) for CRC, the proportion of those al-located to screening who actually attended screening, CRC

incidence in the screening and control groups, CRC stagingin the screening and control groups, and all-cause mortalityin the screening and control groups. We also investigated thephysical harms of follow-up colonoscopy or sigmoidoscopy(e.g. reported bowel perforations and hemorrhages due tothese procedures). The outcomes not explored in this reviewincluded: the disruption screening causes to people’s lives, thestress and discomfort from testing and follow-up investiga-tions, the anxiety caused by false-positive screening results,and the potential advantages of surgery performed earlier(i.e., for early-stage CRCs).

Data ExtractionTwo reviewers (PH and PG) assembled a comprehensive setof articles published between 1989 and 2006 that met theinclusion criteria. These reviewers independently assessedthe eligibility of each article, with any disagreements re-solved through discussion. Data from the RCTs were inde-pendently extracted onto standardized critical appraisal formsby two reviewers (PH and EW). Abstracted data included thestudy citation, study objectives, study design, method of ran-domization and blinded assessment of mortality, length ofthe study and follow-up, number of participants (includingnumber of withdrawals), participant characteristics, descrip-tion of FOBT regime, characteristics of FOBT (i.e., sensi-tivity, specificity, and positive predictive value [PPV]), com-pliance with screening (both overall and for each screeninground), number of CRC cases, number of CRC deaths, all-cause mortality, number of follow-up diagnostic procedures(i.e., colonoscopies and sigmoidoscopies), compliance withfollow-up diagnostic procedures, and stage of cancer.

The RCTs identified were independently assessed for theirquality by two reviewers (PH and EW) using the criteria rec-ommended by the Cochrane Collaboration (19). A methodto generate the sequence of randomization was regarded asappropriate if it allowed each study participant to have thesame chance of receiving each intervention and the inves-tigators to not predict which treatment was next (e.g., tableof random numbers or computer generated). Disagreementsabout quality were resolved through discussion with PG.

Statistical AnalysisThe primary analysis used the group subjects that were orig-inally randomized to “intention to screen,” whether or notthey were ever screened. Data were entered in Review Man-ager (RevMan) 4.2 (20). RevMan allows the user to enterprotocols as well as complete reviews, including text, char-acteristics of studies, comparison table, and study data, andto perform meta-analysis of the data entered (20, 21). Todetermine the effect of screening on CRC mortality, we es-timated the relative risk (RR) and risk difference for eachtrial, and then overall, using fixed and random effects mod-els and tested for the heterogeneity of effects using the χ2

test in RevMan 4.2. The I2 statistic describes the percentageof the χ2 statistic for heterogeneity across studies that aredue to real variations in effect rather than chance (22). As

Updated Cochrane Systematic Review of CRC Screening Using FOBT 1543

analysis by intention to screen would underestimate any realeffect in those attending screening, as a secondary analysis,we adjusted for screening in individual trials using a previ-ously published method (23). Essentially, this method beginswith the intention-to-treat effect RRRi (the RR reduction)and adjusts this based on the proportion of compliance, Pc,that is, the adjusted RRRi is RRRa = RRRi/Pc. Note that ifthe RRRi is 0 (i.e., no effect), then the RRRa is also 0, thatis, the method does not create spurious effects. The varianceis also adjusted (increased) and, hence, the confidence limitswidened. For the meta-analysis, the RRRa is calculated foreach of the individual trials, and then the effect is combined.

Search ResultsThe search (conducted in February 2005) retrieved a total of4,565 citations. Of these, 1,087 abstracts were duplicates andwere excluded, leaving 3,478 abstracts for further considera-tion. In total, 112 relevant abstracts were reviewed in detail,with 21 articles retrieved for possible inclusion in the review.Eleven publications relating to four RCTs were considered inthe updated review. An updated search of the literature (con-ducted in February 2006) found a further 317 citations, butno articles were relevant for inclusion in the review.

RESULTS

Study CharacteristicsThe four RCTs were reported in 11 published articles (Funen24–26, Goteborg 27, 28, Minnesota 29–31, and Nottingham32–34). Four articles (24, 27, 29, 32) were included in the pre-vious Cochrane Review (18). These trials involved 327,043participants in Denmark, Sweden, the United States, and theUnited Kingdom.

Three trials (Funen, Goteborg, and Nottingham) performedbiennial screening. The three-arm Minnesota trial evaluatedboth annual and biennial screenings. The age of participantsranged from 45 to 80 yr for the trials, except for the Goteborgtrial, where the participants were aged 60–64 yr (see Table 1).The length of follow-up ranged from 11.7 to 18.4 yr, and thenumber of potential screening rounds varied between 2 and 11rounds. Three trials randomly allocated individuals or house-holds identified from general practice records or populationregisters to invitation with screening with Hemoccult (Not-

Table 1. Screening Frequency, Age Range of Participants, Length of Follow-Up, Number of Screening Rounds, Percentage of ParticipantsAttending First and At Least One Round of Screening

Screen Age Range Length of No. of Attending First At LeastStudy Country Frequency (yr) Follow-Up (yr) Screening Rounds Screening (%) One Round (%)

Funen Denmark Biennial 45–75 17 9 66.8 NRGoteborg Sweden Biennial 60–64 15.5 2 63.0 70.0

Annual/Minnesota U.S. Biennial 50–80 18 11 (A) NR 75.0 (A)

6 (B) 78.0 (B)Nottingham U.K. Biennial 45–74 11.7 6 53.4 59.6

A = annual screening; B = biennial screening; NR = not reported.

tingham) or Hemoccult-II (Funen and Goteborg) or to controlgroups. One trial (Minnesota) allocated people to either an-nual or biennial screening (Hemoccult) or to the control grouponly after they had agreed to participate in the trial.

All four trials (Funen, Goteborg, Minnesota, and Notting-ham) used an adequate randomization procedure resulting incomparable study groups. Given that FOBT was completedby the trial participants at home, blinding of the participantsto the intervention was not possible. Mortality analyses wereby “intention to treat” for the Funen, Minnesota, and Not-tingham trials. This is not specifically stated for the Goteborgtrial. Blinded, standardized assessment for mortality was per-formed for all four RCTs. The Funen randomized trials in-cluded deaths from the complications of treatment for CRCin the CRC mortality analyses. This is not specifically statedfor the other trials.

CRC MortalityCombining the four RCTs show that screening results ina statistically significant relative reduction in CRC mortal-ity of 16% (fixed and random effects models: RR 0.84,95% confidence interval [CI] 0.78–0.90). This overall re-sult combined the annual and biennial groups in the Min-nesota trial as there was little heterogeneity among the effects(χ2 test for heterogeneity 1.85, df = 3, P = 0.60, I2 = 0%)(see Fig. 1).

SENSITIVITY ANALYSIS. Combining the trials that usedonly biennial screening (Minnesota, Nottingham, and Gote-borg) demonstrated a 15% (fixed and random effects model:RR 0.85, 95% CI 0.78–0.92) relative reduction in CRC mor-tality and a slightly greater CI (χ2 test for heterogeneity 0.21,df = 3, P = 0.90, I2 = 0%).

REDUCTION IN MORTALITY ADJUSTED FOR ATTEN-DANCE. The percentage of participants in the screeninggroups who completed at least one round of screening rangedfrom 60 to 78% (see Table 1). Compliance with screening washigher for the Minnesota trial than for the European trials.Hemoccult screening continued to be offered to all screeningparticipants in most trials, regardless of previous attendance.In the Funen study, only subjects who participated in the firstround of screening (67%) were invited to subsequent screen-ing rounds, hence the compliance with Hemoccult testing was

1544 Hewitson et al.

Figure 1. Effects of screening with Hemoccult on mortality from CRC (fixed effects model).

very high (91–94%) (18). This may affect the generalizabilityof the findings of the Funen study.

The noncompliance rate for the trials ranged from 33 to46% for the first screening and between 22% and 40% for atleast one round of screening (with the exception of the Funenstudy). When the RR is adjusted for attendance to at least oneround of screening in the RCTs, the overall predicted relativemortality reduction is 25% (RR 0.75, 95% CI 0.66–0.84) forthose who are screened (details of the analysis are availableon request).

Individual Trial ResultsDeaths attributed to CRC have been published for three ofthe four trials (mortality results for the Goteborg trial weresupplied by the authors). The RRRi for CRC mortality varyfrom 13 to 33% (see Table 2).

The Minnesota trial reported a 33% reduction (RR 0.67,95% CI 0.51–0.83) in CRC mortality for the annual screeninggroup and a 21% reduction (RR 0.79, 95% CI 0.62–0.97)in CRC mortality for the biennial screening group at 18 yrof follow-up (30). The Goteborg trial has not yet publishedthe mortality findings (27). However, the unpublished results(Haglind 2005, personal communication) indicate that therewas a 16% reduction (RR 0.84, 95% CI 0.78–0.90) in CRCmortality for biennial screening after 15.5 yr of follow-up.The Nottingham trial reported a 13% reduction (RR 0.87,95% CI 0.78–0.97) in CRC mortality for biennial screeningafter 11 yr of follow-up (34). The Funen trial reported a 16%

Table 2. Number of CRC Deaths, Mortality Incidence Ratio, and Mortality Reduction for the Included Trials

No. of CRC Deaths Incidence RatioMortality

Study Screening Group Control Group Screening Group (py) Control Group (py) Reduction (%)

Funen 363/30,967 431/30,966 0.84/1,000 1.00/1,000 16Goteborg 252/34,144 300/34,146 NR NR 16Minnesota (A) 121/15,570 177/15,394 0.67/1,000 1.00/1,000 33Minnesota (B) 148/15,587 (as above) 0.79/1,000 (as above) 21Nottingham 593/76,466 684/76,384 0.70/1,000 0.81/1,000 13

A = annual screening; B = biennial screening: NR = not reported; py = person years.

reduction (RR 0.84, 95% CI 0.73–0.96) in CRC mortality forannual screening and an 11% reduction (RR 0.89, 95% CI0.78–1.01) in CRC mortality, including deaths attributed tocomplications from treatment, for biennial screening at 17 yrof follow-up (26).

All-Cause MortalityThe total number of deaths from all causes (including CRCmortality) and the incidence ratios are shown in Table 3.

Combining the four RCTs did not show any significantdifference in all-cause mortality between the screening andcontrol groups (fixed effects model: RR 1.00, 95% CI 0.99–1.02; random effects model: RR 1.00, 95% CI 0.99–1.01)(see Fig. 2). There was no important heterogeneity amongtrials (χ2 test for heterogeneity 1.96, df = 3, P = 0.59, I2 =0%).

SENSITIVITY ANALYSIS. Furthermore, when excludingdeaths from CRC from the analysis, no significant change innon-CRC mortality between the screening and control groupswas observed (fixed and random effects models: RR 1.01,95% CI 1.00–1.03, P = 0.17). There was no important het-erogeneity among trials (χ2 test for heterogeneity 1.42, df =3, P = 0.70, I2 = 0%).

Incidence and StageAs population screening results in earlier cancer diagno-sis, we would expect an excess of CRCs detected initially

Updated Cochrane Systematic Review of CRC Screening Using FOBT 1545

Table 3. All-Cause Mortality and Incidence Ratio for the Included Trials

No. of Deaths From All Causes All-Cause Mortality Incidence Ratio

Study Screening Group Control Group Screening Group (py) Control Group (py)

Funen 12,205 12,248 28.3/1,000 28.4/1,000Goteborg 10,591 10,432 NR NRMinnesota 10,449 5,213 340–342/1,000 343/1,000Nottingham 20,421 20,336 24.18/1,000 24.11/1,000

NR = not reported.

in the screening groups. With further follow-up, the decreas-ing number of CRC cases occurring over time between thescreening and control groups was evident in most of the in-cluded trials (see Table 4). The Minnesota (CRC cases: an-nual screening 417, biennial screening 435), Goteborg, andNottingham trials all reported a lower number of CRC casesin comparison to the control groups. This suggests that otherfactors, rather than chance (see 35–37), may have contributedto the reported lower number of CRC cases in the screeninggroups. There is also support for the reduction in the inci-dence of CRC following colonoscopic polypectomy (38–41).The reduced number of CRC cases for the screening groupsmost likely reflects the efficacy of colonoscopic polypectomyin preventing adenomas from developing into CRC.

As would be expected with screening, all four trials re-ported more early-stage CRCs (Dukes A) and fewer late-stageCRCs (Dukes C and D) in the screening groups compared tothe control groups. This favorable shift in CRC staging oc-curred across the trials, although the proportion of cancersthat were actually screen-detected (e.g., excluding intervalcancers, repeat screening, and cancers detected in nonrespon-ders to the screening invitation) was fairly low (23–46% forDukes A reported in two of the included trials) (24, 32) (seeFig. 3).

FOBT AccuracyFor two of the trials (Funen and Nottingham), the Hemoccultslides were not rehydrated, resulting in a low test positivityrate (0.8–3.8%) and a higher PPV for CRC (5.0–18.7%).

Figure 2. Effects of screening with Hemoccult on all-cause mortality (fixed effects model).

The test positivity rate for rehydrated slides (Goteborg andMinnesota) was 1.7–15.4% and the PPV was lower at 0.9–6.1% (see Table 5).

The sensitivity of the Hemoccult test, defined as the pro-portion of all CRCs detected during screening (the sum ofscreen-detected cancers and interval cancers within 1 or 2yr of screening), varied from 55 to 57% for the nonrehy-drated slides (81% for a small sample of participants inthe Minnesota trial) and from 82 to 92% for the rehydratedslides.

Further Investigation RatesThe cumulative risk of being invited for further investiga-tion (either colonoscopy or flexible sigmoidoscopy) follow-ing a positive FOBT was 2.6% (1,977/76,466) for the Not-tingham trial, 5.3% (1,647/30,762) for the Funen trial, and6.4% (2,180/34,144) for the Goteborg trial. The cumulativerisk of being invited for further investigation following a pos-itive FOBT is not reported for the Minnesota trial. However,the number of positive screening tests that were followed byan adequate examination (either colonoscopy or flexible sig-moidoscopy and DCBE) was reported as 8,663 for the annualand 5,170 for the biennial screening groups (31). The Min-nesota trial has reported that 38% of the annual group and28% of the biennial group underwent one or more colono-scopic examinations (29), although this was clarified as 31%for the annual group after 13 yr of follow-up in a later pub-lication that did not include the percentage for the biennialgroup (31).

1546 Hewitson et al.

Table 4. Number of CRC Cases and Incidence Rate of CRC Cases for the Screening and Control Groups

No. of CRC Cases Incidence Rate of CRC Cases

Study Screening Group Control Group Screening Group (py) Control Group (py)

Funen 889/30,967 874/30,966 2.06/1,000 2.02/1,000Goteborg 252/34,144 300/34,164 NR NRMinnesota 852/31,157 507/15,394 32–33/1,000 39/1,000Nottingham 1,268/76,466 1,283/76,384 1.51/1,000 1.53/1,000

NR = not reported.

Around 9 in 10 patients with a positive FOBT participatedin further testing. The main investigation was colonoscopy(three of the trials), with only the Goteborg trial using flex-ible sigmoidoscopy and DCBE. Adverse outcomes were re-ported in detail for three of the trials (Goteborg, Minnesota,and Nottingham). The rate of perforation during colonoscopy(Minnesota and Nottingham) is approximately 1 in 1,400(9/13,720 = 0.0007). In the Goteborg trial, 3 participants(out of 2,108 participants) who received follow-up by flex-ible sigmoidoscopy had perforations of the sigmoid colon(28). Three complications (out of only 190 procedures)were also reported for participants undergoing colonoscopyin the Goteborg trial (two perforations and one bleedingcomplication).

DISCUSSION

The findings of the present updated review are consistentwith the Cochrane review published in 1998 (13). Theseresults provide further support for the use of FOBT as a

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

FunenScrn

FunenCntrl

GotbgScrn

GotbgCntrl

NottScrn

NottCntrl

Minn-AScrn

Minn-BScrn

MinnCntrl

Per

cen

tag

e o

f P

arti

cip

ants

Dukes DDukes CDukes BDukes A

Figure 3. Distribution of CRC by Dukes stage for included studies.

population-screening modality to reduce CRC mortality. Thecombined evidence from the RCTs indicates that screeningwith Hemoccult reduces mortality from CRC by about 16%.After adjusting for nonattendance to screening, the overallrelative reduction in mortality was approximately 25%. Al-though the RCTs varied in the selection and age of theirstudy populations, screening intervals, conditions of Hemoc-cult testing and slide processing, length of follow-up, andattendance for screening, the relative reduction in CRC mor-tality with screening is consistent across the trials.

There was no significant difference in all-cause mortalitybetween the screening and control groups (RR 1.00, 95% CI0.99–1.02, P = 0.56). Some commentators have suggestedthat all-cause mortality, rather than disease-specific mortal-ity, is the more appropriate end point for cancer screeningtrials and that the failure to identify a reduction in all-causemortality questions the efficacy of introducing a population-based program (42–44). However, others have suggestedthat all-cause mortality is an inefficient outcome, given thatthe cancer screening trials are not powered to detect such

Updated Cochrane Systematic Review of CRC Screening Using FOBT 1547

Table 5. Rehydration of Slides, Positivity Rate, Sensitivity, and PPV for CRC and Adenomas

Study Rehydration Positivity Rate (%) Sensitivity (%) PPV (CRC) (%) PPV (Adenoma) (%)

Funen No 0.8–3.8 55.0 5.2–18.7 14.6–38.3Goteborg Yes 1.7–14.3 82.0 NR NR

No 1.9 NR NR NRMinnesota Yes 3.9–15.4 92.2 0.9–6.1 6.0–11.0

No 1.4–5.3 80.8 5.6 NRNottingham No 1.2–2.7 57.2 9.9–17.1 42.8–54.5

NR = not reported.

differences and that the intervention is targeted to a diseasethat causes only a small proportion of overall deaths (45–47).

A recent meta-analysis of three of the included CRC tri-als reported that FOBT had no impact on overall mortalityand a statistically significant increase in deaths in non-CRCcauses for the screened population (16). However, this re-view found no significant difference between screened andnonscreened populations after excluding CRC deaths fromall-cause mortality (RR 1.01, 95% CI 1.00–1.03, P = 0.2). Inthe included RCTs, CRC mortality was approximately 2.8%and 3.3% for the screened and nonscreened populations, re-spectively. A complete analysis of the specific reasons fordeath (e.g., coronary heart disease, stroke, etc.) contributingto the all-cause mortality or the mean life years gained by thescreened versus control populations would help to determineif people who have survived CRC die of other causes several(or more) years later. These comparisons would be valuablefor clarifying the reduction of disease-specific mortality inrelation to all-cause mortality.

The majority of trials reported that the PPV of Hemoc-cult for CRC was fairly low, suggesting that over 80% ofall positive tests were false-positives. Investigation of thesefalse-positive participants may have resulted in some nega-tive psychosocial consequences (e.g., stress and anxiety) anda small chance of significant adverse consequences from thediagnostic test (e.g., risk of bleeding, bowel perforation, oreven death). However, the potentially high false-positive ratefor CRC must also be tempered by the PPV for adenomas (1cm or over). The removal of adenomas may reduce the like-lihood of the development of CRC in the future (38–41), al-though this has not been definitively demonstrated and wouldbe dependent on the size and number of adenomas identifiedat colonoscopy (48). The reduction in mortality coupled withthe potential significant decrease in the incidence of CRC(adenomas identified by screening) should be sufficient tocounter concerns regarding the high false-positive rates ob-served for FOBT screening.

This review was restricted to randomized trials that utilizedrepeat invitations to CRC screening. Another high-quality,nonrandomized study has also demonstrated similar results.The Burgundy study (49, 50) is a large-scale controlled trialusing biennial screening (Hemoccult) in people aged 45–74yr. After six screening rounds (and 11 yr of follow-up), thetrial reported a 16% reduction in CRC mortality and a 23%reduction for people attending at least one round (43). Sev-

eral studies utilizing immunochemical FOBT are currentlyunderway (51–53) and are expected to contribute to identi-fying the most accurate, patient-preferred, and cost-effectivescreening test for use in the general population. The accuracyof other variants of FOBT for CRC screening requires fur-ther investigation, particularly in comparison to the existingtests.

CONCLUSIONS

The estimate of mortality reduction from the updated RCTs isconsistent. FOBT screening is likely to be of some benefit inreducing CRC mortality in particular population groups (e.g.,older adults, given the increased incidence of CRC with age).Assuming a constant reduction in RR, the mortality benefitof screening will be highest in populations at greater risk ofCRC death, while the harmful effects of screening are likelyto be independent of this risk (54). Indeed, increasing screen-ing benefit with increasing population risk of CRC death wasobserved in the screening trials. The included trials also in-dicated that there was a general shift toward identifying CRCin the earlier stages within the screening groups (e.g., Dukesstage A) in comparison to the control groups. However, thisis to be expected, given that screening studies would iden-tify earlier-stage lesions because of lead and length time bi-ases. Other benefits of screening that were not explored indetail include the reduction in CRC incidence through de-tection and removal of colorectal adenomas, and potentially,less invasive treatment of identified early-stage CRC. Someareas of CRC screening also require further research. Contro-versy still surrounds the most effective screening modality forthe detection of CRC in the population (9–11). Clarificationof whether FOBT (either guaiac or immunochemical), flex-ible sigmoidoscopy (alone or in combination with FOBT),or colonoscopy will provide the greatest benefits over thepotential risks is required.

ACKNOWLEDGMENTS

We would like to thank the authors of the Funen and Min-nesota trials for providing data for the review. We would espe-cially like to thank Dr. Lindholm and Professor Haglind forproviding data and comments on the Cochrane review. Wewould also like to thank Cancer Research UK for supportingthis work.

1548 Hewitson et al.

Reprint requests and correspondence: Paul Hewitson, Senior Re-search Officer, Department of Primary Health Care, University ofOxford, Old Road Campus, Headington, Oxford OX3 7LF, UnitedKingdom.

Received April 2, 2007; accepted Fabruary 13, 2008.

REFERENCES

1. U.S. Cancer Statistics Working Group. United StatesCancer Statistics: 2003 Incidence and Mortality. Depart-ment of Health and Social Services, Centers for DiseaseControl and Prevention, and National Cancer Institute,2006.

2. Office of National Statistics. Mortality statistics cause: Re-view of the registrar general on deaths by cause, sex and agein England and Wales 2004. Office for National Statistics,2005.

3. U.S. Preventive Services Task Force. Screening for col-orectal cancer: Recommendations and rationale. Ann In-tern Med 2002;137:129–31.

4. Smith RA, Cokkinides V, Eyre HJ. American Cancer Soci-ety guidelines for the early detection of cancer. CA CancerJ Clin 2006;56:11–25.

5. Winawer S, Fletcher R, Rex D, et al. Colorectal cancerscreening and surveillance: Clinical guidelines and ratio-nale. Update based on new evidence. Gastroenterology2003;124:544–60.

6. Rhodes JM. Colorectal cancer screening in the UK: Jointposition statement by the British Society of Gastroenterol-ogy, the Royal College of Physicians, and the Associa-tion of Coloproctology of Great Britain and Ireland. Gut2000;46:746–8.

7. Vernon SW. Participation in colorectal cancer screening:A review. J Natl Cancer Inst 1997;89:1406–22.

8. Subramanian S, Klosterman M, Amonkar MM, et al. Ad-herence with colorectal cancer screening guidelines: A re-view. Prev Med 2004;38:536–50.

9. Walsh JME, Terdiman JP. Colorectal cancer screening.JAMA 2003;289:1288–96.

10. Pignone M, Rich M, Teutsch SM, et al. Screening for col-orectal cancer in adults at average risk: A Summary of theevidence for the U.S. Preventive Services Task Force. AnnIntern Med 2002;137:132–41.

11. Bond JH. The case for direct colonoscopy screeningfor colorectal cancer. Am J Gastroenterol 2006;101:263–5.

12. Anderson WF, Guyton KZ, Hiatt RA, et al. Colorectal can-cer screening for persons at average risk. J Natl Cancer Inst2002;94:1126–33.

13. Towler B, Irwig L, Glasziou P, et al. Screening forcolorectal cancer using the faecal occult blood test,Hemoccult (archive). Cochrane Database Syst Rev 1998;(2):CD001216. DOI: 10.1002/14651858.CD001216.

14. Ekelund G, Janzon L. Feedback on “Screen-ing for colorectal cancer using the faecal oc-cult blood test, Hemoccult (archive)”. CochraneDatabase Syst Rev 1998;(2):CD001216. DOI:10.1002/14651858.CD001216.

15. Gotzsche PC. A missing outcome. Feedback on “Screen-ing for colorectal cancer using the faecal occultblood test, Hemoccult (archive)”. Cochrane DatabaseSyst Rev 1998;(2):CD001216. DOI: 10.1002/14651858.CD001216.

16. Moayyedi P, Achkar E. Does fecal occult blood testing

really reduce mortality? A reanalysis of systematic reviewdata. Am J Gastroenterol 2006;101:380–4.

17. Moayyedi P. Colorectal cancer screening lacks evidenceof benefit. Cleveland J Clin Med 2007;74:545–54.

18. Hewitson P, Glasziou P, Irwig, et al. Screening for colorec-tal cancer using the faecal occult blood test, Hemoccult.Cochrane Database Syst Rev 2007;(1):CD001216.

19. The Cochrane Collaboration. The Cochrane handbook ofsystematic reviews and interventions 4.2.5 [updated May2005].

20. Review Manager (RevMan) [Computer program]. Version4.2 for Windows. Copenhagen: The Nordic Cochrane Cen-tre, The Cochrane Collaboration, 2003.

21. Cochrane Collaboration’s Information Management Sys-tem. About RevMan. 2007. The Information ManagementSystem. Available at: www.cc-ims.net/. Accessed October9, 2007.

22. Higgins JPT, Thompson SG, Deeks JJ, et al. Measur-ing inconsistency in meta-analyses. BMJ 2003;327:557–60.

23. Glasziou PP. Meta-analysis adjusting for compliance: Theexample of screening for breast cancer. J Clin Epidemiol1992;79:488–94.

24. Kronborg O, Fenger C, Olsen J, et al. Randomised studyof screening for colorectal cancer with faecal-occult-bloodtest. Lancet 1996;348:1467–71.

25. Jorgensen OD, Kronborg O, Fenger C. A randomised studyof screening for colorectal cancer using faecal occult bloodtesting: Results after 13 years and seven biennial screeningrounds. Gut 2002;50:29–32.

26. Kronborg O, Jorgensen OD, Fenger C, et al. Randomizedstudy of biennial screening with a faecal occult blood test:Results after nine screening rounds. Scand J Gastroenterol2004;39:846–51.

27. Kewenter J, Brevinge H, Engaras B, et al. Results of screen-ing, rescreening, and follow-up in a prospective, random-ized study for the detection of colorectal cancer by fecaloccult blood testing. Results for 68308 subjects. Scand JGastroenterol 1994;29:468–73.

28. Kewenter J, Brevinge H. Endoscopic and surgical compli-cations of work-up in screening for colorectal cancer. DisColon Rectum 1996;39:676–80.

29. Mandel JS, Bond JH, Church TR, et al. Reducing mortalityfrom colorectal cancer by screening for fecal occult blood.Minnesota Colon Cancer Control Study. N Engl J Med1993;328:1365–71.

30. Mandel JS, Church TR, Ederer F, et al. Colorectal can-cer mortality: Effectiveness of biennial screening for fecaloccult blood. J Natl Cancer Inst 1999;91:434–7.

31. Mandel JS, Church TR, Bond JH, et al. The effect of fe-cal occult-blood screening on the incidence of colorectalcancer. N Engl J Med 2000;343:1603–7.

32. Hardcastle JD, Chamberlain JO, Robinson MH, et al. Ran-domised controlled trial of faecal-occult-blood screeningfor colorectal cancer. Lancet 1996;348:1472–7.

33. Robinson MH, Hardcastle JD, Moss SM, et al. The risksof screening: Data from the Nottingham randomised con-trolled trial of faecal occult blood screening for colorectalcancer. Gut 1999;45:588–92.

34. Scholefield JH, Moss S, Sufi F, et al. Effect of faecal occultblood screening on mortality from colorectal cancer: Re-sults from a randomised controlled trial. Gut 2002;50:840–4.

35. Lang CA, Ransohoff DF. Fecal occult blood screening forcolorectal cancer. Is mortality reduced by chance selectionfor screening colonoscopy? JAMA 1994;271:1011–3.

36. Church TR, Ederer F, Mandel JS. Fecal occult blood

Updated Cochrane Systematic Review of CRC Screening Using FOBT 1549

screening in the Minnesota study: Sensitivity of the screen-ing test. J Natl Cancer Inst 1997;89:1440–8.

37. Ederer F, Church TR, Mandel JS. Fecal occult bloodscreening in the Minnesota study: Role of chance detectionof lesions. J Natl Cancer Inst 1997;89:1423–8.

38. Winawer SJ, Zauber AG, Ho MA, et al. Prevention of col-orectal cancer by colonoscopic polypectomy. N Engl J Med1993;329:1977–81.

39. Citarda F, Tomaselli G, Capocaccia R, et al. Efficacy instandard clinical practice of colonoscopic polypectomy inreducing colorectal cancer incidence. Gut 2001;48:812–5.

40. Singh H, Turner D, Xue L, et al. Risk of developingcolorectal cancer following a negative examination: Evi-dence for a 10-year interval between colonoscopies. JAMA2006;295:2366–73.

41. Brenner H, Arndt V, Sturmer T, et al. Long-lasting re-duction of risk of colorectal cancer following screeningendoscopy. Br J Cancer 2001;85:972–6.

42. Black C, Haggstrom DA, Welch HG. All-cause mortalityin randomised trials of cancer screening. J Natl Cancer Inst2002;94:167–73.

43. Strabanek P. Breast cancer screening with mammography.Lancet 1993;341:1531.

44. Gotzsche PC, Olsen O. Is screening with mammographyjustifiable? Lancet 2000;355:129–34.

45. Church TR, Ederer F, Mandel JS. Re: All-cause mortalityin randomized trials of cancer screening. J Natl Cancer Inst2002;94:861.

46. Gail MH, Katki HA. Re: All-cause mortality in randomizedtrials of cancer screening. J Natl Cancer Inst 2002;94:862.

47. Tabar L, Duffy SW, Yen M-F, et al. All-cause mortalityamong breast cancer patients in a screening trial: Supportfor breast cancer mortality as an end point. J Med Screen2002;9:159–62.

48. Avidan B, Sonnenberg A, Schnell TG, et al. Newoccurrence and recurrence of neoplasms within 5years of a screening colonoscopy. Am J Gastroenterol2002;97:1524–9.

49. Faivre J, Dancourt V, Lejeune C, et al. Reductionof colorectal cancer mortality by fecal occult bloodscreening in a French controlled study. Gastroenterology2004;126:1674–80.

50. Faivre J, Tazi MA, El Mrini T, et al. Faecal occultblood screening and reduction of colorectal cancer mor-tality: A case-control study. Br J Cancer 1999;79:680–3.

51. Grazzini G, Castiglione G, Ciabattoni C, et al. Colorec-tal cancer screening programme by faecal occult bloodtest in Tuscany: First round results. Eur J Cancer Prev2004;13:19–26.

52. Shi-Rong L, Su-Li T, Zi-Tao W, et al. Application of fe-cal occult blood test in consecutive screening of colorectalcarcinoma for natural population. World Chin J Digestol2004:12:137–9.

53. Nakajima M, Saito H, Soma Y, et al. Prevention of ad-vanced colorectal cancer by screening using the immuno-chemical faecal occult blood test: A case-control study. BrJ Cancer 2003;89:23–8.

54. Glasziou PP, Irwig LM. An evidence based ap-proach to individualising treatment. BMJ 1995;311:1356–9.

CONFLICT OF INTEREST

Guarantor of the article: Paul Hewitson.Specific author contributions: Paul Hewitson conducted theliterature search, assessment of abstracts, critical appraisaland data extraction of identified articles, main analyses, andwriting of the report. Paul Glasziou was an author of the orig-inal review and contributed to the assessment of abstracts,main analyses, and writing of the updated review. BernieTowler was the lead author of the original Cochrane report.Eila Watson contributed to the critical appraisal and data ex-traction of identified articles and comments on the report.Les Irwig was the author of the original review and providedcomments on the updated review.Financial support: Cancer Research UK provided supportto Paul Hewitson and Ella Watson for the study.Potential competing interests: None.