Objective Most colorectal cancer (CRC) screening programmes are nowadays based on faecal immunochemical testing (FIT). Eligible subjects often use oral anticoagulants (OACs) or non-steroidal anti-inflammatory drugs (NSAIDs), which could possibly stimulate bleeding from both benign and premalignant lesions in the colon. The aim of this meta-analysis was to study the effect of OACs and NSAIDs use on FIT performance.
Design A systematic search was conducted until June 2017 to retrieve studies from PubMed, Embase, MEDLINE, Web of science, Cochrane Central and Google Scholar. Studies were included when reporting on FIT results in users versus non-users of OACs and/or NSAIDs in average risk CRC screening populations. Primary outcome was positive predictive value for advanced neoplasia (PPVAN) of FIT in relation to OACs/NSAIDs use. Values were obtained by conducting random-effect forest plots.
Results Our literature search identified 2022 records, of which 8 studies were included. A total of 3563 participants with a positive FIT were included. Use of OACs was associated with a PPVAN of 37.6% (95% CI 33.9 to 41.4) compared with 40.3% (95% CI 38.5 to 42.1) for non-users (p=0.75). Pooled PPVAN in aspirin/NSAID users was 38.2% (95% CI 33.8 to 42.9) compared with 39.4% (95% CI 37.5 to 41.3) for non-users (p=0.59).
Conclusion FIT accuracy is not affected by OACs and aspirin/NSAIDs use. Based on the current literature, withdrawal of OACs or NSAIDs before FIT screening is not recommended. Future studies should focus on duration of use, dosage and classes of drugs in association with accuracy of FIT to conduct more specific guideline recommendations.
- colorectal cancer screening
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Previous studies and a meta-analysis of guaiac-based faecal occult blood test screening showed inconclusive data on the accuracy of faecal blood tests in oral anticoagulant (OAC) and non-steroidal anti-inflammatory drug (NSAID)/aspirin users. Some studies suggest an increased accuracy of faecal blood tests by stimulating neoplastic lesions to bleed, while others suggest a decreased accuracy by stimulating benign lesions to bleed.
What are the new findings?
This is the first meta-analysis of faecal immunochemical testing (FIT) in OAC and NSAID/aspirin users. Our results show that the use of OACs and aspirin/NSAIDs do not affect the positive predictive value for advanced neoplasia (PPVAN) in FIT colorectal cancer screening.
The PPVAN of pooled OAC users was 37.6%, and the PPVAN of non-users was 40.3%. For aspirin/NSAID users, the PPVAN was 38.2%, whereas PPVAN of non-users was 39.4%.
How might it impact on clinical practice in the foreseeable future?
Screening guidelines do not provide recommendations on the use of OACs or aspirin/NSAIDs at time of FIT sampling. Given the significant proportion of subjects using these drugs within colorectal cancer screening and the renewing scientific evidence on this topic, guideline recommendation can be made.
Worldwide, most colorectal cancer (CRC) screening programmes are now based on faecal immunochemical testing (FIT).1 In the European Union, FIT-based CRC screening programmes have an average FIT positivity rate (PR) around 6.2% and a positive predictive value for advanced neoplasia (PPVAN) between 35% and 55% and are thereby more accurate than those for older, guaiac-based faecal occult blood tests (gFOBT).2–5 PPVAN of FIT depends on gender, FIT cut-off and participation in previous screening rounds. It is affected by false-positive results from bleeding sources other than colorectal neoplasia.6 7 Several studies suggest the use of oral anticoagulants (OACs) or non-steroidal anti-inflammatory drugs (NSAIDs) as a possible contributor to the false-PR of faecal blood tests. These studies hypothesise that OACs/NSAIDs could stimulate other, benign lesions to bleed and thereby decrease PPVAN.8–10 In contrast, these drugs may in theory also increase the tendency of neoplastic lesions to bleed and thus increase PPVAN.11 12 Results of a previous meta-analysis and systematic review were inconclusive.13 14 However, most studies at that time were performed with gFOBT and not with the currently practised FIT.1
Until today, clinicians lack clear recommendations. This is remarkable given the widespread use of CRC screening tests and the frequent use of OACs and NSAIDs in the target population of subjects aged 50 years and above.15 16 Moreover, discontinuation of anticoagulant therapy is not without risk in terms of (re)occurrence of cardiovascular events, and discontinuation should thus be considered with care.17
Therefore, this meta-analysis aimed to evaluate the PPVAN and positive predictive value for CRC (PPVCRC) in OACs and NSAIDs users compared with non-users in an average risk FIT-based CRC screening population. Second, we assessed PRs, sensitivity/specificity and negative predictive values (NPVs) when possible. Subgroup analyses were performed with respect to patient and drug characteristics when possible.
We conducted a systematic review and meta-analysis of published trials and abstracts following the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses.18 Additionally, the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) checklist was used, containing specifications for the reporting of a meta-analysis of observational studies in epidemiology.19
In collaboration with the Medical School Library of the Erasmus University in Rotterdam, the Netherlands, a systematic search was conducted until June 2017 to retrieve studies that reported on FIT performance in OACs or NSAIDs users versus controls. PubMed, Embase, MEDLINE, Web of science, Cochrane Central and Google Scholar were used as potential sources. The search was conducted using controlled vocabulary supplemented with key words (online supplementary S1). First, two independent reviewers (SAVN and FERV) screened the selected studies by title and abstract. Studies were excluded if they did not correspond with the inclusion and/or exclusion criteria that are stated below. Furthermore, full text of the selected publications were examined by the same authors. Discrepancies were discussed with a third party (MCWS). References of the retrieved studies were manually searched to locate any additional studies.
Supplementary file 1
Studies were included if they met the following criteria: (1) population-based one-sample FIT screening in an average risk population (>40 years old), (2) subjects were screened with FIT, while taking an OAC or NSAID, with subsequent colonoscopy in case of a positive faecal occult blood test; and (3) control group included patients who were screened by means of FIT, not taking OAC or NSAID, and also undergoing colonoscopy in case of a positive faecal occult blood test. The following studies were excluded: (1) those that used gFOBT instead of FIT; (2) systematic reviews and meta-analyses; and (3) editorials/letters.
Primary outcome was the pooled positive predictive value (PPV) of FIT for detecting advanced neoplasia (PPVAN) in patients using any OACs and for aspirin/NSAIDs alone compared with non-users.
Secondary outcomes were the pooled PR of FIT, the pooled NPV and sensitivity and specificity of FIT for advanced neoplasia (AN) and CRC during OACs/NSAIDs use versus no use.
Advanced adenomas (AAs) were defined as adenomas >10 mm, or with villous histology, or high-grade dysplasia. CRC was considered to be the case when malignant cells were observed beyond the muscularis mucosa. AN comprised AA and CRC. Pooled OACs included use of vitamin K antagonists, platelet aggregation inhibitors and novel OACs. NSAIDs were not further specified. We converted units for FIT positivity cut-off into micrograms (µg) of haemoglobin (Hb) per gram of stool for each study when other units were practised.
Data were extracted by the same authors (SAVN and FERV) according to previously stated variables (online supplementary S2). When data in the published studies were not conclusive for our analyses, authors were contacted by mail and/or telephone for additional data.
The sensitivity, specificity, PPV, NPV and PR with corresponding 95% CI were calculated for each study in case data were available. Pooled relative risks (RRs) were obtained by a random-effect forest plot using an inverse-variance estimator, in which an RR smaller than 1 reflects a higher PPV in users versus non-users. An RR greater than 1 implies a lower PPV in users versus non-users.20 Heterogeneity among studies was measured by calculating the inconsistency index (I²). Heterogeneity levels can range from 0% to 100% (maximum heterogeneity), with greater than 25%, 50% and 75% being low, moderate and high heterogeneity, respectively.21
Publication bias was assessed by constructing funnel plots. Assessment of methodological quality of observational cohort studies and case–control studies was carried out using the Ottawa-Newcastle Scale.22 This scale scores quality of design, content and ease of use directed to the task of performing and interpreting meta-analyses results. A star system has been developed in which a study is judged on (1) selection of study groups, (2) comparability of groups and (3) the ascertainment of either the exposure for case–control studies or the outcome of interest for observational studies. The outcome ranges from 0 (low) to 9 (high) stars. Assessment of quality of evidence was carried out using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE).23 Two authors (SAVN and FERV) independently assessed study quality.
Review Manager V.5.3 was used for all analyses. Forest plots were conducted in R V.3.4.2.
After removal of duplicates, we identified 2.022 studies through the electronic database search (figure 1). We excluded 1.970 studies after screening titles and abstracts. Of the remaining, 52 were examined by full-text review. Forty-four studies were excluded. We included six studies in full and two published abstracts in our meta-analysis.24–31
Baseline characteristics of the included studies are shown in table 1. Eight observational cohort studies and one case–control study were included. Seven studies were performed in Europe and one in Asia. The cut-off for a positive FIT ranged between 2 µg and50 µg Hb/g faeces. Pooled analyses of different types of OACs were applied in the included studies.24 27–29 Additionally, separate analyses were made for aspirin.24–26 29–31 One study provided data on NSAIDs, and these users were pooled with aspirin users.31 All studies contained data to calculate PPVAN. Two studies additionally included data on sensitivity, specificity and NPV.30 31 Another two studies contained data on PR of FIT.26 27 Two studies comprised the same screening cohort, yet subgroups for medication use were defined differently in both studies.26 27 For our analyses on pooled OACs, we used the most recent published data.27 For separate analysis for aspirin/NSAID use, we used the published data on the aspirin group.26 A summary of primary and secondary outcomes per study are presented in table 2. On methodological quality, studies scored between six and eight stars (out of a maximum of nine) according to the Newcastle-Ottawa Scale (online supplementary S3). According to the GRADE guidelines, quality of evidence for our analyses scored ‘low’ (online supplementary S4). Heterogeneity between studies for pooled OAC analysis was scored as ‘low’. Separate analysis on aspirin/NSAIDs scored ‘moderate’ (figures 2 and 3). No publication bias was found when funnel plots were conducted (online supplementary S5).
Pooled OAC use versus no use
Positive predictive value for advanced neoplasia
Our meta-analysis composed pooled data on 633 OAC users and 2930 non-users, all FIT-positive patients. Users provided a PPVAN of 37.6% (95% CI 33.9 to 41.4) compared with a PPVAN of 40.3% (95% CI 38.5 to 42.1) for non-users. The forest plot shown in figure 2 showed no significant difference (p=0.75).
Positive predictive value for CRC
Two studies provided data on CRC with pooled OAC use comprising 336 users and 802 non-users.24 29 Pooled OAC users provided a PPVCRC of 5.7% (95% CI 3.7 to 8.7) compared with 6.2% (95% CI 4.8 to 8.1) for non-users.
Aspirin/NSAID use versus no use
Positive predictive value for advanced neoplasia
Pooled data for aspirin/NSAID use identified 463 users and 2438 non-users in FIT-positive patients. Users yielded a pooled PPVAN of 38.2% (95% CI 33.8 to 42.9) compared with 39.4% (95% CI 37.5 to 41.3) for non-users. The forest plot shown in figure 3 revealed no significant difference (p=0.59).
Positive predictive value for CRC
Two studies provided data on CRC and the use of aspirin/NSAIDs.24 29 Together they provided a pooled PPVCRC of 6.9% (95% CI 4.3 to 10.7) in aspirin/NSAID users, compared with 5.3% (95% CI 3.9 to 7.2) for non-users.
The PR of FIT was calculated in one cohort.27 An overall PR of 6.3% was observed. When acenocoumarol was used, PR of FIT was 9.3% versus 6.2% for non-users.
Subanalysis of aspirin alone was associated with a PR of 7.3%, compared with PR of 7.1% for non-aspirin antiplatelet agents.26 In patients undergoing dual antiplatelet therapy (DAPT), PR of FIT was 22.2% compared with 6.3% for non-users (OR 3.5; 95% CI 1.7 to 7.3). Also, the number of AN found in the DAPT subgroup was higher than in non-users (OR 2.8; 95% CI 1.1 to 7.2).
Sensitivity and specificity
No data were available on sensitivity and specificity of FIT in pooled OAC users.
One study assessed sensitivity and specificity in aspirin/NSAID users.31 Sensitivity for AN was 15.8% for users, compared with 34.2% for non-users (p=0.097). Specificity for AN was significantly lower for aspirin/NSAID users; 89.1% compared with 92.1% for non-users (p=0.049). NPV showed no significant difference; 95.0% for users, compared with 96.1% for non-users (p=0.338).
Another study showed a sensitivity of 70.8% for aspirin users alone, compared with 35.9% for non-users (p=0.001). Specificity was 85.7% for aspirin users compared with 89.2% for non-users (p=0.13). NPV was 96.2% for aspirin users, compared with 92.3% for non-users (p=0.05).30
Duration of drug use
One study made a distinction based on the median duration of aspirin use.25 Two categories were formed: a median use of ≤5 years and ≥5 years. A total of 49 patients using aspirin ≤5 years provided a PPVAN of 61.2% (95% CI 47.2 to 73.6) compared with 52 aspirin users ≥5 years providing a PPVAN of 38.5% (95% CI 26.5 to 52.0) (p=0.03).25
Type of FIT used
Seven studies used a quantitative FIT.24–30 One study used a qualitative FIT.31 When the study with a qualitative FIT was excluded, no changes in pooled results were seen (pooled PPVAN in users of OAC: 39.6% vs 44.1% in non-users, RR: 0.99 (95% CI 0.89 to 1.11, p=0.44). Furthermore, five out of the eight studies included used the OC-sensor.24 26–29 After excluding the three studies that used another FIT brand, no alterations in pooled results were seen (pooled PPVAN in users of OAC: 37.8% vs 42.4% in non-users, RR: 1.00 (95% CI 0.87 to 1.14), p=0.99).25 30 31
FIT cut-off used
Different cut-offs were used; most studies vary between a cut-off level of 10–20 µg Hb/g faeces.24–29 Two studies used a cut-off of, respectively, 2 µg and 50 µg Hb/g faeces.30 31 If these two outlier cut-offs were left out, no alterations in pooled results were seen (pooled PPVAN in users of OAC: 39.9% vs 45.8% in non-users, RR: 0.97 (95% CI 0.87 to 1.09), p=0.64).
This is the first systematic review and meta-analysis to determine the PPVAN of FIT in relation to OACs or NSAIDs use. Our results show that the use of OACs or aspirin/NSAIDs do not affect the PPVAN in FIT CRC screening. The PPVAN of pooled OAC users was 37.6% versus 40.3% in non-users. For separate analyses on aspirin/NSAID users, the PPVAN was 38.2%, whereas PPVAN of non-users was 39.4%. Based on current literature, the withdrawal of OACs or aspirin/NSAIDs during FIT screening is not recommended.
Our data are supported by previous work that pooled data on warfarin use during faecal occult blood test screening. Results showed no alterations in PPV of colorectal AN.13 However, included studies were performed on gFOBT and not on FIT. Another meta-analysis compared accuracy of FIT and gFOBT screening if OACs or NSAIDs were used.32 They showed a decrease in PPVAN in gFOBT screening and no significant difference in PPV of FIT. Hence, only one study on FIT screening was included in this meta-analysis.29 FIT and gFOBT differ in their interaction with Hb. Guaiac-based tests interact with the haem part of Hb, and immunochemical tests detect the globin portion of Hb. The latter does not survive passage through the upper gastrointestinal tract, and therefore, FIT has a proven superior accuracy for colon or rectum bleeding compared with gFOBT.2 3 For this reason, it is to assume that effects of OACs and NSAIDs could act differently in both tests. Growing literature on FIT screening helped to perform the current meta-analysis based on the today’s practised FIT. Our results support the previous suggestion that OACs and aspirin/NSAIDs do not affect PPVAN of FIT.
Only one cohort provided data on PR of FIT in which a higher PR was seen in users compared with non-users.26 27 As already hypothetically stated, this could be due to possible stimulation of bleeding from lesions in the colon (both benign and (pre)malignant). More so, the use of DAPT showed an even more strong effect on increased PR, supporting the literature on DAPT and its stimulating effect on lower gastrointestinal bleedings.33 Bearing in mind the similar PPV for users and non-users (or even a greater PPV in the case of DAPT users), this could presume the stimulation of premalignant lesions to bleed and causing a beneficial effect of OAC and aspirin/NSAID use by having more true FIT positives in users.
One study used a qualitative test (ie, providing a positive or negative result without specific blood count) (Hemosure test kit) and calculated a PPVAN of 20.0% for aspirin/NSAID users, compared with 7.5% for non-users.31 In our meta-analysis, these results act as an outlier compared with other study outcomes. When left out of our analysis, no evident effects on pooled PPVAN of users versus non-users were seen.
In our meta-analysis, all included studies applied a one-sample FIT. There is one study evaluating FIT performance and the use of antithrombotics in a two-sample FIT screening showing also that OAC use do not affect FIT performance.34
Globally, CRC screening guidelines focus mostly on age range of screening, time intervals, multiple test options and follow-up diagnostics. Although specific subgroups are discussed (eg, different ethnicities and individuals with a family history of CRC), OAC/NSAID users are left out.35 36 Given the significant proportion of subjects using these drugs and the renewing scientific evidence on this topic, guideline adjustments should be considered. Although this has been an ongoing discussion,37 still no recommendations were made in the latest update of the US Multi-Society Task Force CRC screening guidelines.35
Certain limitations have to be addressed in order to add specific recommendations. First, cut-off points of FIT were varying and overall relatively low. The use of different cut-off points of FIT affects accuracy of FIT. An increase in faecal Hb concentration cut-off is associated with higher PPV.6 Second, no subgroup analyses on age, gender, type of drugs or duration of drug use could be performed since the number of studies was too low. It was already pointed out that separate analysis on duration of drug use could play an important part in FIT performance.25
In conclusion, OACs and aspirin/NSAID use do not affect the PPV of FIT in CRC screening. Based on current literature, withdrawal of OACs and/or NSAIDs before FIT sampling is not recommended. However, subgroup analyses on subject and drug characteristics should be performed in order to conduct specific guideline recommendations, and PR of FIT in relation to the PPV should be taken into account.
The authors of this systematic review and meta-analysis would like to acknowledge the contribution of Wichor Bramer, biomedical information specialist of the Erasmus University Medical Center, for performing the systematic literature search.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.