Article Text
PDF
Abstract
Objective Implementation of faecal immunochemical tests (FIT) as a triage test in primary healthcare may improve the efficiency of referrals without missing cases of colorectal cancer (CRC). We aim to summarise the performance characteristics of FITs for CRC in symptomatic patients presenting to primary healthcare.
Design We performed a systematic literature review of Medline and EMBASE databases from May 2018 to November 2020. Previous related systematic searches were also adapted to this aim and completed with reference screening. We identified studies performed on adult patients consulting for abdominal symptoms in primary care which reported data such that the FIT diagnostic performance parameters for CRC could be obtained. Bivariate models were used to synthesise available evidence. Meta-regression analysis was performed to evaluate the causes of heterogeneity.
Results Twenty-three studies (69 536 participants) were included (CRC prevalence 0.3%–6.2%). Six studies (n=34 691) assessed FIT as rule in test (threshold of ≥150 µg Hb/g faeces) showing a sensitivity of 64.1% (95% CI 57.8% to 69.9%) and a specificity of 95.0% (95% CI 91.2% to 97.2%). A threshold of 10 µg/g (15 studies; n=48 872) resulted in a sensitivity of 87.2% (95% CI 81.0% to 91.6%) and a specificity of 84.4% (95% CI 79.4% to 88.3%) for CRC. At a 20 µg Hb/g faeces threshold (five studies; n=24 187) less than one additional CRC would be missed per 1000 patients investigated compared with 10 µg Hb/g faeces threshold (CRC prevalence 2%).
Conclusion FIT is the test of choice to evaluate patients with new-onset lower gastrointestinal symptoms in primary healthcare.
- clinical decision making
- colonoscopy
- colorectal cancer
- endoscopy
- stool markers
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information. Data obtained from the systematic review and meta-analysis are included in the article and online supplemental material.
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Colorectal cancer (CRC) detection in symptomatic patients is a challenge for healthcare systems given the low specificity of symptoms. This results in overuse of colonoscopy resources and delay in diagnosis.
Faecal immunochemical test (FIT) may be effective in the stratification of CRC risk in patients with abdominal symptoms seen in primary healthcare.
What are the new findings?
A 150 µg Hb/g of faeces threshold identifies more than half of CRC with high specificity.
In low CRC prevalent populations, CRC risk in patients with faecal haemoglobin <10 µg/g of faeces equals the risk of colonoscopy severe complications and the CRC risk in asymptomatic subjects.
How might it impact on clinical practice in the foreseeable future?
The evaluation of patients consulting with new-onset lower gastrointestinal symptoms in primary healthcare with FITs enables rational use of the available resources.
In the near future, we will have to address two questions: how to detect FIT negative CRC and whether FIT evaluation in symptomatic patients improves CRC prognosis.
Introduction
A significant percentage of colorectal cancers (CRCs) are diagnosed in symptomatic patients, after the implementation of CRC screening programmes.1 Unfortunately, most symptoms are non-specific at presentation as they are shared among non-malignant conditions and different types of cancer, which produces additional difficulties and delay in diagnosis.2 Moreover, concordance between patient-reported and doctor-reported symptoms is low,3 and most patients with abdominal symptoms do not have significant colorectal disease.4
In the last few years, evidence has proven that faecal immunochemical tests for haemoglobin (FIT) may be effective in evaluating patients with abdominal symptoms to identify patients at low risk of CRC.5 Furthermore, the amount of faecal haemoglobin (f-Hb) detected has been shown to be related to severity of disease,6 and constitutes a better CRC risk predictor than demographic (age and sex), clinical (presence of symptoms) and family history or lifestyle factors.7
For these reasons, the National Institute for Health and Care Excellence recommended (DG30) in 2017 the use of FIT to guide referral for suspected CRC in patients without rectal bleeding who complain with certain unaccounted for low-risk symptoms.8 Furthermore, implementation of FIT as a triage test in primary care with appropriate safety netting may improve the efficiency of referrals without missing cases of relevant bowel disease. This is even more important nowadays, in regions with additional capacity issues due to the COVID-19 pandemic.9
Notwithstanding the above, a recent systematic review revealed that few countries recommend FIT in primary healthcare as an adjunct to clinical assessment.10 There is limited evidence of the use of FITs in this setting as most studies supporting DG30 recommendation were performed in secondary care.5 This could increase the concerns of general practitioners to use FITs to aid their decision-making process when dealing with a patient with symptoms suggesting CRC. Several studies have recently been published evaluating FIT in symptomatic patients seen in primary healthcare. We, therefore, aim to perform a systematic review to assess the diagnostic accuracy of FIT for CRC detection in patients presenting with recent onset gastrointestinal symptoms in primary healthcare, with special interest on the clinical effectiveness for triaging referrals in this setting.
Materials and methods
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement to conduct and report this systematic review.11
Data sources and searches
MEDLINE (via PubMed) and EMBASE (via Ovid) databases were searched from May 2018 to November 2020 without restrictions on language or publication status. Published search strategies in related systematic reviews were consulted and updated.5 12 The reference lists of all relevant articles extracted were also reviewed to identify additional potentially interesting articles following an iterative process. Furthermore, we also included all studies identified by previous systematic reviews that satisfied the inclusion criteria of this research (online supplemental appendix 1).
Supplemental material
Study selection
Three authors (CT-S, NPV and CS-G) independently screened titles and abstracts and assessed full text articles of studies considered relevant. We included any cohort study which met all the following criteria: (1) adult subjects (older than 18 years) consulting for abdominal symptoms in primary healthcare, (2) FIT diagnostic performance parameters for CRC and/or significant colonic lesions (SCL) available and (3) thresholds used to determine a positive result expressed as micrograms of haemoglobin per gram of faeces (µg Hb/g). We included in this systematic review studies that reported colon evaluation (either by endoscopic or imaging techniques) or or longitudinal follow-up of controls with medical records or cancer registry and a minimum monitoring time of 3 months as reference standard. A previous study showed that the different follow-up periods (3, 6, 12 months) did not affect FIT diagnostic performance for CRC detection.13
We excluded studies if two by two tables with absolute numbers of true positive, true negative, false positive and false negative (FN) test results could not be constructed. Case–control studies, conference abstracts, studies with hospital inpatients and those including screening, or mixed (with and without symptoms) population were also excluded. Those studies conducted on symptomatic patients who were recruited in colonoscopy units, were only included if authors explicitly state that they were performed on patients referred solely from primary healthcare facilities.
Outcome assessment
Our primary and secondary outcomes were FIT diagnostic performance estimates to detect CRC and SCL respectively, at a cut-off value of limit of detection (LoD), 10 µg Hb/g faeces, 20 µg Hb/g faeces and 150 µg Hb/g faeces.
Data extraction
Two reviewers (CT-S and NPV) extracted data and extractions were verified by a second reviewer (CS-G). Any disagreement was consulted with a third reviewer (NPV/JC). In addition to test performance outcome measures, information on study details (author, year of publication, aim and setting, period of recruitment and type of cohort), participant characteristics (inclusion and exclusion criteria, demographic characteristics, symptoms, acceptability defined as the proportion of participants who returned a FIT sample), target reported (prevalence of CRC and SCL as well as the definition used), FITs characteristics (brand, analyser used, f-Hb concentration used as threshold) and reference standard used (bowel examination and follow-up length when applicable) were considered relevant.
Quality assessment
The potential risks of bias were evaluated for each study included using the Quality Assessment of Diagnostic Accuracy Studies 2 tool (QUADAS-2).14 An inverted funnel scatterplot was used to detect publication bias.
Statistical analysis
To avoid threshold effect, studies were classified by f-Hb threshold for a positive test result. Threshold effect is a specific cause of heterogeneity in meta-analyses of diagnostic test accuracy. It occurs when different criteria (cut-off values or thresholds) are used between studies to assess whether a test result is positive or negative. We calculated pooled estimates of sensitivity, specificity and likelihood ratios using a bivariate random-effects model when at least four studies with similar characteristics were available.15 When necessary, a hierarchical summary receiver operating characteristic (ROC) curve presenting summary estimates of sensitivities and specificities along with their corresponding 95% CI and prediction region was also generated for each subgroup of studies.16 If this approach was not possible, a random-effects model was applied following DerSimonian’s method and summary sensitivity and specificity estimates were reported by plotting a summary ROC curve using DerSimonian and Lair’s model.17 We evaluated the diagnostic yield of FIT according to the CRC prevalence and the post-test probabilities of CRC assessed through Fagan nomograms. We calculated the number of necessary colonoscopies to find one CRC (number necessary to scope, NNS), and the number of missed CRC per 1000 patients with an f-Hb value below a chosen threshold.
The percentage of total variation across studies attributable to heterogeneity rather than chance was assessed statistically using the inconsistency index I2, and values greater than 50% represent substantial heterogeneity.18 Threshold effect was assessed through Spearman’s rank correlation (p<0.1 was considered to be statistically significant).
FIT brand, the location where the patient was recruited (colonoscopy unit or primary health facility) or the reference standard used to follow-up on patients with negative FIT results, are variables which may affect the assessment of FIT accuracy. Thus, when the number of studies allowed, we performed a bivariate random-effects meta-regression to evaluate the impact of these variables on our results. Visual inspection of ROC space was used to enable identification of those studies with major differences from each subgroup based on threshold and sensitivity analysis was performed after removing keynote outliers when those differences can be accounted for through bias. We used Stata V.14.0 (StataCorp), and MetaDisc software for statistical analyses.19
Patient and public involvement
We consulted a European association of CRC patients and their relatives during the development of the study protocol (https://europacolonespana.org) to assess the general public acceptability as well as any concern about using FITs as a triage tool for symptomatic patients with suspected CRC in primary care. Feedback was used to select the most relevant information collected in this systematic review from a general public point of view. These data will be included in a friendly designed poster to be shown in primary care centres, patient association websites and disseminated through press releases.
Results
Study selection
The literature search in MEDLINE and EMBASE identified 4620 potentially relevant articles, of which 170 full-text articles were evaluated and 22 articles met the inclusion criteria (figure 1 and online supplemental appendix 2). The reasons for excluding the articles were as follow: secondary literature (41), studies mixing symptomatic and asymptomatic subjects (76), research performed outside primary healthcare (97), uncertainty with the index test (53), the reference standard (21) or the outcome definition (13). These were supplemented by one article from a manual search published 1 month from the search date, providing a total of 23 studies included in this systematic review (table 1 and online supplemental table 1).13 20–41 Furthermore, additional information of the same patients evaluated in the studies of Khan et al,38 and Chapman et al,31 is respectively reported in another two secondary published studies.42 43 Partial information reported in the studies of Högberg et al,40 and McSorley et al,37 can also be found in other studies included in this review,29 34 and this has been considered in the quantitative synthesis.
Characteristics of the studies included in the meta-analysis
Summary of evidence search and selection.
Study characteristics
The total number of participants was 69 536. Sample sizes ranged from 178 to 15 789. Median age ranged from 58 to 72 years and the proportion of women from 49.0% to 64.6%. CRC and SCL prevalence ranged from 0.3% to 6.2% and 2.6% to 31.0%, respectively. Twelve studies provided information on the FIT’s accuracy for SCL detection. SCL definition varied widely between the different studies. Most authors defined it as the sum of CRC plus high risk and/or advanced adenoma plus Inflammatory bowel disease,13 21–24 27 29–32 35 36 although diverticulitis, significant diverticular disease or complicated diverticular disease were also included in three studies.22 26 31 With respect to the reference standard for CRC and/or SCL, colonoscopy was performed as reference standard in 100% of patients recruited in colonoscopy units and in a variable percentage in patients recruited in primary healthcare. In this subgroup of studies, the CRC diagnosis was based either on different bowl imaging investigations (colonoscopy, CT colonography, plain CT and/or sigmoidoscopy) or in follow-up with a variable length of time (3 to 36 months). Full details of these studies are shown in table 1 and online supplemental table 1).
Quality assessment
Overall results of the quality assessment from the 23 articles are reported in figure 2 by means of the QUADAS-2 instrument. Eight studies were retrospective in design.13 27 28 33 34 37 39 40 Of the 13 articles using longitudinal follow-up as reference standard, eight articles were at high risk of bias because they used heterogenous monitoring periods less than 2 years.13 20 24 27 30 31 35 39 Ten articles had a high risk of selection bias, as their cohorts had either been recruited in colonoscopy units or comprised solely of patients referred to colonoscopy, thus having an increased risk of CRC.21 22 25 26 28 32 36–38 41 Two studies included frozen stool samples,22 28 and another two, which assessed the accuracy of a quantitative FIT (HM-JACKarc), evaluated more than one sample for each patient and considered a positive result if any of those samples had a positive outcome.13 27 One study collected the stool sample for FIT through a digital rectal examination.38 A number of studies had ‘patient selection’ applicability concerns. In many cases, a low proportion of patients who were either invited or agreed to participate in the study were included in the analysis.21 22 25 26 36 41 Other studies also had a very low number of patients.13 20 23 28 30 32
Quality Assessment of Diagnostic Accuracy Studies (QUADAS).
Overall accuracy of FIT based on positivity threshold to detect CRC
The LoD value depended on the FIT brand used and ranged from 2 µg Hb/g faeces to 7 µg Hb/g faeces.The overall pooled sensitivity and specificity of FITs for CRC for studies which used the LoD as threshold (11 studies; n=41 338 patients) were 93.4% (95% CI 88.0% to 96.4%) and 76.9% (95% CI 67.7% to 84.0%), respectively. Sensitivity for CRC decreased from 87.2% (95% CI 81.0% to 91.6%) for studies with a threshold of ≥10 µg Hb/g faeces (15 studies; n=48 872 patients) to 84.1% (95% CI 78.6% to 88.4%) for studies with a threshold ≥20 µg Hb/g faeces (five studies; n=24 187 patients), and specificity increased from 84.4% (95% CI 79.4% to 88.3%) to 86.6% (95% CI 75.6% to 93.1%). Furthermore, six studies (n=34 691 patients) evaluated the diagnostic accuracy of FIT with a threshold of ≥150 µg Hb/g faeces showing a sensitivity and specificity of 64.1% (95% CI 57.8% to 69.9%) and 95.0% (95% CI 91.2% to 97.2%), respectively (table 2). Online supplemental appendix 3 shows pooled sensitivity and specificity for FIT studies based on a cut-off value (online supplemental file).
Diagnostic accuracy parameters based on quantitative FIT threshold concentration to detect CRC and significant colonic lesion
Evaluation of heterogeneity
We found substantial heterogeneity between studies when calculating the summary performance estimates of FITs for CRC using the bivariate model (table 2 and online supplemental appendix 3). The type of reference standard used (colonoscopy or follow-up), the place of recruitment (primary care facility or colonoscopy unit) and CRC prevalence (CRC <3% or CRC ≥3%) were significant predictors of heterogeneity for both sensitivity and specificity. Moreover, FIT brand (OC-Sensor or HM-JACKarc) was also a significant predictor of heterogeneity for specificity (figure 3). However, the magnitude of change for the pooled summary estimates and their confidence intervals in each subgroup was small (online supplemental table 2). Online supplemental figure 1 shows ROC space plots. When we removed keynote outliers in the sensitivity analysis the magnitude of change between the summary estimates and their confidence intervals in each subgroup based on a cut-off value was again small. However, pooled sensitivity estimates were more homogeneous. Instead, pooled specificity estimates remained with high heterogeneity (table 2).
Forest plot of multiple univariable meta-regression and subgroup analyses. Meta-regression to assess the effect of covarying factors on summary measures of performance: ‘Brand’, yes: OC-sensor versus no: HM-JACKarc; ‘CRCprev’, yes: CRC prevalence <3% vs no: CRC prevalence ≥3%; ‘primary care’, yes: recruitment performed in primary care facilities versus No: recruitment performed in colonoscopy units; ‘colonoscopy’, yes: follow-up performed as reference standard versus no: colonoscopy performed as reference standard.
Threshold effect was unsurprisingly detected in the subgroup of studies using cut-off values at LoD due to the differences in the threshold defined by each brand. Moreover, a threshold effect was also found in the subgroup of studies at a cut-off value of ≥150 µg Hb/g faeces. In addition to explicit threshold effect, implicit threshold effect may arise due to several biases (ie, different spectrum of patients) which may determine differences in sensitivity and specificity between studies. Once outliers were removed, heterogeneity related to implicit threshold effect in this subgroup was also controlled (table 2).
Diagnostic yield for CRC
Figure 4 shows the expected NNS and the number of missed CRC per 1000 patients according to the CRC prevalence expected in primary care (1%–5%). The post-test probabilities of CRC were assessed through Fagan nomograms (online supplemental figure 2). As an example, the number of missed CRC per 1000 patients if a patient has a ‘negative’ FIT result in population with a CRC prevalence of 2% is expected to increase from four to five when using the threshold of 20 µg Hb/g faeces instead of 10 µg Hb/g faeces. On the other hand, at the same CRC prevalence, the NNS is expected to decrease from ten to four if the 150 µg Hb/g faeces threshold is used instead of 10 µg Hb/g faeces.
Number of patients necessary to scope to find one CRC and number of missed CRC per 1000 patients. Figures are calculated according to the post-test probabilities of CRC assessed by means of Fagan nomograms for different thresholds and CRC prevalence. CRC, colorectal cancer; FIT, faecal immunochemical tests.
Overall accuracy of FIT based on positivity threshold to detect SCL
The overall pooled sensitivity and specificity of FITs for SCL for studies which used the LoD as threshold (seven studies; n=22 624 patients) were 70.4% (95% CI 68.4% to 72.3%) and 78.4% (95% CI 77.8% to 78.9%), respectively. If we consider all SCLs as target instead of solely CRC, FIT sensitivity decreased from 87.2% (95% CI 81.0% to 91.6%) for studies with a threshold of≥10 µg Hb/g faeces (fifteen studies; n=48 872 patients) to 69.1% (95% CI 60.5% to 76.5%) at the same threshold (seven studies; n=20 407 patients), and specificity increased from 84.4% (95% CI 79.4% to 88.3%) to 87.2% (95% CI 83.4% to 90.2%). Furthermore, three studies (n=20 528 patients) assessed the diagnostic accuracy of FIT with a threshold of ≥150 µg Hb/g faeces showing a sensitivity and specificity of 35.9% (95% CI 33.8% to 38.1%) and 97.5% (95% CI 97.3% to 97.8%), respectively (table 2). SCL prevalence ranged widely between 4.4% and 13.6% anticipating high heterogeneity when assessing summary sensitivity and specificity FIT estimates for SCL detection, which combined with reduced number of studies restricted the possibility of subgroup analysis (table 2 and online supplemental appendix 3).
Publication bias
Online supplemental figure 3 shows various funnel plots where each study is represented by one point on the plot drawn based on the natural logarithm of its diagnostic OR (dOR) (x axis) and the value of its standard error (y axis). The existence of a symmetric figure around an axis traced by the pooled dOR value suggests the absence of publication bias.
Discussion
Statement of principal findings
Our results confirm that FITs are the test of choice to evaluate patients with new-onset lower gastrointestinal symptoms in primary healthcare. The high sensitivity for CRC shown at the 10 µg Hb/g faeces threshold means that any result below has a negative predictive value for CRC greater than 99.6%–99.9% at CRC prevalence most commonly reported in primary healthcare. The risk of CRC detection in patients with a negative FIT equals the risk of colonoscopy-associated side effects and the CRC prevalence in asymptomatic adults aged 50–69.44 45 Moreover, the minor differences between sensitivities for CRC shown at 10 µg Hb/g faeces and 20 µg Hb/g faeces thresholds mean that if we choose the higher threshold, less than one additional CRC would be missed per 1000 patients investigated. Finally, pooled estimates of sensitivity for CRC suggest that more than 60% of CRC would be identified at a f-Hb threshold of 150 µg Hb/g faeces. This threshold has recently been proposed in several large studies as a rule in criteria for urgent evaluation.13 31 36 37 39 41 Furthermore, the NNS range is between 2 and 7 for a CRC prevalence between 1% and 3% at this threshold, which constitutes an appropriate criterion for colonoscopy prioritisation.
Strengths and weaknesses
This is the first systematic review and meta-analysis evaluating the diagnostic performance of FIT in symptomatic patients in primary healthcare. The high number of patients included and consistency in relation to previously published systematic reviews in various settings reinforce the validity of these findings.5 46 However, studies included in this systematic review are not free from bias, which could affect our results. On the one hand, verification bias arises in diagnostic and prognostic studies when the reference test may have been performed preferably in those patients with ‘positive’ index tests, as occurs in those studies performed on patients recruited in primary care facilities. Besides, we have found a large heterogeneity in the reference standard used, the length of follow-up in case bowel imaging was not performed and, finally, in the definition of SCL across the studies included. This finding highlights the need of common definition both for a reference standard for CRC diagnosis and for SCL.
Conversely, those studies performed on patients recruited in colonoscopy units are at risk of clinical spectrum bias, because they could lack representation of the whole clinical spectrum of CRC in the study population. Instead of presenting vague symptoms, patients from those studies are more likely to have developed specific symptoms related to advanced stages, and therefore higher f-Hb concentration.6 In both cases, sensitivity could be overestimated. Furthermore, the low ratio between eligible patients and those included in the final analysis may bring risks of representativeness in a number of studies. Although this could also involve selection bias, it would be necessary to compare the characteristics between both subgroups to know in which way the evaluation of FIT diagnostic performance estimates could be affected.
However, these biases may not have a significant impact on the results of this meta-analysis. As stated previously in the methods section of this manuscript, all patients who did not undergo colonoscopy as a reference standard were monitored and any CRC causing symptoms would worsen in the following months leading to diagnosis even if the FIT test proved to be a FN.13 IWe are aware that a short follow-up period could overestimate FIT sensitivity for CRC as long as patients with a positive result would perform a confirmation test. However, in the heterogeneity analysis the magnitude of change for the pooled summary estimates related to the reference standard used was small (online supplemental table 2), suggesting that the reference test had little effect on the diagnostic performance. Furthermore, despite the high risk of selection bias, the patients included in this meta-analysis should be representative of the population consulting in primary care whose clinical situation constitutes a cause of concern for their physician, which is the clinical spectrum where FIT should prove useful.
As expected, this meta-analysis showed high heterogeneity when calculating pooled estimates of specificity. This is because f-Hb can be detected in a number of benign and malignant conditions other than CRC.47 The major variability in the prevalence of some of these conditions (ie, non-steroidal anti-inflammatory drugs enteropathy), together with the absence of randomised or consecutive sampling in most studies included in this review determine the presence of heterogeneity. Instead, those conditions which could account for the presence of f-Hb over the detection limit should only affect FIT sensitivity to detect CRC by serendipity.48 Thus, after removing those studies with higher selection bias, pooled estimates of sensitivity revealed low heterogeneity.
Strengths and weaknesses in relation to other studies
This systematic review could not detect information to compare the accuracy of quantitative and qualitative tests. Several studies offered information on the precision of different brands of qualitative tests, but different cut-off points were used and there are not enough studies to perform an analysis by the different subgroups. However, it is interesting to highlight that several qualitative FIT brands with diverse cut-off values were indirectly compared in the study of Högberg et al,40 which shows that sensitivity to detect CRC was always higher than 80% despite cut-off values ranging between 2 µg Hb/g faeces and 50 µg Hb/g faeces. This information, combined with sensitivity evaluated at a cut-off value of 150 µg Hb/g faeces, and the minor differences between pooled estimates of sensitivity and specificity assessed at 10 µg Hb/g faeces and 20 µg Hb/g faeces for any demographic subgroup,33 strongly suggest that f-Hb should be evaluated for any patient who has been requested a colonoscopy for symptom evaluation to effectively handle the colonoscopy waiting list, as priority. We specifically evaluated the 150 µg Hb/g faeces threshold because several studies have evaluated recently this cut-off due to its reduced number of positive results, high specificity and positive predictive value.13 31 36 37 39 41 The likelihood of cancer increases with increasing f-Hb concentrations, and consequently, FIT could be used to rule-in cancer or prioritise patients for investigation.36
This systematic review cannot recommend any specific quantitative FIT assay either. Although the meta-regression analysis suggests statistically significant differences between OC-Sensor and HM-JACKarc at a cut-off value of 10 µg Hb/g faeces, these are clinically irrelevant and could be partially justified by the different methodology used in the design of their respective studies. Moreover, to the best of our knowledge only one study in this setting directly compared both FIT brands on the same patients,43 and although large variations were found between the different devices, the correlation of the f-Hb results between both was gradually higher as the threshold was increased; 91.7% at a cut-off value of 10 µg Hb/g faeces. Thus, considering that approximately 90% of CRC may be detected above that threshold, it is unlikely that further information will show clinically significant differences between both FIT assays.
The small number of studies conducted in primary care, together with heterogeneity makes it difficult to evaluate publication bias. However, it is unlikely that the most important conclusions of this review will be refuted by additional data. At the time of writing this manuscript, another two studies reporting information are available and their results are in line with this work’s conclusions.49 50
Unanswered questions and future research
Three relevant questions remain to be answered and are critical in the implementation of FIT in primary healthcare. The first is related to the ‘FIT negative CRCs’. It is relevant to know what the factors are that account for a negative result, either related to the patient or to the CRC, to reduce the FN results. The information available is limited to description of the characteristics of the 47 CRC with a negative result in five studies.13 29 37 41 42 On the other hand, if FIT-based strategies are implemented, it is necessary to establish a safety netting strategy to avoid delays in CRC diagnosis that could worsen the prognosis. A re-evaluation of the symptoms and referral to secondary healthcare in case they persist could be a reasonable option until we have further evidence.51 We have additional potential options: CRC prediction models and the combination of noninvasive biomarkers including the microbiota, but these options are complex and not validated in primary healthcare.52–55
The second question is related to the effect of FIT on CRC prognosis. The main objective of any diagnostic strategy is to improve the prognosis of the disease detected. The information regarding CRC prognosis detected after a positive FIT in symptomatic subjects in primary healthcare is still limited. We have evidence from two retrospective studies that suggest that CRC survival is improved in cancers detected through an FIT-based strategy when compared with a clinical evaluation strategy.1 56 The reason for these findings is not clear but could be related either to a reduction in diagnostic delay or, on the contrary, to an opportunistic CRC screening. A specifically designed study is, however, required to answer this relevant question.
Another relevant implication is the effect of a screening programme in the evaluation of patients with symptoms in primary care. On one hand, including FIT in primary care can facilitate opportunistic screening, increasing inequities in the health system and reducing its efficiency.57 On the other hand, the establishment of a population-based CRC screening programme reduces the risk of CRC among the population that adheres to it. This point raises the hypothesis that the diagnostic approach in patients with recent onset gastrointestinal symptoms could be different if they are invited and adherent to a population-based CRC screening programme.58
Conclusion
In this systematic review and meta-analysis, we confirmed that implementation of FIT as a triage test in primary care may improve the efficiency of referrals. Thus, FIT is the test of choice to evaluate patients with new-onset lower gastrointestinal symptoms in this setting. Use of this test as ‘rule in’ at a cut-off value of 150 µg Hb/g faeces identifies more than half of CRCs using few resources while an f-Hb concentration below 20 µg Hb/g faeces rules out more than 85% of CCR at the expected prevalence in this setting (1%–3%). However, appropriate safety netting is still necessary.
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information. Data obtained from the systematic review and meta-analysis are included in the article and online supplemental material.
Ethics statements
Patient consent for publication
Ethics approval
The study required no Ethics Committee approval as long it is a systematic review and meta-analysis study and no human subject was directly involved.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors NPV and JC conceived and designed the research. NPV, CT-S, NdV-B and CS-G performed data acquisition. NPV analysed and interpretated data. NPV and JC drafted the article or made critical revisions related to important intellectual content of the manuscript. All the authors gave their final approval of the version of the article to be published.
Funding This work was financed by Spain’s Carlos III Health Care Institute by means of project PI17/00837 (Co-funded by European Regional Development Fund/European Social Fund 'A way to make Europe'/'Investing in your future').
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.