Approaches to Diagnosing IBS
Approaches to Diagnosing IBS
This study has examined the accuracy of symptom-based diagnostic criteria, biomarkers, psychological markers or combinations thereof, in making a diagnosis of IBS. The Rome III criteria, the current gold standard for the diagnosis of IBS, have only been validated in one study to date, and performed modestly and similar to the other symptom-based diagnostic criteria that have been described previously, with a positive LR >3 and a negative LR of approximately 0.4. Proposed biomarkers, with the exception of abnormal sigmoid muscularis propria thickness in female patients, intestinal mucosal endocrine cells, and faecal VOMs and a combination of FC and intestinal permeability, all examined in single studies, appeared to perform no better than available symptom-based diagnostic criteria. The accuracy of psychological markers was also similar. Combining symptoms, biomarkers and/or psychological markers in various permutations seemed to perform better generally in diagnosing IBS, and with a significantly greater pooled positive LR compared with symptom-based criteria.
Strengths of this study include a comprehensive search strategy, including a recursive search of the bibliographies of all eligible studies, and searching of conference proceedings to identify any potential studies published that may not have been included in the original search of the medical literature. This resulted in the identification of a wide range of potential methods for diagnosing IBS; specifically four different symptom-based diagnostic criteria evaluated in seven studies, 11 biomarkers evaluated in 12 studies, four psychological markers evaluated in two studies, and five different combinations of symptoms, biomarkers and/or psychological markers evaluated in six studies. Pooling the data in some of our analyses resulted in a study population of 1800 patients or more for each of the symptom-based diagnostic criteria, and >1000 patients for the Kruis statistical model. Furthermore, this is the first study that has attempted to summarise data from all available methods, including novel approaches, to diagnose IBS.
There are some limitations to this study. When data were pooled, results varied between individual studies that evaluated the same diagnostic method in some analyses, which may be partly explained by differences in study design, recruitment, setting, and country and differences in the reference standard used for diagnosing IBS. However, we used a random effects model when pooling study data in all these analyses, to provide a more conservative estimate of diagnostic accuracy. The cut-offs we used to define presence of IBS for each of the diagnostic tests assessed in this meta-analysis was imposed by the reporting of the authors of the original studies. This is less relevant for studies employing diagnostic criteria, such as the Manning criteria, because we were able to obtain data for several thresholds, but is an issue for studies using laboratory tests, such as faecal chromogranins or VOMs, or the measures of psychological affect, which were not always used at the threshold recommended by the original authors. In addition, the pooled IBS prevalence of all studies was high at >40%, as the majority of studies were conducted in referral populations in secondary care, meaning that some of the findings may not be applicable to a primary care setting, where the majority of patients are diagnosed and managed, as the prevalence of IBS may well be lower. The inclusion of case–control studies may lead to an overestimation of the diagnostic performance of the test being examined, compared with studies using a clinical cohort, because these are subject to spectrum bias as the study design often omits mild cases that are difficult to diagnose. Finally, 14 of the 22 eligible studies were judged as high risk of bias, or had other applicability concerns, when assessing quality using the QUADAS-2 tool, highlighting the limitations of some of the studies.
In terms of our use of LRs, the advantage of these is that they provide the clinician with an intuitive feeling on ruling in or ruling out a given disease based on the magnitude of the positive and negative LRs respectively, and are less likely to be influenced by prevalence of the disease of interest than predictive values. However, there are some disadvantages to using LRs. In particular, LRs are not often quoted in the medical literature, the main reason being that clinicians are more used to dealing with probabilities as in the case of predictive values, whereas LRs express their results in terms of odds. Another disadvantage is that LRs can have wide confidence intervals, particularly in rare diseases with a low prevalence. However, this is of less relevance to IBS, which has a prevalence of 10–20% in the general population. We have provided sensitivities, specificities, and positive and negative predictive values for each of the diagnostic tests applied in Table S3 http://onlinelibrary.wiley.com/doi/10.1111/apt.13283/suppinfo.
Guidelines for the management of IBS recommend making a positive diagnosis of IBS based on symptoms, and discouraging a 'diagnosis of exclusion' approach. Symptom-based diagnostic criteria were developed to aid in this, and therefore avoid unnecessary and potentially invasive investigations. However, one of the most consistent findings of this study is the modest performance of all the available symptom-based criteria in identifying IBS. This comparable performance between the symptom-based criteria is perhaps not surprising, considering they are derivatives of each other, and therefore share the same strengths and weaknesses.
In general, the performance of biomarkers in the studies we identified was similar to symptom-based criteria, which is disappointing considering their potentially expensive nature. In some cases, the biomarkers would not be considered useful as a test outside of a tertiary referral centre, due to the invasive nature or complexity of the test applied. Furthermore, a number of the studies that assessed the accuracy of biomarkers used healthy volunteers as controls, whereas a biomarker that differentiated IBS from other organic disorders, in which the symptoms are likely to overlap with those of IBS, would be more clinically useful. Sigmoid muscularis propria measurement using trans-vaginal ultrasound, appeared to perform well with a positive LR of 15. However, this study had a number of limitations, including the failure to exclude other causes of abnormal muscularis propria thickness, such as colorectal cancer, inflammatory bowel disease or diverticular disease, only 27 patients in the study population having a confirmed diagnosis of IBS, and the generalisability of the results, given that the test was applied in female patients only. In addition, the results have yet to be validated by other investigators, despite the study being published 8 years ago. Duodenal mucosal CgA cell quantification also performed well in differentiating IBS from health, but only in a single study, and the test is invasive. In addition, the effect of coeliac disease, duodenitis, duodenal ulcers or inflammatory bowel disease on numbers of CgA cells in the duodenum has not been studied, and therefore further work in this area is required before any definitive conclusion can be made from this study. Faecal VOMs showed some promise in differentiating IBS from active inflammatory bowel disease in one small study, but again results of this study will require validating by others.
Given the known association of IBS with somatic symptoms and mood disorders demonstrated by others, it is perhaps surprising that the performance of psychological markers were, in general, no better than that of symptom-based criteria. One possible reason that studies evaluating psychological markers performed poorly, was that these predominantly used markers of anxiety and depression to predict presence of IBS. In a recently conducted study, where somatisation data were collected from more than 4000 patients referred to secondary care for investigation of their symptoms, mean somatisation scores and mean number of somatic symptoms reported were significantly higher in 840 patients with IBS compared with 2137 patients without IBS, but who still reported GI symptoms. These findings support those of Spiller et al., where somatisation was found to be superior to either anxiety or depression. Somatisation as a marker may therefore be more accurate in predicting the presence of IBS than other measures of psychological well-being. More focused measures of psychological affect such as worry or rumination, which may also be associated with IBS, could provide fruitful avenues for future research. An additional issue is the fact that the two studies reporting on the accuracy of psychological markers differentiated IBS from health, and therefore whether psychological markers can accurately discriminate between IBS and organic gastrointestinal disorders is unclear. This would seem less likely, as there is evidence to suggest that many organic GI disorders are also associated with psychological impairment.
Combinations of symptoms, biomarkers and/or psychological markers seemed to perform better, generally, in the six studies that assessed the accuracy of these approaches, and were superior to symptom-based criteria in terms of the pooled positive LR. This may be because IBS is a complex, heterogeneous disorder, for which there is no single unifying explanation, and for which numerous mechanisms have been proposed. Combining symptoms or examination findings, biomarkers and/or psychological markers may therefore be a more useful approach to diagnosing IBS, and perhaps points the way forward for future iterations of the Rome process. However, this approach will likely result in increasing complexity of the diagnostic test, and may reduce its practicality in a clinical setting. This is probably the main reason that the Kruis statistical model was never adopted widely, despite its apparently superior performance in predicting IBS accurately. Any future set of diagnostic criteria would also need to be easily administrable and interpretable in a primary care setting, and this again probably explains why the combination of FC, intestinal permeability ratio, and Rome I criteria did not progress beyond a research setting, despite its accuracy.
Our findings highlight that existing diagnostic approaches are unable to define IBS with any great accuracy. An alternative strategy may be to use latent class analysis methods. These are designed to use available information to create an appropriate pattern of measurements that most closely represents the latent construct IBS, and define groups, or classes, which could serve as a way of improving the accuracy of methods used to diagnose IBS, and therefore help to discriminate between IBS and non-IBS symptom profiles. However, at the time of writing, there are only a few instances of this type of modelling being used in functional GI disorders. These complex statistical methods have the potential to explore a combination of patient-reported, endoscopic and biochemical variables, to develop a stronger predictor of IBS, which could then be employed in everyday practice in the clinic by exploiting technological advancements, such as smartphone applications. These novel methods to aid the diagnosis of IBS could also be used to assess clinically meaningful dependent variables, such as response to therapy, health care consumption and quality of life.
In conclusion, this meta-analysis has shown that symptom-based diagnostic criteria, biomarkers and psychological markers perform only moderately well in diagnosing IBS, and in the case of biomarkers many of these are potentially expensive or invasive, and are not yet practical for clinical application. Combining symptoms with markers of organic disease or psychological affect, for instance using latent variable models, may represent the best way forward in improving the accuracy of diagnosing IBS.
Discussion
This study has examined the accuracy of symptom-based diagnostic criteria, biomarkers, psychological markers or combinations thereof, in making a diagnosis of IBS. The Rome III criteria, the current gold standard for the diagnosis of IBS, have only been validated in one study to date, and performed modestly and similar to the other symptom-based diagnostic criteria that have been described previously, with a positive LR >3 and a negative LR of approximately 0.4. Proposed biomarkers, with the exception of abnormal sigmoid muscularis propria thickness in female patients, intestinal mucosal endocrine cells, and faecal VOMs and a combination of FC and intestinal permeability, all examined in single studies, appeared to perform no better than available symptom-based diagnostic criteria. The accuracy of psychological markers was also similar. Combining symptoms, biomarkers and/or psychological markers in various permutations seemed to perform better generally in diagnosing IBS, and with a significantly greater pooled positive LR compared with symptom-based criteria.
Strengths of this study include a comprehensive search strategy, including a recursive search of the bibliographies of all eligible studies, and searching of conference proceedings to identify any potential studies published that may not have been included in the original search of the medical literature. This resulted in the identification of a wide range of potential methods for diagnosing IBS; specifically four different symptom-based diagnostic criteria evaluated in seven studies, 11 biomarkers evaluated in 12 studies, four psychological markers evaluated in two studies, and five different combinations of symptoms, biomarkers and/or psychological markers evaluated in six studies. Pooling the data in some of our analyses resulted in a study population of 1800 patients or more for each of the symptom-based diagnostic criteria, and >1000 patients for the Kruis statistical model. Furthermore, this is the first study that has attempted to summarise data from all available methods, including novel approaches, to diagnose IBS.
There are some limitations to this study. When data were pooled, results varied between individual studies that evaluated the same diagnostic method in some analyses, which may be partly explained by differences in study design, recruitment, setting, and country and differences in the reference standard used for diagnosing IBS. However, we used a random effects model when pooling study data in all these analyses, to provide a more conservative estimate of diagnostic accuracy. The cut-offs we used to define presence of IBS for each of the diagnostic tests assessed in this meta-analysis was imposed by the reporting of the authors of the original studies. This is less relevant for studies employing diagnostic criteria, such as the Manning criteria, because we were able to obtain data for several thresholds, but is an issue for studies using laboratory tests, such as faecal chromogranins or VOMs, or the measures of psychological affect, which were not always used at the threshold recommended by the original authors. In addition, the pooled IBS prevalence of all studies was high at >40%, as the majority of studies were conducted in referral populations in secondary care, meaning that some of the findings may not be applicable to a primary care setting, where the majority of patients are diagnosed and managed, as the prevalence of IBS may well be lower. The inclusion of case–control studies may lead to an overestimation of the diagnostic performance of the test being examined, compared with studies using a clinical cohort, because these are subject to spectrum bias as the study design often omits mild cases that are difficult to diagnose. Finally, 14 of the 22 eligible studies were judged as high risk of bias, or had other applicability concerns, when assessing quality using the QUADAS-2 tool, highlighting the limitations of some of the studies.
In terms of our use of LRs, the advantage of these is that they provide the clinician with an intuitive feeling on ruling in or ruling out a given disease based on the magnitude of the positive and negative LRs respectively, and are less likely to be influenced by prevalence of the disease of interest than predictive values. However, there are some disadvantages to using LRs. In particular, LRs are not often quoted in the medical literature, the main reason being that clinicians are more used to dealing with probabilities as in the case of predictive values, whereas LRs express their results in terms of odds. Another disadvantage is that LRs can have wide confidence intervals, particularly in rare diseases with a low prevalence. However, this is of less relevance to IBS, which has a prevalence of 10–20% in the general population. We have provided sensitivities, specificities, and positive and negative predictive values for each of the diagnostic tests applied in Table S3 http://onlinelibrary.wiley.com/doi/10.1111/apt.13283/suppinfo.
Guidelines for the management of IBS recommend making a positive diagnosis of IBS based on symptoms, and discouraging a 'diagnosis of exclusion' approach. Symptom-based diagnostic criteria were developed to aid in this, and therefore avoid unnecessary and potentially invasive investigations. However, one of the most consistent findings of this study is the modest performance of all the available symptom-based criteria in identifying IBS. This comparable performance between the symptom-based criteria is perhaps not surprising, considering they are derivatives of each other, and therefore share the same strengths and weaknesses.
In general, the performance of biomarkers in the studies we identified was similar to symptom-based criteria, which is disappointing considering their potentially expensive nature. In some cases, the biomarkers would not be considered useful as a test outside of a tertiary referral centre, due to the invasive nature or complexity of the test applied. Furthermore, a number of the studies that assessed the accuracy of biomarkers used healthy volunteers as controls, whereas a biomarker that differentiated IBS from other organic disorders, in which the symptoms are likely to overlap with those of IBS, would be more clinically useful. Sigmoid muscularis propria measurement using trans-vaginal ultrasound, appeared to perform well with a positive LR of 15. However, this study had a number of limitations, including the failure to exclude other causes of abnormal muscularis propria thickness, such as colorectal cancer, inflammatory bowel disease or diverticular disease, only 27 patients in the study population having a confirmed diagnosis of IBS, and the generalisability of the results, given that the test was applied in female patients only. In addition, the results have yet to be validated by other investigators, despite the study being published 8 years ago. Duodenal mucosal CgA cell quantification also performed well in differentiating IBS from health, but only in a single study, and the test is invasive. In addition, the effect of coeliac disease, duodenitis, duodenal ulcers or inflammatory bowel disease on numbers of CgA cells in the duodenum has not been studied, and therefore further work in this area is required before any definitive conclusion can be made from this study. Faecal VOMs showed some promise in differentiating IBS from active inflammatory bowel disease in one small study, but again results of this study will require validating by others.
Given the known association of IBS with somatic symptoms and mood disorders demonstrated by others, it is perhaps surprising that the performance of psychological markers were, in general, no better than that of symptom-based criteria. One possible reason that studies evaluating psychological markers performed poorly, was that these predominantly used markers of anxiety and depression to predict presence of IBS. In a recently conducted study, where somatisation data were collected from more than 4000 patients referred to secondary care for investigation of their symptoms, mean somatisation scores and mean number of somatic symptoms reported were significantly higher in 840 patients with IBS compared with 2137 patients without IBS, but who still reported GI symptoms. These findings support those of Spiller et al., where somatisation was found to be superior to either anxiety or depression. Somatisation as a marker may therefore be more accurate in predicting the presence of IBS than other measures of psychological well-being. More focused measures of psychological affect such as worry or rumination, which may also be associated with IBS, could provide fruitful avenues for future research. An additional issue is the fact that the two studies reporting on the accuracy of psychological markers differentiated IBS from health, and therefore whether psychological markers can accurately discriminate between IBS and organic gastrointestinal disorders is unclear. This would seem less likely, as there is evidence to suggest that many organic GI disorders are also associated with psychological impairment.
Combinations of symptoms, biomarkers and/or psychological markers seemed to perform better, generally, in the six studies that assessed the accuracy of these approaches, and were superior to symptom-based criteria in terms of the pooled positive LR. This may be because IBS is a complex, heterogeneous disorder, for which there is no single unifying explanation, and for which numerous mechanisms have been proposed. Combining symptoms or examination findings, biomarkers and/or psychological markers may therefore be a more useful approach to diagnosing IBS, and perhaps points the way forward for future iterations of the Rome process. However, this approach will likely result in increasing complexity of the diagnostic test, and may reduce its practicality in a clinical setting. This is probably the main reason that the Kruis statistical model was never adopted widely, despite its apparently superior performance in predicting IBS accurately. Any future set of diagnostic criteria would also need to be easily administrable and interpretable in a primary care setting, and this again probably explains why the combination of FC, intestinal permeability ratio, and Rome I criteria did not progress beyond a research setting, despite its accuracy.
Our findings highlight that existing diagnostic approaches are unable to define IBS with any great accuracy. An alternative strategy may be to use latent class analysis methods. These are designed to use available information to create an appropriate pattern of measurements that most closely represents the latent construct IBS, and define groups, or classes, which could serve as a way of improving the accuracy of methods used to diagnose IBS, and therefore help to discriminate between IBS and non-IBS symptom profiles. However, at the time of writing, there are only a few instances of this type of modelling being used in functional GI disorders. These complex statistical methods have the potential to explore a combination of patient-reported, endoscopic and biochemical variables, to develop a stronger predictor of IBS, which could then be employed in everyday practice in the clinic by exploiting technological advancements, such as smartphone applications. These novel methods to aid the diagnosis of IBS could also be used to assess clinically meaningful dependent variables, such as response to therapy, health care consumption and quality of life.
In conclusion, this meta-analysis has shown that symptom-based diagnostic criteria, biomarkers and psychological markers perform only moderately well in diagnosing IBS, and in the case of biomarkers many of these are potentially expensive or invasive, and are not yet practical for clinical application. Combining symptoms with markers of organic disease or psychological affect, for instance using latent variable models, may represent the best way forward in improving the accuracy of diagnosing IBS.
Source...