- Top of page
- MATERIALS AND METHODS
This study is a systematic review and meta-analysis of studies using nucleic acid-based techniques to detect Mycobacterium avium paratuberculosis (MAP) in patients with Crohn's disease (CD) compared with controls. Database searches were conducted and risk difference estimates were calculated using meta-analysis. Fifty-eight studies were reviewed, 47 of which were included in the analysis. The pooled estimate of risk difference from all studies was 0.23 (95% confidence interval [CI], 0.14–0.32) using a random effects model. Similarly, MAP was detected more frequently from patients with CD compared with those with ulcerative colitis (risk difference 0.19, 95% CI, 0.10–0.28). Year of study, assay type, and inclusion of children explained some but not all of the observed heterogeneity. The data confirms the observation that MAP is detected more frequently among CD patients compared with controls. However, the pathogenic role of this bacterium in the gut remains uncertain. Our analysis demonstrates that there is an association between MAP and CD, across many sites, by many investigators, and controlling for a number of factors; however, this association remains controversial and inconclusive. Future studies should determine whether there is a pathogenic role.
The histopathological and clinical similarities between human chronic granulomatous enteritis,1, 2 intestinal tuberculosis, and animal paratuberculosis3 has led to the suspicion that Mycobacterium avium subspecies paratuberculosis (MAP) has a role in the etiology of Crohn's disease (CD). The literature on the pathogenesis, pathophysiology, and clinical management of CD has evolved considerably over the last decade due to advances in molecular biology and genetics. Despite considerable medical debate, the etiology of CD remains an enigma. However, there is a consensus that CD results from a sustained immune response arising from an environmental stimulus in a susceptible host.4 MAP has been postulated to be a potential stimulus at least in a subset of patients.5
Unlike Mycobacterium tuberculosis, MAP is fastidious and slow-growing, with nutritional dependence for the iron chelating compound mycobactin.5 In vitro culture techniques have improved significantly in the last decade. Grant et al6 compared Mycobacterium Growth Indicator Tube (MGIT) and BACTEC culture media for the recovery of MAP from milk and found both media were able to detect 10 organisms per mL of milk. However, despite these improvements in culture media it remains a difficult organism to isolate from humans, for unclear reasons that may include mycobacterial dormancy or low bacillary burden. Similarly, serological assays are unlikely to provide the evidence required, as they are not reliable for the diagnosis of Johne's disease, a disease of animals caused by MAP.7
A number of studies have reported the more frequent detection of the insertion sequence IS900 in CD patients compared with controls using polymerase chain reaction (PCR) assays.5, 8 In contrast, other authors found no association between CD and the detection of IS900.9, 10 The 2 main methods that have been used include PCR and in situ hybridization (ISH) assays. Although other IS900-like elements have been described in environmental mycobacteria,11, 12 the entire IS900 gene is unique to MAP.13 The short fragments of the IS900 sequence targeted in various studies may not be unique, as evidence of partial cross-hybridization between M. avium strains has been reported in AIDS patients.14 Multiplex PCR simultaneously targeting several fragments of the IS900 gene has been developed to overcome this difficulty.15 In addition, crosshybridizing elements can be distinguished from IS900 by optimized sample processing and through the design of primer pairs.16
Various reviews have been conducted examining the MAP hypothesis that suggest that although there is insufficient evidence, a causal link in a subset of CD patients cannot be excluded.17 There has, however, been no formal systematic review and meta-analysis assessing all the studies using PCR and ISH assays. We present a systematic review of all published studies and, where appropriate, meta-analysis of studies using PCR and ISH to detect MAP in patients with CD and controls.
- Top of page
- MATERIALS AND METHODS
The results suggest that there is sufficient evidence for the presence of MAP in the gut of patients with CD. A summary risk difference of 0.23 (95% CI, 0.14–0.32) was calculated, which remained consistent after sensitivity analysis excluding various sets of studies. Earlier studies have suggested that MAP may also be found among controls but recent investigations using appropriate extraction techniques and more sensitive assays have revealed a significantly higher level of detection by PCR assays in CD patients compared with controls.16, 20, 27, 28, 29
Interpreting the Findings
The pathogenic role of MAP in CD is uncertain. Our analysis reveals a high strength of association with a reasonable level of consistency among recently conducted studies. However, for any agent to be considered a cause of a disease, evidence that the infection precedes the pathology is essential. The studies in this review do not provide evidence of temporality in the association.
The strength of the association and consistent detection of MAP DNA does not prove causation, and may arise from prior ingestion of nonviable organisms in water or milk. However, the detection of MAP DNA in the blood of CD patients8 may suggest that a viable form of the organism is present in CD. This study by Naser et al did not report the detection of MAP in intestinal tissue, nor explore the possibility that the results are a secondary phenomenon due to increased gut permeability or the inability of macrophages in CD to kill MAP. The increased gut leakiness hypothesis is supported by the more frequent detection of other organisms in CD patients compared with controls30, 31 and suggest a lack of specificity of MAP detection. For instance, the risk difference for Bacteriodes and Escherichia coli from the Fujita et al31 study was 0.5 and 0.4, respectively. In contrast, the frequent detection of MAP in CD compared with UC argues against increased gut permeability as the sole explanation.
Subgroup analyses were undertaken since IS900-based ISH has been discredited as a nonspecific tool.32 Therefore, the results of PCR assays were considered more reliable. Similarly, the analysis was restricted to those studies investigating intestinal tissue. This did not alter the findings.
It has been suggested that most of the positive results have been reported by a limited number of laboratories.33 However, data from this review indicates that positive results have been observed by several laboratories in different parts of the world. Low assay sensitivity used by certain laboratories may contribute to this observation, whereby only groups that have developed sufficiently sensitive tests are successful. Conversely, many studies detect MAP DNA in the control patients. This may be because MAP is potentially common in the environment. Alternatively, this may suggest that it is not pathogenic to all exposed individuals.
Strengths and Weaknesses
A comprehensive search strategy was used in this review. However, despite this, significant evidence of publication bias was found, suggesting that not all studies were identified.
In addition to presenting summary measures alone, meta-regression analysis was used to explore heterogeneity. Factors included in the meta-regression model were agreed a priori. However, limited availability of data hampered the exploration of sources of heterogeneity, for example, the effect of factors such as contamination in the meta-regression analysis. Nevertheless, this review investigated important factors related to the variation in the results of published studies.
This review does not systematically identify and critique other studies that may be important in the understanding of the role of MAP in CD such as the use of antimycobacterial therapy. This is the subject of a Cochrane review34 that showed that antibiotics may have a role in maintaining remission. The results of the largest randomized controlled trial to date does not support the use of antibiotics in CD.35
The aim of a review should not be limited to producing a summary estimate of divergent results but also to explore the reasons for the observed differences. Heterogeneity is inevitable in a meta-analysis of observational studies. Furthermore, the studies were carried out over a long time period during which laboratory methods evolved. A significant positive association with year of study with more recent studies more likely to have positive results was identified. This is very likely a proxy measure of changes over time in the assay techniques. The sensitivity of assays improved over the last 2 decades due to a combination of better extraction methods and enhanced detection limits. A significant positive association was observed when studies including children were compared with those with an adult population. It has been suggested that if the initial MAP infection occurs in childhood, its detection in studies among children25 would be more likely. This implies that earlier in the course of the disease, MAP is present. However, others have found no difference in the detection of MAP based on age using PCR.36 Other factors considered a priori to be potential causes of heterogeneity were not statistically significant.
Due to the hypothesized paucibacillary nature of MAP infection in CD, it is possible that there are very low numbers of infecting organisms, below the detection limits of the assay. Limiting the studies in the meta-analysis to those that have detected MAP in at least 1 sample either from a patient with CD, UC, or a control did not alter the results.
The exclusion of the study with a different target sequence23 did not alter the results. A number of studies included immunosuppressed patients. The results of this review showed no association with reported use of immunosuppressive agents in the study population; however, the majority of studies did not explicitly look for an association. Authors reporting negatives results have argued that contamination is a major factor in the detection of MAP.24 Unfortunately, due to variation in techniques for reducing contamination and limitations of the data available in published articles, this factor was not assessed. The use of appropriate controls and the meticulous steps taken by most of the studies suggest that this is unlikely to be the explanation for the detection of MAP.
There are other factors that may be associated with differences in the ability of the studies to detect MAP that we did not investigate. These include limitations of sampling techniques, failure to process fresh tissue samples (fixing results in fragmentation of DNA), and the role of concomitant use of antimicrobial drugs. Tissues from operative procedures provide more quantity and reduce sampling errors, but it is unclear whether patients with early lesions are more likely to have MAP compared with those that have developed complications requiring surgery.37 Sampling error due either to low abundance of MAP in tissues or lower numbers in nongranulomatous tissue has been suggested as an explanation for studies failing to detect MAP. Laser capture microdissection techniques have been used to specifically isolate granulomas that were subsequently assayed, resulting in the detection of MAP more frequently in CD patients compared to controls.38 However, the authors of the same report detected the DNA of other bacteria more frequently in CD tissue compared to controls using the same technique.39 Furthermore, other studies24, 36 that include intestinal granulomas and lymph nodes failed to detect MAP, while some detected the organism in various sites including noninflamed tissue.41 This meta-analysis did not show that investigating granulomas is a significant source of heterogeneity. Considerable heterogeneity remained after controlling for the effect of all the factors investigated.
It is unlikely that similar studies using PCR assays are going to change the current state of evidence. Our analysis demonstrates that there is an association between MAP and CD, across many sites, by many investigators, and controlling for a number of factors. While the summary effect measures from this study show an association between CD and MAP, this association remains controversial and inconclusive. Future studies should determine whether there is a pathogenic role. Such studies should also assess the evidence of temporality and specificity of the association.