CLINICAL, ENVIRONMENTAL, AND ENDOSCOPIC FACTORS
Clinical and endoscopic variables are important medium and long-term markers in the prediction of complicated disease course in IBD. A young age at onset or pediatric presentation is an important risk factor for developing complicated disease behavior, and even disease extension in a large proportion of patients, as reported in a recent French population based-cohort study[6,7].
In CD, an initial need for steroid use (OR: 3.1, 95% CI: 2.2-4.4), an age below 40 years (OR: 2.1, 95% CI: 1.3-3.6), and the presence of perianal disease (OR: 1.8, 95% CI: 1.2-2.8) were associated with the development of disabling disease in the study by Beaugerie et al[8]. The positive predictive value of disabling disease in patients with 2 and 3 predictive factors of disabling disease was 0.91 and 0.93, respectively. More recently, authors from New Zealand[3] have shown that > 70% percent of CD patients had inflammatory disease at diagnosis, with the proportion of patients with complicated disease increasing over time. Progression towards complicated disease was more rapid in those with small bowel than colonic disease location, (P < 0.001), and perianal disease was a significant predictor of change in CD behavior [hazard ratio (HR): 1.62, P < 0.001]. Similarly, perianal lesions, the need for steroids to treat the first flare and ileo-colonic location, but not age below 40 years were confirmed as predictive markers for developing disabling disease (according to the predefined criteria) at 5 years[9]. In the same study, stricturing behaviour (HR: 2.11, 95% CI: 1.39-3.20) and weight loss (> 5 kg) (HR: 1.67, 95% CI: 1.14-2.45) at diagnosis were independently associated with the time to development of severe disease. In addition, small bowel location and stricturing disease were predictors for surgery in a long term follow-up study[10]. Finally, terminal ileal location (P < 0.001), stricturing (P = 0.004), penetrating behavior (P < 0.001), and age younger than 40 years (P = 0.03) at diagnosis were independent risk factors for subsequent surgery in a prospective 10-year follow-up study by the IBSEN group[11].
An important environmental factor that may possibly be important in determining changes in disease behavior in CD is smoking. In CD, smoking was reported to be associated with disease location: most, but not all, studies report a higher prevalence of ileal disease and a lower prevalence of colonic involvement in smokers[12,13]. A recent review[13] and previous data have demonstrated that smoking, when measured up to the time-point of disease behavior classification, was associated more frequently with complicated disease, penetrating intestinal complications[12,14,15], and greater likelihood to progress to complicated disease, as defined by the development of strictures or fistulae[13], and a higher relapse rate[16]. In addition, the risk of surgery as well as the risk for further resections during disease course was also noted to be higher in smokers in some studies[12,17] and in a recent meta-analysis[18]. The need for steroids and immunosuppressants was found to be higher in smokers compared to non-smokers[19]. In a recent paper by Aldhous et al[20] using the Montreal classification, the harmful effect of smoking was only partially confirmed. Although current smoking was associated with less colonic disease, smoking habits at diagnosis were not associated with time to development of stricturing disease, internal penetrating disease, perianal penetrating disease, or time to first surgery.
In one study by Cosnes et al[21], immunosuppressive therapy was found to neutralize the effect of smoking on the need for surgery. The importance of an early aggressive medical strategy in determining long term disease phenotype is further supported by a recent Hungarian study[22,23] where the authors have reported that early azathioprine (AZA)/biological therapy reduced the risk for disease behavior change and delayed the time to the first operation in CD in both smokers and non-smokers. The most convincing data to support a benefit from early use of AZA however, comes from the pediatric literature[24], where in a randomized controlled trial in 55 children, the early use of 6-mercaptopurine (6-MP) was associated with a significantly lower relapse rate (only 9%) compared with 47% in controls (P = 0.007). Moreover, the duration of steroid use was shorter (P < 0.001) and the cumulative steroid dose was lower at 6, 12, and 18 mo (P < 0.01).
In addition, in a recent withdrawal study by the Groupe D’Etude Thérapeutique Des Affections Inflammatoires Du Tube Digestif (GETAID) group[25], authors have provided evidence for a benefit of long term AZA therapy beyond 5 years in patients with prolonged clinical remission. Finally, in a recent controlled randomized prospective trial[26], AZA administered for 12 mo together with metronidazole for 3 mo was more effective in preventing endoscopic postoperative recurrence assessed at 12 mo, compared to metronidazole alone in patients previously only minimally exposed to AZA.
Fewer factors are available in ulcerative colitis (UC). One of the important factors is disease extent, identified by previous studies and also a recent 10 years population-based inception cohort study[27]. In the same study, an age at the time of diagnosis > 50 years was associated with reduced hazard ratio (0.28) for subsequent colectomy. In addition, one-fifth (69/288) of patients with proctitis or left-sided colitis had progressed to extensive colitis. In contrast, pediatric UC was more aggressive. Disease course was characterized by disease extension in 49% of patients during in a 6.5 years population-based cohort study[7]. A delay in diagnosis of more than 6 mo and a family history of IBD were associated with an increased risk of disease extension, with odds ratios of 5.0 (1.2-21.5) and 11.8 (1.3-111.3), respectively. The cumulative rate of colectomy was 8% at 1 year, 15% at 3 years, and 20% at 5 years. Of note, however, surgical rates have shown a wide geographical variation with significantly lower rates reported from the Mediterranean region and Eastern Europe in earlier studies[22,28].
Furthermore, in contrast to CD, smokers with adult UC have been reported to run a more benign disease course compared to non-smokers[29]. Flare-ups, hospitalization rates, the need for oral steroids, and colectomy rates, were reported to be lower, while age at onset was older in smokers compared to non-smokers, though not in all studies. Relapse rates were lower in patients who began smoking after the diagnosis of UC was made. In concordance, in a recent Europe-wide population-based cohort[30], the relapse rate was lower (HR: 0.8, 95% CI: 0.6-0.9) in smokers compared to non-smokers, while being higher in women. In addition, in a meta-analysis of several large series with a total of 1489 UC patients, the risk for colectomy was lower (OR: 0.57, 95% CI: 0.38-0.85) in current smokers compared to non-smokers[31].
Finally, it is still unclear whether mucosal healing prevents the development of complications, although its importance was investigated in several studies using heterogeneous definitions. A well-known example is the study by Allez et al[32] where severe endoscopic activity was associated with an increased risk for colectomy in CD after 1, 3 and 8 years of follow-up. This was only partially confirmed in a more recent population-based follow-up study by the IBSEN group, where mucosal healing at 1 year in UC was associated with a lower subsequent colectomy risk, but mucosal healing in CD was not a significant predictor for resective surgery at 5 years, but the risk of future steroid use was decreased[33]. Similarly, in a very recent follow-up report the beneficial effect of early complete mucosal healing was confirmed in early CD. Complete mucosal healing in patients with early-stage CD was associated with significantly higher steroid-free remission rates 4 years after the start of the therapy[34]. Similarly, mucosal healing at 6 mo following the start of scheduled infliximab maintenance therapy was associated with lower need for major abdominal surgery during follow-up in a study from the Leuven group[35].
BIOMARKERS
Biomarkers can be divided into long-term markers (e.g. genetics and serology) influencing the clinical phenotype and short-term markers that may be used for the prediction of flares.
In 2001, the identification of nucleotide oligomerization domain2 (NOD2)/caspase activation recruitment domain15 (CARD15) on chromosome 16 in the IBD1 region, as a candidate gene for CD[36,37], has stimulated further research focusing on novel genetic determinants of susceptibility and phenotype in IBD. The presence of a NOD2/CARD15 mutation increases the risk for CD by 1.4-4.3-fold in heterozygous patients and 17.6-44.0-fold in homozygous and compound heterozygous patients. Of importance, reports exist of homozygous individuals who are disease-free. However, various geographical differences were noted. A much lower prevalence of these mutations was reported in, for example African American, Chinese and Japanese[38] populations. NOD2/CARD15 is a good example for case-control phenotype-genotype correlations, since in earlier studies the three common NOD2/CARD15 mutations were associated with ileal disease and fibrostenosing behavior. On the contrary, in colonic and fistulizing disease, they were relatively less frequent[39,40]. There are many controversial findings, such as the association between the OCTN TC risk haplotype and penetrating disease (OR: 1.474, 95% CI: 1.028-2.114, P = 0.035) in one study[41] or non-fistulizing non-fibrostenotic disease (OR: 1.57), colonic disease (OR: 1.31), an earlier age at the onset of disease, and reduced need for surgery (OR: 1.38) in another[42], or the association between the presence of DLG5 R30Q variant and steroid refractory disease that could not be confirmed yet[43]. Finally, in a recent study by Weersma et al[44] the same type of statistical analysis was performed in a little more complex genetic dataset by using both novel markers (e.g. ATG16L1, IL23R) and earlier (NOD2, IBD5 and DLG5) genetic markers. In this study, patients with CD with a more severe disease course, surgeries or an age of onset below 40 years had more risk alleles compared to non-stricturing, nonpenetrating behaviour (P = 0.0008), no operations (P = 0.02) or age at onset greater then 40 years (P = 0.028). However, the association became insignificant after a longer disease duration, confirming that many other factors (e.g. clinical variables, disease phenotype at diagnosis, medical therapy) contribute to the long term evolution of disease phenotype.
Although in some of these studies the authors corrected their findings for disease duration using logistic regression, these are all static correlations between phenotype and genotype. In 2009, we finally observed new examples of genotype-phenotype association scans, where the authors have performed a complete follow-up of the patients and so the association between time-dependent clinical outcomes and genotype could be more preciously evaluated in multivariate analysis using Cox regression. Moreover, the authors realized that clinical parameters should also be included in the complex testing. A fine example of how this should be performed was recently published by the Leuven group[45]. Of course, the presentation of the results is now much more complex. For example, homozygosity for the rs1363670 G-allele in a gene encoding a hypothetical protein near the IL12B gene was independently associated with stricturing disease behavior and with shorter time to strictures, especially in patients with ileal involvement, or alternatively male patients carrying at least one rs12704036 T-allele in a gene desert had the shortest time to non-perianal fistula. This landmark study with its complex analysis is, however, bringing genetics closer to being used in routine clinical practice.
Genetic markers may also be used to predict the efficacy and/or side effects of medical therapy, that is pharmacogenetics. This use is also expected to be complex and include certain clinical predictors. The importance of using conventional genetic testing might be highlighted by the well-known thiopurine S-methyltransferase (TPMT) genotype/activity testing. TPMT limits the production of 6-TGNs by converting 6-MP to 6-thioruric acid and 6-methylmercaptopurine[46], which may lead to myelotoxicity. Population studies have shown that the distribution of TPMT activity is trimodal: 0.3%-0.5% of the population has low to absent activity (TPMTL/TPMTL), around 10% have intermediate activity (TPMTL/TPMTH), and approximately 90% inherit normal to high enzyme activity (TPMTH/TPMTH)[46]. In this regard, a correlation between the TPMT genotype and enzyme activity has been proven. Approximately 5% of the white population carries one or more variant TPMT alleles, with more than ten variant alleles reported[47]. The functional consequences of alleles *2, *2A, *3B, and *3C, accounting together for more than 90% of mutant alleles, have been extensively characterized. Nevertheless, it is clear that there are many other causes of myelotoxicity. This was accurately demonstrated by Colombel et al[48], who found that only 27% of patients had a documented low TPMT activity. Other confounding genetic and environmental factors include, for instance, the patient’s age, renal function, AZA formulation, coadministration of mesalazine (a reversible TPMT inhibitor) and allopurinol (XO inhibitor). Thus, the determination of TPMT activity is not an exclusive test to rely on. It may only be helpful in identifying a certain group of high risk patients but as the negative predictive value is rather low, it is not beneficial in ruling out possible side effects. Also, as the prevalence of double carriage of variant TPMT alleles is as low as 1/300, continuous monitoring of red blood cell counts remains mandatory in clinical practice. A different approach was recently reported by the Leuven group[49]. The authors analysed gene expression signatures in the mucosa of patients with UC and were able to identify predictive panels of genes for (non-)response to infliximab. Further studies of the pathways involved should allow a better understanding of the mechanisms of resistance to infliximab therapy in UC.
Similarly, serological markers represent medium to long term markers of disease phenotype and progression. The phenotype and disease course of CD are associated with the presence and extent of the serologic response targeted against various microbial antigens. In patients with conventional anti-Saccharomyces cerevisiae antibodies (ASCA+, IgG and/or IgA) or atypical perinuclear antineutrophil cytoplasmic antibodies (P-ANCA-) phenotype, small bowel involvement (with or without colonic disease) is more typical than the pure colonic disease (60%-80% vs 30%-45%)[50,51]. ASCA positivity predicts a more aggressive disease course with a higher rate of complications. ASCAs have been associated with stricturing and penetrating disease and a higher risk of small bowel resection[51,52]. Some studies also suggest that ASCA positivity is associated with an earlier onset of disease. The presence of anti-OmpC in adult CD patients is associated with an increased prevalence of the penetrating form only[53], while in children, both the penetrating and stenosing forms[54] are more frequent. Moreover, antibody positivity may lead to a more aggressive course of disease and a higher risk for surgical interventions. Like ASCA and anti-OmpC, anti-I2 also appears to be associated with an increased risk for complications in adult CD patients. It is an independent risk factor for the development of the stenosing form and the need for surgical interventions[54]. Recent research has shown that the anti-CBir1 antibody is associated with ileal involvement in adult CD patients independently from other serological markers, and it predisposes for the development of both stenosing and penetrating forms[55]. Interestingly anti-Cbir1 was more frequent in young children with CD (age: 0-7 years)[56]. In a very recent study, an association was also reported between the presence of pancreatic autoantibodies and perforating perianal disease, arthritis, ocular, and cutaneous manifestations[57]. In addition to qualitative correlations, quantitative correlations with serological responses are also present. The number of antibodies produced against microbial antigens in CD also showed a positive correlation with the severity of the disease course[51].
Similarly, the presence and magnitude of the more newly developed antiglycan antibodies (including the most recent anti-laminarin and anti-chitin markers) in CD was associated with a more aggressive disease phenotype in CD, as well as a younger age at onset, ileal involvement, stricturing/penetrating disease behavior, and risk for surgery[58-60]. In the latter study, gASCA IgG and ASCA IgG were comparable for the diagnosis and determination of disease phenotype in CD; however, each test recognized an additional patient population that was missed by the other, indicating a partially complementary role. The differentiation between colon only CD and UC was also better[58,61].
In prospective pediatric cohorts Dubinsky et al[54], found that the presence and magnitude of immune responses to microbial antigens (ASCA, anti-OmpC, anti-I2 and anti-CBir1) are significantly associated with more aggressive disease phenotype. The risk of developing penetrating and/or stricturing CD was increased in those individuals with immune response against all four microbial antigens compared to seronegative cases. Moreover, they demonstrated that the time to develop a disease complication during the 18 mo of follow-up period is significantly faster in those children who have a serologic response against at least one antigen, recently confirmed by the same group in a much larger IBD population[62].
In addition, the occurrence of atypical P-ANCA in UC is associated with a characteristic clinical appearance and represents a distinct subgroup, which is often characterized by specific human leucocyte antigen markers. These patients have a higher probability of developing a severe left-sided UC that is more resistant to treatment than the usual case. The disease has a more aggressive course requiring surgery earlier in the course of the disease[63]. The presence of atypical P-ANCA identifies a subgroup of CD patients characterized by “UC-like” colitis; the inflammation usually involves the left side of the colon and the response to therapy is generally good. The atypical P-ANCA in CD patients is associated with a later age of onset and a relatively smaller incidence of complications, such as stricture and/or perforation[64].
Of note however, one should be aware of the fact that the serology profile is not a stable phenotype; seroreactivity may change during the course of follow-up. This is at least suggested by studies done in patients with celiac disease where seropositivity was lost several months after the introduction of a gluten free diet[65].
IBD follows an alternating disease course and both CD and UC are characterized by periods of remission and relapse. However, disease flare-ups occur in a random way and are often unpredictable. Based on the half-life and changes over time serum laboratory and fecal markers may be best used for the prediction of short or medium term relapses. Erythrocyte sedimentation rate (ESR) and C reactive protein (CRP) are traditional non-specific markers of inflammation. In addition, a number of neutrophil-derived proteins present in stool samples have been studied, including fecal lactoferrin, lysozyme, elastase, myeloperoxidase, and calprotectin.
ESR is defined as the rate at which erythrocytes migrate through the plasma. Inevitably, ESR will depend on the plasma concentration and on the number and size of the erythrocytes. Conditions such as anemia, polycythemia, and thalassemia affect ESR. Compared with CRP, ESR will peak much less rapidly and may also take several days to decrease, even if the patient’s clinical condition or the inflammation is ameliorated. Although CRP is non-specifically upregulated in most inflammatory diseases, including IBD, there is remarkable heterogeneity in the CRP response between CD and UC. Whereas CD is associated with a strong CRP response, UC has only a modest or no CRP response in both adult and children IBD populations[66]. This is an important fact to keep in mind when using CRP in clinical practice. There is no good explanation for this heterogeneity.
In general, patients with severe disease more often have abnormal inflammatory markers, compared with patients without or with only a low-grade inflammation. This has been shown in one of the early studies by Fagan et al[67], both CRP and ESR correlated well with disease activity, yet the correlation was better for CRP. In addition, the correlation of laboratory markers with disease activity has been shown to be much stronger for CD than for UC. However, a wide range of CRP values was observed and no clear cut-off values exist between mild to moderate (10-50 mg/L), moderate to severe (50-80 mg/L), and severe disease (> 80 mg/L). Therefore, the comparison of individual CRP values with previous values in any given patient is of great clinical importance. In addition, CRP showed acceptable correlation with endoscopic and histological activity in CD[68]. For UC, again, this correlation was weaker. In contrast, more recent data from Australia suggested that a subgroup of patients with active ileal disease, low body mass index, and previous ileocecal resection was characterized by low CRP[69].
In CD, a number of studies have investigated a panel of laboratory markers in predicting clinical relapse. Later, Boirivant et al[70] prospectively followed 101 outpatients with CD. Half of the patients had an elevated CRP value and this correlated well with clinical activity. Approximately one-third of CD patients presented with active disease despite normal CRP and one-third had high CRP levels but clinically inactive disease. The likelihood of relapse after 2 years was higher in the patients with an increased CRP compared with those who had a normal CRP value. More recently, the GETAID group[71] proposed a simple biological score for predicting short-term relapse in CD. Multivariate analysis selected two markers predictive of relapse: CRP > 20 mg/L and an ESR > 15 mm/h. The relative risk of short-term relapse for patients with a positive score compared to those with a negative score was 8.0 (95% CI: 2.8-22.9). The score’s sensitivity and specificity was 89% and 43%, respectively, with a negative predictive value of 97%, suggesting that normal CRP and ESR could almost certainly rule out a relapse in the next 6 wk. Much less data exist on the value of laboratory markers in assessing disease course and outcome in UC. The well known, prospective study from Oxford evaluated 49 severe UC patients treated with hydrocortisone and/or cyclosporine (n = 49). On day 3, a daily stool frequency of > 8 or 3-8 with increased CRP (> 45 mg/L) predicted with 85% certainty the need for colectomy[72]. Only one relatively small study investigated the role of hs-CRP in IBD[73]. Ninety CD, 70 UC patients, and 160 controls were investigated. The coefficient of correlation between hs-CRP and the disease activity score was weak but similar in both UC (0.26) and CD (0.36).
Finally, a recent population based study by the IBSEN group has demonstrated[74] that CRP at diagnosis or during follow-up was able to predict medium term outcome as measured by the need for colectomy or resective surgery. In patients with UC with extensive colitis, CRP levels above 23 mg/L at diagnosis predicted an increased risk of surgery (OR: 4.8, 95% CI: 1.5-15.1) during the first 5 years of follow-up. In patients with UC, CRP levels above 10 mg/L after one year predicted an increased risk of surgery during the subsequent 4 years (OR 3.0, 95% CI: 1.1-7.8). A significant association between CRP levels at diagnosis and risk of surgery was also found in patients with CD with terminal ileitis (L1), where the risk increased when CRP levels were above 53 mg/L in this subgroup (OR 6.0, 95% CI: 1.1-31.9).
A change in CRP following therapy serves as a good parameter in assessing the effectiveness of the drug on the underlying inflammation. A decrease in CRP, in response to therapy, is objective evidence that the drug has a beneficial effect on intestinal inflammation, even in patients with small changes in their symptoms. On the other hand, persistently raised CRP indicates therapeutic failure in controlling mucosal inflammation. This was clearly demonstrated by the different response rates to anti-TNF antibodies in patients with CD. In some studies, a high baseline (5-10 mg/L) CRP value before the start of therapy was associated with a higher response compared to patients with lower CRP[75,76]. These findings raise the question of whether CRP should be included in patient selection for future clinical trials, at least for selected drugs. Furthermore, high CRP (> 20 mg/L) was found to be an independent predictor for relapse after AZA withdrawal in patients on AZA therapy for longer than 42 and 63 mo[25,77].
Fecal calprotectin was also shown to predict relapse in CD. Calprotectin, a 36 kDa calcium- and zinc-binding protein, is probably the most promising marker for various reasons. In contrast with other neutrophil markers, calprotectin represents 60% of cytosolic proteins in granulocytes. The presence of calprotectin in feces can therefore be seen as directly proportional to neutrophil migration to the gastrointestinal tract. Fecal calprotectin is a very stable marker (stable for more than 1 wk at room temperature) and is resistant to degradation, making it an attractive option. Other neutrophil protein markers, include fecal lactoferrin, polymorphonuclear-elastase, and S100 A12. They are also more specific to intestinal rather than systemic inflammation, but their clinical role has been less extensively evaluated in IBD.
In the study by Tibble et al[78], calprotectin levels of 50 μg/g or more predicted a 13-fold increased risk for relapse. In another study[79], 38 CD and 41 UC patients were investigated. All patients were in remission for a mean duration of 5 mo. A baseline level of calprotectin of 150 μg/g or more was predictive for a relapse in the next year. Although sensitivity was high for both CD (87%) and UC (89%), specificity was much lower in the case of CD (43%) compared to UC (82%). Similarly, in a recent study, fecal calprotectin and lactoferrin were able to differentiate organic colorectal disease and were associated with active disease at endoscopy, both in UC (78% and 75%, respectively) and in CD (87% and 82%, respectively)[80]; however, specificity was low overall. It is thus difficult to define cut-off values based on these studies. In addition to serum markers, fecal calprotectin also correlates well with endoscopic and histological activity in patients with UC and CD, and elevated calprotectin levels normalize once the inflammation has resolved in both adults and children[81,82].
Lastly, calprotectin, at a cut-off value of 200 mg/L, at 3 mo post-surgery, was useful in predicting endoscopic post-surgical recurrence in asymptomatic patients with a sensitivity of 63% and a specificity of 75%[83].
