Introduction: why should not all patients with Crohn's disease receive the same therapy?

Personalised use of therapeutic options has many facets: it may imply an individualised selection of medication, also the decision on a more or less intense follow-up for a given patient, and more recently, individual adaptation of therapeutic dosing by serum drug monitoring was added to the personalised management of Crohn's disease (CD).

The main reason for not applying one and the same management for CD is the complexity of phenotypes: the extent, localisation and disease severity may vary substantially from patient to patient, and may change over time in a given patient. This means that the available drugs will only have their optimal effect when used in the right indication and patient population.

Nevertheless there are some common recommendations when it comes to managing patients with CD. First, it is important to get a complete map of the disease before any therapy is started. Magnetic resonance (MR) enterography and ileocolonoscopy have become the gold standard for assessing small and large bowel and MR imaging (MRI) of the pelvis is the method of choice for the assessment of perianal CD. Second, the goal of all treatments should be remission and mucosal healing without steroids, and third, this goal should be achieved in a timely manner. Recent clinical trials in CD have clearly illustrated that treatment with anti-tumour necrosis factor (TNF) agents in combination with azathioprine early in the disease course results in higher steroid-free remission rates and higher mucosal healing rates.1 Nevertheless, intolerance, contraindications, lack of response, interest and timing of pregnancy, and concerns around infection and/or malignancy risks means that the choice of treatment is not always as easy as it seems.

Individualised choice of treatment: prognostic factors

Once a diagnosis of CD is made, the best therapeutic strategy needs to be chosen to reach the aforementioned goals. This can go from no treatment in rare cases, to one induction course of steroids in patients with very mild disease, to maintenance medical treatment in most patients, and even primary surgery in patients with very limited stenotic disease and/or complications.2

Medical maintenance treatment can imply an accelerated step-up approach or a more proactive top-down strategy. A number of prognostic factors have been proposed to aid in this decision process. Clinical factors that have been associated with more aggressive CD include young age at diagnosis, the presence of deep (colonic) ulcers, extensive small bowel involvement, significant weight loss and perianal disease3––8 (table 1). In two independent studies using population-based cohorts from New Zealand and Olmsted County, perianal involvement at diagnosis was confirmed as predictive of complicated disease behaviour and of more rapid progression of CD.7 ,8 In the original study by Beaugerie et al,3 the necessity to start steroids at diagnosis was also indicative of a disabling disease course in the subsequent 5 years, but this decision highly depends on the physician and on regional habits. In contrast, an older age at diagnosis and absence of rectal (and perianal) involvement would indicate more mild disease6 (table 1).

Although several of these factors have been confirmed in retrospective studies, patients often have a combination of mild and aggressive prognostic factors, which complicates the therapeutic decision. Also, most of the parameters published about disabling disease included patients attending specialty inflammatory bowel disease centres who may not be representative of a typical patient, and therefore all patients with disabling features may not require combination therapy.

Molecular prognostic markers: hope or reality?

Molecular (genetic, serological, immunological and/or tissue) markers have been the subject of study but have not yet found a place in aiding the therapeutic decision process, in contrast to other fields such as oncology, in which molecular markers have demonstrated clinical utility in predicting response to chemotherapy. So is the response to cetuximab, a monoclonal antibody to epidermal growth factor receptor in metastatic colorectal cancer, influenced by the KRAS mutation status, as the benefit of cetuximab seems limited to patients with KRAS wild-type tumour?9 Likewise, germline mutations may also correlate with clinical outcome to chemotherapy. A subanalysis of a large phase III study with bevacizumab (Avastin), a humanised monoclonal antibody to vascular endothelial growth factor (VEGF), in metastatic pancreatic cancer showed that overall survival and progression-free survival were influenced by single nucleotide polymorphisms (SNPs) in the tyrosine kinase domain of the VEGF receptor 1.10 Genome-wide association study (GWAS) technology has been successfully applied to identify genetic risk variants for flucloxacillin-induced hepatotoxicity. With only 51 cases of flucloxacillin-induced liver injury and 282 controls, a clear association was seen for the major histocompatibility complex region at rs2395029, a marker in linkage disequilibrium with HLA-B*5701.11 Likewise, a GWAS in 85 subjects with statin-induced myopathy and 90 controls identified common variants in SLCO1B1 that were strongly associated with an increased risk of statin-induced myopathy.12

In CD, NOD2/CARD15 remains not only the gene with the largest effect size but also with the most convincingly replicated association with disease phenotypes. Mutations have been associated with terminal ileitis, with a shorter time to onset of structuring disease and the need for surgery.13–15 Specifically, Seiderer et al16 demonstrated a strong risk increase for 1007 fs homozygotes for ileal stenoses and risk for surgery. The same group confirmed these findings in a prospective study.17 In a large Dutch cohort, Weersma et al18 described a higher OR for more severe CD, a greater need for surgery and a younger age at onset with increasing number of risk variants.

Why have genetic studies in CD not been more successful in translating the identified mutations to clinical applications? The SNPs tested are mostly known susceptibility markers for IBD compared with healthy controls, and might thus be more generally applicable to all patients with IBD and less useful to differentiate patients. It is also hypothesized that CD genetics consists of disease susceptibility genes/loci on the one hand, and disease-modifying genes/loci on the other. Work from the international IBD genetics consortium underscores this idea: reanalysis of GWAS data in function of disease behaviour (mild versus aggressive disease) identified a number of SNPs that specifically predisposed a person to a more aggressive disease course in CD. However, the same SNPs were not associated with the disease overall in the original GWAS.19

For serology, higher antimicrobial (OmpC, anti-flagellin cBir1, anti Pseudomonas i2) and/or antifungal (anti saccharomyces cerevisiae antibodies (ASCA), anti laminaribioside antibodies (ALCA), anti mannobioside antibodies (AMCA), anti chitobioside antibodies (ACCA)) antibody titres have been found in patients with fistulising and/or stenosing CD.20–22 Also these markers have not been widely implemented in clinical practice given their limited availability, the fact that they are expensive and above all the lack of prospective studies showing the stability of these markers. The observed increased antibody responses in patients with CD with longer disease duration highlight the importance of prospective longitudinal studies.22 One relatively small paediatric longitudinal study reported a relation between the presence and titre of antibodies (ASCA, anti-OmpC, anti-I2, anti-CBir1) at a time when no fistula or strictures were clinically apparent, and the subsequent development of complicated disease behaviour during a median follow-up of 18 months.23 Similar conclusions were drawn by Rieder et al24 in an adult population.24

In oncology, gene expression profiling has been successfully used to identify transcriptional signatures that predict several aspects of disease behaviour, including risk of metastasis and response to chemotherapy.25 ,26 These gene expression-based biomarkers have also been translated into clinical practice. In autoimmune and inflammatory conditions, such techniques have generally not detected signatures with equivalent prognostic utility. We studied colonic mucosal gene expression to provide a predictive response signature for infliximab (IFX) treatment in ulcerative colitis. A combination of five differentially expressed genes (osteoprotegerin, stanniocalcin 1, prostaglandin-endoperoxide synthase 2, interleukin (IL)-13 receptor α2 and IL-11) separated responders from non-responders with 95% sensitivity and 85% specificity.27

Researchers from the University of Cambridge reported a transcriptional signature in separated CD8 T cells which predicted prognosis in patients with CD.28 Interestingly, the same signature was previously also found to predict disease prognosis in systemic lupus erythematosus and in patients with antineutrophil cytoplasmic antibody associated vasculitis.29 There was a higher incidence of relapsing disease in the subgroup of patients who had an elevated expression of genes involved in antigen-dependent T-cell responses, including signalling initiated by IL-7 and T-cell receptor ligation. The fact that this signature could be found across several autoimmune and immune-mediated diseases suggests that, although being distinct autoimmune and inflammatory conditions, the course of these diseases may be influenced by common pathways. These subgroups, which can be identified by measuring expression of only three genes, raise the prospect of individualised therapy and suggest new potential therapeutic targets in autoimmunity. A prospective study to confirm these findings is ongoing.

Optimising treatment: role of measuring drug levels and metabolites

When to measure metabolites and drug levels?

In the absence of studies showing that drug-level monitoring and adaptation based on the levels leads to improved outcomes and/or reduced costs, routine measurements of TLs and ADAs cannot be recommended yet (figure 1). A post hoc analysis of ACCENT1 suggests that measuring TLs early after induction has a prognostic value.43 The authors re-analysed the sustained response rates in function of week 14 serum IFX TLs and showed that the median week 14 IFX TL was 4.0 μg/ml in the sustained responders compared with only 1.9 μg/ml in patients without sustained response (p=0.0331). A week 14 IFX TL of 3.5 μg/ml discriminated best the sustained response with sensitivity, specificity, positive and negative predictive values of 54%, 72%, 39% and 82%, respectively. We previously also showed that in patients who develop significant antibodies to infliximab (ATIs) during follow-up, the IFX levels 4 weeks after the first infusion were significantly lower than in patients who did not develop ATIs on follow-up.44 Therefore, it may be proposed that obtaining a TL early after induction and optimising early on if needed can avoid development of antibodies with subsequent loss of response.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Proposed algorithm of when to perform anti-tumour necrosis factor (TNF) drug monitoring. IBS, irritable bowel syndrome.

Definitively, when patients treated with anti-TNF are reporting symptoms and signs of inflammation suggestive of loss of response, determining TLs, and when no TLs are detected measuring ADAs, is helpful. TLs should always be measured just before the next dose (just prior to starting the next infusion). As proposed in figure 1, patients with undetectable TLs and low ADAs should be optimised with the same drug. There does not seem to be a preferred way of dose optimisation and decreasing the interval between infusions, increasing the dose, or adding an immunosuppressive drug have all been associated with increasing levels and recapturing response.45 A recent study reported that, in patients with detectable antibodies to IFX, adding an immunosuppressant decreased the antibodies and restored IFX TLs and clinical response.46 Patients with undetectable TLs and high ADAs, however, should be switched to a drug within the same class. In the case of symptoms with adequate TLs, endoscopy or MR enterography should be checked, and if this confirms significant lesions, therapy should be stopped and a switch to a drug with a different mode of action, if at all possible, is the preferred option. However, a wide implementation of TL and ADA determinations in clinical practice requires a rapid assay that can be performed at home by the patient or at the clinic or infusion centre. Meanwhile, the available assays need to be standardised and validated. There is also a need to improve current assays within economic margins.47