Evaluation of disease behaviour by genotype is complex and may change throughout the course of disease. Dissecting pediatric- and adult-onset CD represents a model to identify further associations between NOD2 gene variants and phenotypes and may help in selecting individual treatment regimes for patients with CD. We here demonstrate greater influence of NOD2 variants in pediatric-onset CD patients towards higher disease activity, risk for underweight, osteoporosis and more intensive immunosuppressive therapy. This is the first study identifying an influence of NOD2 variant alleles in patients with CD on osteoporosis and an age of onset dependent risk for steroid dependent or refractory course, long-term use of immunosuppressive drugs.
In this retrospective single center study in a tertiary center we report a high prevalence of CD patients carrying at least one NOD2 variant allele with 44% in the pediatric- and 42% in adult-onset CD similar to other European pediatric multicenter cohort studies with a prevalence of 35 up to 45.6% [12, 23, 24]. In addition, the presence of two NOD2 mutant alleles was found more frequent in the pediatric-onset CD patients as reported by others [23, 25] and may predict complicated disease .
Growth retardation is a frequent specific feature of pediatric-onset CD , which manifest as poor weight gain or short stature and may be the first presenting symptom. We observed a trend in CD patients with NOD2 variants towards underweight at diagnosis in our pediatric-onset cohort analogous to other studies [28–30] describing lower weight and a tendency to growth retardation more or less pronounced in their pediatric cohorts. In addition, we found also significant ongoing underweight in the course of disease in patients with NOD2 variant alleles. Growth failure seems to be associated with disease severity  and high levels of TNFα in pediatric IBD . Indeed, the average disease activity in pediatric-onset CD patients carrying at least one NOD2 variant allele was significantly higher and associated with chronic active or moderate to severe clinical course of disease. However, other studies did not found an influence of NOD2 variants towards higher disease activity [25, 29]. This is explicable by the fact, that in most instances disease activity was educed from hospitalization, need for immunosuppressive drugs, surgery or occurrence of complications but not by PCDAI at onset and during course of disease [25, 28, 29, 32]. The PCDAI-score was only used by one other pediatric study with 65 CD patients, which described an influence of NOD2 variants towards a higher PCDAI . Besides actual history and laboratory values (e.g. haematocrit, erythrocyte sedimentation rate and albumin) the PCDAI score includes also weight loss, growth delay, perirectal disease and extraintestinal manifestation. Therefore a high PCDAI score in our pediatric-onset CD patients with NOD2 variants reflects a more severe disease behaviour indicated by underweight, stricturing and penetrating behaviour. Moreover, an aggressive disease manifestation in pediatric-onset CD patients with NOD2 variants leading to complications and surgical resection is described by Lacher et al.  and Kugathasan et al. [32, 33].
The prevalence of osteoporosis ranges from 7 to 30% and osteopenia from 22 to 70% in pediatric IBD studies [13, 34, 35]. In fact, 30% of our patients with pediatric-onset CD have osteoporosis and 31% osteopenia, some even before corticosteroid treatment confirming data from Walther  and Harpavat et al. . Possible risk factors for pathologic bone density in CD are disease activity, lifetime steroid dosage over 10 g, CD, multiple bowel resections, young age and low body mass index according to a large epidemiological study . We also observed low BMI and a higher disease activity associated with reduced bone density in pediatric-onset CD. Similar Paganelli et al. described an inversely correlation of volumetric BMD with disease activity measured by PCDAI and IL-6 . Other risk factors such as ileal localization and bowel resection did not play a role in our pediatric-onset cohort. Lifetime corticosteroid treatment in children with IBD seems to be less important as previously believed, as children with a chronic non-inflammatory type of disease such as steroid sensitive nephritic syndrome (SSNS) receiving glucocorticoids chronically do have a better bone mineral density than children with CD . Certainly osteoporosis was associated with a steroid refractory or dependent course of disease in the pediatric-onset group. In fact the glucocorticoid dosage in the treatment of pediatric CD may be not as high as in previous older studies; children with CD often receive enteral nutrition therapy and there are more patients with steroid dependent or refractory course of disease in the low BMD group actually switched earlier to a long-term immunosuppressive therapy including azathioprine, methotrexate and anti-TNF-α antibody treatment. Thus the inflammatory state in CD might basically contribute to high-turn over bone mineral loss. It is also known that NOD2 variant alleles induce an inflammatory response e.g. via NF-кB activation . Thus some cytokines with osteolytic activity such as TNFα, IL-1β, IL-6, IFNγ, RANK, RANKL and osteoprotegrin may also be elevated. IL-1β for example stimulates osteoclast activity and enhances mineral bone loss. In fact carriers of IL1β -511b allele have significant lower BMD . Interestingly, CD patients carrying NOD2 variant alleles have an increased risk towards osteoporosis. This is more pronounced in pediatric-onset CD, especially in patients carrying homozygous NOD2 variant alleles. However, we could not prove our hypothesis that osteoporosis is significantly more frequent in pediatric-onset CD patients carrying any NOD2 variant allele may be due to a lack of power. Taken together, these data support screening of osteoporosis in pediatric onset CD, especially in children with NOD2 variants, low BMI and high PCDAI.
We recently reported that CD patients with NOD2 carrier status were more refractory for steroid treatment but responded well to immunomodulators such as AZA/6-MP . This was in contrast to another study that could not find an association between NOD2 carrier status or age of onset of disease and response to steroids . We speculated that treatment response might vary analyzing pediatric and adult CD patient study cohorts regarding NOD2 genotype. To corroborate this hypothesis we now included a pediatric CD cohort and reclassified all CD patients in our adult cohort according to the onset of disease in pediatric and adult onset. Using this approach we found a significant association for steroid failure in CD patients with NOD2 variant alleles and younger age of onset in contrast to Weiss et al. . In addition, the need for long-term immunosuppressive therapy and second line therapy with anti-TNF-alpha agents was also only associated with younger age of onset and the presence of any NOD2 variant allele. Although a positive response to the anti-TNFα agent infliximab was found to be associated with a higher systemic inflammation , the NOD2 genotype was not predictive of treatment outcome in their cohort . The diverging results regarding our study cohort might be due to the subgroup analysis of pediatric-onset CD. The fact that we did not investigate treatment outcome is another disadvantage of this study and its retrospective approach.