• Crohn's disease;
  • social support;
  • recurrence;
  • body mass index


  1. Top of page
  2. Abstract
  6. Acknowledgements


Social support has been found to be protective from adverse health effects of psychological stress. We hypothesized that higher social support would predict a more favorable course of Crohn's disease (CD) directly (main effect hypothesis) and via moderating other prognostic factors (buffer hypothesis).


Within a multicenter cohort study we observed 597 adults with CD for 18 months. We assessed social support using the ENRICHD Social Support Inventory. Flares, nonresponse to therapy, complications, and extraintestinal manifestations were recorded as a combined endpoint indicating disease deterioration. We controlled for several demographic, psychosocial, and clinical variables of potential prognostic importance. We used multivariate binary logistic regression to estimate the overall effect of social support on the odds of disease deterioration and to explore main and moderator effects of social support by probing interactions with other predictors.


The odds of disease deterioration decreased by 1.5 times (95% confidence interval [CI]: 1.2–1.9) for an increase of one standard deviation (SD) of social support. In case of low body mass index (BMI) (i.e., 1 SD below the mean or <19 kg/m2), the odds decreased by 1.8 times for an increase of 1 SD of social support. In case of low social support, the odds increased by 2.1 times for a decrease of 1 SD of BMI. Low BMI was not predictive under high social support.


The findings suggest that elevated social support may favorably affect the clinical course of CD, particularly in patients with low BMI. (Inflamm Bowel Dis 2011;)

Crohn's disease (CD) may affect the entire gastrointestinal tract with discontinuous lesions involving all bowel layers. An irregular disease course with active and inactive periods is characteristic. It is currently thought that the chronic intestinal inflammation results from an aberrant mucosal immune response to microorganisms of the gastrointestinal tract in genetically susceptible individuals. Recent research aiming to elucidate the role of psychological factors for the course of CD was inconsistent.1

Social support includes all aid, encouragement, and benevolence available through relationships and environment. Especially, perceived social support seems to affect health. Low-perceived social support has often been associated with adverse health outcomes, worse prognosis, or mortality, while high-perceived social support was found to be protective and related to better health outcomes and prognosis.2 Social support is thought to promote health either directly under every circumstance (main effect theory) or indirectly by moderating adverse stress effects (buffering hypothesis).3

The present study aimed to elucidate the effect of social support on adverse clinical events in the course of CD based on the following three rationales. First, the debate about main versus buffering effects of social support on health outcomes remains unresolved. For example, high social support has been shown to mitigate the stress response and thus prevent the adverse stress effects on the course of cardiovascular disease, posttraumatic stress disorder, and cancer.4–6 Second, to date hardly any articles focused on the question of whether social support would affect factors that are known to worsen the course of CD. Third, it is a novel aspect of this study to consider social support not merely as a stress modulator, but also as a potential modulator of biological factors contributing to the disease course.

We primarily hypothesized that high social support would lower the odds of adverse events, but also explored interactions between social support and other potential predictors of the course of CD. In case of meaningful interactions, we tested for moderational effects in order to substantiate evidence in favor or against the main effect or buffering hypotheses of social support.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Study Design and Patients

Between July 2006 and February 2008, patients with CD were recruited in the Swiss inflammatory bowel disease cohort through gastroenterologists.7 The sample was sizeable enough to detect a 1.5-fold increase of the odds of an adverse disease course (i.e., flares plus nonresponse to therapy) for a decrease of one standard deviation (SD) of social support with a power of 90%. This calculation followed Power and Precision software assuming initial odds of 0.3 with two-tailed level of significance of 0.01.8 To facilitate generalization of findings to the CD population regularly treated in Switzerland, we applied broad eligibility criteria, which were recurrent symptoms and a confirmed diagnosis of CD. The latter was made by applying Lennard-Jones criteria, which allow for a diagnosis of CD either by radiological, endoscopic, or histological findings, or by surgery.9 The Ethical Committees of the participating study sites approved the study protocol and each patient was enrolled after signing a written informed consent for participating in the study.

At enrolment, gastroenterologists and study nurses performed a thorough medical history and clinical exams at the various study centers (i.e., Basel, Bern, Geneva, Lausanne, St Gallen, and Zurich), as well as regional hospitals and private practices. Baseline characteristics included sex, age, disease duration, and the number of days of previous hospitalization(s) due to CD, as well as current medication, smoking status, and body mass index (BMI). Blood samples were collected to calculate the Crohn's Disease Activity Index (CDAI)10 in order to assess baseline disease activity (BDA). The reasons for selecting these control variables are explained below. Simultaneously, we supplied the participants with self-assessment questionnaires. These were used to assess social support, the main predictor, together with social inhibition (a personality trait) and social diversion (a coping style). Social inhibition and social diversion are—as described below—two potential moderators of social support. Since the time interval between enrolment and completion of those questionnaires was not constant among patients, it was helpful to control for it. After collecting baseline data, patients were followed during 18 months, which is 1.5 times the minimal observational period recommended by the European Crohn and Colitis Organization (ECCO) for prospective studies in inflammatory bowel disease (IBD).11 Baseline measurements were not repeated.

To minimize observer, information, and reporting bias and to optimize data quality, regular educational meetings were performed. During those meetings gastroenterologists and study nurses joined the members of the data center to report any difficulties related to data collection (unclear or too complicated questions to be answered in a reasonable time period). Members of the data center pointed out those parameters they thought particularly susceptible to cause lack of information. Case report forms were adjusted according to these feedbacks. Investigators sent queries do the data center when data were inconsistent or missing.


The main criterion for selecting our outcome measures was their significance for the patients.12 Patients with asymptomatic disease are foremost concerned with a new outburst of symptoms. Study participants were advised to immediately contact the responsible gastroenterologist in such a case in order to verify that new symptoms really reflected a change from subclinical to clinical activity of the disease. Gastroenterologists were encouraged to follow the ECCO guidelines,11 according to which clinical activity is defined by an increase of 100 CDAI points or more. For the final decision, however, we trusted the experience of the gastroenterologists, who—not having access to patients questionnaires—were unaware of psychosocial scores. After excluding alternative explanations for symptoms, including infections, bacterial overgrowth, bile salt malabsorption, dysmotility, and gall stones, changes from subclinical to clinical activity were reported to the data center and coded as “flares.” Flares were first treated with budesonide, sulfasalazine, or systemic corticosteroids. The choice of the medication and its dose depended both on activity, site (ileal, ileocolic, colonic, other), and behavior (inflammatory, stricturing, fistulating) of the disease. A more detailed description of the applied algorithm is provided by the European Panel on the Appropriateness of Crohn's Disease Treatment.13 However, several good reasons applied for encouraging patients to participate in therapeutic decisions such as studying patients in a real life clinical setting, ethical considerations, minimization of drop-outs, and maximization of adherence to treatment.

Once patients have relapsed to clinically active disease, they are concerned whether the medication will succeed in inducing remission. Any need of change from first-line (e.g., budenoside, sulfasalazine) to second-line (e.g., immunosuppressors) or third-line (e.g., antitumor necrosis factor α-agents) medication, as well as a need of higher doses, flags an important adverse event to patients. More aggressive treatment is, besides having more serious side effects such as suppression of the immune system with subsequent infections,14 also more costly. Again, for reasons described in the previous paragraph, the final decision about the necessity of more aggressive medication was left to the gastroenterologists and their patients. There was a general consensus among gastroenterologists participating in the study with regard to the time span after which it was necessary to prescribe more aggressive medication. For instance, thiopurines were given during 16 weeks before a definite decision about response or nonresponse was made.15 If more aggressive medication was necessary, this was communicated to the data center and coded as “nonresponse to medication.”

A major concern of patients is the development of extraintestinal symptoms in addition to gastrointestinal symptoms.12 Extraintestinal manifestations were recorded by gastroenterologists by using an evaluation form they had developed in consensus. The most common extraintestinal manifestations (peripheral arthritis/arthralgia, uveitis/iritis, pyoderma gangrenosum, erythema nodosum, aphtous oral ulcers/stomatitis, ankylosing spondilitis [Bechterew]/sacroileitis, primary sclerosing cholangitis) were specifically recorded, while less common ones (including sensorineural hearing loss, pleuritis, myocarditis, pancreatitis, tendinitis) were summarized under the category “other.” If manifestations such as anemia were considered to be side effects of medication, this was not considered an adverse event. Complications on top of active disease, including strictures, fistulae, malignancies, or any disease-related event requiring hospitalization and surgical intervention, are also unpleasant.14 Extraintestinal manifestations and complications were also reported to the data center and accordingly coded.

Distinguishing between patients who experienced any of these outcome measures, either alone or in combination, from those who did not resulted in two groups. The first included the patients starting in clinical remission and maintaining remission during the entire observation period of 18-months, plus, and this is a novel aspect of the present study, those who started with clinically active disease and successfully remitted without need for more aggressive medication or surgery. Patients in this group at no time since enrolment experienced what they would call relevant deterioration of the disease. The second group contained the patients who could not be maintained in remission, who suffered considerable progression of the disease, and who could not recover with the treatment they had at enrolment. Our outcome was the distinction between those two groups of patients.

To isolate the patients who are currently in remission and to only look if they relapse subsequently might be good for an initial research approach.16 However, obtained results would not reflect the real clinical situation (Berkson's Bias, selection bias [MeSH], impaired generalization of findings). To gain a more detailed insight into patients and their disease course treated by physicians, it is necessary to consider all events with a high clinical impact for the patients.

Main Predictor

Perceived social support was assessed with the ENRICHD Social Support Inventory.17 This self-assessment questionnaire, originally designed to assess the support availability of patients with myocardial infarction, is also applicable to other chronic diseases.18 The original conceptualization of that scale included an extensive literature search of which six items were selected (Table 1). These assess emotional (four items), practical (one item), and informational support (one item), which are rated on a five-point Likert-scale ranging from 1 (“none of the time”) to 5 (“all the time”) resulting in a sum score between 6 and 30 points.

Table 1. Items of the ENRICHD Social Support Inventory
  1. The original items of the ENRICHD Social Support Inventory.17 An additional item asks for current cohabiting with a spouse, adding four points for a positive and two points for a negative answer to the total score. This item is often excluded from the sum score because social support provided by the spouse is also covered by other items.

1. Is there someone available to you who you can count on to listen to you when you need to talk?
2. Is there someone available to give you good advice about a problem?
3. Is there someone available to you who shows you love and affection?
4. Is there someone available to help you with daily chores?
5. Can you count on anyone to provide you with emotional support, such as talking over problems or helping you make difficult decisions?
6. Do you have as much contact as you would like with someone you feel close to, someone you can trust and confide in?

The original version was translated to German and French by two linguistically skilled collaborators. The questionnaire properties were tested in a sample of 1162 cross-sectionally assessed participants of the SIBDCS. Item means varied between 3.58 and 4.23 points, and item SDs between 0.99 and 1.37 points, with an overall respective variance of 0.06 and 0.02, indicating that the items were well weighted. A Cronbach's α of 0.90 indicated good item-item reliability; item-total-scale-correlations ranged between 0.54 and 0.82 and were thus acceptable. The variance shared by the ENRICHD Social Support Inventory and the Depression Subscale of the German and French versions of the Hospital Anxiety and Depression Scale, which had been validated in 199519 respectively in 1989,20 was only 7.4%, showing a good distinction between the assessed constructs (i.e., social support and depression).

Selection and Assessment of Control Variables

The CDAI considers the number of liquid stools per week, abdominal pain, general well being, extraintestinal manifestations, antidiarrheal treatment, abdominal mass, hematocrit change, and weight. The lower the CDAI, the lower the disease activity. While negative values are possible, the upper limit is virtually open. We used the CDAI to control for BDA. This was important because patients did not start at equal levels of BDA. Given that adverse events represent a clinically relevant increase of disease activity, they may depend on initial activity levels. In case of an association between social support and adverse events, we wanted to account for the possibility that high social support did not simply reflect high BDA.

Further relevant control variables were demographics including gender and age. We controlled for intake (yes/no) of six categories of medications, namely, 5-aminosalicylates, sulfasalazine, steroids, immunosuppressors, anti-TNF-α agents, and antibiotics. Hospitalizations related to CD (days) are a proxy indicator of disease behavior (inflammatory, stricturing, fistulizing) and frequency of surgeries. We also included disease duration (years), current smoking status (yes/no), and BMI, as these factors might possibly confound the relationship between social support and the course of CD. A further important confounder, which is reported for the first time in a study on psychological factors in CD, is adherence to therapy. In case of an influence of social support on the disease course, adherence to therapy might be moderated by social support and, if so, a candidate for post-hoc probing. Participants who had no prescription of medication were regarded to be adherent.

In addition to the main predictor (social support), we included two supplementary psychological variables, which supposedly interact with social support, namely, socially inhibited personality style and social diversion oriented coping. Social support might be moderated by social inhibition, because socially inhibited persons may be less likely to accept and seek support and to benefit from it. They might also be less likely to establish social networks to provide them with support. Social diversion might drive individuals to seek social support, potentially maximizing the use made of available social recourses. We assumed that social inhibition would be a negative modulator of social support, whereas social diversion would be a positive modulator.

Briefly, social inhibition was assessed with the Social Inhibition Subscale of the DS-14 Scale to assess Type D (“distressed”) personality.21 Type D persons endorse high levels of negative affect which they are reluctant to express in social interactions. The social inhibition scale contains seven items that are rated from zero to four (total score 0–28); up to two missing items can be replaced by the mean of the completed items. Social diversion was assessed by using the respective subscale of the Coping Inventory for Stressful Situations.22 This scale contains four items ranging from one to five, of which three must be completed; the total score corresponds to the mean of all items. The procedure of translation and validation was the same as for the main predictor. Variance of item means was 0.09 for social inhibition and 0.12 for social diversion. Cronbach's α was 0.87 for social inhibition and 0.80 for social diversion. The Depression Subscale of the German19 and French20 versions of the Hospital Anxiety and Depression Scale shared 14.9% and 5.1% variance, respectively, with social inhibition and social diversion.

The correlation matrix of the control variables showed 29% shared variance between age and disease duration, 7% between age and BMI, 7% to 7.5% between the psychological parameters, and less than 3% in all other cases. It was thus neither necessary nor recommendable to combine them before including them into the equations.

We described the following variables but did not include them into the regression models because their clinical importance for the course of CD has not been established: highest educational level reflecting socioeconomic status, alcohol consumption, and physical exercise. While family history (positive/negative) was shown to increase the risk of disease onset, its role for the disease course is open to question.14

Data Analysis

Data were analyzed using SPSS 15 for Windows (Chicago, IL). Level of significance was set at P < 0.05 (two-sided). To include a case in a model, no variable needed for the model was allowed to be missing (i.e., missing values were not substituted).23 In order to maintain the highest statistical power and to reduce the risk of selective and fitting bias by excluding too many cases from the main models, we excluded the variable instead of the cases if variables contained more than 10% missing values. If exclusion of a variable was necessary, we performed a sensitivity analysis to examine if this variable was likely to affect the results. To control for selective biases (e.g., frequent alcohol consumers might be more likely to skip the variable “alcohol consumption” because they do not want to reveal their drinking habits), we additionally controlled for confounding through nonresponse whenever a variable had more than 10% missing values.

Differences in patient characteristics between the group with adverse events and the group with no adverse events were expressed as mean or percentage differences with asymptotic 95% confidence intervals (CI). Two-sided P-values were computed using independent t-test (age, BMI) and Mann–Whitney U-test, the latter in case of variables showing a skewed distribution (diagnosis duration, hospitalization days related to CD, time to questionnaire completion) or nonmetric quantitative variables (scores). P-values for the comparison in categorical variables between groups were calculated using Fisher's exact test.

We performed binary logistic regression models to estimate the predictive value of social support for the odds of an adverse event by taking the above-mentioned covariates into account, computing odds ratios with 95% CI and respective P-values for all variables. We first modeled the predictive value of covariates without entering social support (Model 1). In a second step we also forced social support into the equation (Model 2). Variables showing relevant changes in their effects after entering social support are likely to interact with social support. In case of significance (α = 0.05) of social support and of at least one covariate in either Model 1 or 2, we computed interaction terms between social support and significant covariates.

Significant interactions were probed for the moderating effect of variables on each other by using the Holmbeck method.24 To conduct a probe of significant interactions, interacting predictors were centered to a mean of zero to reduce multicollinearity between predictors and interaction terms.25 We created two new conditional moderator variables by applying the ±1 SD convention.24 We constructed a high social support group by subtracting 1 SD from social support (assumed moderator) and a low social support group adding 1 SD to social support. This centered the mean equal to 0 in each group. Interaction terms between social support and BMI (assumingly modulated predictor) were calculated for both groups and run in new separate regression analyses to test whether either of the groups differs significantly from 0. We proceeded accordingly for the construction of high and low BMI groups, this time assuming that BMI was the moderator and social support the moderating predictor. The details are described elsewhere.24

Why Was Post-hoc Probing Necessary?

If regression models show a significant interaction between two predictors after controlling for their main effects, this means that the effect on the outcome of at least one of both predictors is moderated by the other predictor. Post-hoc probing provides a deeper insight into the nature of this relationship by revealing if and how strongly predictor 1 inhibits or enhances the effect of predictor 2 and the other way round.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Patient Characteristics

The flowchart (Fig. 1) depicts the inclusion of 458 patients for the analysis. During the observational period of 18 months, 66 (14.4%) patients who were enrolled in clinical remission experienced flares, 25 (5.5%) who were enrolled in a clinically active state did not respond to therapy, 27 (5.9%) suffered from complications, and 9 (2.0%) developed extraintestinal manifestations. Twenty patients experienced more than one category of adverse events (e.g., first flares and later on complications), thus the total of patients with one or more adverse events was 101 (22.1%). The whole sample and the differences between the groups with and without adverse events are described in Table 2. About two-thirds of the sample attended vocational school. Group comparison revealed that patients with adverse events were younger, had lower BMI, lower perceived social support, higher BDA (total sample 72.30 ± 80.39, group with subsequent worsening of the disease 101.98 ± 78.02, group without worsening of the disease 63.91 ± 79.16), and shorter disease duration. Compared to other variables, which yielded between 96% and 100% nonmissing values, adherence to therapy and time to questionnaire completion had at least 13% more (P < 0.001) missing values. The question regarding adherence was answered correctly in 81.2% of the group with adverse events, 83.5% of the group without adverse events (P = 0.653), and in 83.0% of the total sample (e.g., of participants with medication at baseline, 78 did not indicate whether they succeeded in taking their medication as prescribed). The information about questionnaire completion was provided in 88.1% of the adverse events group, in 81.8% of the no event group (P = 0.174), and in 83.2% of the total sample. Adherence to therapy and time to questionnaire completion showed no difference between the group with and without adverse events.

thumbnail image

Figure 1. The number of patients at each stage of the study. Most of the dropouts were patients who did not return the questionnaires.

Download figure to PowerPoint

Table 2. Characteristics and Group Differences of a Sample of 458 Patients with Crohn's Disease
VariableTotal (n = 458)Event + (n = 101)Event – (n = 357)Difference (95%CI)P-value
  1. 95%CI = 95% confidence intervals; baseline disease activity: 458 (100%) nonmissing indications, 20.8% of patients with and 11.8% without events had scores between 150 and 300 (P = 0.032), 1.0% of patients with and 0.6% without events had scores between 301 and 450 (P = 0.527), and none had scores above 450; time difference between assessments of baseline disease activity and social support: 381 (83.2%) nonmissing indications, range = 5.64; 458 (100%) nonmissing indications: social support, disease duration, age, days of hospitalization due to Crohn's disease, gender, medication, and family history; 456 (99.6%) nonmissing indications: body mass index and current smoking status; 454 (99.1%) nonmissing indications: alcohol intake; 453 (98.8%) nonmissing indications: education and social inhibition; 452 (98.7%) nonmissing indications: social diversion; 441 (96.3%) nonmissing indications: sports activities; 380 (83.0%) nonmissing indications: compliance; 9.9% patients with events and 11.5% without events had no medication (P = 0.724); qualitative variables are indicated in means ± standard deviations, quantitative variables in percentages (with absolute frequencies in parentheses). If no unit is specified, quantitative variables are indicated in score points.

Social support24.26 ± 5.5022.65 ± 5.9624.71 ± 5.28−2.06 (−3.26; −0.85)0.001
Baseline disease activity72.30 ± 80.39101.98 ± 78.0263.91 ± 79.1638.07 (20.59; 55.55)<0.001
Age (years)41.64 ± 14.4739.09 ± 15.1042.36 ± 14.22−3.27 (−6.46; −0.73)0.045
Female sex50.2% (230)57.4% (58)48.2% (172)9.2% (−1.8%; 20.3%)0.115
Disease duration (years)12.66 ± 10.4111.09 ± 10.4713.11 ± 10.36−2.01 (−4.31; 0.28)0.028
Hospitalization days2.29 ± 7.613.01 ± 7.102.08 ± 7.740.93 (−0.76; 2.61)0.106
Body mass index23.65 ± 4.6222.14 ± 4.3424.08 ± 4.61−1.94 (−2.95; −0.93)<0.001
5-Aminosalycilates17.0% (78)15.8% (16)17.4% (62)−1.5% (−9.9%; 6.8%)0.767
Sulfasalazin2.2% (10)1.0% (1)2.5% (9)−1.5% (−4.8%; 1.7%)0.472
Steroids23.4% (107)27.7% (28)22.1% (79)5.6% (−3.8%; 15.0%)0.286
Immunosuppressors57.4% (263)52.5% (53)58.8% (210)-6.3% (−17.3%; 4.6%)0.257
Anti-TNFα-agents20.1% (92)25.7% (26)18.5% (66)7.3% (−1.6%; 16.1%)0.122
Antibiotics2.2% (10)3.0% (3)2.0% (7)1.0% (−2.2%; 4.3%)0.465
Smokers43.2% (197)50.0% (50)41.3% (147)8.7% (−2.3%; 19.7%)0.138
Social inhibition9.81 (6.37)10.87 (6.46)9.53 (6.32)−1.35 (−2.77; 0.08)0.056
Social diversion2.16 (0.97)2.08 (0.94)2.18 (0.98)0.10 (−0.12; 0.31)0.415
Adherent participants80.8% (307)81.7% (67)80.5% (240)−1.2 % (−10.8%; 8.5%)0.653
Time difference (months)2.48 (3.79)2.76 (3.56)2.39 (3.87)11.01 (−38.13; 16.11)0.146
 University degree14.6% (66)16.0% (16)14.2% (50)1.8% (−6.3%; 10.0%)0.633
 Tertiary degree15.2% (69)13.0% (13)15.9% (56)−2.9% (−10.6%; 4.8%)0.532
 Vocational school68.2% (309)68.0% (68)68.3% (241)−0.3% (−10.7%; 10.2%)1.000
 No degree2.0% (9)3.0% (3)1.7% (6)1.3% (−2.4%; 5.0%)0.421
Alcohol intake     
 Daily9.5% (43)6.9% (7)10.2% (36)−3.3% (−9.2%; 2.7%)0.440
 Weekly31.7% (144)32.7% (33)31.4% (111)1.3% (−9.2%; 11.7%)0.810
 ≤Monthly58.8% (267)60.4% (61)58.4% (206)2.0% (−8.9%; 13.0%)0.732
Sports activities
 Daily8.4% (37)9.3% (9)8.1% (28)1.1% (−5.4%; 7.7%)0.682
 Weekly49.9% (220)43.3% (42)51.7% (178)−8.4% (−19.8%; 2.9%)0.168
 ≤Monthly41.7% (184)47.4% (46)40.1% (138)7.3% (−4.0%; 18.7%)0.202
Positive family history
 Crohn's dis (CD)12.0% (55)17.8% (18)10.4% (37)7.5% (0.3%; 14.6%)0.055
 Colitis / no CD2.4% (11)3.0% (3)2.2% (8)0.7% (−2.7%; 4.1%)0.713
 No infl bowel dis85.6% (392)79.2% (80)87.4% (312)−8.2% (−15.9%; −0.4%)0.053

Predictors of Adverse Events

Model 1 of Table 3 considers variables that may confound or moderate a potential relationship between social support and disease exacerbation and shows that BDA and BMI predicted disease exacerbation independently of each other. A 1-SD decrease of BMI from the mean (i.e., a BMI of 19 kg/m2) enhanced the odds of adverse events by 1.43 (95% CI: 1.04–1.96, P = 0.029). This effect became only apparent when social support was added to the model (Model 2 of Table 3). More important, elevated levels of social support emerged as an independent predictor of event risk after all covariates had been taken into account. In detail, the odds of experiencing adverse events were 1.50-fold reduced (95% CI: 1.16–1.94, P = 0.002) for an increase of 1 SD of social support, which corresponds to the 0.67-fold increase shown in Table 3.

Table 3. Primary Results: Odds Ratios (95% Confidence Intervals in Parentheses)
Model 1Model 2
Odds RatioP-valueOdds RatioP-value
  1. Odds ratio of disease deterioration for each variable included. If no unit is specified, quantitative variables indicate odds ratios for one standard deviation. Binary variables (yes/no) are distinguished by a “B.” While Model 1 considered all covariates, Model 2 additionally considered social support to test for the unique contribution of social support to event odds after all covariates had been taken into account. Model 2 (454 (99.1%) valid cases): −2xln(likelihood) = 429; Model 1 (454 (99.1%) valid cases): −2xln(likelihood) = 440; −2xln(likelihood ratio Model 2/Model 1) = 11, P < 0.001. As more than 10% of the cases failed to provide the date of questionnaire completion and information about treatment adherence, both of these variables were included in the models as “providing the information” versus “not providing the information” (i.e., as a binary variable). The inclusion of “time to questionnaire completion” as originally intended (time interval between enrolment and questionnaire completion in months) together with all other control variables would result in an odds ratio for one SD of social support of 0.598 (95% CI: 0.442; 0.810, P = 0.001). The inclusion of treatment adherence as originally intended (compliant versus not compliant) would result in an odds ratio of 0.626 (95% CI: 0.474; 0.828, P = 0.001). The inclusion of both as originally intended would result in too high a number of excluded cases.

VariablesSocial support0.666 (0.516; 0.859)0.002
Baseline disease activity1.702 (1.328; 2.179)<0.0011.626 (1.240; 2.132)<0.001
Female sex B1.511 (0.926; 2.464)0.0981.331 (0.802; 2.210)0.268
Age (years)0.990 (0.970; 1.011)0.3490.989 (0.968; 1.011)0.310
Disease duration (years)0.988 (0.961; 1.017)0.4140.981 (0.952; 1.011)0.208
Days hospital CD0.992 (0.964; 1.021)0.5730.990 (0.961; 1.018)0.474
5-Aminosalycilates B1.019 (0.531; 1.958)0.9541.038 (0.524; 2.055)0.915
Sulfasalazin B0.389 (0.043; 3.545)0.4020.395 (0.042; 3.723)0.417
Steroids B0.881 (0.499; 1.555)0.6620.804 (0.445; 1.455)0.472
Immunosuppressors B0.723 (0.440; 1.188)0.2010.726 (0.432; 1.220)0.227
Anti-TNFα-agents B1.327 (0.734; 2.400)0.3491.273 (0.687; 2.359)0.444
Antibiotics B0.969 (0.206; 4.566)0.9680.920 (0.192; 4.397)0.917
Smoking B1.269 (0.776; 2.077)0.3431.398 (0.840; 2.327)0.198
Body mass index0.760 (0.565; 1.023)0.0700.701 (0.510; 0.964)0.029
Positive family history1.391 (0.643; 3.007)0.1211.295 (0.923; 1.819)0.135
Adherence B1.000 (0.137; 7.322)0.5301.099 (0.550; 2.195)0.789
Time difference B1.322 (0.706; 2.477)0.0530.540 (0.253; 1.154)0.112
Social diversion0.944 (0.733; 1.214)0.6531.053 (0.805; 1.376)0.706
Social inhibition1.137 (0.890; 1.453)0.3041.055 (0.814; 1.368)0.686

A significant interaction emerged between BMI and social support in predicting adverse event risk (odds ratio 1.014, 95% CI: 1.002–1.027, P = 0.027). Post-hoc probing revealed that, if social support was low, a decrease of 1 SD of BMI from the mean increased the odds by 2.09 (95% CI: 1.28–3.36, P = 0.003). This suggests that the probability of adverse events increased from 22% in the entire sample to 37% in the group with low social support. In contrast, if social support was high, BMI was not predictive for adverse event risk (odds ratio 1.03, 95% CI: 0.68–1.55, P = 0.899). In the group with low BMI, an increase of social support of 1 SD decreased the odds of adverse events by 1.80 times (95% CI: 1.32–2.46, P < 0.001), but showed no effect on adverse event risk in the group with high BMI (odds ratio 1.00, 95% CI: 0.68–1.45, P = 0.986). There was no significant interaction between social support and BDA in predicting adverse event risk (odds ratio 1.00, 95% CI 1.00–1.00, P = 0.368). Consequently, no moderation effect of social support on BDA was expectable, meaning that BDA acted independently of social support and vice versa.


  1. Top of page
  2. Abstract
  6. Acknowledgements

In a middle-aged sample of patients with CD, we found that social support decreased the odds of subsequent progression of the disease. Low BMI was associated with greater odds for adverse events, and this effect was moderated by social support, which attenuated the risk of adverse events when BMI was low, but not when BMI was high. In addition, BDA at baseline also emerged as an independent predictor of adverse events.

The buffering effects of social support are thought to attenuate the impact of an individual's physical stress reactions. We tested this hypothesis by investigating its effect as a moderator of the adverse impact of low BMI on the disease. Our findings supported that high social support decreases the adverse effect of low BMI, and high BMI softens the adverse effect of low social support. The main-effect model suggests a beneficial effect of social support that is independent of covariates and does not exert a moderating effect on event risk by its association with other prognostic factors. Our data questioned such a main effect, since elevated social support was only protective in patients with low BMI, but did not add to the effect of high BMI.

To our knowledge, this is the first study reporting on social support buffering a biologic parameter that has prognostic value for a physical disease in general and CD in particular. An epidemiological survey of 34,653 patients found both main and buffering effects of social support26 alleviating the pathophysiological effects of traumatic events. A cohort study of 292 patients with rheumatoid arthritis found that the interaction between social support and joint tenderness predicted distress after 4 years. That study discussed a buffering effect of social support on joint tenderness, but actually did not probe for that effect.27 Another cohort study found social support and anxiety to have significant effects on CD activity, but did not test for interaction effects between those two predictors.28

Patients might benefit from higher levels of social support, because the perceived availability of social support modifies the cognitively appraised threat of stress in a way that individuals' coping resources are enhanced. This leads to better moods, more positive emotions, and perceived higher control.29 We do not exactly know why social support buffers the negative effect of low BMI. Low BMI might elicit social support from family members. In addition, elevated levels of social support perhaps modulate many negative biological consequences of adverse health conditions. Social support attenuates hemodynamic and coagulation responses to acute psychosocial stress.30 Further studies may want to test for a favorably modulating effect of social support on physiological changes related to BMI. This might help to provide important novel information about patients needs in terms of tailored interventions.

Patients might benefit from support groups. Such groups were successfully tested in other chronic diseases like cancer, and revealed the importance of available community, information, acceptance, and relief of the family.31 Therapeutic interventions might also focus on social skills and support in existing networks (psychotherapy, interventions). Alzheimer caregivers who underwent a multicomponent intervention reported over 5 years of follow-up higher satisfaction with their social network, including higher emotional support, more visits, and closer relationships in the network.32 Such therapeutic programs might be addressed by future studies.

Low BMI probably reflected maldigestion as a harbinger of clinically manifest increase in disease activity. In line with this, low BMI was found to predict severe forms of CD.33 Computing the BDA by means of the CDAI implicates accounting for body weight, because one of the nine domains considers the difference between actual weight and “standard” weight in standard weight percents. However, BMI and BDA predicted poor disease outcome independently from each other, because the domain “standard weight” contributed to less than 2% of the variance of the CDAI, which is a weak correlation. It is problematic that the CDAI, a most frequently used disease activity score in IBD research, adds factors which cannot be additive because of their different nature. Of a well-balanced scale all domains should contribute equally to and correlate as much as possible with the sum score. We failed to find such quality criteria for the CDAI in Best et al's article.10 The BMI has the advantage of defining an interval that can be regarded as normal, while computing the deviation from the “standard” weight only provides a one-sided distance from an idealized point estimate. Since Best et al's article does not provide a clear definition of “standard” weight, different formulas are commonly used, the most common correlating to over 90% with BMI, resulting in less than 3% shared variance between CDAI and BMI.

A shortcoming of our study was that a high number of patients did not provide the date when they had completed the self-assessment questionnaire, as it might be argued that the scores could have been affected by the delay between assessments of baseline disease activity and social support. However, social support is commonly viewed as a persistent psychological construct and 1-month test–retest reliability of the applied questionnaire was 94%, indicating good consistency over time.34 In addition, controlling for the time interval between assessments of baseline disease activity and social support in patients who had this information available even strengthened the association between social support and disease deterioration (odds ratio 1.67, 95% CI: 1.23–2.26, P = 0.001 versus 1.50, 95% CI: 1.16–1.94, P = 0.002).

We included a sizeable sample allowing adjustment for a range of important control variables. The response rate was 80.2%, which is comparably high. Under a given number of subjects, however, binary logistic regression is limited to the proportion of outcome (adverse event/no event). Skewed proportion increases the risk of over-, under-, and paradoxical model fitting, because of too few degrees of freedom. Adverse events occurred in 22% of our patients; therefore, we selected the number of variables in order to keep an acceptable control of both biased fitting and confounding. A less skewed proportion between patients with and without adverse events would have allowed us to control for other psychological variables. These might also be important for the perception of social support, such as anxiety, depression, and perceived stress. The weak correlations between those variables and social support, however, indicated, assuming an influence of those psychological parameters on the perception of social support, that this influence might be relatively small (15.8% variance shared with depression and 15.6% with perceived stress). Moreover, depression was strongly associated with perceived stress (43.6% shared variance) and thereby interfered with perceived stress when included together in a regression model. The associations between the disease course and other psychological factors, including anxiety, depression, and perceived stress, are being reported separately. Compared to other cohort studies, the number of missing values was relatively small (Tables 2, 3). Consecutive sampling and little restrictive inclusion criteria allowed for a good representation of the referred population. These issues have been discussed more extensively elsewhere.35

We found that elevated levels of social support may favorably affect the course of CD. On closer examination of the interaction between social support and BMI, we found that high levels of one predictor compensated low levels of the other, but that the effects were not multiplicative. Future research may want to investigate the psychophysiologic pathways through which those factors influence the course of CD. The observation that social support may favorably moderate the impact of physiological factors suggest a potential for biobehavioral interventions to improve the course of CD. We would like to encourage future studies on social support groups.


  1. Top of page
  2. Abstract
  6. Acknowledgements

We thank Annette Kocher for editing assistance, the collaborators of the Swiss Inflammatory Bowel Disease Cohort Study for data collection, as well as the patients for their participation.


  1. Top of page
  2. Abstract
  6. Acknowledgements
  • 1
    Cámara RJ, Ziegler R, Begré S, et al; Swiss Inflammatory Bowel Disease Cohort Study (SIBDCS) group. The role of psychological stress in inflammatory bowel disease: quality assessment of methods of 18 prospective studies and suggestions for future research. Digestion. 2009; 80: 129139.
  • 2
    Reblin M, Uchino BN. Social and emotional support and its implication for health. Curr Opin Psychiatry. 2008; 21: 201205.
  • 3
    Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol Bull. 1985; 98: 310357.
  • 4
    Andre-Petersson L, Engstrom G, Hedblad B, et al. Social support at work and the risk of myocardial infarction and stroke in women and men. Soc Sci Med. 2007; 64: 830841.
  • 5
    Gorst-Unsworth C, Goldenberg E. Psychological sequelae of torture and organised violence suffered by refugees from Iraq. Trauma-related factors compared with social factors in exile. Br J Psychiatry. 1998; 172: 9094.
  • 6
    Mehnert A, Lehmann C, Graefen M, et al. Depression, anxiety, post-traumatic stress disorder and health-related quality of life and its association with social support in ambulatory prostate cancer patients. Eur J Cancer Care (Engl). 2009 [Epub ahead of print].
  • 7
    Pittet V, Juillerat P, Mottet C, et al. Cohort profile: the Swiss Inflammatory Bowel Disease Cohort Study (SIBDCS). Int J Epidemiol. 2009; 38: 922931.
  • 8
    Borenstein M, Rothstein H, Cohen J. Power and precision, 1st ed. Englewood, NJ:. Biostat; 2001.
  • 9
    Lennard-Jones JE. Classification of inflammatory bowel disease. Scand J Gastroenterol Suppl. 1989; 170: 26.
  • 10
    Best WR, Becktel JM, Singleton JW, et al. Development of a Crohn's Disease Activity Index. National cooperative Crohn's disease study. Gastroenterology. 1976; 70: 439444.
  • 11
    Stange EF, Travis SP, Vermeire S, et al. European evidence based consensus on the diagnosis and management of Crohn's disease: definitions and diagnosis. Gut. 2006; 55( suppl 1): i115.
  • 12
    Casati J, Toner BB, de Rooy EC, et al. Concerns of patients with inflammatory bowel disease: a review of emerging themes. Dig Dis Sci. 2000; 45: 2631.
  • 13
    Michetti P, Stelle M, Juillerat P, et al. Appropriateness of therapy for active Crohn's disease: results of a multidisciplinary international expert panel-EPACT II. J Crohn's Colitis. 2009; 3: 232240.
  • 14
    Baumgart DC, Sandborn WJ. Inflammatory bowel disease: clinical aspects and established and evolving therapies. Lancet. 2007; 369: 16411657.
  • 15
    Travis SP, Stange EF, Lémann M, et al. European Crohn's and Colitis Organisation. European evidence based consensus on the diagnosis and management of Crohn's disease: current management. Gut. 2006; 55( suppl 1): i1635.
  • 16
    Bitton A, Dobkin PL, Edwardes MD, et al. Predicting relapse in Crohn's disease: a biopsychosocial model. Gut. 2008; 57: 13861392.
  • 17
    Mitchell PH, Powell L, Blumenthal J, et al. A short social support measure for patients recovering from myocardial infarction: the ENRICHD Social Support Inventory. J Cardiopulm Rehabil. 2003; 23: 398403.
  • 18
    Burg MM, Barefoot J, Berkman L, et al. Low perceived social support and post-myocardial infarction prognosis in the enhancing recovery in coronary heart disease clinical trial: the effects of treatment. Psychosom Med. 2005; 67: 879888.
  • 19
    Herrmann C, Buss U, Snaith RP. HADS-D Hospital Anxiety and Depression Scale German version: a questionnaire of assessment of anxiety and depression in somatic medicine, 1st ed. Bern, Switzerland: Huber; 1995.
  • 20
    Razavi D, Delvaux N, Farvacques C, et al. Validation of the HADS French version in hospitalized cancer patients. Rev Psychol Appl. 1989; 39: 295307.
  • 21
    Denollet J. DS14: standard assessment of negative affectivity, social inhibition, and Type D personality. Psychosom Med. 2005; 67: 8997.
  • 22
    Endler NS, Parker JD. Multidimensional assessment of coping: a critical evaluation. J Pers Soc Psychol. 1990; 58: 844854.
  • 23
    Enders CK. A primer on the use of modern missing-data methods in psychosomatic medicine research. Psychosom Med. 2006; 68: 42736.
  • 24
    Holmbeck GN. Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations. J Pediatr Psychol. 2002; 27: 8796.
  • 25
    Aiken LS, West SG. Multiple regression: testing and interpreting interactions, 1st ed. Newbury Park, CA: Sage; 1991.
  • 26
    Moak ZB, Agrawal A. The association between perceived interpersonal social support and physical and mental health: results from the national epidemiological survey on alcohol and related conditions. J Public Health (Oxf). 2010; 32: 191201.
  • 27
    Strating MM, Suurmeijer TP, van Schuur WH. Disability, social support, and distress in rheumatoid arthritis: results from a thirteen-year prospective study. Arthritis Rheum. 2006; 55: 736744.
  • 28
    Persoons P, Vermeire S, Demyttenaere K, et al. The impact of major depressive disorder on the short- and long-term outcome of Crohn's disease treatment with infliximab. Aliment Pharmacol Ther. 2005; 22: 101110.
  • 29
    Uchino BN. Social support and health: a review of physiological processes potentially underlying links to disease outcomes. J Behav Med. 2006; 29: 377387.
  • 30
    Wirtz PH, Redwine LS, Ehlert U, et al. Independent association between lower level of social support and higher coagulation activity before and after acute psychosocial stress. Psychosom Med. 2009; 71: 3037.
  • 31
    Ussher J, Kirsten L, Butow P, et al. What do cancer support groups provide which other supportive relationships do not? The experience of peer support groups for people with cancer. Soc Sci Med. 2006; 62: 25652576.
  • 32
    Drentea P, Clay OJ, Roth DL, et al. Predictors of improvement in social support: five-year effects of a structured intervention for caregivers of spouses with Alzheimer's disease. Soc Sci Med. 2006; 63: 957967.
  • 33
    Loly C, Belaiche J, Louis E. Predictors of severe Crohn's disease. Scand J Gastroenterol. 2008; 43: 948954.
  • 34
    Vaglio J Jr, Conard M, Poston WS, et al. Testing the performance of the ENRICHD Social Support Instrument in cardiac patients. Health Qual Life Outcomes. 2004; 2: 24.
  • 35
    Schoepfer AM, Beglinger C, Straumann A, et al. Fecal calprotectin correlates more closely with the Simple Endoscopic Score for Crohn's disease (SES-CD) than CRP, blood leukocytes, and the CDAI. Am J Gastroenterol. 2010; 105: 162169.