Article Text

Download PDFPDF

Predicting relapse in Crohn’s disease: a biopsychosocial model
  1. A Bitton1,
  2. P L Dobkin1,
  3. M D Edwardes1,
  4. M J Sewitch1,
  5. J B Meddings2,
  6. S Rawal1,
  7. A Cohen3,
  8. S Vermeire4,
  9. L Dufresne1,
  10. D Franchimont1,
  11. G E Wild1,5
  1. 1
    McGill University Health Centre, Montreal, Canada
  2. 2
    University of Alberta, Canada
  3. 3
    Jewish General Hospital, Montreal, Canada
  4. 4
    University Hospital Gasthuisberg, Leuven, Belgium
  5. 5
    Douglas Hospital Center for Studies on Human Stress, Montreal, Canada
  1. Dr A Bitton, McGill University Health Centre, 687 Pine Avenue West, Montreal, Quebec, Canada, H3A 1A1; alain.bitton{at}


Background: Crohn’s disease (CD) is a chronic relapsing inflammatory bowel disorder. Both biological and psychosocial factors may modulate the illness experience.

Aim: The aim of this study was to identify clinical, biological and psychosocial parameters as predictors of clinical relapse in quiescent CD.

Methods: Patients in medically induced remission were followed prospectively for 1 year, or less if they relapsed. Disease characteristics were determined at baseline. Serum cytokines, anti-Saccharomyces cerevisiae antibodies, C-reactive protein (CRP), erythrocyte sedimentation rate and intestinal permeability were measured every 3 months. Psychological distress, perceived stress, minor life stressors and coping strategies were measured monthly. A time-dependent multivariate Cox regression model determined predictors of time to relapse.

Results: 101 patients (60 females, 41 males) were recruited. Fourteen withdrew and 37 relapsed. CRP (HR = 1.5 per 10 mg/l, 95% CI 1.1 to 1.9, p = 0.007), fistulising disease (HR = 3.2, 95% CI, 1.1 to 9.4, p = 0.04), colitis (HR = 3.5 95% CI 1.2 to 9.9, p = 0.02) and the interaction between perceived stress and avoidance coping (HR = 7.0 per 5 unit increase for both scales, 95% CI 2.3 to 21.8, p = 0.003) were predictors of earlier relapse.

Conclusions: In quiescent CD, a higher CRP, fistulising disease behaviour and disease confined to the colon were independent predictors of relapse. Moreover, patients under conditions of low stress and who scored low on avoidance coping (ie, did not engage in social diversion or distraction) were least likely to relapse. This study supports a biopsychosocial model of CD exacerbation.

View Full Text

Statistics from

Crohn’s disease (CD) is a chronic inflammatory bowel disorder of uncertain aetiology. Prevalence estimates ranging from 26 to 199 cases per 100 000 persons in North America and from 8.3 to 214 cases per 100 000 persons in Europe have been reported.1 The clinical course of CD is characterised by relapsing and remitting chronic intestinal inflammation with symptoms that significantly impact health-related quality of life.2 Clinical relapse may occur in 30–60% of patients within 1 year of medically induced remission.3 4 While progress has been made in understanding the pathogenesis of CD, far less is known about the manner in which environmental factors modify the expression of genes in susceptible individuals and modulate the effector function of the intestinal immune system, both of which are key in determining disease onset or relapse.5 6

Psychosocial stressors act through complex neuroendocrine immune pathways and are associated with a diverse set of negative health outcomes.7 While evidence derived from animal models has shown that environmental stressors modulate intestinal inflammation, delineating the relationship between life stressors and inflammatory bowel disease (IBD) activity in humans has been challenging.8 Stress may be defined as the condition in which person–environment interactions lead to a perceived discrepancy between the demands of the situation and the biological, psychological and social resources of the person.9 It may be postulated that in chronic diseases such as CD, how an individual interprets and responds to the environmental challenge determines responses to stress, influences health behaviours and contributes to the neuroendocrine and immune responses that may ultimately affect health outcomes. The coping response has been examined as a modifier of the stress experience in both healthy and disease states.10 To date, data pertaining to stress and coping in CD are sparse, and little is known about the impact of particular coping strategies on disease activity and outcomes.

Determining factors that identify patients at high risk for a relapse represents an unmet challenge in CD. In a seminal article, Engel proposed that a comprehensive understanding of the aetiology and progression of disease must take into account the interactions between psychological and social factors and biological processes.11 This biopsychosocial approach has provided an important framework for studies that have provided compelling evidence for the role of psychosocial processes in the pathogenesis of diverse disorders.1215 However, few studies have used this conceptual paradigm as an approach to study patients with IBD.

The aim of this study was to examine clinical, biological and psychosocial variables concurrently and prospectively to determine which factors predicted clinical relapse in quiescent CD. It was hypothesised that serum C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), anti-Saccharomyces cerevisiae antibodies (ASCAs), serum cytokines and intestinal permeability increase prior to manifestations of clinical symptoms and may be useful predictors of relapse. Furthermore, it was postulated that life stressors and psychological distress lead to earlier relapse and that coping strategies for stress and distress, if effective, would reduce the risk of relapse.


Study design and patient population

Patients with inactive CD were followed for a 1 year period, or less if they relapsed. Patients were recruited between 1998 and 2002 at the McGill University Health Centre (Royal Victoria Hospital and Montreal General Hospital) and the Jewish General Hospital in Montreal, Canada. Consecutive patients seen in gastroenterology outpatient clinics were screened for eligibility. Inclusion criteria were: (1) a diagnosis of CD, based on standard clinical, radiological, endoscopic and histological criteria16; (2) clinical remission for at least 1 month at study entry as defined by a Crohn’s Disease Activity Index (CDAI) of <15017; and (3) age 18–65 years. Exclusion criteria were: (1) symptomatic CD at study entry including CDAI ⩾150 and current complications including acute bowel obstruction, abdominal abscess and gastrointestinal bleeding; (2) significant perianal disease; (3) pregnancy; (4) previous extensive small bowel resection; (5) presence of an ileostomy or colostomy; (6) use of prednisone or budesonide within 30 days of study entry; and (7) antibiotic use at study entry. Patients receiving oral mesalamine, or 6-mercaptopurine (6-MP), azathioprine or methotrexate were excluded if their medication dose had been altered within 30 days (oral mesalamine) or within 3 months (6-MP/azathioprine or methotrexate) before study entry. Patients continued these maintenance drug therapies throughout the study but were instructed not to alter their dosage. No patients were on infliximab or adalimumab as this study was conducted during a time when these agents were not readily available. The protocol was approved by the Institutional Review Boards of the participating hospitals. Written informed consent was obtained from all patients.

Study follow-up

Patients were seen at baseline and every 3 months for a total of 1 year if they remained in remission or for a shorter period if they relapsed. Relapse was defined as a CDAI of >150 with an increase of ⩾70 points from baseline. Patients were instructed to communicate with the research coordinator if they developed symptoms suggestive of an exacerbation, at which time a visit with a study doctor was arranged to confirm relapse. At each visit and at relapse, the CDAI was calculated and blood was drawn for measurement of CRP, ESR and serum interleukin 1β (IL1β), soluble interleukin 2 receptor (sIL2R), interleukin 6 (IL6), interleukin 10 (IL10), tumour necrosis factor α (TNFα) and ASCAs. Urine was collected after overnight ingestion of lactulose/mannitol for intestinal permeability testing. Patients completed a questionnaire for 7 days prior to each visit in order to calculate a visit CDAI. At each visit, adherence with maintenance medication was verified.

Psychosocial questionnaires were completed at baseline and monthly thereafter for 12 months or until relapse. At each 3 month visit, patients were given three sets of questionnaire packets that were completed and returned monthly by mail in preaddressed, stamped envelopes. The research coordinator received these questionnaires and reviewed them for completeness. If questionnaires were not submitted in a timely manner, the patient was contacted and reminded to mail them.

Psychosocial measures

Consistent with the conceptual framework for our study, four psychosocial scales were used to measure stress, psychological distress and coping strategies.

Minor life stress was assessed using the abridged version of the Hassles Scale, which asks that participants rate each of 53 minor events during the past month on a 4-point (0–3) scale.18 19 This validated and reliable version consists of items pertaining to common experiences occurring in the context of work (eg, workload), family (eg, home repairs), health (eg, physical abilities) and the environment (eg, weather). Items are rated based on how irritating, frustrating or distressing they were perceived to be. A cumulated severity of hassles score is derived by summing the ratings given across items, and a higher score indicates more stress.

The Perceived Stress Scale-10, a 10-item scale, was used to measure the extent to which participants appraised situations or events in their life to be stressful during the past month.20 21 It differs from the Hassles Scale in that it assesses how the respondent globally perceives stress (ie, its predictability and controllability) rather than rating events per se. Each item is scored from 0 to 4. A global score is computed ranging from 0 to 40, with higher scores indicating greater perceived stress. This scale, designed for use in community samples, has been shown to have good internal and test–retest reliability.21

The Coping Inventory for Stressful Situations (CISS) was used to assess coping styles.22 23 This 48-item measure asks that respondents indicate on a 5-point scale to what degree they employed the behaviours described. The three general coping styles measured are: (1) task-oriented coping (eg, conscious effort to solve a problem); (2) emotional-oriented coping (eg, feel overwhelmed; blame self); and (3) avoidance-oriented coping (eg, take time off and get away from the situation; social diversion or distraction). There is strong support for the CISS in terms of construct validity, its three-factor structure and reliability.23

The Symptom Checklist 90-R (SCL-90R) was used to measure psychological distress. This 90-item self-report symptom inventory is a validated measure of the psychological symptom patterns of psychiatric and medical patients, during the past week.24 The SCL-90-R generates nine primary symptom dimensions, namely: (1) somatisation; (2) obsessive–compulsive; (3) interpersonal sensitivity; (4) depression; (5) anxiety; (6) hostility; (7) phobic anxiety; (8) paranoid ideation; and (9) psychoticism. The Global Severity Index (GSI) combines the number and intensity of symptoms and is considered to be the optimal summary score.

Biological measures and intestinal permeability

Serum and urine samples were coded so that technicians were blinded to the study patients’ clinical status. CRP was measured using rate nephelometry, and ESR by the Westergren method. Commercially available ELISAs (Roche Diagnostics, Mannheim, Germany) were used to measure the serum levels of IL1β, sIL2R, IL6, IL10 and TNFα according to the manufacturer’s protocol. ASCA was determined using a commercially available ELISA (Quanta Lite ASCA IgG and IgA, Inova Diagnostics, San Diego California, USA) according to the manufacturer’s protocol. All blood samples were collected on the morning of the scheduled visit and the serum was stored at −80°C until the time of assay.

Intestinal permeability was assessed by measuring the urinary excretion of orally administered non-metabolised sugar probes.25 Briefly, a test solution of 5 g of lactulose and 2 g of mannitol in 350 ml of water was consumed at bedtime. Patients were asked not to consume alcohol or non-steroidal anti-inflammatory drugs (NSAIDs) for at least 1 week prior to drinking the test solution. Patients voided prior to drinking the solution. All urine including the first morning void was collected in preweighed containers containing 5 ml of 10% thymol in methanol. Urinary sugars were measured by high-performance liquid chromatography as described previously.26 The lactulose/mannitol ratio was calculated as a measure of intestinal permeability.

Statistical analyses

Analyses were performed using SAS (SAS Institute, Cary, North Carolina, USA). All variables were analysed for their association with time to relapse using univariate Cox regression. Time zero was defined as study entry, and patients were followed-up to relapse or to approximately 12 months or the date of early withdrawal. Multivariate Cox regression was used to identify independent determinants associated with time to relapse while adjusting for potential confounders. The determinants were selected a priori and included clinical, biological and psychosocial, as well as some prespecified interaction terms. Best subset regression, an option of the SAS program for Cox regression, PROC PHREG, was used to guide selection of predictors of time to relapse. This option systematically compares all possible combinations of a given list of factors.

Alternative analyses of the time-dependent data were considered. Every relapse corresponds to a risk set, defined as the data at the relapse time for the relapser and all patients that have not yet relapsed. For the psychosocial data this was modified to exclude a risk set member if the most recent questionnaire was >35 days old (any shorter period would have excluded too many subjects) or <14 days old. The 14-day lag was chosen through the consideration that there could be 14 days during which relapse occurred but had not yet become apparent to the subject or the doctor. For the biological data, we used a 14–92 day time window because those data were recorded at 3-month intervals.

All continuous predictors were analysed as such for Cox regressions. Categorisation of some predictors was only done for illustration, for some univariate results shown in table 2, and for defining survival curves to illustrate effects identified using the multivariate Cox model. The survival curve estimates were computed for time-varying predictors to illustrate the effects identified using the multivariate Cox model. The estimates are similar to Kaplan–Meier estimates, but they account for the 14 to 35 or 92 day time windows prior to each relapse. When time-varying data were not measured during the appropriate time window for a patient in a particular risk set, the overall Kaplan–Meier estimate was assigned to the patient at that time. Otherwise, the curve which patients contributed to changed as their values changed over time. For example, when a patient who previously had CRP <10 mg/l obtained a new value of CRP >10 mg/l, that patient’s data were deleted from the calculation of the CRP <10 mg/l curve beyond the time of the new value, and were added to the calculation for the CRP ⩾10 mg/l curve. Each estimate of the CRP <(or ⩾) 10 mg/l curve has a subtle interpretation: it is the observed percentage of patients who remain in remission at least to the time of the estimate, given that they are predisposed, at baseline, to have a CRP <10 (or ⩾) mg/l recorded 14–92 days prior to that time.

The cut-offs used to define the survival curves for the Perceived Stress Scale and for the CISS avoidance subscale were based on the average median over all time points. A value of 15 for perceived stress (high >15, low ⩽15) and a value of 40 for CISS avoidance (high >40, low ⩽40) ensured that all curves in fig 2 were based on roughly equal numbers of patients. For the Perceived Stress Scale, the cut-off is similar to data based on population norms.21 For CRP, the difference between the curves was maximised at a cut-off of 12 mg/l. We chose 10 mg/l, since that was a frequently used cut-off point from clinical trials in IBD and it still shows a strong difference.

Repeated-measures regression by GEE (generalised estimating equations, using SAS PROC GENMOD) was used to test whether scores on the biological and psychosocial variables decreased significantly over time as in our earlier ulcerative colitis study.27 This analysis is secondary, as it only justifies our careful use of time windows in the Cox regressions and survival curves.


Baseline characteristics

A total of 101 patients (60 females) entered the study. Fourteen patients (14%) were either lost to follow-up or withdrew. These patients’ data were used up to time of withdrawal. Thirty-seven patients (37%) relapsed. Patients who relapsed during the follow-up (relapsers) and non-relapsers were similar in all baseline characteristics (table 1).

Table 1 Baseline characteristics (means and percentages) and univariate HR of relapse

Time to relapse—univariate analysis

The mean values of biological and psychosocial parameters measured during the visit prior to relapse in relapsers and the mean value at 6 months in non-relapsers are presented in table 2. Univariate Cox regression analysis of repeated measures revealed CRP ⩾10 mg/l and ESR to be associated with earlier time to relapse.

Table 2 Biological parameters in relapsers versus non-relapsers (means) and univariate HR of relapse

Time to relapse—multivariate analysis and survival curves

Our multivariate Cox regression model is presented in table 3. All variables listed in tables 1 and 2 as well as selected interactions were considered for this model. Specifically, the interactions between each coping style (ie, emotion-oriented, task-oriented and avoidance coping) and stress were considered based on a priori knowledge that coping can act as a moderator of stress.28 The interaction between gender and number of previous relapses was also analysed given this interaction was found to be predictive of relapse in our previous ulcerative colitis work.29 Characteristics associated with earlier time to relapse included CRP, fistulising disease, colitis and the interaction between perceived stress and avoidance coping. Of the three coping styles considered, the interaction between avoidance coping and stress was the most significant. The other two styles, to a lesser degree, also had a significant interaction with stress. In all three instances, the interaction effect showed such a strong effect that the two main effects were not significant without the presence of the interaction term. Fistulising disease and colitis were determined at baseline.

Table 3 Multivariate time-dependent model

Figure 1 presents the survival curves for the relationship between CRP and time to relapse. Patients having current (14–92 days previous) CRP ⩾10 mg/l were more likely to relapse earlier compared with those with current CRP <10 mg/l. Figure 2, which illustrates the stress–avoidance coping interaction, shows that patients who were under current (14–35 days previous) conditions of low stress and who scored low on the CISS-avoidance subscale were least likely to relapse during the study period. In contrast, patients with the other three combinations (high stress–low avoidance score, high stress–high avoidance score and low stress–high avoidance score) relapsed earlier.

Figure 1 Most recent C-reactive protein (CRP; 14–92 days prior to relapse) and risk of Crohn’s disease relapse. At months 0 and 12, respectively, n = 68 and 41 for CRP <10 mg/l curve, and n = 24 and 11 for CRP ⩾10 mg/l.
Figure 2 Most recent score (14–35 days prior to relapse) on Perceived Stress Scale and Coping Inventory for Stressful Situations (CISS) avoidance subscale and risk of Crohn’s disease relapse. At months 0 and 12, respectively, n = 30 and 5 for high stress–high score avoidance coping curve, n = 23 and 9 for high stress–low score avoidance coping curve, n = 26 and 5 for low stress–high score avoidance coping curve, and n = 19 and 26 for low stress–low score avoidance coping curve.


In this study the biopsychosocial paradigm was used as a framework to identify concurrently a biological marker, clinical parameters and psychosocial variables as predictors of relapse in patients with inactive CD.

Patients with colonic CD were found to relapse earlier. No specific association has been reported between the anatomical location of disease and disease course in population-based studies.30 31 One study did report a higher risk of relapse in patents with colonic involvement, although patients with concurrent small bowel disease may have been included.32 Fistulising disease behaviour (patients with internal fistulas) was also found to be predictive of earlier relapse. Most studies have not evaluated disease behaviour and its effect on relapse following medically induced remission. However, our findings are consistent with the postoperative literature which has shown that patients who undergo bowel resection for a perforating (fistulising) indication are more likely to relapse earlier over time.33 This suggests that this type of disease behaviour is a more aggressive form of CD associated with early recurrence following medically or surgically induced remissions. Shorter duration of remission has been reported as a predictor of relapse in other work, but not in our present study.32 This lack of association may reflect the different patient populations enrolled across studies.

Diverse biological markers involved in the pathogenesis of CD have been proposed as predictors of recurrence, but their value as warning signals warrants further study. In IBD, increased levels of CRP have been associated with active disease.34 Several studies have examined the relationship between CRP levels and CD relapse. One longitudinal study noted that patients in clinical remission with persistently elevated CRP levels were more likely to relapse within 2 years from study entry.35 Another study identified CRP as a predictor of relapse particularly, when included with ESR as part of a biological score.36 In contrast, a separate study failed to find an increase in CRP between 3 months and 1 month prior to a clinical relapse.37 In the present study, CRP levels when measured within 3 months prior to relapse were found to be a significant predictor.

Other serum markers (IL1β, sIL2R, IL6, IL10 TNFα and ASCA) were not found to be associated with relapse. Both IL6 and sIL2R have been reported as predictors of CD recurrence.3840 CRP hepatic production is regulated by IL6, and thus both are highly correlated.41 IL6 may not have been identified as a predictor of relapse because in the multivariate analysis it was accounted for by CRP. sIL2R was reported to be predictive of relapse previously.38 However, this latter study differed from ours in that sIL2R was assessed at a single time point only and included some patients on corticosteroid therapy.

Intestinal permeability has been proposed as a factor involved in the pathogenesis of CD and has been studied as a surrogate marker of mucosal inflammation.42 In contrast to the findings reported in several studies, intestinal permeability was not found to be a predictor of relapse in the present work.25 4345 One possible explanation for our negative finding is that 32% of patients had isolated colonic involvement and that the lactulose/mannitol test does not detect changes in colonic permeability. Other explanations may include differences across studies such as the inclusion of patients with surgically induced remission, the use of corticosteroids (both of which were exclusion criteria herein) as well as varying definitions of relapse. One study using [51Cr]EDTA, however, found no association between intestinal permeability and subsequent relapse.46

Controversy exists regarding the importance of stress in determining relapse in CD as few adequately powered prospective studies have evaluated the impact of stress or coping strategies for stress on subsequent relapse in individuals with quiescent CD.8 47 48 Consistent with other prospective reports which included CD cohorts, neither minor stressful events nor perceived global stress independently predicted relapse in this study.4952

Current stress and coping theory emphasises that coping modulates the effects of life stress and psychological factors on illness experience and its outcome by buffering or enhancing these effects.28 However, coping research in IBD has, until now, been limited to descriptive cross-sectional studies.53 54 In the present study, high stress influenced relapse for all patients. Importantly, an interaction between perceived stress and avoidance coping was found to predict time to relapse. Specifically, under conditions of low stress, patients who were less inclined to employ avoidance coping had the lowest risk for relapse compared with the other three groups. The name of the subscale (avoidance) is somewhat misleading as it does not refer to a type of denial or escaping from the stressor as such, but rather the CISS avoidance coping subscale items address an individual’s engagement in social diversion and distraction activities, to deal with stressful or difficult situations. Thus, persons scoring low on avoidance coping limit distractions of a social or individual nature. Given the social implications inherent in CD, our findings imply that patients at lowest risk for relapse are those who have little stress and restrict activities, thereby avoiding potentially embarrassing situations and/or overextending their limits. In contrast, patients under conditions of low stress who scored high on the avoidance coping subscale may not have been mindful of their chronic condition and subsequently were unable to set appropriate limits. The importance of the interaction between stress and coping is underscored by findings in a prospective study of patients with multiple sclerosis in remission where coping appears to modulate the effects of stress on disease activity as measured by the development of new brain lesions.55

Although not measured in the present work, psychological stress has been shown to elicit alterations in the levels of mediators of the sympathetic nervous system in patients with IBD.8 56 In addition, alterations in monoamines (norepinephrine, dopamine and serotonin) in the intestinal mucosa of patients with IBD have been reported.57 58 Interestingly, data from uncontrolled studies have shown that antidepressants that may act through these pathways may improve somatic symptoms in IBD.59 Future work in this field is required.

The strengths of this study included the longitudinal design with repeated measures in a pure CD patient population. This permitted a multivariate time-dependent analysis and allowed for the measurement of study parameters prior to relapse. One limitation of our study was the use of psychosocial measures that are not disease specific. However, such instruments were not available. Another potential limitation was the use of the CDAI as the outcome used to define relapse. The CDAI relies heavily on self-reported subjective parameters (abdominal pain, number of loose stools, general well-being) and therefore is subject to bias. Moreover, there remains the possibility that the observed increases in CDAI scores were driven by functional gastrointestinal symptoms (irritable bowel) which may be induced by stress and not CD relapse per se.60 However, until more robust parameters reflecting mucosal inflammatory activity are available, this will remain an inherent problem in all clinical studies using this widely accepted index.

Relapse in CD is subject to interindividual and intraindividual variability. Understanding which factors are related to disease exacerbation permits the identification of patients who may benefit from intensive maintenance therapy and would avoid unnecessary medical therapy in low risk individuals. The design of future interventional clinical trials would be enhanced by targeting a more homogeneous patient group based on risk of relapse. A fistulising disease behaviour and colonic site of disease were identified as predictors of earlier clinical relapse. CRP, which is an inexpensive marker and readily available to clinicians, can be useful in the routine management of patients with IBD, and assist in identifying those at risk of relapse. Doctors may inquire about their patients’ stress levels and coping strategies in order to distinguish those who may be at risk of relapse. Patients who score high on stress inventories and use ineffective coping methods may benefit from referral to an appropriate mental health professional for counselling or stress management.61 Although clinical trials to date have shown no net effect of psychotherapy on disease course, psychotherapy may have a beneficial effect on health-related quality of life and coping skills in patients with IBD.62 The present study supports the importance of the biopsychosocial approach in identifying patients at high risk of relapse in CD.


AB was supported by the Fonds de la Recherche en Santé du Québec. The authors would like to thank the research coordinators and nurses Patricia Papalia, Lisa Bonici, Marie-Luce Bernier, Linda Dion, Natalie Pellerin and Julie Rivard; their laboratory assistant Chloé Villani; and Dr S Suissa for statistical support. Presented as an oral presentation, Digestive Disease Week, IMIBD Distinguished Abstract Plenary Session 2005, Chicago, Illinois


View Abstract


  • See Commentary, p 1345

  • Funding: Supported by a research grant from the Crohn’s and Colitis Foundation of Canada.

  • Competing interests: None.

  • Ethics approval: The protocol was approved by the Institutional Review Boards of the participating hospitals.

  • Patient consent: Obtained.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles