SEARCH

SEARCH BY CITATION

Keywords:

  • inflammatory bowel disease;
  • environmental factors;
  • twins;
  • co-twin control study

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References

Background: Genetics and environmental factors are implicated in the etiology of inflammatory bowel disease (IBD). We studied environmental factors in a population-based Swedish-Danish twin cohort using the co-twin control method.

Subjects and Methods: A questionnaire was sent to 317 twin pairs regarding markers of exposures in the following areas: infections/colonization and diet as well as smoking, appendectomy, and oral contraceptives. Odds ratios (OR) were calculated by conditional logistic regression. When confounding appeared plausible, multivariate conditional logistic regression was added. The questions were also divided into topic groups, and adjustment was made for multiple testing within each of the groups.

Results: The response rate to the questionnaire was 83%. In consideration of the study design, only discordant pairs were included (Crohn's disease [CD], n = 102; ulcerative colitis [UC], n > = 125). Recurrent gastrointestinal infections were associated with both UC (OR, 8.0; 95% confidence interval [CI], 1.0–64) and CD (OR, 5.5; 95% CI, 1.2–25). Hospitalization for gastrointestinal infections was associated with CD (OR, 12; 95% CI, 1.6–92). Smoking was inversely associated with UC (OR, 0.4; 95% CI, 0.2–0.9) and associated with CD (OR, 2.9; 95% CI, 1.2–7.1).

Conclusions: The observed associations indicate that markers of possible infectious events may influence the risk of IBD. Some of these effects might be mediated by long-term changes in gut flora or alterations in reactivity to the flora. The influence of smoking in IBD was confirmed.

The etiology of inflammatory bowel disease (IBD) is unknown but is considered to be caused by interplay of genetic and environmental factors. A prevailing hypothesis is that the intestinal inflammation represents an inappropriate immune response to normal luminal bacteria in genetically susceptible individuals.1 The importance of genetics in IBD and especially in Crohn's disease (CD) is supported by the concordance for disease in twins2–6 and linkage with certain genomic loci and, more recently, specific genes.7 However, the rising incidence of IBD8–10 during the last decades with an increased occurrence of colonic CD8–10 emphasizes the etiologic importance of environmental factors. A combination of various exposures may conspire to cause clinical IBD in genetically susceptible individuals. Exposures in early life, such as initial patterns of gut colonization and infectious events, may influence mucosal immunity so that full tolerance to the normal gut flora is not established.11 Later infectious events might trigger or exacerbate inflammation and promote an inappropriate immune response to the normal gut flora.12 Diet may influence gut colonization, which may alter the risk of initiating an immune response to the flora. Diet may also have metabolic influences on immune function and the risk of initiating inflammation. It is proposed that thrombotic events and vasculitis may exacerbate the symptoms of CD, and oral contraceptives have been implicated in this.13 However, apart from smoking being protective against ulcerative colitis (UC) and promoting CD,14,15 data on other factors are inconsistent. This might be because the exposure is subtle and proxy markers are imprecise or that methodologic issues cloud the results.13

Monozygotic twins are genetically identical and share intrauterine environment and to a high extent other, but necessarily not all, external factors during childhood. Dizygotic twins share environmental factors but, on average, only half of the genes. The co-twin control method is suitable to study the etiologic influence of environmental factors.16 Discordant twin pairs are studied, in which the healthy co-twin serves as the control to the diseased twin. Because the discordant pairs, and especially the monozygotic pairs, are matched for genetic and largely for intrauterine and childhood environment,16 this design more robustly controls for confounding by these factors.

The aim of our co-twin control study was to examine markers of exposure before onset of clinical IBD in the following areas: infection/colonization and diet as well as smoking, appendectomy, and oral contraceptive use.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References

Twins

The twins derived from two Swedish and one Danish population-based cohort of twins with IBD, as earlier described.2-4,17 Twin pairs, in which both twins were alive, were sent a questionnaire concerning diagnosis of IBD, general gastrointestinal symptoms, and exposure to environmental factors before diagnosis of IBD. If a twin did not answer the questionnaire after a reminding letter, a telephone interview was made. After consent, the medical notes from twins with IBD or any history of gastrointestinal symptoms were scrutinized to verify or refute the diagnosis of IBD and to characterize the disease phenotypically. For CD, the Vienna classification18 was used, and in UC, age at diagnosis, disease location, and surgery were recorded. Zygosity classification was based on the method applied by the Swedish and Danish twin registries. It relies on questions on childhood resemblance and is accurate.16,19 The study was approved by the ethical committees of Örebro County and Copenhagen.

Questionnaire

The questionnaire, formulated in 1999, consisted of non-open-ended questions covering 25 different topics that had been proposed as risk factors for IBD. Before being sent to the twins, it was circulated for comments among the members of the International Organisation for study of Inflammatory Bowel Diseases. The questionnaire included questions regarding markers of infections/colonization such as early weaning, sanitary conditions, recurrent infections (defined as ≥3/year) including gastroenteritis or respiratory tract infections, tonsillectomy, cholecystectomy, traveling abroad, swimming environment, frequent antibiotic therapy (defined as treatment every or almost every year), pets, childhood diseases including measles, pertussis, rubella, chickenpox, mumps, or scarlet fever as well as vaccinations against measles, rubella, polio, diphtheria, tetanus, pertussis, or tuberculosis before age of 20 years. Dietary habits before diagnosis or in the healthy co-twin before the time of diagnosis in the diseased twin such as consumption of egg, fast food, bread, breakfast cereals, coffee, tea, juice, sugar, fruit, and vegetables were asked for. Smoking habits at diagnosis were registered and were classified as nonsmoking, ex-smoking, or active smoking, which was defined as daily consumption of tobacco for at least 6 months. Data on appendectomy before age 20 and before diagnosis of IBD were recorded. Only appendectomy without concomitant bowel resection was included. Use of oral contraceptives was similarly recorded. Furthermore, shared childhood major stressful events and physical activity before diagnosis were registered.

Statistics

Conditional logistic regression was used to calculate odds ratios (OR) with corresponding 95% confidence intervals (CI) for developing CD or UC according to exposure. P values in Table 2 were estimated using the log likelihood test in logistic regression, and this represents a test of homogeneity. This test will identify the lowest (statistically significant) P value for an association with any of the categories of the independent variable. The analyses were performed with one exposure variable at a time. The conditional logistic regression was performed twice, first including all discordant twin pairs and second with the discordant monozygotic pairs only, according to the co-twin design. The second regression analysis was performed to exclude possible confounding by genetic predisposition. If the regression did not converge because of sparse data, we instead used McNemar analysis for testing the association as well as exact Fischer's CI for OR.20 When associations were observed and confounding between different exposures appeared plausible from a biological perspective, multivariate conditional logistic regression was added. The questions were also divided into topic groups, and adjustment was made for multiple testing within each of the groups. For comparison of age, Mann-Whitney was used. All analyses were conducted using STATA 8.2. (Stata Statistical Software, Release 8.2. College Station, Texas).

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References

Twins

The study population consisted of 218 Swedish and 99 Danish twin pairs. In 263 of these 317 pairs, both twins responded to the questionnaire, giving a pair response rate of 83%. In 22 additional pairs, only one of the twins responded, and these pairs were excluded. In 119 pairs, the diagnosis was CD (concordant monozygotic n = 15, discordant monozygotic n = 32, concordant dizygotic n = l, discordant dizygotic n = 70, and discordant with unknown zygosity n = 1), whereas the diagnosis was UC in 141 twin pairs (concordant monozygotic n = 9, discordant monozygotic n = 45, concordant dizygotic n = 6, discordant dizygotic n = 80, and discordant with unknown zygosity n = 1). In the remaining three dizygotic pairs, one twin had CD and one UC. In consideration of the co-twin control study design, only discordant twin pairs with known zygosity were included (CD = 102 pairs, UC = 125 pairs). In all these pairs, the twins were brought up together to age 15 or older, except for three pairs with UC and two pairs with CD. The median (range) age at study entry was 46 (19–87) years and 46 (19–86) years in twin pairs with UC and CD, respectively. Demographic and clinical data on the diseased twins in the 227 discordant pairs with IBD are presented in Table 1.

Table 1. Demographic and Clinical Characteristics of Discordant Dizygotic and Monozygotic Twins with Inflammatory Bowel Disease from a Population-Based Swedish-Danish Cohort
Ulcerative ColitisNo. of Twins(%) (n = 125)Crohn's DiseaseNo. of Twins (%) (n = 102)
  • *

    * Time from diagnosis to survey.

Sex Sex 
Males62 (50)Males41 (40)
Females63 (50)Females61 (60)
Median age at diagnosis, yr (range)27 (5Y79)Median age at diagnosis, yr (range)25 (11–79)
Location of disease at diagnosis (%) Location of disease at diagnosis (%) 
Proctitis46 (37)Terminal ileum (L1)34 (33)
Left-sided disease28 (22)Colon (L2)34 (33)
Substantial46 (37)Ileocolon (L3)29 (28)
Unknown5 (4)Upper GI (L4)4 (4)
  Unknown1 (1)
Colectomy (%) Behavior (%) 
Yes41(33)Nonstricturing, nonpenetrating (B1)49 (48)
No84 (67)Stricturing (B2)29 (28)
  Penetrating (B3)22 (22)
  Unknown2 (2)
Median disease duration, yr (range)*17 (5Y53)Median disease duration, yr (range)*17 (0Y43)

Ulcerative Colitis

Markers of infections/colonization

Twins with UC more often reported recurrent gastrointestinal infections than their healthy co-twins (OR, 8.0; 95% CI, 1.0–64) (Table 2). Similarly, the diseased twins had more often, although not statistically significantly, been hospitalized for gastrointestinal infections (OR, 2.3; 95% CI, 0.6–9.0). Twins who reported recurrent gastrointestinal infections were younger at diagnosis compared with those who did not; median (range) age 21 (12–33) years versus 27 (5–79) years (P = 0.02).

Table 2. Environmental Factors in Twin Pairs Discordant for Inflammatory Bowel Disease
  Ulcerative Colitis Twins versus Healthy Co-TwinsCrohn's Disease Twins versus Healthy Co-Twins
  Di- and Monozygotic PairsMonozygotic PairsDi- and Monozygotic PairsMonozygotic Pairs
  UC/ControlOR95% CIP ValueUC/ControlOR95% CIP ValueCD/ControlOR95% CIP ValueCD/ControlOR95% CIP Value
  1. Nonresponders for each exposure are not shown and are reflected by missing number of twins in each subgroup.

  2. UC = ulcerative colitis; OR = odds ratio; CI = confidence interval; CD = Crohn's disease; GI = gastrointestinal; NS = not significant; ref = reference; inf = infinity.

Total 125/125   45/45   102/102   32/32   
Markers of infections and colonization before age 20 Recurrent GI infectionsYes9/28.01.0–640.0130/012/35.51.2–250.0091/0 
 No92/1121.0ref 34/421.0ref 81/911.0ref 28/291.0ref 
Hospitalization for GI infectionsYes9/52.30.6–9.00.202/31.00.1Y161.0012/1121.6–920.0011/0 
 No102/1121.0ref 36/391.0ref 82/941.0ref 28/301.0ref 
Recurrent respiratory infections, childhood infections, tonsillectomy, cholecystectomy, vaccinations, traveling abroad, and antibiotics    NS   NS   NS   NS
Mainly swimming in lakesYes54/402.81.1–7.20.01917/170.70.2–2.40.5345/431.10.4–2.70.4719/172.00.4–110.41
 No48/601.0ref 22/211.0ref 43/441.0ref 10/131.0ref 
Swimming in pool, sea, or river    NS   NS   NS   NS
Diet before diagnosis Intake of fruitDaily60/542.90.9–9.40.1222/191.40.1–260.5839/490.20.1–0.90.0528/140.20.01–3.00.14
 Weekly59/493.10.9–10 21/230.70.04–13 48/410.40.1–1.4 18/120.70.1–9.0 
 Less5/121.0ref 1/11.0ref 13/61.0ref 6/41.0ref 
 frequently                
Intake of vegetables    NS   NS   NS   NS
Intake of eggsDaily11/82.70.8–9.00.0264/22.60.4–190.634/60.40.1–2.60.630/00.71
 Weekly97/823.31.3–8.5 32/321.40.3–6.5 65/650.80.4–1.7 17/170.80.2–3.4 
 Less13/241.0ref 7/91.0ref 31/261.0ref 15/131.0ref 
 frequently                
Additional sugar on porridgeYes49/322.41.2–4.90.00817/122.50.8–8.00.1039/351.20.6–2.60.5811/91.20.4–3.90.76
 No74/851.0ref 27/321.0ref 61/611.0ref 21/211.0ref 
Additional sugar on breakfast cereals    NS   NS   NS   NS
Additional sugar in coffee or tea    NS   NS   NS   NS
Coffee, no. of cups per day3+40/580.10.03–0.40.00113/230.10.0–0.744/382.10.7–6.00.3814/14 
 1–241/370.2 0.06–0.817/140.0 0.0–3934/351.60.6–4.7 12/8  
 036/181.0ref 13/61.0ref 20/221.0ref 6/81.0ref 
Intake of tea, different types of bread, cereals, soft drinks juice, and fast food    NS   NS   NS   NS
Smoking habits at diagnosisSmoker25/440.40.2–0.90.0127/160.20.1–1.20.01153/422.91.2–7.10.03815/170.70.1–40
 Ex-smoker21/121.40.6–3.6 10/44.20.5–37 10/82.70.7–10 4/1inf0.2-inf 
 Nonsmoker79/691.0ref 28/251.0ref 39/521.0ref 13/141.0ref 
Appendectomy before age 20Yes3/40.50.1–2.70.411/06/42.00.4–110.411/30.50.04–5.50.56
 No116/1191.0ref 44/441.0ref 82/941.0ref 26/281.0ref 
Oral contraceptive use at diagnosisYes29/280.60.1–2.50.4810/81.00.1–161.0032/291.50.4–5.30.5311/10inf0.2-inf0.50
 No30/261.0ref 11/111.0ref 24/241.0ref 4/51.0ref 
Physical activity before diagnosis Regular physical activity    NS   NS   NS   NS

No differences were found regarding recurrent respiratory tract infections, tonsillectomy, cholecystectomy, or traveling abroad. Occurrence of childhood infections and vaccinations could not be evaluated because of the high frequency of twins not responding to these questions. UC twins more often reported swimming in lakes (OR, 2.8; 95% CI 1.1–7.2). No difference was observed regarding swimming in a pool, sea, or river.

Twenty-nine (12%) twins in the 125 discordant pairs had been treated frequently with antibiotics, without a statistically significant difference between UC twins and healthy co-twins (OR, 2.0; 95% CI, 0.7–5.8). After adjustment for multiple testing, statistical significance remained only for recurrent gastrointestinal infections (P = 0.05).

Diet

Before diagnosis, UC twins consumed more eggs than their co-twins (daily intake: OR, 2.7; 95% CI, 0.8–9.0; weekly: OR, 3.3; 95% CI, 1.3–8.5) (Table 2).

UC twins added more sugar on porridge, but not on breakfast cereals or in coffee or tea, and did not drink more soft drinks or juice. The diseased twins consumed less coffee than the co-twins (≥3 cups/day: OR, 0.1; 95% CI, 0.03–0.4; 1–2 cups/day: OR, 0.2; 95% CI, 0.06–0.8). When stratifying for smoking habits and analyzing nonsmokers only, the associations did not remain statistically significant (≥3 cups/day: OR, 0.22; 95% CI, 0.5–1.0; 1–2 cups/day: OR, 0.46; 95% CI, 0.1–1.9). When coffee consumption and smoking habits were included in a multivariate analysis, similar associations were also observed (>3 cups/day: OR, 0.13; 95% CI, 0.35–0.51; 1–2 cups/day: OR, 0.23; 95% CI, 0.63–0.85; active smoking: OR, 0.36; 95% CI, 0.16–0.85; ex-smoking: OR, 1.38; 95% CI, 0.47–4.1). No statistically significant differences were observed regarding consumption of any of the other specified dietary products (Table 2). After adjustment for multiple testing, statistical significance remained for adding sugar on porridge (P = 0.05) and consumption of coffee (P = 0.01).

Smoking

Healthy co-twins were smokers more often than UC twins at the time of diagnosis (OR, 0.4; 95% CI, 0.2–0.9) (Table 2). Smoking also appeared to reduce the risk of UC in the subgroup of monozygotic twins (OR, 0.2; 95% CI, 0.1–1.2). In contrast, being an ex-smoker slightly increased, although not statistically significantly, the risk for developing UC (OR, 1.4; 95% CI, 0.6–3.6).

Appendectomy

No statistically significant difference was observed between UC twins and the co-twins regarding appendectomy (OR, 0.5; 95% CI, 0.1–2.7) (Table 2).

Oral contraceptives

In 88% of the female twins, data on oral contraceptives were available (Table 2). Twenty-nine of 58 (50%) UC twins versus 28 of 53 (53%) healthy co-twins were users (OR, 0.6; 95% CI, 0.1–2.5).

Physical activities

There was no difference in physical activity between the twin groups.

Crohn's Disease

Markers of infections/colonization

CD twins more often reported recurrent gastrointestinal infections than the co-twins (OR, 5.5; 95% CI, 1.2–25), as well as hospitalization for such infections (OR, 12; 95% CI, 1.6–92) (Table 2). The median (range) age of twins with recurrent gastrointestinal infections was 21 (12–34) years at diagnosis of CD compared with 25 (11–79) years in those who disavowed this (P = 0.07).

There were no differences regarding recurrent respiratory tract infections, tonsillectomy, cholecystectomy, traveling abroad, or swimming habits. As in UC, occurrence of childhood infections and vaccinations could not be assessed because of the high frequency of twins not responding to these items. No difference was observed between CD twins and healthy co-twins regarding swimming habits. Thirty (15%) twins in the 102 discordant pairs with CD reported frequent antibiotic treatment before age 20, without statistically significant differences between CD twins and co-twins (OR, 1.3; 95% CI, 0.6–2.8). After adjustment for multiple testing, statistical significance remained for recurrent gastrointestinal infections (P = 0.04) as well as for hospitalization for such infections (P < 0.01).

Diet

Before diagnosis, twins with CD less often consumed fruits compared with their co-twins (daily intake: OR, 0.2; 95% CI, 0.1–0.9; weekly: OR, 0.4; 95% CI, 0.1–1.4) but no difference was found regarding consumption of vegetables (Table 2), and the association did not remain statistically significant after adjustment for multiple testing. No differences were found in sugar intake or in consumption of egg or any of the other specified dietary products (Table 2).

Smoking

CD twins more often were smokers than the co-twins (OR, 2.9; 95% CI, 1.2–7.1). A similar trend was observed in ex-smokers (OR, 2.7; 95% CI, 0.7–10). These findings could not be confirmed in the monozygotic pairs (Table 2).

Appendectomy

No statistically significant difference was observed between CD twins and unaffected co-twins regarding appendectomy (OR, 2.0; 95% CI, 0.4–11) (Table 2).

Oral contraceptives

Oral contraceptives were used by 56% of the women without any difference between the groups (OR, 1.5; 95% CI, 0.4–5.3) (Table 2).

Physical activities

No differences were seen between CD twins and the co-twins regarding physical activity.

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References

Although the understanding of the genetics of IBD has made major progress,7 the increase in incidence during the 20th century emphasizes the importance of environmental factors. The combination of several environmental exposures or characteristics may conspire to cause clinical IBD in genetically susceptible individuals. Here, we study markers of exposure regarding infections/colonization and diet as well as smoking, appendectomy, and oral contraceptive use before diagnosis of IBD. Previous studies have addressed a broad range of environmental factors, but data are, with few exceptions, conflicting. To avoid many of the problems with previous case-control studies,13,21,22 we conducted a co-twin control study. The methodology can minimize confounding by genetics, internal intrauterine factors, and external childhood environment,16 unless the twins were early separated.

There is evidence that an inappropriate immune response to the enteric flora is involved in the etiology of IBD.1 Colonization of the gastrointestinal tract starts immediately after birth11 and is of great importance for the composition of the permanent flora in adults.23 Furthermore, the intestinal microflora affects the gut-associated lymphoid tissue, and their interaction appears to be essential for development of appropriate regulatory circuits of the immune response.11 Environmental factors influencing the intestinal microflora include infections, diet, and antibiotics.24 In the present study, gastrointestinal infections and dietary habits were associated with IBD. Twins with UC and CD reported recurrent gastrointestinal infections before age 20 more often than their co-twins, and CD twins had more often been hospitalized for such infections. The lower age at diagnosis in those who reported recurrent gastrointestinal infections might reflect the fact that the infection precedes and triggers an inappropriate immune response, leading to clinical IBD or, if in early life, influences establishment of gut colonization and nascent immune tolerance. However, recurrent gastrointestinal infections were reported as three or more gastroenteritis per year before 20 years of age, and more detailed information was not available. Medical notes for the time period prediagnosis were not accessible in all twins. Therefore, the precise time period between recurrent gastroenteritis and diagnosis of IBD is unknown. When the analyses were restricted to twins diagnosed from 20 years of age only, an association between CD and hospitalization for gastrointestinal infections was still observed (OR, infinity; 95% CI, 1.44-infinity), thus supporting our hypothesis. However, the associations between recurrent gastrointestinal infections with or without hospitalization and CD as well as UC did not remain statistically significant. Thus, it cannot be ruled out that the observed risk is caused by early manifestations of IBD. Furthermore, twins reporting recurrent gastrointestinal infections also appeared to be younger at study entry than those who rejected this, median (range) age in CD 37 (29–56) years versus 47 (19–86) years (P = 0.03) and in UC 41.5 (27–60) years versus 47.5 (19–84) years (P = 0.07), increasing the risk of recall bias.

Dietary factors have been proposed as risk factors for IBD.14,15,25 Although the evidence is sparse, it can be hypothesized that diet alters the gut flora and thereby may trigger an inappropriate immune response, leading to IBD. However, dietary factors may have direct metabolic effects. It has recently been hypothesized that sulfur compounds in the diet impair use of butyrate by colonocytes and thus mediate relapse in UC.26 Of the major dietary sources of sulphur,26 only eggs were included in the present study. Interestingly, according to the univariate analysis, UC twins consumed more eggs than healthy co-twins.

Fruits and vegetables have been proposed to be protective for IBD. Yet, the observed associations were inconsistent, and previous data are conflicting.21,27–30 A high sugar intake has been associated with IBD and especially with CD.31,32 However, no consistent pattern of sugar intake could be observed. Dietary studies, aiming at reducing recall bias, report conflicting data.28,33–35 An inverse association between coffee consumption and UC was observed, but the association did not remain statistically significant after adjusting for smoking habits. Previous data are conflicting.27,30,36 There is no established biological mechanism, and any retrospective study is prone to reverse causation.

Recreational swimming in polluted water is related to gastrointestinal infections.37–39 Theoretically, these infections might trigger an inappropriate immune response, leading to clinical IBD. Clustering of CD in unrelated individuals exposed to polluted water and association with frequent use of a swimming pool have also been described.22,40 An association between UC and swimming in a lake but not in other bathing environments was seen. However, this association did not remain statistically significant when adjusting for multiple comparisons, and it needs to be further examined whether this reflects a true causation caused by differences in bacterial counts. No associations with the other studied markers of potential triggering infectious events could be observed.

Card et al.41 recently reported an increased use of antibiotics, 71% in CD patients versus 58% in controls, during 2 to 5 years before diagnosis of CD. We did not find any difference in antibiotic use up to age 20, and antibiotic use was overall less frequent than in the time period studied by Card et al.

In accordance with numerous studies,14,15,42,43 smoking was inversely associated, and ex-smoking appeared to be associated with UC. As expected,14,15,42,43 the opposite pattern was observed in CD, in which smoking was associated with the disease.

Appendectomy appears to protect against UC,30,44,45 even though the literature is not entirely consistent.46 It has recently been shown that only appendectomy before 20 years of age and caused by appendicitis or mesenteric lymphadenitis is inversely associated with UC, in contrast with appendectomy for other indications or at older age.47 In contrast, appendectomy appears to be associated with future CD.30,44,48 The observed ORs for UC and CD in our study are in accordance with the literature but are not statistically significant. This is probably because of the small numbers, as reflected by the wide CIs.

No association between use of oral contraceptives and UC or CD was observed in the twins. Most previous cohort and case control studies have pointed to a moderate association in CD and even less so in UC.49

The strength, but also the weakness, of the study is the co-twin control design. In the co-twin design, the ideal is to exclude genetic confounding by analyzing monozygotic twins only. This is the largest study of twins with IBD derived from a general population. Yet, it was difficult to reproduce the findings from the total twin group in the monozygotic pairs because of the limited numbers. To our knowledge, other unselected twin cohorts with IBD do not exist. Therefore, the material could only be increased by including selected twins with IBD. This approach, however, could cause recruitment bias and was therefore rejected. External controls are sometimes included in afirst step in the co-twin design to evaluate possible influence ofintrauterine or environmental factors in early childhood. Thisstrategy may reduce the number of included twins because of nonresponding controls, especially if non-twin siblings areused. Therefore, we excluded this step. Overmatching is amajor drawback of the co-twin design, and may explain absence of certain associations and sanitary conditions, pets, early weaning and major stressful events during childhood could due to this matching not be evaluated. In considering the close matching and the exclusion of risk-sets in which only one twin in each pair responded the questionnaire, the risk of differential bias is minimized. However, substantial problems with reporting bias and particularly recall bias remain and couldinfluence all observed associations, especially because median time since diagnosis was 17 years in both disorders. Reverse causation is another potential problem for most observed associations and could particularly influence dietary associations. Food preferences may change over time, and themedian time (range) of 0 (0–28) years and 1 (0–19) year between onset of symptoms and diagnosis of UC and CD, respectively, indicate that alterations in dietary preferences could already have occurred at diagnosis. On the other hand, a long prodromal period in IBD may hide true causative associations, and exposure might be of importance only during an age-specific window or a specific time interval before diagnosis.

Because of the unique possibilities offered by a study of twins with IBD, we collected information on a large number of exposures. When so many exposures are examined, there is always the possibility that a statistically significant association may arise by chance. To minimize this risk, we critically evaluated the consistency in the observed associations. Furthermore, we adopted the cautious approach of adjusting for multiple comparisons, although this may lead to underestimation of the “true“ associations when the original level of statistical significance is appropriate.50 Associations that did not remain statistically significant should therefore not be dismissed, but they should be treated with caution. Also, the fact that some of these associations have been reported by others supports that they might not be artifacts of multiple testing.

In summary, using the co-twin design, we studied markers of exposure regarding infections/colonization and diet as well as smoking, appendectomy, and oral contraceptive use before diagnosis of IBD. Recurrent gastrointestinal infections up to age 20 were associated with both UC and CD. These associations indicate that markers of possible infectious events during childhood/adolescence may influence the risk of IBD. Some of these effects might be mediated by long-term changes in gut flora or alterations in reactivity to the flora. In addition, previous reported associations between smoking and IBD were confirmed.

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References

Supported by grants from the International Organization for Study of Inflammatory Bowel Diseases (IOIBD), Bengt Ihre's Foundation, Danish Colitis-Crohn Foundation, Janssen-Cilag's Research Grant, Örebro University Hospital Research Foundation, and Örebro County Research Foundation.

References

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. References