Antibiotic Exposure and the Development of Celiac Disease
Antibiotic Exposure and the Development of Celiac Disease
In this case–control study we linked nationwide histopathology data on individuals undergoing small intestinal biopsy to the Swedish Prescribed Drug Register in order to examine the association between use of antibiotics and CD. We hypothesized a positive association between antibiotic use and CD.
A literature search at PubMed (http://pubmed.gov/) was performed using the following combinations of words as our major search terms: "celiac", "coeliac", "antibiotic" and "antimicrobial".
Between 2006 and 2008, we searched the computerized register of Sweden's 28 pathology departments to identify individuals with CD. In this study CD was defined as small-intestinal villous atrophy (Marsh grade 3). An earlier evaluation has shown that 95% of Swedish individuals with villous atrophy have CD. To examine the context of the association between antibiotic use and subsequent CD we also identified individuals with small-intestinal inflammation (Marsh grade 1–2) but without villous atrophy and individuals with normal small-intestinal mucosa (Marsh grade 0) but with positive CD serology. The biopsies were performed between July 1969 and January 2008. A detailed account of the data collection process has been described elsewhere.
In the current study we used the same dataset described in our previous study of mortality (29,096 individuals with CD, 13,306 individuals with inflammation, 3,719 individuals with normal mucosa but positive CD serology). Data on individuals with normal mucosa and positive CD serology were regional and obtained from the ascertainment areas of eight Swedish university hospitals covering approximately half of the Swedish population. Positive CD serology was defined as a positive IgA or IgG AGA (antigliadin), EMA (endomysial), or TTG (tissue transglutaminase) test less than 180 days before or no later than 30 days after a normal biopsy (and with no prior or subsequent biopsy showing villous atrophy or inflammation). In a recent consensus paper individuals with normal mucosa and positive CD serology were identified as having potential CD.
For each individual undergoing biopsy, the government agency Statistics Sweden identified up to five controls from the population matched for age, sex, calendar period of birth and county of residence. For example, a girl living in the county of Blekinge, diagnosed with CD in 2006 at the age of 13 years; was matched with five 13-year-old girls who were living in Blekinge in 2006. After exclusion of individuals with data irregularities, we identified 228,632 controls (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart of exclusion criteria. CD, Coeliac disease.
Individuals undergoing biopsy and their matched controls were then linked to the Swedish Prescribed Drug Register (established on July 1st 2005). Through this linkage, we identified individuals biopsied between July 1st 2005 and January 29th 2008 (end of the study period). Thus, the final analyses included 2,933 individuals with CD, 2,118 individuals with inflammation, 620 individuals with normal mucosa but positive CD serology and 28,262 controls (Figure 1).
The Swedish Prescribed Drug Register contains prospectively recorded individual data (on e.g. date of dispensing) on more than 99% of all dispensed prescribed drugs in Sweden. Antibiotics in Sweden are not sold over the counter.
We collected data on use of all systemic antibiotics (anatomical therapeutic chemical, ATC code: J01) from July 1st 2005 (launch of the Prescribed Drug Register) through January 29th 2008 (end of the study period), and up to the date of the biopsy (and the corresponding date in matched controls). Antibiotics were grouped into penicillin V, extended-spectrum penicillins, quinolones, macrolides and other systemic antibiotics (Additional file 1).
We used conditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Each stratum (one individual undergoing biopsy and up to five matched controls) was analyzed separately before a summary OR was calculated.
In our main analysis we examined the association between use of any systemic antibiotics and subsequent CD. Early-onset CD (i.e. before the age of 2 years) may have different risk factors compared with late-onset CD. Additionally, antibiotic exposure early in life may have a more profound impact on the composition of the microbiota. Accordingly, we performed stratified analyses by age at CD diagnosis (<2 years, 2–19 years, 20–39 years, 40–59 years and ≥60 years). We also stratified our analyses for sex. Similar sub-analyses were performed for individuals with small-intestinal inflammation and individuals with normal small-intestinal mucosa but positive CD serology. For each of these stratifications we examined for interaction via the inclusion in our models of multiplicative interaction terms, and the use of likelihood ratio tests between models with and without them.
Antibiotics differ in their influence on the intestinal microbiota. In pre-planned sub-analyses we estimated the association between CD and type of antibiotic exposure: penicillin V, extended-spectrum penicillins, quinolones, macrolides and other systemic antibiotics. This grouping of antibiotics has previously been used and is largely based on the ATC classification system where the subgroups indicate the different therapeutic indications of antibacterial agents. To evaluate potential causality we estimated the dose- and time-dependent association between antibiotic use and CD in two separate analyses: (1) when individuals had received 1–2 courses and at least 3 courses of antibiotics and (2) when antibiotics had been prescribed in the year (≤365 days) before biopsy.
Education level has been associated with antibiotic use and may influence the risk of CD diagnosis. In a sub-analysis we therefore adjusted for education using seven predefined education categories determined by Statistics Sweden.
Certain antiparasitic medications have similar pharmacokinetic and pharmacodynamic properties as systemic antibiotics with a strong impact on the gut microbiota. In a post-hoc analysis we therefore examined the relationship between use of any antiparasitic medications (ATC codes P01-P03, e.g. oral tinidazole) and CD, as well as specifically the use of metronidazole and CD.
Individuals with undiagnosed CD have an increased risk of comorbidity, and with that, potentially increased surveillance and probability to receive antibiotic treatment. In a post-hoc analysis we therefore restricted our data to individuals who had not been admitted to a hospital during the study period. Hospital admission data were collected from the national Inpatient Register.
To further reduce the risk of surveillance bias we constructed a variable representing outpatient health care consumption. Hospital-based outpatient care has been recorded nationally in Sweden since January 1st 2001. We calculated the number of hospital-based outpatient visits from birth or start of the registry (whichever occurred latest) until the day before small-intestinal biopsy (or corresponding date in matched controls). We excluded visits in which CD was coded as the main reason for the visit. Individuals then were divided into four groups according to number of visits per year (those with no record of prior hospital-based outpatient care (0); >0 but <1 visit/year; 1- < 2 visits/year; and ≥2 visits/year). Those individuals with no record of hospital-based outpatient care may have undergone initial CD investigation in primary care before undergoing biopsy. For example: A patient A, undergoing biopsy in December 2006, with eight hospital-based outpatient visits in the six years between 2001 (start of registration of outpatient data) and December 2006 (time of biopsy) has an average of 1.3 visits per year (= 8 visits/6 years). In a post-hoc analysis, we added this variable to our statistical model to evaluate whether the association between antibiotic exposure and CD remained.
CD is elicited by dietary gluten and thus virtually nonexistent before the age of six months. To establish whether antibiotic use truly preceded CD, i.e. to evaluate the risk of reverse causation, we performed a sub-analysis of individuals who were exposed to antibiotics before the age of six months. In an additional post-hoc analysis we limited our exposure to antibiotic more than one year (>365 days) before CD diagnosis.
SPSS version 20.0 was used for all statistical analyses.
This study was conducted in accordance with the national and institutional standards and was approved by the Regional Ethical Vetting Board in Stockholm.
Methods
In this case–control study we linked nationwide histopathology data on individuals undergoing small intestinal biopsy to the Swedish Prescribed Drug Register in order to examine the association between use of antibiotics and CD. We hypothesized a positive association between antibiotic use and CD.
Literature Search
A literature search at PubMed (http://pubmed.gov/) was performed using the following combinations of words as our major search terms: "celiac", "coeliac", "antibiotic" and "antimicrobial".
Study Population
Between 2006 and 2008, we searched the computerized register of Sweden's 28 pathology departments to identify individuals with CD. In this study CD was defined as small-intestinal villous atrophy (Marsh grade 3). An earlier evaluation has shown that 95% of Swedish individuals with villous atrophy have CD. To examine the context of the association between antibiotic use and subsequent CD we also identified individuals with small-intestinal inflammation (Marsh grade 1–2) but without villous atrophy and individuals with normal small-intestinal mucosa (Marsh grade 0) but with positive CD serology. The biopsies were performed between July 1969 and January 2008. A detailed account of the data collection process has been described elsewhere.
In the current study we used the same dataset described in our previous study of mortality (29,096 individuals with CD, 13,306 individuals with inflammation, 3,719 individuals with normal mucosa but positive CD serology). Data on individuals with normal mucosa and positive CD serology were regional and obtained from the ascertainment areas of eight Swedish university hospitals covering approximately half of the Swedish population. Positive CD serology was defined as a positive IgA or IgG AGA (antigliadin), EMA (endomysial), or TTG (tissue transglutaminase) test less than 180 days before or no later than 30 days after a normal biopsy (and with no prior or subsequent biopsy showing villous atrophy or inflammation). In a recent consensus paper individuals with normal mucosa and positive CD serology were identified as having potential CD.
For each individual undergoing biopsy, the government agency Statistics Sweden identified up to five controls from the population matched for age, sex, calendar period of birth and county of residence. For example, a girl living in the county of Blekinge, diagnosed with CD in 2006 at the age of 13 years; was matched with five 13-year-old girls who were living in Blekinge in 2006. After exclusion of individuals with data irregularities, we identified 228,632 controls (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart of exclusion criteria. CD, Coeliac disease.
Individuals undergoing biopsy and their matched controls were then linked to the Swedish Prescribed Drug Register (established on July 1st 2005). Through this linkage, we identified individuals biopsied between July 1st 2005 and January 29th 2008 (end of the study period). Thus, the final analyses included 2,933 individuals with CD, 2,118 individuals with inflammation, 620 individuals with normal mucosa but positive CD serology and 28,262 controls (Figure 1).
Antibiotic Use
The Swedish Prescribed Drug Register contains prospectively recorded individual data (on e.g. date of dispensing) on more than 99% of all dispensed prescribed drugs in Sweden. Antibiotics in Sweden are not sold over the counter.
We collected data on use of all systemic antibiotics (anatomical therapeutic chemical, ATC code: J01) from July 1st 2005 (launch of the Prescribed Drug Register) through January 29th 2008 (end of the study period), and up to the date of the biopsy (and the corresponding date in matched controls). Antibiotics were grouped into penicillin V, extended-spectrum penicillins, quinolones, macrolides and other systemic antibiotics (Additional file 1).
Statistical Analyses
We used conditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Each stratum (one individual undergoing biopsy and up to five matched controls) was analyzed separately before a summary OR was calculated.
In our main analysis we examined the association between use of any systemic antibiotics and subsequent CD. Early-onset CD (i.e. before the age of 2 years) may have different risk factors compared with late-onset CD. Additionally, antibiotic exposure early in life may have a more profound impact on the composition of the microbiota. Accordingly, we performed stratified analyses by age at CD diagnosis (<2 years, 2–19 years, 20–39 years, 40–59 years and ≥60 years). We also stratified our analyses for sex. Similar sub-analyses were performed for individuals with small-intestinal inflammation and individuals with normal small-intestinal mucosa but positive CD serology. For each of these stratifications we examined for interaction via the inclusion in our models of multiplicative interaction terms, and the use of likelihood ratio tests between models with and without them.
Antibiotics differ in their influence on the intestinal microbiota. In pre-planned sub-analyses we estimated the association between CD and type of antibiotic exposure: penicillin V, extended-spectrum penicillins, quinolones, macrolides and other systemic antibiotics. This grouping of antibiotics has previously been used and is largely based on the ATC classification system where the subgroups indicate the different therapeutic indications of antibacterial agents. To evaluate potential causality we estimated the dose- and time-dependent association between antibiotic use and CD in two separate analyses: (1) when individuals had received 1–2 courses and at least 3 courses of antibiotics and (2) when antibiotics had been prescribed in the year (≤365 days) before biopsy.
Education level has been associated with antibiotic use and may influence the risk of CD diagnosis. In a sub-analysis we therefore adjusted for education using seven predefined education categories determined by Statistics Sweden.
Post-hoc Analyses
Certain antiparasitic medications have similar pharmacokinetic and pharmacodynamic properties as systemic antibiotics with a strong impact on the gut microbiota. In a post-hoc analysis we therefore examined the relationship between use of any antiparasitic medications (ATC codes P01-P03, e.g. oral tinidazole) and CD, as well as specifically the use of metronidazole and CD.
Individuals with undiagnosed CD have an increased risk of comorbidity, and with that, potentially increased surveillance and probability to receive antibiotic treatment. In a post-hoc analysis we therefore restricted our data to individuals who had not been admitted to a hospital during the study period. Hospital admission data were collected from the national Inpatient Register.
To further reduce the risk of surveillance bias we constructed a variable representing outpatient health care consumption. Hospital-based outpatient care has been recorded nationally in Sweden since January 1st 2001. We calculated the number of hospital-based outpatient visits from birth or start of the registry (whichever occurred latest) until the day before small-intestinal biopsy (or corresponding date in matched controls). We excluded visits in which CD was coded as the main reason for the visit. Individuals then were divided into four groups according to number of visits per year (those with no record of prior hospital-based outpatient care (0); >0 but <1 visit/year; 1- < 2 visits/year; and ≥2 visits/year). Those individuals with no record of hospital-based outpatient care may have undergone initial CD investigation in primary care before undergoing biopsy. For example: A patient A, undergoing biopsy in December 2006, with eight hospital-based outpatient visits in the six years between 2001 (start of registration of outpatient data) and December 2006 (time of biopsy) has an average of 1.3 visits per year (= 8 visits/6 years). In a post-hoc analysis, we added this variable to our statistical model to evaluate whether the association between antibiotic exposure and CD remained.
CD is elicited by dietary gluten and thus virtually nonexistent before the age of six months. To establish whether antibiotic use truly preceded CD, i.e. to evaluate the risk of reverse causation, we performed a sub-analysis of individuals who were exposed to antibiotics before the age of six months. In an additional post-hoc analysis we limited our exposure to antibiotic more than one year (>365 days) before CD diagnosis.
SPSS version 20.0 was used for all statistical analyses.
Ethics
This study was conducted in accordance with the national and institutional standards and was approved by the Regional Ethical Vetting Board in Stockholm.