Microbial Exposure in Infancy and Subsequent Appearance of Type 1 Diabetes Mellitus Associated Autoantibodies A Cohort Study

Transcription

1 Research Original Investigation Microbial Exposure in Infancy and Subsequent Appearance of Type 1 Diabetes Mellitus Associated Autoantibodies A Cohort Study Suvi M. Virtanen, MD, PhD; Hanna-Mari Takkinen, MSc; Bright I. Nwaru, MPhil, PhD; Minna Kaila, MD, PhD; Suvi Ahonen, MSc; Jaakko Nevalainen, PhD; Sari Niinistö, MSc; Heli Siljander, MD, PhD; Olli Simell, MD, PhD; Jorma Ilonen, MD, PhD; Heikki Hyöty, MD, PhD; Riitta Veijola, MD, PhD; Mikael Knip, MD, PhD IMPORTANCE The role of microbial exposure during early life in the development of type 1 diabetes mellitus is unclear. Supplemental content at jamapediatrics.com OBJECTIVE To investigate whether animal contact and other microbial exposures during infancy are associated with the development of preclinical and clinical type 1 diabetes. DESIGN, SETTING, AND PARTICIPANTS A birth cohort of children with HLA antigen DQB1 conferred susceptibility to type 1 diabetes was examined. Participants included 3143 consecutively born children at 2 hospitals in Finland between 1996 and EXPOSURES The following exposures during the first year of life were assessed: indoor and outdoor dogs and cats, farm animals, farming, visit to a stable, day care, and exposure to antibiotics during the first week of life. MAIN OUTCOMES AND MEASURES Clinical and preclinical type 1 diabetes were used as outcomes. The latter was defined as repeated positivity for islet-cell antibodies plus for at least 1 of 3 other diabetes-associated autoantibodies analyzed and/or clinical type 1 diabetes. The autoantibodies were analyzed at 3- to 12-month intervals since the birth of the child. RESULTS Children exposed to an indoor dog, compared with otherwise similar children without an indoor dog exposure, had a reduced odds of developing preclinical type 1 diabetes (adjusted odds ratio [OR], 0.47; 95% CI, ; P =.005) and clinical type 1 diabetes (adjusted OR, 0.40; 95% CI, ; P =.08). All of the other microbial exposures studied were not associated with preclinical or clinical diabetes: the odds ratios ranged from 0.74 to CONCLUSIONS AND RELEVANCE Among the 9 early microbial exposures studied, only the indoor dog exposure during the first year of life was inversely associated with the development of preclinical type 1 diabetes. This finding needs to be confirmed in other populations. JAMA Pediatr. 2014;168(8): doi: /jamapediatrics Published online June 23, Author Affiliations: Author affiliations are listed at the end of this article. Corresponding author: Suvi M. Virtanen, MD, PhD, Nutrition Unit, National Institute for Health and Welfare, PO Box 30, FI Helsinki, Finland (suvi.virtanen@thl.fi). 755

2 Research Original Investigation The hygiene hypothesis that microbial exposure during early life may affect susceptibility to the development of allergic diseases 1 has been extended to investigate whether the hypothesis is true for autoimmune diseases, such as type 1 diabetes. 2 Decreased environmental biodiversity through influence on human microbiota and its immunologic actions was recently implicated 3 in the occurrence of atopic and other chronic inflammatory diseases. Affluent and sparsely populated environments or less microbial diversity of the gut have been associated 4 with the occurrence of type 1 diabetes. Increased maternal and child use of antibiotics was associated 5 with an increased risk of type 1 diabetes in children. We set out to investigate whether contacts with animals or other microbial exposures during infancy are associated with the development of clinical or preclinical type 1 diabetes. Preclinical type 1 diabetes was defined as repeated positivity for both islet-cell antibodies and at least 1 of the 3 other autoantibodies analyzed and/or clinical type 1 diabetes. Methods Study Design and Participants We studied children belonging to the Oulu and Tampere cohorts of the Finnish Type 1 Diabetes Prediction and Prevention study. This multidisciplinary, population-based cohort study examines potential means to predict and prevent the manifestation of type 1 diabetes. 6 Infants born with HLA antigen conferred susceptibility to type 1 diabetes are monitored at 3- to 12-month intervals for diabetes-associated autoantibodies, growth, and environmental exposures until age 15 years or diagnosis of type 1 diabetes. The study procedures were approved by the local ethics committees and the parents provided written informed consent. Participants did not receive financial compensation. The Finnish Type 1 Diabetes Prediction and Prevention Nutrition Study started in Oulu (Northern Botnia province, northern Finland) in 1996 and in Tampere (Pirkanmaa province, south-mid Finland) in Children born between September 2, 1996, and August 31, 2004, at Oulu University Hospital and between October 20, 1997, and September 5, 2004, at Tampere University Hospital were screened for HLA antigen DQB1 conferred susceptibility to type 1 diabetes using cord blood samples. 8 All deliveries in the area of these provinces took place in Oulu and Tampere university hospitals. Infants carrying increased genetic susceptibility (either high risk [HLA-DQB1* heterozygous] or moderate risk [DQB1*0302 x-positive; x indicates homozygosity or a neutral allele]) were invited to participate in the nutrition study (N = 7782). The nutrition study examines the relationship of maternal diet during pregnancy and lactation and the child s diet during infancy and childhood with the development of type 1 diabetes, allergic diseases, and asthma in childhood. Of the children invited, 6078 participated in the assessment of type 1 diabetes end points. Of these, 4075 children who were participating in the dietary follow-up at age 5 years and had not received a diagnosis of clinical type 1 diabetes were invited to take part in the allergy study by completing a modified version of the International Study of Asthma and Allergies in Childhood questionnaire 9 (named in the following allergy questionnaire) that included, in addition to asthma and allergy of the child and parents, questions on several microbial exposures during childhood. A total of 3143 children (77%) returned the allergy questionnaire. The flow diagram of the study participants is presented in the Figure. Microbial Exposure The allergy questionnaire was completed by the parents when the child was aged 5 years. The following microbial exposures reported to have occurred during the first year of life were used in the analysis (all with yes/no response options): indoor/ outdoor dog exposure, indoor/outdoor cat exposure, farm animal (cows, horses, sheep, goats, hens, other) exposure, farming occupation of the family, visit to a stable, and day care attendance. In Finland, dogs that are used for purposes such as hunting usually stay outside the house and may have little or no contact with children. The same is true for cats used to catch mice in a stable. Data on exposure to antibiotics during the first week of life (yes/no) were obtained from the medical birth registries of the 2 hospitals. Sociodemographic and Clinical Characteristics Information on the child's sex, maternal age, maternal professional educational level, and the number of siblings was recorded in a structured questionnaire completed by the parents after delivery. Information on the duration of gestation, mode of delivery, birth weight, and maternal smoking during pregnancy was retrieved from the medical birth registries of the 2 hospitals. Home municipality urbanization level was categorized according to Statistics Finland guidelines as rural, semiurban, and urban. Questions on parental history of asthma and allergic rhinitis and the child s asthma, allergic rhinitis, and atopic eczema were also assessed in the allergy questionnaire. A child s asthma was defined as physician-diagnosed asthma with either any wheezing symptoms or use of asthma medication during the preceding 12 months. Allergic rhinitis was defined as sneezing, nasal congestion, or rhinitis other than with respiratory infections, accompanied by itching of the eyes and tearing during the previous 12 months. Atopic eczema was defined as atopic eczema ever diagnosed by a physician. Outcome Assessment Clinical and preclinical type 1 diabetes were used as outcomes. The latter was defined as repeated positivity for islet-cell antibodies plus for at least 1 of 3 other diabetesassociated autoantibodies and/or clinical type 1 diabetes. The definition used for preclinical type 1 diabetes is closely related to clinical disease: approximately 60% to 70% of children with preclinical diabetes are expected to progress to the clinical disease within the next 8 years. 10 When a child seroconverted to positivity for islet-cell antibodies for the first time, all of the child s preceding (starting from birth) and subsequent serum samples were analyzed for insulin autoantibodies, autoantibodies to glutamic acid decarboxylase, and insulinoma-associated 2 molecule. Islet-cell antibodies were 756 JAMA Pediatrics August 2014 Volume 168, Number 8 jamapediatrics.com

3 Original Investigation Research Figure. Flow Diagram of the Study Population 7782 Children carried increased genetic risk for type 1 diabetes mellitus 3916 Tampere 3866 Oulu 1704 Children did not participate in the assessment of type 1 diabetes end points Imputation analysis 6078 Children had type 1 diabetes end point information available 3175 Tampere 2903 Oulu Sensitivity analysis 2740 Children had pet information available from Tampere University Hospital patient records at age 1 y 32 Children developed clinical type 1 diabetes before age 5 y 592 Children were lost to follow-up or did not participate at age 5 y 86 Children developed clinical type 1 diabetes before age 5 y 1917 Children were lost to follow-up or did not participate at age 5 y 4075 Children were invited to take part in the allergy study at age 5 y 2208 Tampere 1867 Oulu 932 Children did not return the allergy questionnaire Complete case analysis 3143 Children returned the allergy questionnaire 1888 Tampere 1255 Oulu The formation of the study populations for multiple imputation analysis, sensitivity analysis, and complete case analysis is presented. detected by a standard indirect immunofluorescence method. Insulin autoantibodies were quantified with a microassay and glutamic acid decarboxylase and insulinoma-associated 2 molecule autoantibodies with specific radiobinding assays. 11 Maternal autoantibodies transferred through the placenta were excluded from the analyses. In the present study we used follow-up for type 1 diabetes associated autoantibodies and type 1 diabetes up to January 1, Follow-up for clinical diabetes was available for all children who started the study (n = 6078) until January Sensitivity Analysis Data Series For a sensitivity analysis data series we used 2740 participant records, which were based on prospective interviews conducted by study nurses at the 12-month visit and were available only for children observed at Tampere University Hospital. Study nurses recorded on an open-ended table whether the family had indoor animals and, if so, which animals during the first year of the child s life. Indoor animals other than dogs and cats were infrequent and were not included. The other microbial exposures were not available from the sensitivity analysis data series. Statistical Analysis Logistic regression was used to estimate the associations between microbial exposure and preclinical and clinical type 1 diabetes. The possible dependence among siblings was accounted for by using the generalized estimating equations with the sandwich estimator of variance to estimate regression coefficients in logistic regression analyses. 12 Potential confounders adjusted in the models were sex, genetic risk according to HLA-DQB1 (moderate/high), family history of diabetes mellitus (yes/no), mode of delivery (vaginal/cesarean section), place of birth (Tampere/Oulu), parental asthma or allergic rhinitis jamapediatrics.com JAMA Pediatrics August 2014 Volume 168, Number 8 757

4 Research Original Investigation (yes/no), maternal professional educational level (none, professional education/course, secondary professional education, university), maternal age (<25, 25-29, 30-34, and 35 years), home municipality urbanization level (rural, semiurban, urban), and the presence of asthma (yes/no) and atopic eczema (yes/no) in the child by the age of 5 years. The confounders were chosen based on earlier findings from the literature 13 (eg, mode of delivery) or/and being associated with (P <.20) both the exposure and the end point in the present study. Analyses were performed for all observed exposure data (complete case analysis), for exposure data obtained from Tampere University Hospital medical records (sensitivity analysis), and on multiply imputed data (Supplement [eappendix]). Multiple imputation was performed to correct for potential biases associated with missing values in explanatory variables. 14 There was a considerable chance for bias for 2 reasons: (1) the exposure variables were not obtained for the type 1 diabetes cases that were diagnosed before the allergy questionnaire assessment and, according to the study design, were ineligible for further followup; and (2) other dropouts by the age of 5 years. Only 3143 of the 6078 children (51.7%) for whom type 1 diabetes end point information was available had returned the allergy questionnaire, resulting in a high proportion of missing values on the microbial exposures. To make multiple imputations for these variables as effective as possible, surrogate data from Tampere University Hospital medical records were used in the imputation models. Because information on only indoor dog and cat exposure was obtained from the medical records, multiple imputation was done only for those 2 microbial exposures. All other variables involved in the imputation model were also imputed themselves. A detailed description of the imputation process is presented in the Supplement (eappendix). The κ coefficients for the agreement between indoor dog and cat exposure during the first year of life obtained from the allergy questionnaire and participants medical records were calculated. Analyses were performed using SAS, version 9.2 (SAS Institute Inc). Results Complete Case Analysis Table 1 reports the baseline characteristics and allergic diseases of the participating children according to indoor dog and cat exposure in the complete case data. Preclinical type 1 diabetes was present in 5.5% of the children (173 of 3143) and clinical type 1 diabetes was present in 1.6% of the children (51 of 3143) in the complete case data set. The preclinical outcome was based on clinical diabetes in 2 children. The median (range) time of manifestation of preclinical type 1 diabetes was 5.0 ( ) years, and that of clinical type 1 diabetes was 9.2 ( ) years. The median (range) length of follow-up was 10.0 ( ) years for preclinical and 12.2 ( ) for clinical type 1 diabetes. The frequency of preclinical and clinical type 1 diabetes was higher among boys, children carrying high- compared with moderate-risk HLA genotype, those with a family history of type 1 diabetes, and children whose mothers had not received professional education. Clinical type 1 diabetes was more common among children of older mothers. Preclinical type 1 diabetes was more prevalent among children with than without asthma, whereas the opposite was true for atopic eczema (data not shown). The farming occupation of the family, visit to a stable, and day care attendance during the first year of life, as well as the use of antibiotics during the first week of life, were not significantly associated with the outcomes, with adjusted odds ratios (ORs) for preclinical type 1 diabetes varying from 1.15 to 1.56 and for clinical type 1 diabetes from 0.74 to 1.58 (Table 2). Of the indoor and outdoor dog and cat and farm animal contacts during the first year of life, only indoor dog exposure was inversely associated with the development of preclinical type 1 diabetes in the complete case analysis (adjusted OR, 0.47; 95% CI, ; P =.005) (Table 2). Indoor dog exposure had an OR of 0.40 (95% CI, ; P =.08) for clinical type 1 diabetes. The OR for outdoor dog exposure was 1.14 for preclinical and 1.15 for clinical disease; for other animal exposures, ORs varied from 0.93 to 1.02 and from 0.79 to 1.34, respectively (Table 2). Sensitivity Analysis In the sensitivity analysis series of 2740 children from Tampere University Hospital, preclinical type 1 diabetes was present in 6.2% (169) and clinical type 1 diabetes was present in 3.0% (82) of the children. The median (range) time of manifestation was 4.0 ( ) years for preclinical type 1 diabetes and 5.7 ( ) years for clinical type 1 diabetes. The median (range) length of follow-up was 8.1 ( ) years for preclinical and 11.7 ( ) for clinical type 1 diabetes. We obtained data on indoor dog and cat exposure during the first year of life from both the allergy questionnaire (complete case data) and from Tampere patient records (sensitivity analysis data) for 1843 children. Of these, 19.9% (367) were exposed to a dog according to the allergy questionnaire and 18.5% (341) were exposed according to patient records; 47 children were exposed to a dog according to only the allergy questionnaire and 21 according only to patient records. The κ coefficient for the agreement between the methods was The κ coefficient for the agreement of cat exposure was The sensitivity analysis showed similar ORs compared with complete cases analysis for the association between indoor dog exposure and preclinical type 1 diabetes (OR 0.58; 95% CI, ; P =.09) and clinical type 1 diabetes (OR, 0.17; 95% CI, ; P =.09) (Table 3). The proportions of disease outcomes by cat exposure were similar (Table 3) according to both types of analysis. Multiple Imputation Analysis Distributions of preclinical and clinical type 1 diabetes and background characteristics among children with complete and incomplete data on indoor dog and cat exposure during the first year of life are presented in the Supplement (etable 1). After adjustment for potential confounding variables, indoor dog exposure was inversely associated with the development of preclinical type 1 diabetes in the multiple imputation data analysis (OR, 0.54; 95% CI, ; P =.01), but not significantly associated with clinical type 1 diabetes 758 JAMA Pediatrics August 2014 Volume 168, Number 8 jamapediatrics.com

5 Original Investigation Research Table 1. Characteristics of the Study Population According to Indoor Dog and Cat Exposure During the First Year of Life Characteristic Sex Total (N = 3143) No (n = 2461) Indoor Dog Exposure Yes No (n = 592) P Value a (n = 2679) Indoor Cat Exposure Female 1497 (47.6) 1185 (48.2) 272 (46.0) 1278 (47.7) 179 (47.9) Male 1646 (52.4) 1276 (51.8) 320 (54.0) (52.3) 195 (52.1) HLA-DQB1 conferred risk Moderate, DQB1*0302/x b 2527 (80.4) 1980 (80.5) 473 (79.9) 2160 (80.6) 293 (78.3) High, DQB1*02/* (19.6) 481 (19.5) 119 (20.1) (19.4) 81 (21.7) Family history of diabetes mellitus c No 2860 (93.6) 2242 (93.5) 531 (93.2) 2439 (93.5) 334 (93.0) Yes 196 (6.4) 155 (6.5) 39 (6.8) (6.5) 25 (7.0) Missing information 87 Parental asthma or allergic rhinitis No 976 (32.7) 727 (31.1) 215 (38.1) 781 (30.6) 161 (46.1) Yes 2010 (67.3) 1607 (68.9) 350 (61.9) (69.4) 188 (53.9) Missing information 157 Maternal age, y c < (14.8) 337 (13.7) 118 (19.9) 387 (14.5) 68 (18.2) (35.0) 843 (34.3) 231 (39.0) 956 (35.7) 118 (31.6) (31.0) 786 (31.9) 153 (25.8) 827 (30.9) 112 (30.0) (19.2) 495 (20.1) 90 (15.2) 509 (19.0) 76 (20.3) Maternal professional educational level c None 157 (5.1) 108 (4.5) 44 (7.6) 127 (4.8) 25 (6.8) Professional education/course 808 (26.2) 612 (25.2) 177 (30.7) 664 (25.2) 125 (34.1) Secondary professional education 1434 (46.4) 1125 (46.4) 257 (44.5) 1237 (46.9) 145 (39.5) University 690 (22.3) 280 (23.9) 99 (17.2) 607 (23.0) 72 (19.6) Missing information 54 Maternal smoking during pregnancy No 2812 (92.5) 2218 (93.0) 514 (90.5) 2405 (92.8) 327 (90.3) Yes 229 (7.5) 167 (7.0) 54 (9.5) (7.2) 35 (9.7) Missing information 102 No. of siblings c (46.0) 1065 (44.0) 322 (55.7) 1186 (45.1) 201 (55.1) (32.0) 797 (33.0) 164 (28.4) 859 (32.7) 102 (28.0) (13.7) 343 (14.2) 61 (10.6) 368 (14.0) 36 (9.9) (8.4) 213 (8.8) 31 (5.4) 218 (8.3) 26 (7.1) Missing information 59 Mode of delivery Vaginal 2701 (86.5) 2113 (86.4) 510 (87.2) 2295 (86.3) 328 (87.9) Cesarean section 420 (13.5) 333 (13.6) 75 (12.8) (13.7) 45 (12.1) Missing information 22 Place of birth Tampere University Hospital 1888 (60.1) 1508 (61.3) 376 (63.5) 1628 (60.8) 256 (68.4) Oulu University Hospital 1255 (39.9) 953 (38.7) 216 (36.5) (39.2) 118 (31.6) Home municipality urbanization c Rural 400 (12.8) 282 (11.5) 94 (16.1) 310 (11.7) 66 (17.7) Semiurban 305 (9.8) 226 (9.2) 73 (12.5) 253 (9.5) 46 (12.3) Urban 2416 (77.4) 1938 (79.2) 418 (71.5) 2095 (78.8) 261 (70.0) Missing information 22 Yes (n = 374) P Value a (continued) jamapediatrics.com JAMA Pediatrics August 2014 Volume 168, Number 8 759

6 Research Original Investigation Table 1. Characteristics of the Study Population According to Indoor Dog and Cat Exposure During the First Year of Life (continued) Characteristic Gestational age at birth, wk Total (N = 3143) No (n = 2461) Indoor Dog Exposure Yes No (n = 592) P Value a (n = 2679) Indoor Cat Exposure < (6.2) 150 (6.1) 37 (6.4) 159 (6.0) 28 (7.5) (93.8) 2290 (93.9) 545 (93.6) (94.0) 344 (92.5) Missing information 31 Birth weight <2500 g 138 (4.4) 106 (4.3) 29 (5.0) 115 (4.3) 20 (5.4) 2500 g 2983 (95.6) 2340 (95.7) 556 (95.0) (95.7) 353 (94.6) Missing information 22 Asthma d No 2949 (93.8) 2302 (93.5) 564 (95.3) 2516 (93.9) 350 (93.6) Yes 194 (6.2) 159 (6.5) 28 (4.7) (6.1) 24 (6.4) Allergic rhinitis d No 2670 (85.8) 2079 (85.2) 512 (88.1) 2271 (85.7) 320 (86.0) Yes 442 (14.2) 362 (14.8) 69 (11.9) (14.3) 52 (14.0) Missing information 31 Atopic eczema d No 2325 (74.8) 1809 (74.1) 451 (76.8) 1961 (73.8) 299 (80.4) Yes 785 (25.2) 632 (25.9) 136 (23.2) (26.2) 73 (19.6) Missing information 33 a χ 2 and Fisher exact tests were used. b x is not equal to *02, *0301, or *0602. c At the time of the child s birth. d By the age of 5 years. Yes (n = 374) P Value a (Supplement [etable 2]). Indoor cat exposure was not significantly associated with the development of preclinical or clinical type 1 diabetes (Supplement [etable 2]). Discussion In the present study, several farming- and animal contact related factors, day care attendance, and antibiotic use during early infancy were studied. Of these factors, only indoor dog exposure was inversely associated with subsequent development of preclinical type 1 diabetes. This association was consistently seen in both complete case and multiple imputation data analyses. Outdoor dog exposure showed a nonsignificant OR (1.14) with preclinical diabetes outcome. The birth cohort of children with increased genetic susceptibility to type 1 diabetes was closely observed at 3- to 12- month intervals with frequent measurements of the type 1 diabetes end points. The participation and retention rates of the study were good. Although the number of children with clinical type 1 diabetes was relatively small, the number of preclinical end points was moderate. The questions about animal contacts, farming, and day care were retrospectively asked with detailed and validated questions when the child was aged 5 years. We could show in a subpopulation that the reliability of the indoor dog and cat contact data collected at that time was good in comparison with the data collected from patient records when the child was aged 12 months. Participation in our study ended if the child received a diagnosis of type 1 diabetes. Consequently, our complete case data source for animal and other microbial data collected at age 5 years did not include children who developed diabetes at a younger age. To decrease the probability of outcome-dependent bias we conducted the analysis in 2 further data sets: evaluating patient records (sensitivity analysis) available for 1 of the clinics and undertaking multiple imputation for the whole cohort, using partially observed data. Both data sets included children with a diagnosis of type 1 diabetes when they were younger than 5 years (Figure). In the sensitivity analysis data, the ORs for indoor dog and cat exposure were similar to those of complete case analysis and multiple imputation analysis. In sensitivity analysis, we found no indication that the risk associated with dog or cat exposure would have been different among children younger than 5 years compared with older children. The numbers of children with outdoor and farm animal exposure, farming occupation of the family, or early use of antibiotics were low and CIs were wide, making the estimates uncertain. However, the only statistically significant finding in the present study, the inverse association between indoor dog exposure and preclinical type 1 diabetes, can be a spurious finding. We used multiple imputation to try to address the prob- 760 JAMA Pediatrics August 2014 Volume 168, Number 8 jamapediatrics.com

7 Original Investigation Research Table 2. Complete Case Analysis: Microbial Exposure During the First Year of Life and Preclinical and Clinical Type 1 Diabetes Preclinical Type 1 Diabetes Microbial Exposure During the First Year Adjusted OR of Life No Yes (95% CI) a,b,c P Value No Yes Indoor dog No 2310 (80.1) 151 (89.3) 1 [Reference].005 Clinical Type 1 Diabetes 2414 (80.4) 47 (92.2) 1 [Reference] Adjusted OR (95% CI) a,b,c P Value Yes 574 (19.9) 18 (10.7) 0.47 ( ) 588 (19.6) 4 (7.8) 0.40 ( ) Indoor cat No 2531 (87.8) 148 (87.6) 1 [Reference] 2635 (87.8) 44 (86.3) 1 [Reference] Yes 353 (12.2) 21 (12.4) 0.93 ( ) (12.2) 7 (13.7) 1.34 ( ) Outdoor dog No 2676 (93.5) 154 (92.2) 1 [Reference] 2782 (93.4) 48 (94.1) 1 [Reference] Yes 187 (6.5) 13 (7.8) 1.14 ( ) (6.6) 3 (5.9) 1.15 ( ) Outdoor cat No 2750 (96.1) 160 (95.8) 1 [Reference] 2859 (96.0) 51 (100) Yes 113 (3.9) 7 (4.2) 1.02 ( ) (4.0) 0 NA Farm animal No 2749 (96.0) 161 (96.4) 1 [Reference] 2860 (96.0) 50 (98.0) 1 [Reference] Yes 114 (4.0) 6 (3.6) 1.00 ( ) > (4.0) 1 (2.0) 0.79 ( ) Farming occupation of family No 2842 (96.0) 165 (95.4) 1 [Reference] 2958 (95.9) 49 (96.1) 1 [Reference] Yes 119 (4.0) 8 (4.6) 1.48 ( ) (4.1) 2 (3.9) 1.58 ( ) Visit to a stable No 2342 (81.7) 136 (80.5) 1 [Reference] 2437 (81.7) 41 (80.4) 1 [Reference] Yes 524 ( (19.5) 1.16 ( ) (18.3) 10 (19.6) 1.43 ( ) Day care attendance No 2597 (87.4) 149 (86.1) 1 [Reference] 2700 (87.3) 46 (90.2) 1 [Reference] Yes 373 (12.6) 24 (13.9) 1.15 ( ) (12.7) 5 (9.8) 0.74 ( ) Use of antibiotics during first week of life No 2900 (98.3) 167 (97.1) 1 [Reference] 3017 (98.3) 50 (98.0) 1 [Reference] Yes 49 (1.7) 5 (2.9) 1.56 ( ) (1.7) 1 (2.0) 0.85 ( ) Abbreviations: NA, not applicable; OR, odds ratio. a The total population was 3143; the number of participants in the adjusted analysis varied from 2758 to 2854, depending on the microbial exposure used in the model. b Adjusted for sex, genetic risk according to HLA antigen DQB1 (moderate/high), family history of diabetes (yes/no), mode of delivery (vaginal/cesarean section), place of birth (Tampere/Oulu), parental asthma or allergic rhinitis (yes/no), maternal professional educational level (none, professional education/course, secondary professional education, university), maternal age (<25, 25-29, 30-34, 35 y), home municipality urbanization level (rural, semiurban, urban), and child s asthma (yes/no) and atopic eczema (yes/no) by the age of 5 years. c Logistic regression was used in the analyses. lem of dropouts from the initial 6078 children to 3143 at the 5-year follow-up. Multiple imputation relies on an assumption that data are missing at random, that is, the probability of missingness is independent of the unobserved data, given the observed characteristics. We included several characteristics in the imputation model (Supplement [etable 1]), but we may have missed some important factors associated with the missingness mechanism. Although we cannot rule out the possibility that the data are not missing at random, in the sense of standard multiple imputation terminology, we believe that the similarity of the results of the multiple imputation and complete case analyses suggest that the missingness is not a source of significant bias. In the sensitivity analysis the proportion of missing data was, however, smaller than in the complete case data (Figure). The limitations of the study also include a possibility of selection bias, because 22% of the invited families declined to participate. In a Swedish cohort study, 15 the presence of a cat or dog at home or day care during the first year of life was directly associated with positivity for glutamic acid decarboxylase (crude OR, 1.3) and cat exposure with positivity for insulinomaassociated 2 molecule (crude OR, 1.4), but neither of these exposures was related to simultaneous positivity of the 2 autoantibodies. A hospital-based German case-control study 16 did not find evidence that regular contact with a stable (adjusted OR, 1.2; 95% CI, ) or any specific farm animal or pet was associated with type 1 diabetes. The adjusted ORs (95% CI) for regular dog and cat contacts were 0.67 ( ) and 0.86 ( ), respectively, in that study. The ORs for contact with a stable and dogs were relatively close to our results. jamapediatrics.com JAMA Pediatrics August 2014 Volume 168, Number 8 761

8 Research Original Investigation Table 3. Sensitivity Analysis: Microbial Exposure During the First Year of Life and Preclinical and Clinical Type 1 Diabetes a Preclinical Type 1 Diabetes Microbial Exposure During the First Year Adjusted OR of Life No Yes (95% CI) b,c,d P Value No Yes Indoor dog exposure No 2068 (80.4) 144 (85.2) 1 [Reference].09 Clinical Type 1 Diabetes 2140 (80.5) 72 (87.8) 1 [Reference] Adjusted OR (95% CI) b,c,d P Value Yes 503 (19.6) 25 (14.8) 0.58 ( ) 518 (19.5) 10 (12.2) 0.17 ( ) Indoor cat exposure No 2259 (87.9) 150 (88.8) 1 [Reference] 2335 (87.9) 74 (90.2) 1 [Reference] Yes 312 (12.1) 19 (11.2) 0.82 ( ) (12.2) 8 (9.8) 1.37 ( ) Abbreviation: OR, odds ratio. a Tampere patient records for pet exposure were available for children born in Tampere (n = 2740) and were used for sensitivity analyses. b Adjusted for sex, genetic risk according to HLA antigen DQB1 (moderate/high), family history of diabetes (yes/no), mode of delivery (vaginal/cesarean section), place of birth (Tampere/Oulu), parental asthma or allergic rhinitis (yes/no), maternal professional educational level (none, professional education/course, secondary professional education, university), maternal age (<25, 25-29, 30-34, 35 y), home municipality urbanization level (rural, semiurban, urban), and child s asthma (yes/no) and atopic eczema (yes/no) by the age of 5 years. c The number of participants in the adjusted analysis was d Logistic regression was used in the analyses. Indicators of social mixing and crowding, such as day care attendance, having siblings, sharing bedrooms, and regular changes in place of residence, have been associated with a decreased risk of type 1 diabetes in case-control studies. In cohort studies, 20,21 including the present study, attendance at a day care facility was not associated with preclinical or clinical type 1 diabetes. In children aged 3 years and younger, the area per person was directly related to type 1 diabetes in an Italian study. 21 The rising incidence of type 1 diabetes in children in several affluent countries has been linked 22,23 to increasing standards of living and socioeconomic transformation. Children susceptible to type 1 diabetes may have less bacterial diversity and less stable microbiota. 24 Enterovirus infections, especially toward coxsackievirus B, may play a role in the cause of type 1 diabetes. 25,26 There is evidence that the enterovirus load in the gut is larger among individuals with type 1 diabetes than in those without diabetes. 26 A case-control study 5 nested in a population-based cohort linked frequent early antimicrobial use to the development of type 1 diabetes: more than 7 purchases resulted in an OR of 1.7. In the present study, we had information only on the use of antimicrobials during the first week of life and found no indication of any association with type 1 diabetes. Dog contact may affect the human immune system in many ways. Dog contact has been shown 27 to protect children from respiratory infections during the first year of life, and cat and/or dog ownership may reduce the risk of gastroenteritis. 28 Dogs may also induce alterations in home microbial communities and spread infections including helminthes, thus affecting immune regulation. 29 In addition to affecting adaptive immune responses, 30,31 early exposure to dogs may reduce innate immunity responses. 32 Conclusions Most of the microbial factors evaluated in the present study were not associated with preclinical or clinical type 1 diabetes. Early exposure to an indoor dog may protect a child from preclinical type 1 diabetes. This observation needs to be confirmed in additional studies. ARTICLE INFORMATION Accepted for Publication: February 5, Published Online: June 23, doi: /jamapediatrics Author Affiliations: The National Institute for Health and Welfare, Nutrition Unit, Helsinki, Finland (Virtanen, Takkinen, Ahonen, Niinistö); Center for Child Health Research, University of Tampere and Tampere University Hospital, Tampere, Finland (Virtanen, Kaila, Knip); The Science Center of Pirkanmaa Hospital District, Tampere, Finland (Virtanen, Ahonen); School of Health Sciences, University of Tampere, Tampere, Finland (Virtanen, Takkinen, Nwaru, Ahonen); University of Helsinki, Hjelt Institute, Helsinki, Finland (Kaila); Department of Statistics, Faculty of Mathematics and Natural Sciences, University of Turku, Turku, Finland (Nevalainen); Children s Hospital, University of Helsinki, and University Central Hospital, Helsinki, Finland (Siljander, Knip); University of Helsinki, Diabetes and Obesity Research Program, Helsinki, Finland (Siljander, Knip); Department of Pediatrics, Faculty of Medicine, University of Turku, Turku, Finland (Simell); Immunogenetics Laboratory, University of Turku, Turku, Finland (Ilonen); Department of Clinical Microbiology, Faculty of Health Sciences, University of Eastern Finland, Kuopio (Ilonen); School of Medicine, University of Tampere, Tampere, Finland (Hyöty); Department of Pediatrics, University of Oulu, Oulu, Finland (Veijola); Folkhälsan Research Center, Helsinki, Finland (Knip). Author Contributions: Drs Virtanen and Takkinen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Virtanen, Takkinen, Nwaru, Kaila, Nevalainen, Simell, Hyöty, Veijola, Knip. Acquisition, analysis, or interpretation of data: Virtanen, Takkinen, Nwaru, Kaila, Ahonen, Nevalainen, Niinistö, Siljander, Simell, Hyöty, Veijola, Knip. Drafting of the manuscript: Virtanen, Takkinen, Nwaru, Nevalainen. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Takkinen, Nwaru, Nevalainen. Obtained funding: Virtanen, Simell, Hyöty, Veijola, Knip. Administrative, technical, or material support: Virtanen, Kaila, Ahonen, Siljander, Simell, Ilonen, Veijola, Knip. Study supervision: Virtanen, Nevalainen, Veijola, Knip. Conflict of Interest Disclosures: None reported. Funding/Support: The study was supported by the Academy of Finland (grants 44105, 48724, 80846, , , and , and Centre of Excellence in Molecular Systems Immunology and 762 JAMA Pediatrics August 2014 Volume 168, Number 8 jamapediatrics.com

9 Original Investigation Research Physiology Research grant ); the Prevaller Consortium; the European Foundation for the Study of Diabetes (EFSD/Novo Nordisk Partnership and ESFD/Juvenile Diabetes Research Foundation/Novo Nordisk Programme); the Foundation for Pediatric Research; the Tampere Tuberculosis Foundation; the Juho Vainio Foundation; the Yrjö Jahnsson Foundation; Medical Research Funds, Turku University Hospital and Oulu University Hospital; the Juvenile Diabetes Research Foundation; Novo Nordisk Foundation; EU Biomed 2 Program (grant BMH4-CT ); and the Competitive Research Funding of the Tampere University Hospital (grants 9B099, 9C084, 9D081, 9E082, 9F089, 9G087, 9H092, 9J147, 9K149, 9L117, 9M114, 9N086, and 9P057). Role of the Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Additional Contributions: We thank the physicians, research nurses, nutritionists, and laboratory staff of the Type 1 Diabetes Prediction and Prevention Study for their continuous collaboration through the years. We express our gratitude to Eeva Korhonen (Tampere University Hospital), Sirpa Pohjola (Oulu University Hospital), and Katri Jokinen, MSc (University of Tampere), for their expert technical assistance. The contributors did not receive financial compensation. REFERENCES 1. Strachan DP. Hay fever, hygiene, and household size.bmj. 1989;299(6710): Rook GAW. Hygiene hypothesis and autoimmune diseases. Clin Rev Allergy Immunol. 2012;42(1): Hanski I, von Hertzen L, Fyhrquist N, et al. Environmental biodiversity, human microbiota, and allergy are interrelated. Proc Natl Acad SciUSA. 2012;109(21): Chapman NM, Coppieters K, von Herrath M, Tracy S. The microbiology of human hygiene and its impact on type 1 diabetes. Islets. 2012;4(4): Kilkkinen A, Virtanen SM, Klaukka T, et al. Use of antimicrobials and risk of type 1 diabetes in a population-based mother-child cohort. Diabetologia. 2006;49(1): Kupila A, Muona P, Simell T, et al; Juvenile Diabetes Research Foundation Centre for the Prevention of Type I Diabetes in Finland. Feasibility of genetic and immunological prediction of type I diabetes in a population-based birth cohort. Diabetologia. 2001;44(3): Virtanen SM, Kenward MG, Erkkola M, et al. Age at introduction of new foods and advanced beta cell autoimmunity in young children with HLA-conferred susceptibility to type 1 diabetes. Diabetologia. 2006;49(7): Ilonen J, Reijonen H, Herva E, et al. Rapid HLA-DQB1 genotyping for four alleles in the assessment of risk for IDDM in the Finnish population: the Childhood Diabetes in Finland (DiMe) Study Group. Diabetes Care. 1996;19(8): Asher MI, Keil U, Anderson HR, et al. International Study of Asthma and Allergies in Childhood (ISAAC): rationale and methods. Eur Respir J. 1995;8(3): Siljander HT, Simell S, Hekkala A, et al. Predictive characteristics of diabetes-associated autoantibodies among children with HLA-conferred disease susceptibility in the general population. Diabetes. 2009;58(12): Knip M, Virtanen SM, Seppä K, et al; Finnish TRIGR Study Group. Dietary intervention in infancy and later signs of beta-cell autoimmunity. N Engl J Med. 2010;363(20): Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986; 73(1): Cardwell CR, Stene LC, Joner G, et al. Caesarean section is associated with an increased risk of childhood-onset type 1 diabetes mellitus: a meta-analysis of observational studies. Diabetologia. 2008;51(5): Sterne JA, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. doi: /bmj.b Wahlberg J, Vaarala O, Ludvigsson J; ABIS Study Group. Asthma and allergic symptoms and type 1 diabetes related autoantibodies in 2.5-yr-old children.pediatr Diabetes. 2011;12(7): Radon K, Windstetter D, Solfrank S, von Mutius E, Nowak D, Schwarz H-P; Chronic Autoimmune Disease and Contact to Animals (CAT) Study Group. Exposure to farming environments in early life and type 1 diabetes: a case-control study. Diabetes. 2005;54(11): McKinney PA, Okasha M, Parslow RC, et al. Early social mixing and childhood type 1 diabetes mellitus: a case-control study in Yorkshire, UK. Diabet Med. 2000;17(3): Cardwell CR, Carson DJ, Yarnell J, Shields MD, Patterson CC. Atopy, home environment and the risk of childhood-onset type 1 diabetes: a population-based case-control study. Pediatr Diabetes. 2008;9(3, pt 1): D Angeli MA, Merzon E, Valbuena LF, Tirschwell D, Paris CA, Mueller BA. Environmental factors associated with childhood-onset type 1 diabetes mellitus: an exploration of the hygiene and overload hypotheses. Arch Pediatr Adolesc Med. 2010;164 (8): Snell-Bergeon JK, Smith J, Dong F, et al. Early childhood infections and the risk of islet autoimmunity: the Diabetes Autoimmunity Study in the Young (DAISY). Diabetes Care. 2012;35(12): Bruno G, Spadea T, Picariello R, et al; Piedmont Study Group for Diabetes Epidemiology. Early life socioeconomic indicators and risk of type 1 diabetes in children and young adults. J Pediatr. 2013;162(3): , e1. doi: /j.jpeds Patterson CC, Dahlquist G, Soltész G, Green A; EURODIAB ACE Study Group. Europe and Diabetes. Is childhood-onset type I diabetes a wealth-related disease? an ecological analysis of European incidence rates. Diabetologia. 2001;44(suppl 3): B9-B Jarosz-Chobot P, Polanska J, Polanski A. Does social-economical transformation influence the incidence of type 1 diabetes mellitus? a Polish example. Pediatr Diabetes. 2008;9(3, pt 1): Giongo A, Gano KA, Crabb DB, et al. Toward defining the autoimmune microbiome for type 1 diabetes. ISME J. 2011;5(1): Bluestone JA, Herold K, Eisenbarth G. Genetics, pathogenesis and clinical interventions in type 1 diabetes. Nature.2010;464(7293): Oikarinen M, Tauriainen S, Oikarinen S, et al. Type 1 diabetes is associated with enterovirus infection in gut mucosa. Diabetes. 2012;61(3): Bergroth E, Remes S, Pekkanen J, Kauppila T, Büchele G, Keski-Nisula L. Respiratory tract illnesses during the first year of life: effect of dog and cat contacts. Pediatrics. 2012;130(2): Heyworth JS, Cutt H, Glonek G. Does dog or cat ownership lead to increased gastroenteritis in young children in South Australia? Epidemiol Infect. 2006;134(5): Smallwood J, Ownby D. Exposure to dog allergens and subsequent allergic sensitization: an updated review. Curr Allergy Asthma Rep.2012;12 (5): Ownby DR, Peterson EL, Wegienka G, et al. Are cats and dogs the major source of endotoxin in homes? Indoor Air. 2013;23(3): Abraham JH, Finn PW, Milton DK, Ryan LM, Perkins DL, Gold DR. Infant home endotoxin is associated with reduced allergen-stimulated lymphocyte proliferation and IL-13 production in childhood. J Allergy Clin Immunol. 2005;116(2): Lappalainen MH, Huttunen K, Roponen M, Remes S, Hirvonen MR, Pekkanen J. Exposure to dogs is associated with a decreased tumour necrosis factor-α producing capacity in early life. Clin Exp Allergy. 2010;40(10): jamapediatrics.com JAMA Pediatrics August 2014 Volume 168, Number 8 763