The availability of a large, well-characterized population enabled us to examine the relationship between environmental exposures before illness onset and phenotypes in JIIM. We confirmed a number of exposures that had also been seen in prior studies of JDM, particularly the temporal association of respiratory infections preceding illness onset [16
]. We identified for the first time that a number of other non-infectious exposures occurred within 6 months of the first signs of illness, including medications, many of which are potentially myopathic or photosensitizing, immunizations, stressful life events and sun exposure. The main novel findings of this study were differences in some exposures by age at diagnosis, delay to diagnosis, race, disease course and autoantibody phenotypes. For example, children younger than the median age at the time of diagnosis had a higher frequency of documented infections, whereas older children had a higher frequency of stressful life events in the months before illness onset. Patients without a myositis autoantibody had a higher frequency of infections in the 6 months before illness onset than was seen in patients with anti-p155 or anti-SRP autoantibodies, whereas patients with anti-SRP autoantibodies had a greater number of documented exposures than patients without a myositis autoantibody. These findings suggest that environmental exposures may differ by phenotype, and that they could be useful in understanding pathogeneses [1
We found that an infectious illness, particularly a respiratory infection, frequently occurs within several months before juvenile myositis onset, supporting the findings of other studies of exposures temporally associated with the onset of JDM. In one study, a prospective registry of patients within 6 months of illness onset in which data were based on a parent environmental interview and medical record review, respiratory infections were identified within 3 months of illness onset in 57% of patients [16
]. The other, a retrospective cohort with review of medical records by infectious disease specialists, identified infections within 3 months before the first symptoms of JDM in 33–50% of patients, and respiratory infections accounted for 80% of the infections [17
]. The lack of control comparator groups in all of these studies, however, does not enable one to conclude that these exposures differ from a healthy population, nor that they are associated with the onset of illness. While infections, particularly upper respiratory infections, are reported frequently in school age children [32
], a prospective matched cohort of new-onset JDM patients reported a higher frequency of antecedent illness in the JDM patients compared with friend controls from the same geographical region [19
We identified for the first time that a number of other non-infectious exposures also occurred within 6 months of the first signs of illness, including medications, many of which are potentially myopathic or photosensitizing, immunizations, stressful life events and sun exposure. Pachman et al
] noted medication use in >60% of patients, including medications for symptoms of early illness or antibiotics to treat associated infections. A listing of medications taken by patients in the present study and in others includes similar medications (), and we noted that many of the medications could be potentially myopathic or phototoxic [26
]. Drug-induced myositis has been well described with a number of different medications, including d
-penicillamine, lipid-lowering agents, l
-tryptophan and IFN-α [33
]. Myopathic or phototoxic drugs, however, could lead to the first symptoms of myositis. Other environmental factors reported here, including ultraviolet light exposure, emotional stress and heavy weight lifting, have been reported as possible risk factors for adult DM or PM in case-controlled studies [35–38
Almost 40% of the patients in this study had two or more reported exposures within 6 months before illness onset, rather than a single documented exposure. This is consistent with the concept that, just as systemic autoimmune diseases are polygenic [39
], they might also be polyenvironmental, meaning that patients may have more than one exposure before developing the disease. These exposures may also be dependent on gene–gene, environment–environment and gene–environment interactions. In diseases such as cancer, multiple infectious and non-infectious environmental factors have been associated with specific malignancies, and these environmental exposures have been shown to affect the development of disease in different ways, including altering mutagenesis, promotion and direct carcinogenesis [40
]. Synergistic interactions between some of these environmental factors, including viral and non-infectious exposures, have also been seen in certain malignancies [41
]. It is possible, though, that there was a confounding between exposures, such as an infectious illness and the use of antibiotics, as noted by Pachman et al
]. Our data suggest that further investigation of the interaction between environmental exposures may be useful.
It is important to emphasize that the temporal association of environmental exposures with illness onset does not imply causality. For example, certain exposures, such as trauma or weight training, could have occurred after the onset of illness as a consequence of the first unrecognized symptoms of disease, such as fatigue or muscle weakness. Rather, exposures with temporal relationships to disease onset, as were seen in this hypothesis-generating study, constitute a first step for determining which factors may trigger the onset of illness and warrant further investigation. Additional support for a relationship between these exposures and disease pathogenesis could be provided by dechallenge data, which did not exist in this cohort-based study, from laboratory investigations and from case-controlled epidemiological studies [43
]. A case–control study by Pachman et al
] did not find any significant differences in pesticide use, psychological stress or exposure to animals in 80 JDM patients within 6 months of illness onset compared with 63 age-matched geographically similar healthy controls with similar school or daycare experiences, nor was parvovirus found to be an aetiological factor in recent-onset JDM patients compared with age-, gender- and race-matched controls [44
]. However, both of those studies may not been adequately powered to detect differences between the cases and controls. Also, the extent of matching of controls may have obscured differences with JDM patients. For example, in the parvovirus study [44
], the controls were age, race and gender matched to patients, but they were not geographically matched, whereas in the study of Pachman et al
], the healthy controls, frequently age-matched classmates and neighbours, may have been geographically over-matched, but they were not gender or race matched. An appropriately powered prospective case-controlled study is needed to confirm the observations from this and other previous reports.
There are a number of potential limitations in this study. A primary limitation is the absence of a control group. Thus, the frequencies of exposures observed in juvenile myositis patients overall may not differ from healthy control populations and these exposures may not be associated with the onset of illness. In addition, there could be under- or over-reporting of potential exposures, including a selection bias in the patients who had the environmental component of the questionnaire completed. We also found more exposures, including infectious illnesses, in white patients compared with patients in other racial groups. This could potentially be the result of differences in access to health care, resulting in better documentation of such exposures. The somewhat arbitrary period of 6 months before the onset of illness for identification of environmental factors might not be relevant to the initiation of myositis for all exposures. Certain exposures could require a longer period to induce their effects, as has been reported in malignancies, silicosis and other disorders, while for other exposures a shorter time frame might be more relevant [37
]. Also, exposures other than infections, drugs and vaccines were reported in an open-ended manner, and patients were not required to be directly interviewed to obtain information about environmental exposures. We attempted to overcome these possible biases by conducting a formal review of most of the medical records of the study subjects. However, the medical records might also have selection bias by reporting only some of the significant environmental exposures. Certain exposures, such as exposures in the home and use of certain chemicals, are likely not captured uniformly in the medical record by the treating physician. Nonetheless, the fact that our data on infections before illness onset are similar to those of other large cohorts suggests that the quality of the data and reporting are reliable [16
]. Finally, while some of the ORs in our study are large, the CIs may be wide and estimates could be inflated due to relatively small numbers of patients in some groups.
In summary, we have identified a number of environmental exposures, including infectious and non-infectious agents that occurred within 6 months before illness onset, varied by phenotype and may be important in the pathogenesis of JIIM. These findings suggest that a search for a single environmental factor that causes or triggers a single disease as currently defined, such as JIIM, may be unproductive, as patients could have several environmental exposures and these could vary with the disease phenotype that develops. These exposures require confirmation in case-controlled studies to identify whether they are associated with illness onset and whether they play any role in aetiology, yet they suggest focused areas of further research to better understand the environmental factors associated with the onset of JIIM phenotypes and their possible interrelationships with genetic risk factors.