The case series reported by Bridges and colleagues, in this issue of Annals, is another series of patients referred to rheumatologists interested in a potential silicone-rheumatic disease link, for evaluation of musculoskeletal complaints and emphasizes that most of these women do not have evidence of a well-defined connective tissue disease. The case series reported by Cukier and colleagues, also in this issue of Annals, includes a statistical analysis that strongly supports an association between bovine collagen implants and a dermatomyositis- or polymyositis-like syndrome. Their statistical model, however, is based on assumptions that may not be valid, and its robustness cannot be adequately judged given the available data. Well-designed epidemiologic studies are needed to examine the validity of these putative associations; until such data are available, clinicians need to be aware that the benefit-to-risk ratios of these cosmetic surgical procedures for their patients are not yet known.

In 1964, Miyoshi and colleagues [1] reported the first patient with presumed connective tissue disease after a cosmetic procedure. Numerous single case reports and case series of patients with presumed and well-defined connective tissue diseases developing after cosmetic surgical procedures have appeared in the literature during the past 20 years; notable are the reviews by Kumagai and colleagues [2], Weisman and colleagues [3], and the recent comment by Germain [4]. In 1992, I reviewed reports of 85 such cases occurring after augmentation mammoplasty; 50 patients had a confirmed diagnosis of connective tissue disease (24 had systemic sclerosis, 13 had rheumatoid arthritis, 8 had systemic lupus erythematosus, 4 had mixed connective tissue disease, and 1 had localized scleroderma), whereas the remainder had a poorly defined syndrome that was often called "human adjuvant disease".

Concerns about the relation between cosmetic surgical procedures and the development of autoimmune connective tissue diseases, among other conditions, were expressed by the Food and Drug Administration (FDA) in the proposed rules for premarket approval of silicone gel-filled breast prostheses. However, in August 1991, the FDA stated "... there is no conclusive evidence at present that women with breast implants have an increased risk of developing arthritis-like diseases or other auto-immune diseases ..." (FDA Backgrounder, August 1991). Nonetheless, in April 1992, the FDA limited the availability of silicone gel-filled breast prostheses based on the absence of any safety data for these devices. Indeed, Dr. Kessler wrote "... the link, if any, between these implants and immune-related disorders ... is also unknown" [5]. A task force convened by the American Society of Plastic and Reconstructive Surgeons concluded that no data existed to infer a causal relation between silicone gel-filled breast implants and scleroderma-like disorders or other autoimmune diseases [6]. Further, this task force felt that the diagnostic term "human adjuvant disease" was imprecise and should be avoided in future clinical reports.

The clinical epidemiologic approach to assessing the role of cosmetic surgical procedures as causal factors in the development of connective tissue disease begins with a review of clinical observations, including case reports and case series [7]. Based on these initial observations, the second step involves an examination of available data in an attempt to infer a statistical association between exposure and disease. Such analyses are usually hypothesis generating rather than hypothesis confirming, and they provide a basis for further study. At the third step, epidemiologic studies are required to scientifically examine and to establish the link between cosmetic procedures and the development of connective tissue disease, if one exists. Implementing these studies, however, is complicated by the rarity of the individual connective tissue diseases other than rheumatoid arthritis and the rarity of cosmetic surgical procedures.

Bridges and colleagues [8], in this issue of Annals, present a case series of 156 women with silicone breast implants and rheumatic complaints seen by three rheumatologists with an "... interest in silicone and rheumatic disease". The authors divided these women into three groups: 95 with arthralgias, myalgias, and fatigue; 32 with mild arthritis and fatigue; and 29 with "... findings suggestive of a connective tissue disease". The latter group included 13 patients who apparently had scleroderma, 3 with systemic lupus erythematosus, and 1 each with mixed connective tissue disease, polymyositis, and isolated Raynaud phenomenon. An additional 10 patients were classified as having early connective tissue disease based only on the presence of antibodies to the BB' polypeptide, although none of the patients had positive results of tests for antinuclear antibody in clinically important titers or fulfilled American College of Rheumatology criteria for any established connective tissue disease. This report, and that of another case series of 24 patients by Press and colleagues [9], failed to provide any information about the frequency of the development of connective tissue disease among women with silicone breast implants. Further, the likelihood of referral bias in these samples precludes any estimate of risk for women with silicone breast implants.

Cukier and colleagues [10], also in this issue of Annals, report a series of 9 patients with inflammatory myositis, 8 with dermatomyositis, and 1 with polymyositis, which developed after treatment with intradermal bovine collagen implants. Their report extends the descriptive case series to the second step by including an analysis estimating the risk for the development of inflammatory myositis in persons with bovine collagen implants. Using estimates of the number of new collagen-treated patients from July 1980 through June 1988 and the age-, race-, and sex-specific incidence of myositis derived from a hospital-based study in Allegheny County, Pennsylvania, the authors concluded that the incidence of inflammatory myositis, especially dermatomyositis, was statistically greater than expected among collagen-treated patients.

The implications of causality raised by the occurrence of various connective tissue diseases in women who have previously had breast augmentation with silicone gel-filled prostheses or have received bovine collagen implants, especially if a statistical association has been shown, must be carefully examined. Of concern to the epidemiologist in evaluating the relation between an exposure and a disease are the validity of the association, and, if one exists, the strength of the association, the temporal relation between the exposure and the onset of disease, the consistency of the association between studies, the specificity of the association, and the presence of biologic plausibility.

Idiopathic inflammatory myositis, including polymyositis and dermatomyositis, is a disorder of unknown cause(s) [11]; previous epidemiologic studies, however, point to contributions from host and environmental factors [12]. Inflammatory myositis, like other connective tissue diseases, is more common in women than men. A genetic predisposition, marked by an association with human leukocyte class II antigens HLA-DR3 and DRw52, has been reported especially in those patients with antibodies to aminoacyl-transfer RNA synthetases [13]. Evidence to support a role for environmental factors includes seasonal clustering [14] and the presence of increased levels of antibodies to Toxoplasma gondii in adults with polymyositis [15] and Coxsackie B virus in children with dermatomyositis [16]. In addition, an animal model of polymyositis develops after infection of mice with Coxsackie virus B1 Tucson strain [17].

The report by Cukier and colleagues [10], in contrast to that by Bridges and colleagues [8], attempts to examine the question of whether an association exists between an exposure, bovine collagen implants, and the development of a specific connective tissue disease (inflammatory myositis). Although all 9 patients fulfilled proposed criteria for classification of polymyositis and dermatomyositis, muscle biopsies were done in only 6, and the result was equivocal in 1. Tests for antinuclear antibody titers were done for only 5 patients and were positive for only 3 patients. No data are provided about the presence of either myositis-specific antibodies, including anti-Jo1 and anti-signal recognition particle antibodies, or the HLA class II phenotypes of the patients. Of note, 2 of the 9 patients also had had augmentation mammoplasty with silicone implants before receiving bovine collagen implants.

In the report by Cukier and colleagues [10], some of the assumptions underlying the statistical analysis may not be valid; for example, the expected number of cases may have been underestimated. The number of persons estimated to have been exposed to bovine collagen implants was derived from company sales reports, based on the estimated volume of material injected per patient and the proportion of patients who had retreatment; the overall estimate does not include persons who were exposed to only test doses. Inclusion of those persons is essential because two of the nine patients received only test doses of collagen. Thus, the number of persons exposed to bovine collagen implants was probably higher than estimated by the authors, and the expected incidence of inflammatory myositis was probably lower than reported by the authors. The optimal denominator for estimating the incidence of myositis would have been a population-based registry of persons exposed to bovine collagen implants [18].

Another source contributing to the underestimation of expected cases is the incidence rates used in the calculation. Although the authors adjusted the published rates to reflect those in the second decade of the study, 1973 to 1982, the original data show that average annual rates continued to increase over time and reached maximum levels in the 5-year period from 1978 to 1982 at 10.2 per million (95% CI, 1.02 to 19.4) [19]. In addition, the expected number of patients may have been underestimated because of genetic or other differences in the population exposed to bovine collagen implants and that of Allegheny County, Pennsylvania, where the incidence study was done and because of the lack of inclusion by Oddis and colleagues [19] of myositis patients diagnosed outside of the hospital. Underestimation of the expected number of patients would increase the likelihood of finding an association where one does not exist (type I error); the robustness of the author's statistical model cannot be judged without appropriate consideration of these and other potential concerns about the validity of their assumptions.

Even if future studies show that an association exists, the interpretation of that association will not be clear without evidence supporting the biologic plausibility of a causal role of bovine collagen implants in inflammatory myositis. Animal studies of bovine collagen implants have failed to show evidence of the development of inflammatory myositis [20], and patients with myositis have not been shown to produce autoantibodies against native collagen despite the presence of antibodies to myosin [21].

Conclusion

Clearly, the reports of various connective tissue diseases developing in patients after cosmetic surgical procedures including silicone gel-filled breast implants and bovine collagen implants warrant the conduct of well-designed epidemiologic studies. Such study designs should include either nonconcurrent or concurrent prospective studies of persons exposed to the cosmetic procedure and appropriately chosen nonexposed controls, or case–control studies with newly diagnosed cases that fulfill published classification criteria for the disease under study. Indeed, the results of a previous case–control of newly diagnosed cases of polymyositis and dermatomyositis failed to identify an association with bovine collagen implants [22]; that study, however, was limited by insufficient power to examine an uncommon exposure. At least two case–control studies are in progress to examine the association of augmentation mammoplasty with systemic sclerosis; preliminary data from one multicenter study noted a frequency of 0.66% among 741 patients for augmentation mammoplasty before physician-diagnosis of systemic sclerosis [23]. This frequency was not statistically different from the expected frequency of 0.65% to 1% based on data from the 1988 Medical Device Implant Supplement to the National Health Interview Survey [24] and other sources. Data derived from epidemiologic studies are required to accurately assess the benefit-to-risk ratio of these cosmetic surgical procedures.