Antinuclear Antibodies and Silicone Breast Implants
A number of investigators have explored the frequency and titers of antinuclear antibodies (ANAs) in women with or without signs and symptoms of illness and silicone (almost always gel-filled) breast implants. These investigators have compared findings of ANAs in women with implants to findings in control groups of women, some matched by age, some concurrent, some historical, some healthy and some with similar rheumatic-like signs and symptoms. Some investigators have reported increased frequency and increased titers of ANAs in women with implants and have speculated about the implications of this finding for autoimmune disease or other immunological reactions that might result from exposure to silicone breast implants. Other investigators have not found any difference in ANA titers between otherwise similar groups of women with or without breast implants and have concluded that there is no relationship between implants and autoantibodies. Many of these studies suffer from methodological flaws, discussed below, which call into question their results or interpretations (Bar-Meir et al., 1995; Blackburn et al., 1997; Bridges et al., 1993a, 1996; Brunner et al., 1996; Claman and Robertson, 1994, 1996; Cuéllar et al., 1995a,b; Edworthy et al., 1998; Freundlich et al., 1994; Gabriel et al., 1994; Karlson et al., 1999; Lewy and Ezrailson, 1996; Miller et al., 1998; Ostermayer-Shoaib et al., 1994; Park et al., 1998; Peters et al., 1994a, 1997; Press et al., 1992; Rowley et al., 1994; Silverman et al., 1996b; Smalley et al., 1995c; Solomon, 1994a; Tenenbaum et al., 1997a; Teuber et al., 1993; Vasey et al., 1994; Young et al., 1995b, 1996; Zazgornik et al., 1996).
Technical Considerations Necessary to Interpret Reported ANA Titers
An understanding of the prevalence and titers of ANAs in various subsets of the population of normal individuals, and of the effects that different technical factors may have on measurement of these antibodies, is important to the interpretation of ANA results from study populations of symptomatic and asymptomatic women with silicone breast implants. The type of tissue that is used as the substrate (antigen) in the performance of the ANA test can affect the frequency of positive tests at all titers. In the mid-1980s and early 1990s, laboratories used frozen sections of mouse or rat liver as the test substrate. More recently, human tissue culture cells, such as HEp-2 cells, have replaced animal liver (or kidney) sections. Prevalence and titers of ANAs are higher when tissue culture cells, rather than mouse liver, are used as a substrate (Fritzler et al., 1985; McCarty et al., 1984). Hollingsworth et al. (1996) reported results of comparative ANA testing of 106 healthy females and 91 healthy males. There were more positive tests at all titers when HEp-2 cells were used than when rat liver was used. Also, the prevalence was nearly twice as high in normal women as in men (Hollingsworth et al., 1996). Measurements in controls using mouse or rat liver (or other tissues) cannot be compared to values from women with breast implants using human tissue culture cells.
There are differing views on what titer and what intensity of fluorescence constitute a positive ANA test. Reports of ANA prevalence in women with breast implants have not always used the same criteria to define positivity. Furthermore, differences in the performance of the ANA test and in the results obtained are common across different laboratories. These factors make it essential that studies use the same laboratory at the same time for ANA assays of experimental and control women. Strongly positive ANA tests in women with implants who have connective tissue disease undoubtedly reflect the disease. These positive ANA tests should not be ascribed to the presence of the implants. Moreover, studies from investigators with a special interest in autoantibodies or a particular connective tissue disease may have attracted a nonrepresentative group of women with implants and with connective tissue disease.
Antinuclear Antibodies in Normal Individuals
In 1997, an important analysis of ANAs in normal sera was carried out by a subcommittee of the International Union of Immunological Societies' Standardization Committee (Tan et al., 1997). The objective of this study, carried out in 15 laboratories from around the world, was specifi-
cation of the range of ANA titers in normal individuals and in patients with connective tissue disease. Centers for Disease Control and Prevention (CDC) reference sera for ten specific ANAs and normal serum samples in four adult age categories were provided to the laboratories. At 1:40 dilution, 31.7% of normals had a positive ANA test; at 1:80, 13.3%; at 1:160, 5%; and at 1:320, 3.3%. Results did not differ significantly across the four age categories. This absence of increasing ANA positivity with age has been reported, though infrequently, by others (Juby et al., 1994). In addition, a group of patients with soft-tissue rheumatism was studied. At a titer of 1:40, 38.5% were positive, and 23.1% were positive at a titer of 1:80. This is the kind of group whose ANA results should be compared with those reported for symptomatic women with implants who do not have a defined connective tissue disease. In contrast, nearly all patients with systemic lupus erythematosus (SLE) were positive at greater than 1:40 dilution, with the vast majority positive at 1:320. This study also found a wide variation in results across laboratories, even when the same test antigens and the same definition of positivity were used. Presumably, this was due to a number of factors, including different reagents, different skills and training of technicians, and variation in the number and sources of control sera, among others. The authors concluded that normal individuals could be successfully (but not perfectly) differentiated from those with SLE, scleroderma, and Sjogren's syndrome. Moreover, the cutoff titer of 1:160 appeared to provide an acceptable balance between sensitivity and specificity.
Gender and age affect the prevalence of ANAs in the normal population. Women are more commonly ANA positive than men (Fritzler et al., 1985; Thomas and Robinson, 1993). Most studies have shown that the prevalence of ANAs also increases with age. Slater et al. (1996) analyzed the results of ANA testing in a consecutive sample of 1,010 patients over a ten-month period. Patients 64 years of age or older had a statistically higher prevalence of ANAs. Of interest is that 23% of patients with neurologic symptoms had a positive ANA test, as did 10% of patients with constitutional symptoms. In this study, the overall false-positive rate for ANAs was 79%, that is, 79% of positives occurred in persons without defined disease (Slater et al., 1996). The age effect is also illustrated by a study of 3,492 Australians, representing more than 90% of the population of a small town, in whom the frequency of positive ANAs was 1.5% in the 31-40 year age group, but increased progressively to 13.5% in the 71-75 year age group (Hooper et al., 1972). Additional reports of high ANA values in normal populations or, particularly, increasing values with age include those of de Vlam et al. (1993); Fritzler et al. (1985); Ruffatti et al. (1990); and Xavier et al. (1995). These reports and the report by Tan et al. noted earlier cite frequencies of positive ANA tests ranging up to 31%
depending on the age of the population and the titers that are used as criteria for positivity. Some studies have also suggested that women with breast cancer may have elevated ANAs (Klajman et al., 1983; Turnbull et al., 1978); in a Japanese population, a prevalence of 15% was reported (Imai et al., 1979).
The committee finds that ANA testing is a complex and variable technology. The performance of studies and the interpretation of results require consideration of a number of variables and confounding factors, as noted above. In addition, some studies of ANA prevalence in women with silicone breast implants suffer from inadequate reporting of the details of the test technology and criteria for a positive test; inadequate descriptions of the signs, symptoms, and presence of defined connective tissue disease in some of the test populations; biased ascertainment of women with implants; inadequate descriptions or absence of control populations; uncertainty regarding adequate blinding of those who perform the assays; and questions of proper matching of control groups, among others. The result has been studies with conflicting findings, which often have flaws suggesting that the results are questionable or, through lack of information, cannot be interpreted with confidence.
Antinuclear Antibodies in Women with Silicone Breast Implants
The committee reviewed 30 studies from the scientific literature on the subject of ANAs in women with silicone breast implants. Although information in some of these was substantially lacking, and some were also extensions of previous reports, they are discussed briefly for the sake of completeness. Press et al. (1992) studied 24 women referred for rheumatic complaints who had either silicone breast implants (N = 22), silicone injections (N = 1) or both (N = 1). A defined connective tissue disease was diagnosed in 11 of the 24 patients. There also were four patients with fibromyalgia, four with chronic fatigue syndrome, four with myalgias, and one with arthralgia. An ANA titer of 1:80 or more was found in 91% of the patients with defined connective tissue disease. ANAs were present in 54% of the group with symptoms but with no connective tissue disease. This study did not include three important control groups: (1) women with implants without rheumatic complaints or a defined connective tissue disease; (2) age-matched normal women without implants; and (3) symptomatic age-matched women without implants. The presence of a defined connective tissue disease in nearly half of the women studied and of rheumatic signs, symptoms, and conditions associated with ANAs in the remainder make the data very difficult to interpret (Press et al., 1992).
Bridges et al. (1993a) found that 22% of 156 symptomatic women with implants had an ANA titer of 1:80 or more; rheumatoid factor was found in 9%. Of 95 of these patients who had joint and muscle pain, 15% had positive ANAs; 22% of 32 who had joint swelling were positive; and 45% of 29 who had a defined connective tissue disease were positive. From a review of the literature and 401 patients presented in abstracts at the 1992 American College of Rheumatology Meeting, Bridges identified 5% of mostly symptomatic patients with silicone breast implants and specific autoantibodies, such as anti-Ro and anti-La, anti-Scl-70, anti-Sm, and anti-dsDNA. Positive ANA titers were found in 25% of a sex- and age-matched control group of 174 fibromyalgia patients. In contrast, only 8% of 12 women with implants without rheumatic complaints had ANA titers of 1:80 or more (Bridges et al., 1993a). Bridges and Lorden (1993) also failed to find significant elevations of ANAs in women with breast implants and sicca syndrome. In a subsequent report, Bridges et al. described 500 consecutive women with implants referred to rheumatologists, 25 age-matched controls, 25 asymptomatic women with implants, and 100 women with fibromyalgia. These women were tested for ANAs as previously reported. It is not clear whether there was any overlap of patients in the two reports, however. Of the symptomatic women with implants, 150 of 500 (30%) were positive. Positive ANA tests were obtained from 8% of the controls; 28% of the asymptomatic women with implants, and 25% of the fibromyalgia patients. A speckled or nucleolar pattern of ANA fluorescence was more frequently, but not more significantly, found in women with implants, as were very high titers in women with implants (7%) compared to normal controls (0%). This study concluded that patients with implants and those with fibromyalgia were significantly more likely to be ANA positive than normal women without implants (Bridges et al., 1996).
Teuber et al. (1993) studied 57 consecutive patients, self-referred in 1992 because of concerns about their implants. After the exclusion of eleven women with previous exposure to bovine collagen, 46 women with implants were age matched with 45 control women from the same geographical area. Implants were in place for a mean of 13.5 years, primarily for augmentation. Many of the women with implants were symptomatic, and one patient had SLE; one primary biliary cirrhosis; one small vessel vasculitis; and two had polyarthritis with ANAs. ANAs were measured using HEp-2 cells and were considered positive at titers of 1:80 or more. Of the 57 patients, 16 (35%) had a positive ANA and 11% had titers of 1:320 or more. This study was carried out only on patients preselected for symptoms or a defined connective tissue disease, and data on the prevalence of ANAs in the 45 normal control women were not reported (Teuber et al., 1993). In 1994, the same group of investigators reported on
epitope mapping of antibodies to collagen found in women with implants. This study included most of the women previously reported, except those with defined connective tissue disease. ANAs were found in 19 of 70 patients at a titer of 1:80 or more. Again, no data on controls were reported (Rowley et al., 1994).
ANA testing of 200 consecutive self-referred patients concerned about their silicone gel breast implants was reported by Peters et al. (1994a). Controls were 100 consecutive women without implants referred for a variety of plastic surgery procedures. A group of 29 patients with surgically confirmed ruptured silicone gel implants was also studied. Specific autoantibodies were studied if ANAs were positive. Testing was on HEp-2 cells, and a titer of 1:100 was considered positive. Of the 200 implant patients, 4 had a defined connective tissue disease. A positive ANA test was found in 26.5% of 200 breast implant patients, 28% of controls, and 17.2% of 29 patients with ruptured implants. Titers of ANAs did not differ among the groups. Controls in this study were from the same plastic surgery practice, all patients were examined by rheumatologists, and ANAs were similar in experimental and control groups. However, the patients were a self-referred group, and ANA positivity in the control group was higher than usual (Peters et al., 1994a). This Toronto group also reported a follow-up of 100 consecutive patients requesting explantation of their gel-filled breast implants, many of whom were presumably included in the 1994 report. The same control group of women appears to have been used. In this group of highly concerned, self-selected, more symptomatic women, four of whom had connective tissue disease and ten of whom had fibromyalgia, the prevalence of ANAs was 24%, compared to 28% in the control group, and titers did not differ between the groups. ANA results after explantation would have been of interest, although the authors report that, with one exception, their patients' clinical status remained the same or deteriorated rather than improved over more than two years of observation after removal of their implants (Peters, 1997).
Claman and Robertson (1994) studied 131 women with breast implants and 19 healthy controls. The women with implants included 38 who felt healthy, 82 who had various symptoms, and 11 who had defined connective tissue diseases. ANA assays used HEp-2 cells. Fluorescence of 1+ or greater and a titer of 1:256 or more were considered positive. No positive ANAs were found in healthy women; 18% of asymptomatic women, 26% of symptomatic women, and 64% of women with implants and a defined connective tissue disease were positive. There was no correlation between presence of a positive ANA and type of implant, indication for implantation, duration of implantation or implant rupture. The controls in this study were not age matched, and there were only 19 healthy controls to compare with 131 implanted women with an array of
different health conditions (Claman and Robertson, 1994). These authors reported an extension of their work in 1996, in which they tested 37 healthy control women and compared them with 75 asymptomatic women with silicone breast implants. ANA technology was the same except that a titer of 1:80 was considered positive. In this report, 3% of controls and 27% of women with implants were positive, a significant difference between the control and experimental groups (Claman and Robertson, 1996).
Cuéllar et al. (1995a) analyzed data from 300 patients (298 women) referred with a variety of musculoskeletal and other complaints. Of these, 24 had saline implants, the rest were gel-filled, but results were not sorted by type of implant. A positive ANA was defined as fluorescence of 2+ or greater on HEp-2 cells at a titer of 1:40 or more. By these criteria, 123 of 265 patients tested (46.4%) had positive ANAs. Of these patients, 185 were diagnosed as having a defined or undifferentiated connective tissue disease, fibromyalgia, or other rheumatic-like conditions. The report does not allow assignment of ANA results to any of the poorly defined categories of patients, so the data are difficult to interpret (Cuéllar et al., 1995a).
In a follow-up, the same group reported a cohort of 813 patients (810 women), including the 300 reported earlier. This report also identified 264 women without breast implants referred for ANA testing who were approximately equally divided between those with fibromyalgia and those with soft-tissue rheumatism. These women served as a control population. It is uncertain whether or how they were matched, although the mean ages of the experimental and control groups were very close. ANA assays were done using mouse kidney or HEp-2 cells. A positive test was considered fluorescence of 2+ or greater and a titer of 1:64 or more for mouse kidney and 3+ or greater fluorescence at a titer of 1:40 or more for HEp-2 cells. Most patients, 470 (57.8%), were positive using human cells as substrate, and 244 (30%) patients were positive using mouse kidney as substrate. It is not clear how the use of these different technologies was distributed between the patient and the control groups. Positive ANAs were found in 7.6% of the control population. This report also describes a selected group of women. Very little information is available on the health conditions of these women, but presumably this population consists of a number of implant patients with various connective tissue and other rheumatic-like diseases as did the original 300-patient cohort (Cuéllar et al., 1995b). As a result, conclusions on the relationship between implants and ANAs cannot be drawn from these reports.
A group of 3,380 women with silicone breast implants, self-, physician-, or attorney-referred to two physicians for evaluation relative to injury claims, was tested by a commercial testing laboratory (Roche Biomedical Labs, Burlington, N.C.) using HEp-2 cells. The prevalence of
ANAs with titers of 1:40 or more was 22%. A historical, non-concurrent, unmatched control with ANA prevalence of 5% was used. These patients complained of fatigue (83%), joint and muscle pain (83%), chest or breast pain (81%), burning of the extremities, numbness (80%), and other symptoms. The authors also found a correlation of ANAs with duration of implantation and with serum immunoglobulin G concentrations (Lewy and Ezrailson, 1996). This study reports a highly selected group of symptomatic women with low-titer ANAs and no appropriate control group.
Bar-Meir et al. (1995) studied the prevalence of 20 different antinuclear autoantibodies in the sera of 116 women with breast implants of whom 11 were caucasian. Of these women with implants, 30 had a history of implant rupture and 29 had contractures. They were compared with 134 age- and sex-matched controls. The prevalence of 15 of 20 specific ANAs was significantly higher in the silicone breast implant group than in controls. This study has incomplete data on the race of the controls, no information about how subjects were selected, and no specificity controls. The occurrence of more than twice as many patients with anti-La as with anti-Ro antibodies is unusual (Mattioli and Reichlin, 1974; Wasicek and Reichlin, 1982). Likewise, the presence of anti-Sm and anti-Scl-70 antibodies, despite the fact that only one patient each had SLE and scleroderma, suggests that there may have been false positives. In a subsequent abstract report, these U.S. and Israeli groups compared 86 asymptomatic women with implants with these 116 women and reported increased titers of 13 antinuclear autoantibodies in 2%-13% of the asymptomatic group. They concluded that anti-La antibodies were increased in both groups (Zandman-Goddard et al., 1996).
Zazgornik et al. (1996) studied 36 nonselected women with implants (77% gel filled) and 36 age-matched controls in Vienna, Austria. A positive ANA, defined as a titer of 1:80 or more using mouse liver substrate, was found in 8% of controls and 33% of women with implants (a third of these were saline or PVP [polyvinylpyrrolidone] filled). Only one of the women with implants had clinical musculoskeletal symptoms, but 29 were breast cancer patients (versus only 2 of the controls), which may have affected the results of this small study (Zazgornik et al., 1996).
Lack of information on ANA technology, deficient controls, use of less stringent (low titer) or undefined criteria for ANA positivity, and biased selection of patients handicap the interpretation of studies by Freundlich et al. (1994); Ostermayer-Shoaib et al. (1994); Silverman et al. (1996b); Smalley (1995c); Solomon (1994a); and Vasey et al. (1994). Ostermayer-Shoaib et al. (1994) reported a prevalence of ANAs in symptomatic patients with breast implants of 36%. ANA assay methods are not described, nor is there information about the titers of these antibodies or their frequency in control sera. This cohort was apparently largely self- or
attorney referred (Ostermayer-Shoaib et al., 1994). Silverman et al. (1996b) studied 3,184 consecutive symptomatic, self- and attorney-referred patients with breast implants, 40 age-matched controls, 37 asymptomatic women with breast implants, and 200 consecutive fibromyalgia patients without implants from Arizona and California. ANAs were performed using HEp-2 cells. Positives were determined at titers of 1:80 or more. The prevalence of ANAs was 35% in symptomatic women with implants, 28% in fibromyalgia patients, 3% in asymptomatic women with breast implants and 5% in healthy controls. The small number of control patients, the potential selection bias, and the absence of mean age and other demographic data are problems in this study. The study results suggest that ANAs may be more closely associated with fibromyalgia, which is also suggested by the results of Claman and Robertson and Bridges (Silverman et al., 1996b). Freundlich et al. (1994) reported 50 self- or attorney-referred women with breast implants who had medical complaints. Some of these patients had defined connective tissue disease. Positive ANAs were detected in 24% of those patients. The titer used to define a positive test was not stated (Freundlich et al., 1994).
A similar study was reported by Vasey et al. (1994) in which 50 patients with symptoms and some (ca. 20%) with definite rheumatic disease were studied. No control group was assembled. An ANA titer of 1:20 was considered positive. Of these women with implants, 13 (26%) were positive (Vasey et al., 1994). Another uncontrolled study of symptomatic women mainly attorney referred was reported by Solomon (1994a) in which 44 of 176 (25%) women with implants had an ANA titer of 1:40 or more. Smalley et al. (1995c) described the presence of ANAs in sera from 231 symptomatic women with silicone gel implants. Symptoms included fatigue, arthralgias, myalgias, joint swelling, sicca syndrome, hair loss, atypical rashes, poor memory, and other symptoms ''similar to and consistent with those reported by Bridges et al. (1993a)." ANAs were assayed on HEp-2 cells and tested at an initial titer of 1:40. Positive ANAs were found in 27.7% of patients. No information was provided regarding the presence of a defined rheumatic disease in these patients or whether they were examined by a rheumatologist, and there was no concurrent control group (Smalley et al., 1995c).
Young et al. (1995b, 1996) reported ANA titers in a study of human leukocyte antigen (HLA) types in four groups of women: Group I, 77 symptomatic women (primarily myalgia or arthralgia and fatigue) with breast implants; group II, 37 asymptomatic women with implants; group III, 54 healthy female volunteers; and group IV, 31 women with fibromyalgia with implants. ANA assays were performed using HEp-2 cells. A titer of 1:40 or more was considered positive. ANA titers of 1:80 or more were found in 24% of 58 symptomatic breast implant patients tested. Only
5 of 37 asymptomatic women with implants were studied, and they were negative. Sera from 11 of 31 women with fibromyalgia were tested, and 2 (18%) were positive. There was no relationship between ANA positivity and implant rupture or capsular contracture. The authors noted that there was a patient with SLE in group I. This study is of interest primarily for HLA and other data. The ANA tests were not part of the study and were carried out at patient (or patient insurance) expense; thus, there were no control sera available for ANA testing (Young et al., 1995b, 1996).
Brunner et al. (1996) surveyed 581 women who had had breast reconstruction or augmentation with implants in Munich, Germany. Of 239 who responded, six had an ANA titer of 1:80 or more. On clinical evaluation, 33.4% of the women were found to have some rheumatic symptoms (Brunner et al., 1996). In a study reported by Blackburn et al. (1997) of a small, selected group of women with implants, 70 patients referred for complaints or concerns were examined by rheumatologists, and a series of immunologic studies was performed; 5 of 58 (8.6%) patients tested had a positive ANA titer (1:40 or more). A control group was not assembled. Clinically, 3 of the 70 patients had a defined connective tissue disease (Blackburn et al., 1997).
One-half of approximately 370 consecutive silicone breast implant recipients attending a rheumatology practice between April and September 1994 were asked to enroll in a study, as were 40 nonrecipients with defined connective tissue disease. A significant number of women (N = 259, 70%) declined to participate as did 20 of the patients with connective tissue disease without implants. A control group of 23 women was recruited. The 95 participating implant recipients who did not meet criteria for a defined connective tissue disease were classified into four groups by severity of symptoms, from limited or mild to moderate or advanced, and by functional capacity, from normal to incapacitated. More than 90% of the women with implants reported fatigue and arthralgias, however, suggesting that those who participated were a selected group. There were 34 patients with limited symptoms and normal functional capacity, 26 with mild symptoms and functional incapacity, 16 with moderate symptoms and functional incapacity, and 19 with advanced symptoms and functional incapacity. Women with ANA titers of 1:40 or more in the four groups were 18, 19, 0, and 32%, respectively. Of controls, 4 (17%) were positive, and 14 (70%) of the patients with connective tissue disease without implants were positive. Overall, 26% of implanted women (compared to 17% of controls) had positive ANA tests. This study used an experimental group that appears selected (30% participation), used a control group of non-age-matched clinic personnel, had no symptomatic control group, and did not find a statistically significant difference between the women with implants and the controls (Tenenbaum et al., 1997a).
Park et al. (1998) studied 110 women in Scotland with breast implants for augmentation, of whom none had ANA titers of 1:40 or more, compared with three positive women in a group of 128 age-matched controls attending the plastic surgery outpatient clinic. In a group of 207 women with implant reconstructions after cancer mastectomy, 5 were positive compared with 4 of 88 control women who had had mastectomy for cancer without reconstruction. In this appropriately controlled study without apparent selection bias, no increase in ANA positivity was found in the women with breast implants even though human tissue culture cells were used and a titer of only 1:40 or more was considered positive (Park et al., 1998).
A Prospective Study
A recent publication by Miller et al. describes a study carried out from 1985 to 1998 of 414 women with saline- or gel-filled implants for reconstruction or augmentation. The study was designed to monitor any changes over time in the immune status of patients with breast implants by measuring rheumatoid factor, antistreptolysin O titers, C-reactive protein, and ANAs. Only the 218 patients who had these assays performed both before and after implantation are reported. Patients were divided approximately equally between into two groups, a gel implant and a saline implant group. The mean duration of follow-up for the gel implant group was 5.77 years (range 1 to 25 years), and the mean duration of follow-up for the saline implant group was 1.99 years (range 0 to 3.64 years). Positive ANAs were found in 3.57% of the silicone gel implanted patients before, and 4.16% after, implantation. In the saline implanted group, 14.15% were positive preoperatively, and 16.98% after implantation, a difference that is not statistically significant. The other measurements also were not significantly different statistically before and after implantation. The difference in values in the silicone gel and saline implant patients may reflect a change in the method of performing the ANA test, that is, using different substrates, as noted earlier. This cannot be confirmed because the methods section of the report does not describe the method for ANA assay or the criteria for a positive test. The more recent use of HEp-2 cells, which result in higher titers, may coincide with the more recent use of saline implants, however. This study is the only one of its kind discovered by the committee. The failure to obtain before-and-after samples on almost half the patients suggests a somewhat ad hoc approach, and the lack of information on ANA testing technology is an important omission. The follow-up period, especially for the saline group, was critically short also. Nevertheless, this report provides important prospective information on antinuclear autoantibody reactions to the stimu-
lus of gel and saline breast implants in patients with a previously defined ANA status. The study also presents an interesting reverse perspective (i.e., changes after placement of implants) from the scattered, anecdotal reports of changes, some increasing and some decreasing, in ANA test results after implant removal. The similarity of results with saline and gel implants is also notable (Miller et al., 1998).
Ana and Autoantibody Analysis in Epidemiological Cohort Studies
The medical records of all women with silicone breast implants in Olmsted County, Minnesota, were reviewed by Gabriel et al. (1994). Patients were divided into those with prophylactic mastectomy (90), those with cancer mastectomy (125), and those who had implants for augmentation (534). Almost all implants were gel filled. A positive ANA test was found in 11 of a cohort of 749 women with implants and in 27 of 1,498 healthy controls (a rate ratio of 0.86, 95% confidence interval [CI], 0.42-1.70), but titers were not reported. This is a thorough review of an unbiased cohort of women by a highly regarded epidemiologic group. However, ANA determinations were performed over a number of years, probably by differing technologies and undoubtedly by different personnel, and may not be comparable (Gabriel et al., 1994).
Edworthy et al. (1998) carried out a cohort study of women with cosmetic silicone breast implants between 1978 and 1986 identified through the procedural codes of the Alberta, Canada, Health Registry. Controls were women having other cosmetic surgery. Of the women with implants, about 22% had received only saline-filled silicone implants. Of those women who provided blood samples, 324 of 1,426 (22.7%) had a positive ANA by criteria of greater than 1+ fluorescence and a titer of 1:40 or more, compared with 162 of 649 (25.6%) of the control women. A significant fraction of the total 9,200 implantations recorded were lost because of missing addresses and refusal to participate, and this study was carried out during active litigation by claimants of silicone injury. These factors raise questions of potential selection in the experimental group. Nevertheless, this study has a large cohort of women with implants and controls without known selection bias who were evaluated clinically and for ANAs in a blinded fashion (Edworthy et al., 1998).
A further analysis of the cohort from the Nurses' Health Study reported by Sánchez-Guerrero et al. (1995a,b) and discussed in Chapter 8, was carried out by Karlson et al. (1999). In 1988, 32,826 women from the group of 121,701 female nurses enrolled in the Nurses' Health Study in 1976 provided 30 ml blood samples each. From the response to a 1992 questionnaire sent to all women in the study, 1,863 women with breast implants of all kinds placed before the date of blood collection were iden-
tified as described by Sánchez-Guerrero et al. (1995a,b). Of 288 women from this cohort of 1,863 women with breast implants who did not have connective tissue disease or cancer and who had provided a blood sample, 200 women were chosen randomly. Their average age was 53.6 years at the time of blood sample, and their clinical status was unknown.
Antinuclear antibodies, specific autoantibodies, including anti-ssDNA, anti-dsDNA, anti-Sm, anti-RNP, anti-SSA (Ro), anti-SSB (La), anti-SCL-70, rheumatoid factor, and antisilicone antibodies (according to Rosenau et al., 1996) were measured in the women with implants and compared with measurements in four age-matched control groups taken from the Nurses' Health Study cohort that had provided blood samples. These groups consisted of 200 healthy women without breast implants; 100 women with insulin dependent diabetes mellitus (presumably exposed to silicone, see discussion in Chapter 4, Chantelau et al., 1986; Collier and Dawson, 1985); 100 women with defined connective tissue disease; and 100 women with at least one documented symptom or sign of connective tissue disease. Further testing for complement (C3, C4), C-reactive protein, quantitative IgG, IgA and IgM and anti-thyroglobulin, anti-cardiolipin and anti-microsomal antibody was carried out on a random 25% of all tested women. All tests were performed by blinded laboratory personnel.
A positive ANA assay, defined as a titer equal to or greater than 1:40 using HEp-2 cells, was observed in 14% of the women with implants6% (95% CI, 1-13%) lower than in the healthy controls (20%) and 26% (95% CI, 14-36%) lower than in the control women with connective tissue disease (39%). Positive ANA assays in the women with diabetes were not significantly different (15%), and in women with connective tissue disease signs and symptoms were higher (37%). Antibody to ssDNA was significantly higher (p = 0.02) in women with implants (41%) than in healthy women (29%), a 12% (95% CI, 3-21%) difference. All other antibodies, complement, and immunoglobulin levels were not elevated in women with implants compared to healthy control women without implants. No antisilicone antibodies were detected in any of the 700 women. Additional studies of 17 women with silicone breast implants and definite or self-reported connective tissue disease were not revealing, except that this group had significantly (p = 0.02) lower anti-ssDNA (12%) than the original 200 women with breast implants. There were no differences among women with saline, gel, polyurethane, or double-lumen implants (Karlson et al., 1999).
Antibody to ssDNA is found in normals at low prevalence and is elevated in a variety of connective tissue diseases and inflammatory or infectious conditions. The significance of this antibody in these women is unknown. Clinical data are not available, elevations were not confirmed
in women with implants and self-reported or definite connective tissue disease, and (in an abstract) Simpson et al. (1994) reported no difference in anti-ssDNA between 40 symptomatic women with silicone breast implants and 40 healthy control women without implants. This study provides evidence against an association between silicone breast implants and ANAs or other autoantibodies, except ssDNA. This cohort appears to be unbiased, laboratory examinations were carried out according to carefully defined techniques with attention to quality control and in blinded fashion, and several appropriate control groups from the same large group of women were analyzed in an equally blinded and careful way at the same time (Karlson et al., 1999).
Specific Antinuclear Autoantibodies in Normals and the Predictive Value of a Positive Autoantibody Test
In 1989, a study of autoantibodies in 506 sera randomly selected from 5,000 blood samples from healthy women of childbearing age in the Negev that were drawn for a study of viral antibodies was reported. Sixty women were found to have autoantibodies against at least one of ten nuclear antigens, including anti-ssDNA, anti-dsDNA, anti-La, anti-Ro, anti-Sm, and anticardiolipin. Of these women, 57 were evaluated after five years to see whether an autoimmune disease had developed. None was found to have overt autoimmune disease although 7 of the 57 had symptoms associated with rheumatic diseases such as Raynaud's phenomenon, arthritis, or multiple abortions (Yadin et al., 1989). One study showed that ANA-positive healthy women rarely (less than 1%) develop SLE, and another found that no women with highly positive ANAs developed overt disease and only 12% had any symptoms after five years of follow-up (Aho et al., 1992; Schoenfeld and Isenberg, 1989). Anticentromere antibodies are rarely present in normal individuals (Lee et al., 1993; Rothfield, 1998, unpublished data on 898 patients), but they may be found in patients with Raynaud's phenomenon. The presence of either antitopoisomerase I or anticentromere antibodies was associated with a 163-fold increased chance of developing a connective tissue disease. Gaither et al. (1987) reported that about 17% of normal sera have elevated titers of anti-Ro by enzyme linked immunosorbent assay (ELISA).
In a study potentially relevant to the finding of ANAs in symptomatic women with breast implants, Clegg et al. (1991) using mouse liver and kidney and HEp-2 cells and titers of 1:32 or more, reported 205 patients with either isolated Raynaud's phenomenon or unexplained polyarthritis, or undifferentiated connective tissue disease and found ANA positives of 55, 57 and 59%, respectively. When seen five years later, these
patients generally remained the same or, in a few cases had remitted, but about 20% had progressed to defined connective tissue disease (Williams et al., 1998), indicating that frequent elevated ANAs may suggest future health problems in some groups of patients. In general, as discussed here, ANAs are also present in normal women and in the elderly, and can be associated with a number of conditions as a laboratory finding without heralding the onset of autoimmune disease.
Specific Antinuclear Antibodies and Rheumatoid Factor in Women with Implants
Peters et al. (1994a) studied anti-Ro, anti-La, anti-Sm, anti-Scl-70 and anti-RNP antinuclear autoantibodies using an ELISA technique, but no details of the antigens used are given. In ANA-positive women with implants, anti-Ro was positive in 33.9%, anti-La in 11.3%, anti-Scl-70 in 20.7%, anti-Sm in 9.4%, and anti-RNP in 5.6%. Comparable results in ANA-positive control women were 35.7, 3.5, 10.7, 7.3, and 3.5%, respectively. These data are most unusual in the very high prevalence of (and differences in) anti-Ro and anti-La in patients and controls and in the very high prevalence of anti-Scl-70 and anti-Sm in controls and patients without SLE or scleroderma (Peters et al., 1994a). The strengths and weakness of this study have been noted earlier.
In a study of 231 symptomatic women with silicone gel breast implants, Smalley et al. (1995c) found that 5.2% had anti-DNA antibodies using an ELISA test and 20.5% had rheumatoid factor by nephelometry. No controls were reported. The authors interpreted Western blot bands at 105 and 70 kDa (kilodaltons) as being anti-Scl-70 (antitopoisomerase-I). Other bands were interpreted as reacting with other autoantigens such as Ku and Ro. There were no supporting data in which the sera reacting with certain bands were reacted with specific autoantigens. The authors did not use standardized reference sera containing specific autoantibodies such as anti-opoisomerase I or anti-Ro as controls, which are available from the CDC.
Cuéllar et al. (1995b) reported a study of 470 ANA-positive sera in which other autoantibodies, anti-dsDNA, anti-Ro, anti-La, and anti-Scl-70, were studied. Anti-dsDNA was found in six patients by indirect immunofluorescence and in six by ELISA. Anti-Ro was found in six, anti-La in four, anti-RNP in one and anti-Scl-70 in one. Anticentromere antibodies were seen by indirect immunofluorescence on HEp-2 cells in five patients. The difficulties in evaluating this study were noted earlier. Lewy and Ezrailson (1996) found anti-dsDNA autoantibodies in 17.13% of their patients, but no details of methods used were given, and there was no control group. In the report of Bridges et al. (1996), 19 patients were
positive for anti-Ro, and of these, 9 were also positive for anti-La. One patient each had anti-Ro and anti-Scl-70 antibodies. Five patients had anti-RNP antibodies, one had anti-dsDNA antibodies, and one had anti-Sm antibodies. None of the normal controls, asymptomatic breast implant patients, or fibromyalgia patients was positive. This report used controls of three types, although they were not age matched. A high ANA prevalence was reported in fibromyalgia patients, and there was no information on the presence of defined connective tissue disease in the 500 symptomatic breast implant patients. Very high, specific antinuclear autoantibody prevalences were found in the previously discussed study by Bar-Meir et al. (1994). In fact the percentage of women with breast implants and at least one autoantibody (84%) was strikingly higher than in the other reports cited here.
Gabriel et al. (1994) found a similar prevalence of rheumatoid factor in implant patients and controls (both 0.5%); no difference in prevalence of antimicrosomal antibodies was found. In their large multisite study, Silverman et al. (1996b) found 13.5% of 392 ANA positive women with implants to have anti-cardiolipin antibody. They also found low frequencies of anti-DNA (0.6%), anti-Sm (0.2%), anti-RNP (0.6%) antibodies and rheumatoid factor (3.7%). Anticentromere and antithyroid microsomal antibodies were present in 4.6 and 11.7%, respectively. Young et al. (1995b) found that 10% of their 58 symptomatic patients with implants had positive rheumatoid factor. Similar results were reported by Bridges et al. (1993a). In the study by Edworthy et al. (1998), two implant patients and seven control women were anti-Ro positive, one implant patient was anti-Ro and anti-La positive, and one control was anti-La positive. Anti-DNA antibodies were found in five implant patients and five controls; Anti-Scl-70, anti-Sm, and anti-RNP antibodies were not found. Rheumatoid factor was present in 5.5% of women with implants and 7.4% of control women (Edworthy et al., 1998). This is an important study because it reviewed a very large cohort and control group, the serologic tests were performed blinded, and clinical assessments were done by rheumatologists who also were blinded. No difference in specific autoantibodies was found between controls and silicone breast implant patients.
In the study of Karlson et al. (1999) discussed earlier, the authors compared specific autoantibodies to ssDNA, dsDNA, Sm, RNP, Ro, La, SCL-70, cardiolipin, thyroglobulin, and microsomes in 200 women with implants to the same autoantibodies in four control groups. Except for anti-ssDNA, there was no significant difference in autoantibodies between women with implants and 200 healthy controls or 100 women with diabetes. A number of autoantibodies were more prevalent in women with defined connective tissue disease or signs and symptoms of connective tissue disease, as might be expected. This study provides evidence against
an association of autoantibodies and silicone breast implants (except anti-ssDNA). The study examined antibodies in a large, random cohort with identified techniques, blinded laboratory personnel and four appropriate concurrent control groups (Karlson et al., 1999). In general, the reports discussed earlier of ANAs in women with implants describe ANAs that are not reactive to defined autoantigens, and specific antinuclear autoantibodies when evaluated are not found.
As noted at the outset, studies of ANAs in women with silicone breast implants are subject to a number of weaknesses. Results reported have varied from positive to negative in a number of experimental groups, including women with saline or gel implants and those with connective tissue disease, an array of symptoms and disabilities, fibromyalgia, or no symptoms. No differences between saline and gel implants emerge from these studies, but results are not always reported by type of implant. A number of different control groups have been reported including historical, concurrent, asymptomatic or healthy, with fibromyalgia, with soft tissue rheumatism, with connective tissue disease, and with diabetes, and they have been assessed using different ANA technologies and criteria for positivity. Studies with no controls at all are essentially case reports. Even though some may report large numbers of women, they offer only weak evidence. Different results of testing for defined antinuclear autoantibodies have also been reported. Even theoretically well designed, prospective studies (Miller et al., 1998) have problems, such as short follow-up, failure to include significant portions of the potential experimental group, and no description of the testing technology. The fact that a positive ANA test is not a disease diagnosis should also be kept firmly in mind.
The cohort studies, however, add strength to the evidence against an association between silicone breast implants and ANAs or other autoantibodies. The committee concludes that the data in support of a finding of increased prevalence, higher titers, or different profiles of antinuclear antibodies in women with gel- or saline-filled silicone breast implants compared to control women without breast implants are insufficient or flawed. The weight of the better-quality evidence suggests the lack of an association between silicone breast implants and positive ANAs. Although there are fewer data on specific autoantibodies, they also suggest no association and are insufficient to support a finding of increased prevalence or different profiles of specific autoantibodies in women with silicone breast implants.