Association between Bovine Collagen Dermal Implants and a Dermatomyositis or a Polymyositis-like Syndrome

Methods

Case Finding

A case was defined as any person with definite or probable dermatomyositis or polymyositis [7] who had a history of bovine collagen injection(s) before their diagnosis. Case finding was achieved through various methods, including patients (seen by coauthors JC and JS) at the Doctors Center and Arthritis Clinic of Houston (three cases), self-referred patients (one case), and patients found after review of adverse reaction reports received by the manufacturer (five cases). Because no reports after June 1988 were available for review, the study period was limited to July 1980 through June 1988.

Determining Expected Dermatomyositis or Polymyositis Case Ratio

Annual estimates of the number of new patients treated with collagen and their demographic characteristics were obtained from the manufacturer. These data were combined to produce age-, race-, and sex-specific monthly estimates of the collagen-treated population during the study period. The test and treatment dates for each case patient, together with the date of diagnosis, were used to derive an average latency period for that patient. As each of the nine case patients was added to the series, the period for that case patient was combined with those of all previous case patients to produce the mean latency period for the group up to that point in time. The cumulative number of patients who had received collagen injections to that point in time was determined from the projected 1-month treatment cohorts. These figures were combined to derive the effective total patient-years of observation for collagen recipients through the date of diagnosis for each case patient. Age-, race-, and sex-specific incidence rates of dermatomyositis or polymyositis were obtained from a 20-year, hospital-based study by Oddis and colleagues [3], in which the incidence during 1973 to 1982 more than tripled compared with 1963 to 1972. Because the second decade of this study is closer in time to our 1980 to 1988 study period, the incidence rates were adjusted upward to produce an overall rate of 8.6 instead of their reported 20-year rate of 5.5 case patients/million population per year. These adjusted rates were then applied to the patient-years of observation to arrive at the number of dermatomyositis or polymyositis cases expected at the point in time when each of the observed case patients was diagnosed. From these data, the standardized incidence ratio, the 95% confidence interval, and the Poisson probability were calculated for cumulative observed and expected pairs.

Oddis and colleagues [3] reported that, of the 115 persons with adult dermatomyositis or polymyositis (not including patients who also had other types of connective tissue disease [overlap patients] or patients whose cases were associated with malignancy), 88 (77%) had polymyositis and 27 (23%) had dermatomyositis. In our series, only 1 case patient (11%) had polymyositis, whereas 8 case patients (89%) had dermatomyositis. To evaluate the importance of dermatomyositis alone in the series, the single patient who had polymyositis was excluded. The incidence rates from the Oddis and colleagues [3] study were modified to reflect the proportion of dermatomyositis case patients with respect to polymyositis and dermatomyositis combined. The number of dermatomyositis case patients, expected at the point in time when each of the eight observed case patients were diagnosed, was then calculated using the method described above.

Monte Carlo Simulation

The third phase of the analysis used a Monte Carlo simulation to further assess the apparent temporal association between bovine collagen dermal implants and the development of dermatomyositis or polymyositis. The term Monte Carlo simulation refers to any experiment in which random numbers are drawn (either using a table of random numbers or using computer generation) to simulate sampling from a population or probability distribution [20]. These techniques are often used to simulate experiments that either cannot be done in the laboratory, are mathematically intractable, or would require prohibitively expensive equipment [21-23]. Although the simulation results are only an approximation of the true value, any desired degree of accuracy may be obtained by selecting a large enough number of trials [24].

The first step of the analysis involved the calculation of the maximum number of patients who could have been observed at intervals from 1 to 96 months after collagen exposure. Because the collagen-treated patients were not under any form of systematic medical follow-up, these calculations were made only to determine the theoretical maximum number of dermatomyositis or polymyositis case patients expected for this select population. Using these assumptions, patients treated early in the study period contributed many patient-months of observation, whereas those treated in the last month were considered to have been observable for only a single month. These data were summed and converted to patient-years of observation within each of the 96 observation periods. The age-, race-, and sex-specific dermatomyositis or polymyositis incidence rates were applied to the patient-years of observation to arrive at the number of patients with dermatomyositis or polymyositis expected to occur in each age group and within each of the 96 periods. Expected case patients from each age group were summed within each postexposure observation period to produce a single total for each of the 96 periods. This distribution was then converted to a probability distribution by dividing each element by the sum for the entire distribution. The corresponding data for expected dermatomyositis case patients were calculated by adjusting the rates from the Oddis and colleagues [3] study to reflect the proportion of dermatomyositis case patients with respect to polymyositis and dermatomyositis combined. The monthly numbers of expected case patients were aggregated into 6-month intervals, and these data were plotted against the observed cases.

The Monte Carlo simulation generated one million sample groups with nine items per group. Each of the nine items (corresponding to the nine observed case patients) received a value ranging from 1 to 96, based on the probability distribution for expected cases. The nine values from each sample group were combined to produce an average score, corresponding to the mean latency period for that particular sample. The number of trials resulting in each possible average score was counted, and the total number of average scores less than or equal to the average mean latency period for the nine reported case patients was determined. This number, divided by the total number of trials (for example, one million), was then interpretable as the empirically derived probability of finding, on the basis of chance alone, a sample of nine dermatomyositis or polymyositis case patients having a mean latency period less than or equal to the latency period observed in our study group.

Antibody Assays

Serum samples were available from one patient (patient 7) for collagen antibody analysis in our laboratory. An enzyme-linked immunosorbent assay (ELISA) was used with native, pepsin-modified chick type I collagen and human types I and III collagens as test antigens [25, 26]. Based on previous experience, titers of > 1.0 log₂ dilutions were considered positive. The positive control, used as a reference for the assays, was a patient with rheumatoid arthritis who had titers of 3.0, 6.0, and 5.5 to chick type I collagen and human types I and III collagens, respectively. Negative controls consisted of two collagen implant patients who had titers of zero to all three test antigens.

Data Analysis

Most of the data analysis was done on a Standard 286-12 microcomputer using SuperCalc 4 software. Each of the calculating spreadsheets used was constructed specifically for the application. The Monte Carlo simulation was run on an IBM 3081 mainframe computer using a customized FORTRAN program with the linear congruential pseudorandom number generator, RAND, written by Law and Kelton [27].

Results

Patients

Patient 1

A 47-year-old white woman developed a positive reaction within hours of a test dose of Zyderm given on 20 April 1981 (Table 1). The test site remained pruritic for several hours, and redness and swelling persisted for several days. These results were overlooked, and treatments were administered on 22 June 1981 and 13 July 1981. Local reactions developed almost immediately at the injection sites. Six to 8 weeks after the last treatment, the patient complained of facial redness, itching, and scaling. The rash spread to the entire body. During the next 2 months, the patient developed muscle weakness, facial swelling, and fever. In late November 1981, the rash became more typical of dermatomyositis. During this period, the antinuclear antibody titer showed a faintly positive, speckled pattern at a 1:10 dilution. On 18 January 1982, the patient was hospitalized because of marked muscle weakness in all extremities. The creatine kinase level was increased at 4.33 to 5.33 kat/L (normal, 0.42 to 2.08 kat/L; 260 to 320 U/L [normal, 25 to 125 U/L]), and the electromyogram was consistent with the diagnosis of dermatomyositis or polymyositis. A test for anti-Zyderm antibodies, obtained after the patient had received prednisone for approximately 4 months, was negative by radioimmunoassay (done by Collagen Corp.). Despite plasmapheresis (2 to 3 times/wk) and cyclophosphamide therapy, severe weakness required assisted ventilation and intensive care for 4 months followed by residence in an intermediate facility or hospital until follow-up was lost in September 1986.

Patient 2

A 40-year-old white woman received a test dose of Zyderm on 9 March 1982 that was negative (Table 1). Subsequently, she received Zyderm treatments on 19 April, 3 May, 19 July, and 23 August 1982. Transient erythema and periorbital edema were observed after each treatment. Five days after the last treatment, she complained of painful, swollen joints in the hands, wrists, and ankles, accompanied by a fever of 38.3 C (101 F). By September 1982, symptoms had progressed to severe proximal muscle weakness and myalgias. A rash characteristic of dermatomyositis was noted in conjunction with diffuse erythema over the arms and legs, and there were bilateral knee effusions. Both the muscle biopsy and the electromyogram were indicative of dermatomyositis or polymyositis, and the patient was hospitalized 4 days later. Anti-Zyderm antibodies were positive by radioimmunoassay at 251 counts/min (cpm) (normal, 0 to 157 cpm). Six plasmapheresis treatments, beginning on 12 October 1982, resulted in considerable improvement. The patient received prednisone (10 to 25 mg/d), and during a follow-up period extending for 4 years she had mild persistent weakness of the proximal muscles of her lower extremities. Creatine kinase levels ranged from 1.33 to 159 kat/L (80 to 9540 U/L). During this time, cardiomegaly along with pericardial and pleural effusions were noted, and the characteristic dermatomyositis rash persisted. The antinuclear antibody titer was 1:160.

Patient 3

A 38-year-old white man received a Zyderm skin test on 21 February 1984 (Table 1). After a negative response to the test injection, he had three treatment sessions during the period from 27 March 1984 to 26 April 1984, receiving multiple injections of Zyderm for facial acne. By 19 May 1984, he had a generalized proximal muscle weakness and a rash characteristic of dermatomyositis. Increased muscle enzyme levels and results of an electromyogram and a muscle biopsy were consistent with dermatomyositis or polymyositis, and therefore he started receiving prednisone therapy.

Patient 4

A 53-year-old white woman had a negative response to a Zyderm skin test in November 1983 (Table 1). Subsequently, she received Zyderm implants in three treatment sessions between 8 December 1983 and 15 January 1984. Within 1 month of the last injection, erythema and induration were observed at the treatment sites. On 13 February 1984, a consulting dermatologist noted the heliotrope rash of dermatomyositis. The rash persisted, and, in August 1984, the patient had proximal muscle weakness, fever, fatigue, and dyspnea on exertion. The creatine kinase level was 11.2 kat/L (671 U/L), and the aldolase level was 834 kat/L (50 U/L [normal, 2 to 8 U/L]). Electromyogram and muscle biopsy results supported the diagnosis of dermatomyositis or polymyositis. Bilateral pleural effusions were seen on x-ray. Anti-Zyderm antibody levels were increased (320 cpm).

Patient 5 (Index Case)

A 39-year-old white woman received a test dose of Zyderm on 29 July 1983 (Table 1). On 15 October 1983, after the test site had remained negative for more than 2 months, she received Zyderm implants. Within 48 hours, the patient developed redness, swelling, and tenderness at the treatment sites along with a similar reaction at the previously normal test site. By 72 hours after treatment, she complained of generalized malaise, headache, chills, facial swelling, and severe abdominal and back pain along with peripheral joint discomfort and myalgias. A serum sample, obtained on 12 December 1983, was positive for anti-Zyderm antibodies (506 cpm).

The injection sites remained erythematous and swollen during the next several months, exacerbated by dietary exposure to beef products. During the next 6 to 8 months, she had malaise, easy fatigability, myalgias, muscle wasting, weight loss, photosensitivity, hair loss, heat intolerance, and depression. By October 1985, she had developed a rash on her face and hands that was characteristic of dermatomyositis, along with proximal muscle weakness. A creatine kinase level of 26.5 kat/L (1590 U/L), a muscle biopsy, and an electromyogram confirmed the diagnosis of dermatomyositis.

An antinuclear antibody titer, done in 1980 to evaluate a single episode of mild stomatitis, was negative. When an antinuclear antibody titer was obtained on 3 April 1985, it was positive at greater than 1:2560 with a speckled pattern. Subsequent studies confirmed the high titer, but a precipitin panel was negative. Anti-Zyderm antibodies were still present in serum samples from 11 April 1985 (ELISA titer of 1:320, Collagen Corp.) and from 2 August 1985 (ELISA titer > 1:100). She was treated with prednisone from 18 October 1985 until 17 October 1986. Azathioprine was added on 18 December 1985 and hydroxychloroquine on 16 June 1989; she is still receiving both drugs. Bilateral silicone breast implants had been placed in May 1981.

Patient 6

A 48-year-old white woman received a Zyderm skin test on 7 November 1985 that was negative (Table 1). Zyderm implants were administered on 10 December 1985. Within 24 hours, she developed persistent redness, pain, and swelling at the treatment sites. In March 1986, the previously negative test site developed a similar reaction, and the patient reported flu-like symptoms along with proximal muscle weakness. After development of a characteristic skin rash, a diagnosis of dermatomyositis was made. A radioimmunoassay for anti-Zyderm antibodies, done in April 1986, was positive at 640 cpm. The patient had many exacerbations of her cutaneous signs and symptoms, not only at the treatment sites but also at the test site, after ingestion of beef products, use of cosmetic creams containing collagen, and after exposure to heat or stress. The patient also had substantial hair loss in late 1986, and she developed significant weakness in her neck muscles in early 1987. Several increased creatine kinase levels, obtained during the period from April 1986 through April 1987, averaged 6.95 kat/L (417 U/L) (n = 6; range, 2.85 to 14.9 kat/L [171 to 892 U/L]). Although normal electromyograms were reported in 1986 and 1987, a muscle biopsy taken from the right vastus lateralis in late 1988 showed a nonspecific myopathy with no inflammatory infiltrates. Her symptoms continued to worsen, and she began having dysphagia that resulted in several aspiration events in 1989. She was placed on prednisone therapy beginning in March 1989 for approximately 18 months, until her dermatomyositis was quiescent.

Patient 7

A 36-year-old white woman received a test dose of Zyplast on 1 May 1987 that produced a test site reaction consisting of marked redness, tenderness, and swelling that precluded treatment (Table 1). By 7 May 1987, she had panfacial edema and erythema along with severe fatigue and generalized malaise by the end of the month. In early July 1987, she had profound weakness and recurring fevers of 37.8 C to 38.9 C (100 F to 102 F). By August 1987, creatine kinase levels were increased with values up to 262 kat/L (15 700 U/L); the aldolase level was 5668 kat/L (340 U/L), and a rash developed characteristic of dermatomyositis. The electromyogram and muscle biopsy results supported this diagnosis. In September 1987, she was started on prednisone (60 mg/d), and, in October, methotrexate (15 mg/wk) was added. By February 1989, the prednisone had been reduced to 30 mg daily and her immunosuppressive therapy was switched to azathioprine (100 mg/d).

We detected IgG antibodies to collagen, using native chick type I as a test antigen, in a serum sample obtained in February 1991 (ELISA titer, 4.5 log₂ dilution, normal, 1). IgG antibodies were also identified when native human type I or type III collagen was used as a test antigen (titer, 4.0 for both antigens); they were still present in a sample obtained in March 1992 (titer, 3.0 for both antigens).

Patient 8

A 43-year-old white woman developed redness and swelling at the test site within 24 hours after a skin test on 4 May 1987 (Table 1). No treatment was given, but in October 1987 the patient reported muscle weakness. The creatine kinase level was 50.0 kat/L (3000 U/L), and an electromyogram test result was typical of polymyositis, prompting prednisone therapy (60 mg daily). By December 1987, the creatine kinase level had decreased to 8.34 kat/L (550 U/L), and the weakness was considerably improved. In January 1988, after her prednisone dose had been reduced to 10 mg daily, her symptoms worsened and the creatine kinase level increased to 36.7 kat/L (2200 U/L). Increases in the prednisone dosage failed to improve her symptoms and methotrexate (7.5 mg weekly) was added. Because it was ineffective, prednisone was discontinued in July 1988. One month later, the patient developed diffuse interstitial pneumonia and cardiomyopathy, possibly related to methotrexate, and she died.

Patient 9

A 39-year-old white woman received a skin test with Zyplast on 1 March 1988 (Table 1). The test site was negative, but on 27 March 1988 she started having myalgias with severe muscle weakness. On 1 April 1988, Zyplast treatments were administered. Shortly thereafter, the patient developed erythema, tenderness, and induration at the treatment sites that persisted into May 1988. Clinical evaluations between 4 April 1988 and 6 April 1988 showed a dermatomyositis rash, a proximal muscle weakness, a creatine kinase level of 34.3 kat/L (2060 U/L), an aldolase level of 183 kat/L (11 U/L), an electromyogram indicative of dermatomyositis or polymyositis, an antinuclear antibody titer of 1:320 with a speckled pattern, and a negative precipitin panel. Treatment with prednisone and methotrexate was initiated, and she is still receiving these drugs. Medical history included bilateral silicone breast implants in 1979.

Clinical Summary of Case Patients

Nine patients with a dermatomyositis or polymyositis-like illness were found among the approximately 345 000 patients receiving collagen dermal implants for cosmetic purposes during the period from July 1980 through June 1988. Seven of the patients had a combined total of 15 treatment visits (range, 1 to 4 visits per patient) (Table 1). The remaining two cases occurred in patients who had positive skin test reactions and, therefore, did not receive implants. The average age of the patients at diagnosis was 42.6 years (range, 36 to 53 years); all were white, and eight of the nine patients were female. Seven patients met criteria [7] for definite dermatomyositis, 1 patient for probable dermatomyositis, and 1 patient for probable polymyositis. None of the patients could be classified as having myositis that occurred with other connective tissue disease (overlap) or having myositis associated with malignancy. The interval from the mean of collagen exposures to diagnosis for the nine patients averaged 6.4 months (range, 0.7 to 24.9 months). Eight patients were diagnosed within 8 months of their last collagen treatment; the remaining patient (patient 5) was not referred for a rheumatologic workup until she had been ill for about 20 months. Systemic symptoms were noted an average of 43.2 days after the last collagen injection (29.9 days for the eight patients with dermatomyositis). The average time from the onset of systemic symptoms to diagnosis of dermatomyositis or polymyositis was 4.16 months. In the patients for whom the information was clearly described, positive responses to the skin test or local inflammatory reactions or both at the injection sites were observed. Serum antibodies to collagen were increased in four of the five patients tested; the patient with a negative result (patient 1) had been receiving prednisone therapy for approximately 4 months.

Statistical Analysis

When all the patients with dermatomyositis or polymyositis were compared with the estimated contemporaneously treated population, a statistically increased (P < 0.007) incidence was noted beginning with the second patient. If all collagen-treated patients were observed for a period comparable to that spent on observing the patients with dermatomyositis or polymyositis (for example, 6.4 months), only 1.78 cases would have been expected by the end of the study period. With nine observed patients, the resultant cumulative, standardized incidence ratio was 5.05 (CI, 2.31 to 9.59, P < 0.0001). A similar analysis focusing on dermatomyositis alone produced significant results (P < 0.01) beginning with the first case. Only 0.43 patients with dermatomyositis were expected; the observed eight patients produced a standardized incidence ratio of 18.8 (CI, 8.1 to 37.0, P < 0.0001).

Five of the women who had dermatomyositis or polymyositis were 35 to 44 years old; the remaining three women were 45 to 54 years old. All the women were white. The resulting age-specific incidence rates for these two age groups were 103 and 65.3 patients/million population per year, respectively. These rates show a considerable increase above those reported by Oddis and colleagues [3] for the same age groups, which were 5.5 and 10.9, respectively. Consequently, for these two age groups among white women, the standardized incidence ratio for dermatomyositis or polymyositis was 10.4 (CI, 4.5 to 20.6, P < 0.0002).

If all 345 000 patients treated during the study period had been followed continuously from their average time of exposure until June 1988, they would have a theoretical upper limit of approximately 1 020 502 patient-years of observation. Based on these data, the maximum number of patients with coincidental dermatomyositis or polymyositis would be about 10.0, with 1.67 case patients expected in the first 6 months after collagen exposure and 1.49 patients for the interval between 7 and 12 months after exposure. Figure 1 shows that the observed cases occurred much sooner after treatment than expected. Therefore, simple comparison of the 9 observed case patients with the net total of 10.0 expected patients overlooks the important temporal relation.

When the dermatomyositis case patients were examined in a similar manner, the maximum number expected would be about 2.35, with only 0.39 cases expected in the first 6 months after exposure and 0.35 cases in the second 6 months after exposure. These data are plotted in Figure 2, where the temporal shift is even more apparent. The observed eight dermatomyositis case patients represent a 3.4-fold increase over the expected maximum total of 2.35 patients (standardized incidence ratio, 3.4; CI, 1.5 to 6.7; P < 0.003).

As expected (under the Central Limit Theorem of statistical inference), the frequency distribution of sample averages from the Monte Carlo simulation (figure not shown) appeared to be a standard normal distribution with a mean of 25.4 months and a standard deviation of 6.30 months. The average post-treatment observation period of 6.4 months is more than 3 SD less than the expected mean. In a run consisting of one million sample groups, with nine items per group, an average period of 6.4 months or less was observed fewer than 72 times (CI, 57 to 91; P < 0.0008).

Discussion

Immunologic Responses to Bovine Collagen

On the basis of histologic studies in rodents, it was anticipated that a rapid invasion of host fibroblasts would occur into the collagen implant, followed by deposition of new collagen and neovascularization [28]. However, Zyderm was completely resorbed in humans within about 3 months, without substantial colonization by fibroblasts or deposition of new collagen [29]. Zyplast seemed to provoke a more intense inflammatory response [29, 30] that was followed by tissue fibroplasia and new collagen deposition [29-31]. Although results with Zyplast were somewhat superior to Zyderm, additional injections were required every 4 to 6 months to maintain a satisfactory result [31].

Candidates are routinely screened by skin testing [10]. Approximately 2.5% to 4% of apparently normal persons have a positive response, which indicates preexisting immunologic sensitivity to bovine collagen [32-34] and precludes treatment [12, 13, 34-40]. An additional 1% to 9.8% of patients who were initially skin test negative may develop hypersensitivity reactions after treatment [13, 31, 34, 38-43]. Adverse skin reactions have been correlated with an increased frequency of the HLA-B5 and HLA-DR5 antigens, together with an absence of HLA-DR4 antigen, suggesting that the process is under immunogenetic control [16].

Initial test- or treatment-site reactions include edema, erythema, induration, local pruritus, and local or generalized urticaria, which last an average of 4 months (range, 1 to 18 months) [13, 40]. Fifteen percent to 45% of patients with adverse treatment-site reactions may develop intermittent erythema and induration at the treatment sites that may recur for up to 3 years [40]. Such responses, which can last from hours to weeks, may occur spontaneously or may be triggered by exposure to sun or heat, consumption of alcohol, onset of menses, exercise, flare-ups of hay fever, and exposure to other stimuli that produce peripheral vasodilatation; they may also be triggered by the ingestion of beef products [12, 13, 40]. Six percent to 20% of patients who have local adverse reactions to either test or treatment implantation have reported systemic signs or symptoms, including fever, a flu-like syndrome, myalgias, arthralgias, or swelling of face, hands, arms, ankles, or feet [12, 13, 34, 36]. Persons having delayed-type hypersensitivity reactions to Zyderm or Zyplast generally have increased levels of circulating antibodies to the implant material [34, 39, 41, 43, 44]. Biopsy specimens from injection sites showing adverse reactions are typically described as diffuse granulomatous nodules or palisading foreign body granulomas [34, 35, 40, 41].

The finding of substantial antibody responses in approximately 3% to 5% of persons exposed through test or treatment injections, combined with histopathologic evidence of granuloma formation (and, therefore, a cell-mediated immune response), suggest that, for certain persons, injectable bovine collagen has a distinct immunogenic potential [41].

Bovine Collagen and Autoimmune Disease

In a 1982 questionnaire survey of 434 dermatologists, Castrow and Krull [12] received responses from 316 physicians who administered bovine collagen implants. From a collagen-treated patient population of approximately 7000, they reported 1 patient with dermatomyositis, 1 patient with an antinuclear antibody conversion who did not have signs or symptoms of autoimmune disease, and 14 patients with self-limited arthritis and arthralgia. Chaouat and colleagues [15] and Jarrett and Roguska-Kyts [14] reported separate patients who had transient autoimmune-like polyarthritis with onset occurring within 1 week after bovine collagen injections. In a 6-year investigation consisting of 4318 patients who received Zyderm skin tests only and an additional 5109 patients who received skin tests followed by Zyderm treatments, Cooperman and colleagues [13] reported 33 patients with arthralgia alone or in combination with various other local or systemic reactions or both, including myalgia. DeLustro and colleagues [43] reported eight patients who had substantial rheumatologic disease among those receiving bovine collagen, including three patients with rheumatoid arthritis, two with dermatomyositis, one with polymyositis, and two with other autoimmune diseases. In these studies [13, 43], the data were interpreted as being within the expected range. The three patients with dermatomyositis or polymyositis described in the two studies above are included among the nine patients in this study.

Reports [34, 39, 42-45] that antibodies to collagen do not crossreact with human collagen argue against the possibility that bovine collagen plays a role in the development of dermatomyositis or polymyositis in a select group of sensitive patients. This apparent lack of crossreactivity may, in part, be due to the fact that the anti-implant antibody levels reported by Cooperman and Michaeli [42, 45] were so low that partial crossreactivity might have been undetectable [42]. Furthermore, rather than using the more generally accepted normal range, defined by the mean value 2 SD (thus approximating the 95% CI), Cooperman and Michaeli [42, 45] as well as DeLustro and colleagues [34] used the mean value 3 SD in defining the normal range of antibodies to human collagen. This rather restrictive definition may have resulted in an excessive amount of type II error. We were able to analyze serum from patient 7. Antibodies to chick type I, human type I, and human type III collagen were clearly present in this serum, indicating that species crossreactivity does exist.

Study Limitations

Although we have shown a close temporal association between bovine collagen injections and the onset of a dermatomyositis or polymyositis-like illness, the number of patients identified so far is relatively small. Because the case-finding methods used depended primarily on adverse reaction reporting to the manufacturer and referral of patients (both of which are passive in nature), it is likely that case ascertainment is incomplete. This conclusion is substantiated by Rogers and colleagues [46], who studied the adverse drug event reporting practices of 1121 Maryland physicians. They found that only 18% of the observed adverse drug events that were moderate to severe were ever reported to any source and, of those, only 45% were reported to the product manufacturer. Assuming that physicians report adverse medical device events with similar frequencies, as many as 90% of these events may go unreported to the manufacturer.

Included in the Oddis and colleagues study [3] were 86 (48%) definite case patients, 56 (32%) probable cases, 28 (16%) possible cases, and 7 (4%) cases that did not meet the Bohan and Peter criteria [7] on first admission. Also included were 21 (12%) patients who had myositis and also other connective tissue disease (overlap) and 20 (11%) patients who had myositis associated with malignancy. In our case series, there were seven (78%) definite case patients, two (22%) probable cases, and no overlap cases or cases associated with malignancy. Adjusting the incidence rates to account for the differences would decrease the expected number of case patients by as much as 43% and increase the resultant standardized incidence ratios by as much as 75%. The fact that all of our case patients were white also differs from the usual black predominance for dermatomyositis or polymyositis [2, 3].

The necessity for using the Oddis and colleagues study [3] to derive expected incidence rates for dermatomyositis or polymyositis complicates the interpretation of our results. First, their study was hospital based, and no information exists for the frequency of dermatomyositis or polymyositis in an outpatient setting. Second, there is no assurance that the dermatomyositis or polymyositis incidence in their area, Allegheny County, Pennsylvania, is representative of the general incidence in the United States. Third, our study extends 6 years beyond Oddis and colleagues' [3] analysis. The incidence rates could have changed during this period, resulting in an inaccurate estimation of the expected incidence in our assessment.

Two of our patients with dermatomyositis or polymyositis had a history of silicone breast implants, placed 4 and 9 years before their diagnoses. An unproven condition, human adjuvant disease has been reported in more than 100 patients who have received silicone or paraffin breast implants [47-52]. The average latency period between mammoplasty and the development of human adjuvant disease is generally about 15 years (range, 2 to 25 years) [51]. However, human adjuvant disease generally simulates systemic sclerosis (scleroderma), rheumatoid arthritis, systemic lupus erythematosus, or mixed connective tissue disease [47-52]. No patients with dermatomyositis and only one patient with polymyositis after breast augmentation have been reported [47].

Elimination of the two patients who received silicone breast implants from the case series actually increased the standardized incidence ratio for dermatomyositis or polymyositis to 6.14 (CI, 2.47 to 12.6; P < 0.0002). The standardized incidence ratio for dermatomyositis was similarly increased to 24.1 (CI, 8.86 to 52.5; P < 0.0003). This occurred because one of the patients eliminated had the delayed diagnosis of 24.9 months, which had artificially extended the mean observation period after exposure to 6.4 months. With this patient eliminated, the period was decreased to 4.6 months for the remaining seven patients with dermatomyositis or polymyositis (six patients with dermatomyositis), and the effective total patient-years of observation was reduced from approximately 175 000 to 113 000. Thus, although the number of observed patients was decreased, the number of expected patients was decreased even more, and the standardized incidence ratio increased.

Conclusion

Because the observed number of patients with dermatomyositis or polymyositis is approximately what would be expected for more than 1 020 000 patient-years of observation, it could be argued that these patients represent only the expected coincidental cases for this population. However, such an argument would necessitate making the unrealistic assumption that all 345 000 collagen-treated patients have been monitored for the development of dermatomyositis or polymyositis from the time of first exposure through June 1988 and that any such cases (coincidental or not) would be reported as an adverse reaction to the manufacturer. Furthermore, our analysis has shown that such an assessment overlooks a number of other key factors in these cases. First, rather than being diagnosed at an average of 25.5 months after collagen exposure (as would have been expected), they were diagnosed an average of only 6.4 months after exposure. Second, approximately 3% of persons tested will have positive skin test reactions, and another 5% of patients treated will have adverse treatment site reactions. Of 9 reported patients with dermatomyositis or polymyositis, 3 patients with test-site reactions were observed compared with 0.27 expected patients (standardized incidence ratio, 11.1; CI, 2.3 to 32.5; P < 0.003). Similarly, 8 patients with treatment-site reactions were observed compared with 0.45 expected patients (standardized incidence ratio, 17.8; CI, 7.7 to 35.0; P < 0.0001). These proportions of positively reacting patients would be highly improbable if the observed patients simply represented the coincidental, expected patients with dermatomyositis or polymyositis in the collagen-treated population. Third, when dermatomyositis is evaluated separately, the 8 observed patients were already 3.4 times more than expected using the unrealistic assumptions of 100% follow-up and reporting as mentioned earlier.

Even though dermatomyositis or polymyositis case ascertainment to date is probably incomplete and the use of historic and regional incidence data is a hindrance, our data suggest that patients receiving collagen dermal implants are at increased risk for developing a dermatomyositis or a polymyositis-like syndrome. It would seem prudent to reassess the cosmetic use of injectable bovine collagen implants in a context of risk versus benefit.