Epidemiology of Breast implants

[Guest guest]

EPIDEMIOLOGY

>

> INTRODUCTION

> In the largest controlled epidemiology study(697) conducted to date,

> the researchers, a team from Harvard Medical School, Brigham & Women's

> Hospital, and the Harvard School of Public Health, found a

> statistically significant relationship between breast implants and the

> development of connective tissue disease.(698)That study, supported by

> research grants from Dow Corning Corporation, also found a

> statistically significant relationship between breast implants and

> other connective tissue disease (including mixed).(699) The study also

> found a number of relationships to be of "borderline" statistical

> significance, including the relationship between implants and

> rheumatoid arthritis,(700) implants and Sjogren's

> syndrome,(701)implants and poly or dermatomyositis,(702) and implants

> and scleroderma.(703) Nonetheless, the manufacturers have continued to

> tout the "controlled study" body of literature as "proving that

> silicone breast implants do not cause autoimmune disease, connective

> tissue, or various symptoms."(704)

>

> In truth, and as outlined by Professor Sander Greenland in his

> testimony before the Panel,(705) all of the existing controlled

> epidemiology papers are flawed and the "epidemiologic evidence is

> highly inconclusive." Dr. Greenland's criticisms were made without any

> review of the internal documents that show the manufacturers'

> involvement in study methodology, data collection, and analysis. When

> the flaws in the studies are considered and the results closely

> examined, the implanted population has elevated symptoms of disease.

>

> Multiple biomarker studies have shown elevated levels of

> autoantibodies, blood and tissue silicon levels, and signs of immune

> disturbance in women with implants. The evidence is strong that there

> are at least some identifiable subsets of women particularly

> susceptible to silicone induced signs, symptoms, and disease, and of

> course, the large studies average out the high effects in the

> subgroups with the low effects in the protected or less susceptible

> groups. The only epidemiological study of women exposed to inhaled

> silica found a strong association with UCTD and a variety of immune

> system biomarkers, as well as with classical CTD's. Three recent

> studies on people with silicone implants have found an association

> with UCTD, with severe atypical rheumatic diseases, and with atypical

> neurological conditions such as idiopathic progressive neuropathy and

> inner ear disease. These studies are discussed in detail later in this

> section, but the picture of both silica induced and silicone induced

> immune-mediated signs, symptoms and diseases with an atypical

> presentation is coherent and detailed: auto-immune phenomena and

> immune system perturbances are common in a significant subset of

> exposed persons, and the serious symptoms and end-point disease states

> are related to the known immunopathology. To consider only the small

> set of "controlled studies" cited by the manufacturers is to miss the

> bigger picture.

>

> In fact, the manufacturers' interpretation of the controlled studies

> flies in the face of accepted epidemiological interpretation which,

> when applied to the breast implant literature, shows a consistent

> pattern of increased risk of disease in the implanted population.

> Merely because a particular result does not reach a level of

> "statistical significance" does not mean that the results are of no

> value. As Dr. Greenland told the Panel,

>

> tatistical nonsignificance means only that the null hypothesis of

> no effect is statistically compatible with the data. . . . It does not

> mean that other hypotheses, for example, in this case, doubling of the

> rates are incompatible with the data. It does not even mean that the

> null hypothesis is the hypothesis most compatible with the data.(706)

>

> Dr. Greenland's opinion is consistent with the opinions of other

> leading epidemiologists.

>

> Dr. Greenland is the co-author, along with Dr. Rothman, of the

> soon-to-be published Second Edition of Dr. Rothman's widely-respected

> book Modern Epidemiology. Dr. Rothman has long been a critic of

> unquestioned reliance on P values. In a 1978 letter to the editor he

> wrote:

>

> [t]he decision-making mold, emphasizing the procedure of making a

> choice, fails to characterize the extent of difference between groups.

> The P value, which is the final common pathway for nearly all

> statistical tests, conveys no information about the extent to which

> two groups differ or two variables are associated. Highly

> "significant" P values can accompany negligible differences (if the

> study is large), and unimpressive P values can accompany strong

> associations (if the study is small). P values, therefore, are not

> good measures of the strength of the relation between study variables.

> P values serve poorly as descriptive statistics.(707)

>

> Dr. Rothman continued by providing an example:

>

> [c]onsider a study of a new drug, B, which is compared with a standard

> treatment, A, with the following results:

>

>

>

> OUTCOME

> TREATMENT

>

> Successful 7 14

> Unsuccessful 13 6

> 20 20

>

>

> OUTCOME TREATMENT Successful 7 14 Unsuccessful 13 6 20 20

>

> A two-tailed test of the null hypothesis that the two treatments are

> equally effective gives P = 0.06 with Fisher's exact procedure. These

> data might be reported in several ways. One way, the least

> informative, would be to report that the observed difference is "not

> significant." Somewhat more informative would be to report the actual

> P value; an inequality such as P>0.05 is not much better than "not

> significant," whereas P = 0.06 gives the P value explicitly, rather

> than degrading the measure into two arbitrary ranges, significant and

> not significant. An additional improvement would be to report P (2) =

> 0.06, denoting the use of a two-tail, rather than one-tail, P value.

> Any P value, however, fails to convey the descriptive finding that 70

> per cent were treated successfully with B as compared with 35 per cent

> with A, a difference in proportion of 35 per cent. A 95 per cent

> confidence interval for the difference in proportion ranges from - 1

> to +71 per cent (the positive values favor treatment . The position

> of the lower limit of the confidence interval corresponds to the

> outcome of the significance test: the interval includes zero

> difference, which corresponds to labeling of the observed treatment

> difference as "not significant." Quantitatively, the lower bound of

> the 95 per cent interval extending just slightly beyond zero

> difference corresponds to a P value slightly greater than 0.05. The

> full confidence interval, however, indicates that these data, though

> compatible with no real difference between the treatments, are equally

> compatible with B as being markedly more effective; the range of

> possibilities consistent with the data generally suggests a greater

> efficacy for treatment B. The complete confidence interval summarizes

> the findings clearly and unambiguously.(708)

>

> Finally, he concluded:

>

> [e]mphasis on confidence intervals puts into proper perspective

> statistical concerns that have occasionally befuddled researchers. One

> example is the controversy over the Yates correction, which sometimes

> has a modest effect on the P value. The equivalent of a Yates

> correction in calculating a confidence interval would be a slight

> change in the boundary of the interval. The precise position of the

> boundary of the interval, however, is not relevant to appropriate

> interpretation. The boundary of a confidence interval depends also on

> the degree of confidence that is arbitrarily selected; 90 per cent, 95

> per cent and 99 per cent intervals all differ in width, but rarely

> would different interpretations emerge from consideration of these

> different confidence intervals, because the precise location of the

> limits to the interval is of little practical consequence. Rather, it

> is the approximate position of the interval as a whole on its scale of

> measurement that governs the interpretation. Significance testing, on

> the other hand, is equivalent to funneling all interest into the

> precise location of one boundary of a confidence interval. In the

> past, journals have encouraged the routine use of tests of statistical

> significance; I believe the time has now come for journals to

> encourage routine use of confidence intervals instead.(709)

>

> A review of even the controlled studies of classically defined

> scleroderma, discloses the correctness of Dr. Rothman's position

> since, in virtually every case, the confidence intervals, while they

> may include the number one, demonstrate that a full analysis of the

> upper limits is consistent with a causative relationship. In his

> presentation, Dr. Greenland displayed a slide depicting the confidence

> intervals of the various studies:

>

> 1st Author Year Design Outcome Article RR est. (95% CI)

> Burns 1996 CaCo Sclero 1.30 (0.27-6.23)+

> Englert 1996 CaCo Sclero 1.00 (0.16-6.16)+

> 1994 Cohort Sclero 0 (0-78)

> Sjögren's infin. (0.051-infin.

> Goldman 1995 Cross Sclero 0 (0-2.05)

> -sect. Sjögren's 0.66 (0.13-3.42±

> Myositis 0 (0-3.71)

> Other/mixed 0 (0-2.68)

> Hennekens 1996 Cross Sclero 1.84 (0.98-3.46)

> -sect. Sjögren's 1.49 (0.97-2.28)

> Myositis 1.52 (0.97-2.37)

> Other/mixed 1.30 (1.05-1.62)

> Hochberg 1996 CaCo Sclero 1.10 (0.54-2.24)

> Liang 1996 CaCo UCTD 2.27 (0.67-7.71)

> Wolfe 1995 CaCo Fibromyalgia

> Osteoarthritis controls: 1.21 (0.18-8.29)

> Community controls: 2.67 (0.49-14.7)

>

>

>

>

>

> Crude scleroderma summary RR (1-6): 1.34 (0.87-2.06)

>

> homogeneity p-value=0.66

>

> + Silicone gel cases only

>

> ± Based on 1 case; excludes case that occurred before Implant

>

> Thus, as depicted above taking the worst example for the plaintiffs

> shows that the confidence intervals are consistent with the postulate

> of causation.

>

>

>

>

>

> B. A LARGE SAMPLE OF THE GLOBAL SETTLEMENT DATA DEMONSTRATES AN

> INCREASED INCIDENCE OF DISEASE AND SYMPTOMS IN EXPOSED WOMEN

> Extrapolations from a large sample of the claims submitted under the

> original global settlement demonstrate dramatic increases in the

> incidence of certain diseases in the population of implanted women.

> These data are not self-reported, but are based on the diagnosis of

> board certified physicians using recognized disease criteria and

> approved by the breast implant manufacturers.

>

> The history of the original global settlement is lengthy and complex.

> Beginning in December, 1992, representatives of plaintiffs and certain

> manufacturers began to meet to discuss a way to resolve the breast

> implant litigation. Finally, in late summer 1993, the parties reached

> a statement of principles and the manufacturers began negotiating

> among themselves as to how to fund the agreement. When the agreement

> was presented to the Court in the spring of 1994, Judge Pointer

> granted preliminary approval and established a notice and comment

> period and scheduled a fairness hearing for August, 1994. The

> settlement was ultimately approved by Order dated September 12, 1994.

> Under the terms of the agreement, the manufacturers were to provide

> funding for a Disease Compensation Program and for certain Designated

> Funds. The Disease Compensation Program would have provided

> significant payments to women who satisfied the Disease and Disability

> Definitions while the Designated Funds would have provided

> compensation or reimbursement for medical evaluations, explantation,

> and rupture payments.

>

> As the claims were submitted, it quickly became obvious that the money

> allocated to the settlement would be woefully inadequate to pay claims

> at the agreed upon levels. While the settlement had allowed for that

> possibility and permitted claimants to return to litigation in the

> event the payments were "ratcheted," it became apparent that due to

> the severity of the ratcheting -- payments in all likelihood would

> have been less than 5 % of the original amounts. Massive numbers of

> claimants would have rejected the settlement and attempted to pursue

> their claims in litigation. The manufacturers had the right to

> withdraw from the settlement if they deemed that too many women had

> rejected the offer. Faced with the eventuality of massive opt outs and

> a walk away by the manufacturers, the Court allowed a Revised

> Settlement Program to be offered to women who could either accept the

> offer or reject it and pursue their claims through the tort system.

>

> The current importance of both the original global settlement and the

> Revised Settlement Program is the use of negotiated disease

> definitions(710) to determine whether a woman qualified for

> compensation. The same definitions have been used to evaluate claims

> under both settlements. The compensable diseases included: (1)

> systemic sclerosis/scleroderma; (2) systemic lupus erythematosus; (3)

> mixed connective tissue disease (MCTD)/overlap syndrome; (4) poly and

> dermatomyositis; (5) primary Sjogren's syndrome; and (6) two forms of

> atypical disease -- one neurological and the other connective tissue.

> For the first five categories, the disease definitions employed

> recognized disease criteria, but allowed for atypical presentations or

> a "like disease." The atypical diseases tracked a similar approach to

> diagnosis but attempted to capture what clinicians were observing in

> the implanted population. In addition to meeting the disease

> definitions, in order to qualify for compensation a woman had to

> fulfill the disability criteria. Depending on the disease claimed, the

> disability criteria were based either on the severity of the disease

> (e.g. major organ involvement for SLE) or a functional capacity test

> (e.g. inability to perform vocation, avocation and self-care or the

> ability to do so but only with regular or recurring pain).

>

> For purposes of this discussion, it is important to understand that a

> woman could qualify in two ways. She had to present the claims office

> with either her complete medical records so as to permit the claims

> office to make an appropriate diagnosis and disability determination

> or she had to present a statement from a Qualified Medical Doctor --

> generally a physician holding a board certification in an appropriate

> area -- together with appropriate medical records so that a claim

> could be evaluated. Thus, this was not a case of self-reporting;

> instead, all claims had to be based either on the claims office's

> review of the records or on the certification of a board certified

> physician that a woman met the disease and disability criteria.(711)

>

> When the tremendous number of claims became obvious, and in an effort

> to understand the extent of claims submitted, the Court directed the

> claims office to perform a sample of the claims to find out how many

> claims would likely qualify for compensation and to estimate the

> extent of "ratcheting" that would be necessary. In performing the

> sample, the Court(712) estimated that a total of 248,500 claims filed

> by domestic registrants were postmarked by the September 16, 1994

> deadline for current claimants. That total included 191,400 claims

> received on or before September 16 and an additional 57,100 claims

> postmarked by that date. The Court drew a random sample of 2994 claims

> from the first 191,400. The sample included the claims of 1164 women

> who were attempting to qualify as current claimants -- women who

> claimed to be sick as of September 1994 -- and 1830 "future" claimants

> -- implanted women who registered to protect their right to make a

> claim in the event they became sick in the future.

>

> The 1164 current disease claims were reviewed by a team of specially

> trained nurses and were categorized as "approved," "minor deficiency,"

> "major deficiency," and "unknown as to extent of deficiency." It was

> assumed that the "minor" deficiencies were in fact so minor (a missing

> signature or omission of a key word in the diagnosis) that virtually

> all could and would be cured. Of the 1164 claims, 239 were approved as

> submitted, 610 had minor deficiencies, 306 had major deficiencies, and

> 9 fit in the unknown category. Combining the approved claims and

> claims with minor deficiencies gave a total of 849 "valid" claims.

>

> In order to properly evaluate the information that the Court provided,

> plaintiffs retained Dr. Shanna Swan to review the statistics and

> provide additional details.(713) Dr. Swan's extrapolation suggested

> that of the original 248,500 claims, 96,611 would be current

> claimants. Of the 96,611 current claimants, 70,467 would have valid

> claims (19,837 approved plus 50,630 minor deficiencies). The

> extrapolation is particularly interesting when applied to two

> categories of claims: systemic sclerosis and lupus.

>

> With respect to systemic sclerosis, the sample contained the claims of

> four women whose scleroderma claims were approved and one whose

> scleroderma claim had a minor deficiency. These were the only cases

> included for purposes of extrapolation and may therefore result in an

> underestimate since it does not account for any women who "opted out"

> (elected not to participate) or any women who failed to register.(714)

> Dr. Swan estimated the number of scleroderma cases in the total

> population as follows:

>

> p = Proportion of valid claims with SS = 5/2,679(715) = 0.19%. Dr.

> Swan then subtracted the projected number of claims with "unknown"

> outcome to get the projected number of valid claims:

>

> N = Projected number of valid claims = 248,500 - (25,398 + 746)(716) =

> 222,346. She then calculated the standard deviation:

>

> SD = [(p x (1-p))/N]1/2 = 0.0008. Thus, the estimate p +/- SD is given

> by 0.19% +/- 0.08%. She then multiplied p x N to get the expected

> number of scleroderma cases among all claims (with known outcome):

>

> Total SS = N x p = 0.19% x 222,356 = 415.(717) Thus, the anticipated

> 70,467 valid claims would contain 415 claims for scleroderma.

>

> That number, 415 scleroderma cases, far exceeds the number anticipated

> in the population. Based on the information that she provided in her

> work, Dr. Swan estimated that about 75 new cases of scleroderma should

> have been diagnosed in implanted women under the null hypothesis of no

> increased risk.(718) Assuming that 75 cases are expected in the

> population, a total of 415 cases in all implanted women represents a

> relative risk for scleroderma of 5.5. Using similar calculations for

> the systemic lupus erythematosus claims, Dr. Swan calculated a

> relative risk of between 12.2 and 24.3. Thus, based on the claims

> submitted pursuant to strict criteria that the manufacturers had

> agreed to, there is a dramatic increase in the incidence of defined

> disease in the implanted population.

>

> C. PUBLISHED STUDIES ARE CONSISTENT WITH INCREASED RATE OF DISEASE AND

> SYMPTOMS IN THE IMPLANTED POPULATION

> In 1984, Kumagai, et al., published an article(719) which presented

> the clinical findings of 18 patients and a review of 28 additional

> cases that had been reported in the Japanese literature. The authors

> placed the patients in two major groups -- one group of patients

> (N=24) with definite connective tissue disease(720)while the other

> group (N=22) had what had been termed "human adjuvant disease." This

> was defined as some symptoms and signs of disease that were suggestive

> of, but not diagnostic for, connective tissue disease. They found that

> the incidence of scleroderma in their population was three times that

> expected to occur naturally and they noted that their scleroderma

> patients were not "entirely typical." Instead, they had lower

> incidences of antinuclear antibodies and esophageal hypomotility than

> expected. Those findings, however, did not dissuade the authors from

> the correctness of their conclusions since virtually all

> occupationally or environmentally caused cases of scleroderma had some

> unusual presentation. The importance of the paper, then, is that the

> authors noted the atypical symptoms in an implanted population and it

> remains highly relevant today in assessing the question of an

> association between implants and disease.

>

> If the disease process that women with implants undergo includes

> primarily an atypical one, then it does no good to conduct

> epidemiology studies looking for classical presentations of connective

> tissue diseases. Instead, it is of greater benefit to examine symptoms

> and symptom complexes. While the manufacturers persist in saying the

> controlled studies "prove" there is no causation, examination of the

> studies reveals increased disease symptoms in the implanted

> population. For example, the Giltay paper(721)studied 287 women who

> received silicone breast prostheses between 1978 and 1990. They were

> age matched with women who had aesthetic surgery in the same year.

> With the exception of pleuritis, the number of "cases" with symptoms

> exceeded the number of "controls" with symptoms. Their results appear

> in the following table:

>

> Symptoms Controls (N=210) Cases

> (N=235)

> Odds

> Ratio

> Confidence

> Interval

>

> Painful joints > 3 months 18 46 2.6 1.45 - 4.64

> Joint swelling > 1 week 10 14 1.27 .005 - 2.92

> Regularly burning eyes 15 37 2.43 1.29 - 4.57

> Oral ulcers > 3 weeks 2 4 1.80 .33 - 9.93

> Raynaud's 7 12 1.56 .60 - 4.04

> Pleuritis 5 4 .71 .18 - 2.68

> Proteinuria 4 8 1.81 .54 - 6.11

> Skin abnormalities 4 20 5.05 1.71 - 14.97

>

> The same results can be seen in the paper.(722) Dr.

> and her colleagues compared 749 implanted women with 1798 controls who

> were age matched and who had undergone a medical evaluation within two

> years of the date of breast implantation in the controls. Their

> results, in instances in which the symptoms in the implant cases

> exceeded those in the controls, are presented in the following table:

>

> Symptom Implant for any reason (N=749) No implant (N=1498)

> Any arthritis(723) 25 (times 2 = 50) 39 (divided by 2 = 19.5)

> Sicca(724) 33 (times 2 = 66) 50 (divided by 2 = 25)

> Morning stiffness 30 (times 2 = 60) 35 (divided by 2 = 17.5)

> Serositis(725) 18 (times 2 = 36) 21 (divided by 2 = 10.5)

>

> Thus, the women with implants experience symptoms at a higher rate

> than women without implants.

>

> While discussed in the prior section on "Immunology," it is worth

> highlighting again the biomarker studies because of their importance

> to this issue. The findings are compelling. With the publication of

> the first paper, in 1992, that looked at the serology of women with

> implants(726) and continuing through the more recent work, there is

> increasing evidence that women with breast implants have elevated

> autoantibodies. The Press paper is particularly significant because of

> its finding that women with more defined diseases had higher titer

> antinuclear autoantibodies than did women with atypical presentations,

> though both groups had elevated levels.

>

> The results that Press found are consistent with those from Dr.

> Claman, whose study was funded by manufacturers.(727) He summarized

> his findings in the Immunology of Silicones and reported that he had

> found positive high titer ANA's in 27% of 75 apparently healthy women

> with implants compared to 3% of controls. Dr. Claman concluded, ". . .

> the fact that the findings were similar [for two groups of exposed

> women whose sera were analyzed by different laboratories], namely that

> implanted women had a significantly higher prevalence of positive ANA

> tests, lends credence to the concept that this is a meaningful and not

> a chance result."

>

> Most recently, an independent group of researchers from Vienna

> confirmed Dr. Claman's results.(728) They determined that 33% of the

> implanted women, including women who considered themselves "healthy,"

> had positive ANA's with a titer of at least 1:80. Only 8% of the

> controls had similar findings and the difference between the two

> groups reached a level of statistical significance. The results led

> the authors to conclude, ". . . elevated ANA levels and the occurrence

> of autoantibodies detected by immunoblotting suggest considerable

> autoimmune reactions in these patients."

>

> The findings of all of these authors, reporting on the results of

> controlled studies in which implanted women were compared to

> unimplanted women, are very significant to this Panel's deliberations

> and demonstrate, again, that exposure to silicones has led to immune

> system reactions.

>

> Also of particular interest is a 1995 abstract from Kayler and

> Goodman.(729) They apparently synthesized the epidemiology literature

> using seven then-existing studies which met their inclusion criteria

> for meta-analysis. They concluded as follows:

>

> The pooled odds ratio for NON-CTD arthralgia following breast implant

> augmentation was significantly increased (OR-1.74, 95% CI, 1.19 to

> 2.55), corresponding to a cohort-based relative risk of 2.10 (95% CI,

> 1.48 to 2.91). From a clinical perspective, the probability that a

> breast implant recipient's nonspecific rheumatic complaints are

> attributable to the implants is approximately 52% (95% CI, 32% to

> 66%). Although no significant association was found between silicone

> breast implants and ANY CTD (OR-0.68, 95% CI, 0.90 to 1.36) or CTD-SCL

> (OR-0.49, 95% CI, 0.11 to 1.43).

>

> Meta-analysis suggests that breast implant recipients are at

> substantially increased risk for arthralgia without other findings of

> connective tissue disorders. Given the strength and robustness of this

> association, consistent with both basic science and pooled

> explantation literature, it seems reasonable to accept a causal

> relationship and consider removal of silicone breast implants in women

> experiencing arthralgia.

>

> Thus, there is epidemiologic evidence of an association.

>

> The recent Friis paper lends further support to the evidence on

> association.(730) While the paper has serious deficiencies, the paper

> nonetheless found more than a doubling of the risk for muscular

> rheumatism, fibrositis, and myalgia in the cohort of women who had

> received implants. The fact that similar increases were detected in

> the cohort that had undergone breast surgeries with implants may be

> attributable to the deficiencies in the study, deficiencies including

> the fact that the entire study was based only on hospital discharge

> records.

>

> Two other recent works that examined non-breast prosthetic devices and

> occupational silicone exposure demonstrated an increased risk of

> disease in the exposed population. One of those, a 1996 abstract by

> Laing, et al.(731) demonstrated that the odds ratio was elevated for

> an association between breast implants and UCTD. (OR 2.27; 95% CI,

> 0.67-7.71). For all silicone containing implants (including breast

> implants) there was a statistically significant association. (OR 2.98,

> 95% CI, 1.45-6.13).

>

> The findings of Greenland and Finkle(732) are consistent. They

> examined the possibility of an association between non-breast

> prosthetic implants and selected malignant neoplasms, connective

> tissue disorders, and neurological diseases. Their data showed "clear

> associations of prosthetic implants and subsequent diagnosis of

> certain arthritic and neurologic disorders."

>

> The silica epidemiology literature is also revealing. The studies on

> long term disease endpoints in workers exposed to silica contain

> remarkable parallels to the silicone epidemiology, from biomarker

> studies of immune activation to connective tissue diseases,

> hematopoietic cancers, and other consequences of chronic immune

> stimulation and silicic acid overload.

>

> First, Dr. Virginia Steen's doubts about the strength of the

> association between silica and scleroderma should be resolved by two

> new studies from Germany. Looking at all mineworkers in East Germany

> from 1978 to 1991 (1995 in the uranium group), divided into

> non-overlapping groups, and controlling for silica exposure levels,

> epidemiologists from the German Federal Institute of Occupational

> Health reported:

>

> [t]he results of both surveys show a strong association between

> scleroderma and exposure to dust with a high quartz content,

> especially for men with silicosis but also for non-silicotic men.(733)

>

>

> Overall, combining all data from both studies, the relative risk was

> 10.4. It was 23 for silicotics and 6 for non-silicotics, with the

> lower limit of the confidence interval at 3.0.

>

> The Steenland, Goldsmith 1995 review article(734) set out clearly the

> studies on scleroderma and other connective tissue diseases through

> 1994. In addition to the new East German mineworker study, the most

> important studies of disease not covered in the Steenland, Goldsmith

> review are the scouring powder plant worker study from Spain(735) and

> the German study of atypical and classic lupus in male uranium

> miners.(736) Of the Spanish workers, 88% were women and they developed

> more scleroderma than lupus and also evidenced a variety of elevated

> autoimmune and immune perturbance biomarkers. Most importantly,

> however, a huge portion of them developed atypical connective tissue

> disease, classified by the authors as UCTD. The relevance of these

> observations to the data recently reported by Schottenfeld on UCTD in

> women with silicone implants is obvious.

>

> The German mine workers were all men, but still an atypical disease

> pattern emerged, albeit not as strongly as in the Spanish women. Heavy

> dust exposure was significantly associated with increases in both

> classic and atypical lupus, and the same autoimmune and immune

> perturbations were seen as well.

>

> The Steenland, Goldsmith review cites other studies finding

> significant associations between occupational silica dust exposure and

> rheumatoid arthritis, lupus, "musculo-skeletal disease," and renal

> disease. These are all mortality studies, without record review,

> without clinical data pre-death and, without blood or tissue samples.

> Obviously, there could well have been a much greater incidence of

> atypical disease and symptoms that went uncatalogued and thus

> undetected.

>

> One other important and relevant disease end-point in the silica

> epidemiology literature is hematopoietic cancers: lymphomas and

> myelomas, and other immune system cancers. Drs. Brinton and Brown

> acknowledged this month that whether silicone causes such cancers is

> an open question of "heightened concern" to many scientists and

> physicians, with substantial evidence pointing to the likelihood of at

> least some exacerbation or acceleration of the stepwise loss of

> control of a clone of lymphocytes or their blood marrow precursors in

> susceptible animals and humans.

>

> Again, for the women, the silica literature is not encouraging. Among

> the details of the Steenland and Brown mortality study of South Dakota

> gold miners(737) are data supporting this conclusion in the paper:

> "[m]ultiple cause analysis also showed a significant excess of

> diseases of the blood and blood-forming organs." Thus, despite the

> manufacturers' effort to claim the epidemiology "proves" their case,

> controlled studies show a relationship between exposure and symptoms.

>

>

>

>

>

> D. THE "EPIDEMIOLOGY" STUDIES ON WHICH THE MANUFACTURERS RELY SHARE

> COMMON LIMITATIONS THAT UNDERMINE THEIR RELIABILITY

> It is presumptuous, at best, for lawyers to tell this panel about

> epidemiology. Acknowledging, however, the risk, plaintiffs do feel

> that there are certain terms and concepts that may be of use to the

> panel in evaluating the epidemiology issues. Among these are the

> concepts of risk, p values, and the use of one-sided vs. two-sided

> tests.

>

> 1. The Choice of Statistical Methods Influenced The Outcome Of A

> Study.

> To begin, when a researcher undertakes to design a protocol for an

> epidemiology study, the researcher makes a number of decisions at the

> outset -- decisions that will impact the results. Obvious among these

> is the decision as to what the study will examine. For example, in the

> context of breast implants, the researcher will formulate the "null

> hypothesis." That would be something like the following: breast

> implants do not have any effect on the development of either symptoms

> or disease in women. The second likely decision will be the form of

> the study. That is, is it a case-control study; is it a cohort study;

> is it a cross-sectional survey? A researcher having made that

> decision, may next turn to the statistical issues.

>

> Among these statistical issues is the question of the appropriate

> confidence interval and whether the data will be tested using a

> one-sided test or a two-sided test. While there are studies in a

> variety of areas that use 80, 85, 90, or even 99 % confidence

> intervals, the interval seen most in the implant literature is 95 %.

> That means that 95 out of 100 times that the study is performed, the

> results will be the same. The decision on one-sided vs. two-sided is

> an important one because it is that decision that dictates what the

> study will examine. For example, if the null hypothesis is that breast

> implants do not affect the development of either symptoms or disease

> in implanted women, a two-sided test will examine two propositions.

> The first is that breast implants increase the risk of symptoms or

> disease in implanted women and the second is that breast implants

> prevent women from developing symptoms or disease. A one-sided test of

> the same null hypothesis would only test the proposition that breast

> implants increase the risk of disease.

>

> The one-sided/two-sided decision is also important because it dictates

> how the risk of error is spread. As the diagram on the following page

> demonstrates, if a researcher selects a two-sided test, then the risk

> is split between the two alternative hypotheses. On the other hand,

> with a one-sided test, all of the risk is on one side, as it should be

> when no one has hypothesized a prophylactic effect.

>

> With the exception of the Strom paper(738), each of the breast implant

> controlled epidemiology papers, to the extent it can be determined on

> the face of the paper, employed a two-sided analysis. The authors of

> the Strom paper used a 90% confidence interval and explained their

> decision as follows: "[w]e used 90% confidence intervals, equivalent

> to a one-tailed statistical test, because it is inconceivable that

> breast implants could protect against SLE." As one of the other breast

> implant researchers explained, "[a] one sided test will put all the

> power of the statistical test toward detecting a risk. A two-sided

> test splits the power between the two (detecting a risk and detecting

> a protective affect)."(739) One must, however, wonder why the

> two-tailed approach was used since the "choice of a two-tailed test

> indicates that there is equal concern [with both an increase and a

> decrease in risk]."(740) As one group of authors has explained in the

> asbestos setting, a setting analogous to this one, "f we are

> looking at the health of workers in an asbestos factory, we are

> interested in whether or not their health is worse than average. No

> one has suggested that working in such a factory can actually be good

> for people, so we need not consider the possibility that these workers

> lose fewer days from work than all other factory workers . . . we are

> using a one-tailed test."(741)

>

> Quite appropriately, virtually every researcher who submitted a

> proposal to the manufacturers for funding proposed the use of a

> one-sided test. The manufacturers, however, consistently rejected that

> approach and demanded the use of a two-sided test. For example, in

> 1992, Dr. Sherine submitted a protocol to the Plastic Surgery

> Education Foundation(742) for what ultimately became the so-called

> "Mayo Clinic Study."(743) In her protocol, at page 8, she discussed

> the statistical analysis that she proposed which specified the use of

> a one-sided test. When the paper was published, and without

> explanation, the data were evaluated using a two-sided test.(744)

>

> Similarly, when Helen Englert, the author of the Australian

> scleroderma studies, questioned the power of her work to detect an

> increased risk of scleroderma, she wrote to Ralph Cook, the chief Dow

> Corning epidemiologist:

>

> Ralph, I was partly expecting you to have called all the project off

> after I gave you the power calculations, because I imagine Dow Corning

> would like an earlier rather than later answer to this dilemma. I

> suppose I could have played around with the relative risks as well to

> see how low this could get while still maintaining a power of 80-90%.

> You will notice I used a two-tailed analysis in my power calculations,

> two-tailed as a worse case scenario. However, I think it not

> unreasonable that a one-tailed test should be used because we're not

> really interested in the question does augmentation mammoplasty

> protect against the development of scleroderma, only that it increases

> it. Would you as an epidemiologist agree or disagree? I'd be

> interested in your opinion. If we and other independent folk agreed

> then this would bump up the power of this Sydney study a little

> higher.(745)

>

> Dr. Cook did not, however, agree. He responded that he, personally,

> did not like one-tailed tests of significance.(746)

>

> In fact, even the Hennekens paper(747) was proposed as a one-sided

> test, though when published, the analysis was two-sided. Specifically,

> the protocol, submitted to Dow Corning for review in June, 1992,

> stated, "[g]iven the estimates presented above, and using a one-sided

> test with alpha = 0.05, as detailed in Table 1 we will be able to

> detect the following with 80% power:"(748)When published, the paper

> contained the following: "[f]or each RR, a 95% confidence interval

> (CI) was calculated. All P values cited are two-sided."(749)

>

> Through discovery, Plaintiffs have obtained preliminary

> information(750) about some studies, the results of which are not yet

> published. For example, in response to a proposal from B. Teel

> to conduct a "Breast Implant and Rheumatic Disease" study, Dow Corning

> epidemiologist Cook wrote (criticizing the researcher's proposal for a

> one-sided alpha):

>

> n view of the possibility that disease may be either greater or

> less than expected, I would like to see this approached as a

> two-tailed phenomenon and the results reported out as point estimates

> and 95% confidence intervals to the extent possible. (751)

>

> The importance of the change from a one-sided test to a two-sided test

> is significant. In evaluating Table 1 in the Hennekens study, for

> example, the published results for Sjogren's Syndrome, poly and

> dermatomyositis, and scleroderma if recalculated to a one-sided

> P-value are as follows:

>

> Disease Relative Risk P value (two-sided) P value (one-sided)

> Sjogren's syndrome 1.49 .067 .0335

> Dermato/polymyositis 1.52 .068 .034

> Scleroderma 1.84 .060 .030

>

> Thus, when using a one-sided P-value, all of the above diseases reach

> a level of statistical significance at the 95% confidence interval

> level. And, if that were done, then, as a matter of mathematical

> proof, all of the confidence intervals would be greater than 1.0.

> Thus, the manufacturers' early decisions as to how studies would be

> conducted -- decisions apparently contrary to those that the

> researchers themselves initially proposed -- had a significant impact

> on the outcome of the studies.(752)

>

> 2. The Serious Design Limitations In The Epidemiology Studies

> Conducted To Date Undermine Their Reliability.

> Leaving aside for the moment the question of whether any of the

> controlled studies relied upon by the manufacturers were looking for

> the right disease or symptoms, each of the controlled epidemiology

> studies suffers limitations in design so serious as to undermine the

> value of the study in assessing whether there is an increased risk of

> disease from exposure to silicone breast implants. Perhaps the most

> concise criticisms are those found in the Silverman, Kessler review

> article.(753) While acknowledging that the studies "vary widely in

> quality," they listed certain common problems:

>

> 1. sample sizes inadequate to rule out rate outcomes;

>

> 2. study methods inappropriate for detecting atypical syndromes;

>

> 3. poor choice of comparison group; and

>

> 4. inadequate duration of follow-up or information-gathering

> techniques that may have biased the detection of implants or clinical

> outcomes.

>

> In addition to those weaknesses, there are other weaknesses in the

> controlled epidemiology studies that run across all of the studies.

> These include the lack of power to detect certain rare conditions,

> failure to use appropriate controls, small sample size, and the

> potential for bias and misclassification. The manufacturers and their

> consultants recognized these design issues, yet they funded the

> studies, apparently manipulated the results of some of them, and now

> zealously overstate their importance.

>

> As an example of the criticisms made of the studies in advance of

> funding decisions were criticisms made by Dr. Philip Cole, a

> well-known epidemiologist who served, and continues to serve, as a

> consultant to Dow Corning. When asked to review the Hochberg

> protocol(754), Dr. Cole said, "The major strengths of this proposal

> are: (1) The commitment of the investigator to the area of the

> rheumatic diseases. This, combined with an interest in epidemiology

> make him a suitable person for a study such as this. However, Dr.

> Hochberg is not an experienced research epidemiologist and his

> proposal and questionnaire evidence this. This study could profit

> immensely from sophisticated epidemiologic input."(755) His criticism

> continued and called the proposal "severely underdeveloped, perhaps

> even shoddy."(756)

>

> Dr. Cole's criticisms were not solely addressed to the Hochberg

> proposal. He also criticized the Michigan scleroderma proposal, "In my

> view, the study is fundamentally overblown."(757) He continued: "I am

> concerned by the large contributions to be made by junior members of

> the investigative team. Neither Dr. Gillespie nor Dr. Laing has any

> experience in epidemiologic research, and Dr. and Ms. Gentry

> (now Dr. Carol Burns) are still in training."(758) Dr. Cole, however,

> was not alone in his criticisms. Some of the authors were particularly

> self-critical, most often when describing the "power" of their

> studies.

>

> As the Panel knows, the "power" of a study is its ability to determine

> an increased risk of disease. The larger the number of subjects and

> the more prevalent the disease in the general population, the greater

> the power of the study. By their own admission, many of the studies

> lack the power to detect an increased risk of disease. For example,

> the authors of the Weisman paper(759) acknowledged that their sample

> size was "clearly inadequate" to demonstrate the absence of an

> association between implants and systemic sclerosis since systemic

> sclerosis is an "uncommonly encountered disorder."

>

> Likewise, the Strom paper(760) examined 195 lupus patients and

> compared them to age and sex-matched controls. While the authors claim

> their study was based on a "very large" group, that is not the case

> and, necessarily, limits the study's ability to detect an increased

> incidence of disease.

>

> The collaborators in the paper(761) were similarly candid in

> this regard. In their discussion of their overall results, they

> acknowledged that the answer to the question of a "cause and effect"

> relation between implants and disease and symptoms "would require a

> well-conducted prospective cohort study using a multicenter controlled

> design with follow-up of a large population of women for at least 10

> years." As to their findings with respect to scleroderma, they wrote:

>

> . . . we had limited power to detect an increased risk of rare

> connective tissue diseases, such as systemic sclerosis. Indeed, we

> calculated that it would require a sample of 62,000 women with

> implants and 124,000 women without implants, followed for an average

> of 10 years each, for a doubling of the relative risk of this

> condition to be detected.(762)

>

> They concluded:

>

> [o]ur results, therefore, cannot be considered definitive proof of the

> absence of an association between breast implants and connective

> tissue disease. Other controlled studies of this question are

> on-going.(763)

>

> By their own assessment, then, the authors recognized the limitations

> in their work.

>

> and her co-authors were similarly critical of some of the

> prior studies. They commented that, while other studies had looked for

> association, the studies were "limited by substantial methodological

> flaws." These flaws included:

>

> . . . uncontrolled, retrospective designs, lack of objective

> validation of the survey replies, very small samples, inadequate

> statistical power, large losses to follow-up, and the use of referral

> based cohorts.

>

> The Hennekens group also noted problems with other studies:

>

> . . . even the large cohort studies have included relatively few women

> with breast implants (maximum number 1183 women), especially those

> having had implants for longer duration . . . .

>

> The power of the -Guerrero paper is similarly limited, though

> the authors themselves do not raise that issue. In fact, of the 1,183

> women with implants, only 32 received the screening questionnaire on

> connective tissue disease. And, in the proposal submitted to secure

> funding for the study, one of the authors said the study was only

> sufficient to detect of relative risk of 3.0 or more.(764)

>

> Tied to the "power" issue is the question of latency. The studies to

> date have lacked the ability to detect symptoms because they have

> failed to address the question of latency. Alternatively, the studies,

> when examined closely, demonstrate either a positive correlation when

> latency is addressed or acknowledge that it is an issue that is

> unresolved.

>

> One of the papers on which the manufacturers rely is the Schusterman

> study.(765) In that study, the authors made broad pronouncements about

> the inability of their study to detect any increased risk of disease.

> But the follow-up was only from two to six years following

> implantation.

>

> When latency has at least been considered, as in the Wells paper,(766)

> for example, the authors noted that women who had their implants in

> longer were more likely to experience symptoms. They wrote:

>

> We conclude that our data show the expected pattern of higher

> frequency of symptoms among those operated on early, although

> limitations in sample size make it difficult to find statistically

> significant relationships.(767)

>

> Thus, these researchers at least recognize the significance of a

> latency consideration.

>

> Despite having included at least one woman who had only been implanted

> for thirty days, the Hennekens paper(768) did examine duration of

> implant and, while the author claims to have seen no "consistent

> pattern" based on duration, their data, when the P-values are

> re-computed to reflect a one-sided analysis, shows statistically

> significant increases in connective tissue disease, rheumatoid

> arthritis and Sjogren's Syndrome in women implanted for more than 10

> years. In all but one instance (lupus), the relative risks exceed one

> and, as reflected in the following table, approach or exceed a 95%

> degree of certainty in three instances.

>

> DURATION OF IMPLANTS 10 YEARS

>

> DISEASE EXPOSED CASES RELATIVE RISK P (2-sided) P (1-sided)

> Any Connective Tissue 78 1.23 .072 .036

> Rheumatoid Arthritis 40 1.32 .084 .042

> Sjogren's Syndrome 11 1.95 .028 .014

>

> It is important to remember that the manufacturers experienced the

> greatest number of sales in the mid to late 1980s. As one

> manufacturer's sales figures demonstrate, of the approximately 778,000

> gel implants that the company sold between 1962 and 1992, fully 27%

> were sold in the years 1984 - 1989.(769) The other manufacturers' data

> are similar.

>

> The importance of latency is two-fold. First, none of the studies

> provides information on the distribution of women based on duration of

> implant. Second, the time between implant and data gathering has been

> insufficient to allow for the development of symptoms in the more

> recently exposed population.

>

> Despite these acknowledged flaws, the manufacturers continue to rest

> their entire defense on the strength of the controlled studies. In

> doing so, however, they consistently misinterpret the results.

>

> For example, in responding to Dr. Greenland's presentation to the

> Panel, the manufacturer's witness, Dr. Ory, commented with respect to

> the Hochberg study that "the odds are seven to one that the relative

> risk is 1.1 rather than 2.23."(770) That statement is incorrect for

> two reasons - one statistical, the other epidemiologic. Dr. Ory's

> statistical error lies in his misinterpretation of the results as

> referring to probabilities or odds of relative risks. In fact, the

> maximum likelihood estimate and the accompanying confidence limits are

> based on probabilities of data given relative risk values, not on

> probabilities of relative risks given the data. While in ordinary

> English, "likelihood" and "probability" mean the same thing, in

> statistics they are very different. In statistics, the "likelihood of

> relative risk given the data" equals the "probability of the data

> given the relative risk." In this technical sense, "likelihood" is the

> reverse of "probability." Likelihood ratios may be far from the

> probability ratios they are confused with because such probability

> ratios depend on prior probabilities and bias probabilities that are

> not accounted for by standard computations.

>

> The epidemiologic error that Dr. Ory makes is his assumption that the

> statistics apply only under the additional assumption that the study

> was methodologically perfect. In particular, the statistics assume

> that people were randomly allocated to have or not have implants, were

> randomly sampled for study, and were measured for time of implant and

> disease outcome with no errors. One or more of these assumptions is

> patently absurd for all the studies.

>

> The chart that follows is taken from Table 2 in the Silverman, Kessler

> paper(771) and lists some of the more fundamental design flaws in the

> various controlled studies.

>

> Study, Year (Reference)

> Design/Data Collection Study Group Comparison Group Design Issues

> Goldman, et al.,

> 1995 (68)

>

>

>

>

>

> [Record No. 6164]

> Cross-sectional/

> retrospective record review

> Rheumatology/primary care practice - 721 patients with defined

> rheumatoid arthritis or other connective tissue disease

> Rheumatology/primary care practice - 3508 patients without defined

> rheumatoid arthritis or connective tissue disease Potential for

> incomplete ascertainment of implant status and clinical information

>

>

> "Control" population likely to include symptomatic patients

>

>

>

>

>

> Did not consider atypical syndromes

>

>

>

>

>

> 85% of implants consisted of silicone gel; 11%, unknown; 4%, saline

>

> Weisman, et al., 1988 (66)

>

>

> [Record No. 1175]

> Cohort/mailed questionnaire survey 378 silicone breast implant

> recipients (response rate, 33%) None No comparison groups studied

>

>

> Poor response rate

>

>

>

>

>

> Fibromyalgia and atypical syndromes not considered outcomes of

> interest

>

>

>

>

>

> Sample size inadequate to detect rare outcomes

>

> Wells, et al., 1994 (69)

>

>

> [Record No. 1752]

> Cohort/mailed questionnaire survey 826 silicone breast implant

> recipients (response rate, 59% 310 other patients having cosmetic

> surgery (response rate, 46%) Poor response rate

>

>

> Exposed/non-exposed groups differ in age and time since surgery

>

>

>

>

>

> Sample size inadequate to detect rare outcomes

>

> Giltay, et al., 1994 (70)

>

>

> [Record No. 1731]

> Cohort/mailed questionnaire survey 287 silicone breast implant

> recipients (reported response rate, 82%; actual response rate, 63%)

> 287 patients having cosmetic surgery without silicone ex-posure

> matched by age and year of surgery (reported response rate, 73%;

> actual response rate, 56%) Sample size inadequate to detect rare

> outcomes

>

>

> Poor response rate

>

> Schusterman, et al., 1993 (71)

>

>

> [Record No. 1691]

> Cohort/

> medical record review and telephone follow-up

> 250 patients receiving silicone breast implants after mastectomy (2%

> lost to follow-up) 353 patients having autogenous reconstruction after

> mastectomy (3% lost to follow-up) Sample size and follow-up inadequate

> to detect rare outcomes with long latencies

> , et al., 1994 (72)

>

>

> [Record No. 1765]

> Cohort/medi-

> cal record database review (Mayo study)

> 749 Olmsted County, MN, residents with breast implants (78% silicone,

> 5% saline, 10% polyurethane, 7% silicone and saline) 1498 Olmsted

> County, MN, residents who did not receive implants Sample size

> inadequate to detect rare outcomes

>

>

> Would not detect atypical syndromes or symptoms complexes not recorded

> in medical record

>

>

>

>

>

> Patients seeking care elsewhere could be lost to follow-up

>

> -Guerrero, et al., 1995 (73)

>

>

> [Record No. 1819]

> Cohort/mailed questionnaire (Nurses' Health Study) 1883 study

> participants with breast implants (876 silicone gel, 170 saline, 67

> double lumen, 14 polyurethane, 56 unknown) 86,318 study participants

> without implants History of connective tissue disease symptoms sought

> onlyin participants reporting diagnosed rheumatologic,

> musculoskeletal, or connective tissue disease (6% of study group)

> Dugowson, et al. 1992 (75)

>

>

> [Record No. 2919]

> Case-control/

> questionnaire survey

> 349 patients with rheumatoid arthritis (response rate, 86%) 1,456

> "similarly aged women" Brief abstract

>

>

> Response rate for controls not specified

>

> Strom, et al., 1994 (76)

>

>

> [Record No. 0654]

> Case-control/

> telephone survey

> 195 patients with systemic lupus erythematosus from Philadelphia

> (response rate, 76%) 143 age- and sex-matched controls (response rate,

> 78%) Second control group from another study used for statistical

> tests

>

>

> Small sample size results in low power

>

> Hochberg, et al., 1994 (78)

>

>

> [Record No. 3246]

> Case-control/

> telephone interviews and self-administered questionnaires

> 860 patients with scleroderma at several centers (669 with

> "informative data") 2,061 age- and race-matched controls Brief

> abstract

>

>

> Type of implant not specified

>

>

>

>

>

> Response rate for controls not specified

>

>

>

>

>

> Case-patients and controls queried differently about exposure

>

> Englert and , 1994 (80)

>

>

> [Record No. 1745]

> Case-control/

> telephone interviews

> 315 patients with scleroderma from Sydney, Australia (251 women) 371

> patients in Sydney general practice (289 women) Implant history

> ascertained differently for case-patients and controls

>

>

> Type of implant not specified

>

>

>

>

>

> Initial size of selected case - control groups not specified

>

> Burns, C.G. (79)

>

>

> [Record No. 1727]

> Case-control/

> questionnaire survey

> 461 patients from Michigan with probable or definite scleroderma

> (response rate, 82%; 59% considered in analysis) 1,184 community

> controls Controls had greater likelihood of exposure to implants

>

>

> Some case-patients excluded from analysis without exclusion of

> corresponding controls

>

>

>

>

>

>

>

>

> a. As An Example, a Detailed Review of the -Guerrero Paper

> Reveals Fundamental Errors That Render The Conclusions Unreliable

> While the need for brevity suggests that it is probably not wise to

> provide a detailed criticism of each of the studies, some are so often

> cited as demonstrating the lack of a causative relationship that a

> closer examination is warranted. One such study is that performed by

> -Guerrero and his colleagues.(772) The study, published in

> June, 1995 in The New England Journal of Medicine, was the focus of

> considerable media attention and has been hailed in courtrooms as

> supporting the manufacturers' theme that breast implants cannot cause

> disease. The study examined women in the Nurse's Health Study cohort,

> which was assembled in 1976.

>

> By way of background, the cohort, when started, consisted of

> approximately 120,000 married registered nurses who lived in

> California, Connecticut, Florida, land, Massachusetts, Michigan,

> New Jersey, New York, Ohio, Pennsylvania, and Texas. Their age at the

> beginning of the cohort ranged from 30 to 55. Every two years, the

> women completed a questionnaire which sought information on a variety

> of health conditions. After the FDA hearings on silicone breast

> implants in late 1991 and early 1992, the investigators added a breast

> implant component to the study so that, in 1992, the questionnaire

> asked for the first time whether the participants had ever had breast

> implant surgery or silicone, paraffin, or collagen injections. 88,153

> participants responded to the 1992 questionnaire and, after a

> supplemental questionnaire, the investigators determined that 1,183

> women had received implants (876 silicone gel; 170 saline; balance

> type unknown).(73)

>

> In August, 1992, the investigators mailed a separate screening

> questionnaire on connective tissue disease to only those participants

> who had been diagnosed with, and had reported on prior questionnaires,

> a classically defined connective tissue disease diagnosed before June

> 1, 1990. Excluded from the mailing were women who had been diagnosed

> with and reported fibromyalgia. In fact, only those women who, on

> questionnaires between 1980 and 1990, reported that they had been

> diagnosed with systemic lupus erythematosus or rheumatoid arthritis or

> who had written in a diagnosis of a connective tissue disease under

> the space "other major illness" were sent the screening

> questionnaires. Thus, despite the fact that the 1992 questionnaire

> asked about a physician's diagnosis of scleroderma, polymyositis,

> dermatomyositis, and Sjogren's syndrome, a woman answering that she

> had been diagnosed with one of those conditions would not have

> received the screening questionnaire unless she had also reported that

> diagnosis prior to June 1, 1990. While 5,024 women responded with a

> self-report of connective tissue disease, investigators were only able

> to confirm that 516 had a connective tissue disease and only three of

> them had implants.(774)

>

> As the Silverman Kessler(775) review article summarized:

>

> However, the investigators sought information on signs and symptoms

> suggestive of connective tissue disease only in the group reporting a

> diagnosis of a defined connective tissue disease and not in the entire

> cohort or a random sample thereof. The 94% of study participants who

> were not given the opportunity to complete a 'screening questionnaire'

> were considered, for the purposes of the data analysis, to have had no

> self-reported symptoms, an assumption that probably resulted in an

> underestimate of the prevalence of such symptoms. Therefore, although

> the study provides additional evidence against a markedly increased

> risk for certain connective tissue disorders in implant recipients, it

> could not address whether symptoms of atypical connective tissue

> disorder may occur more frequently in such recipients.

>

> Thus, the study's power was quite limited.

>

> b. The Paper Contains Similar Limitations

> As discussed earlier, the authors of the paper(776)

> acknowledged the power limitations of their study. In addition, there

> are other design limitations. These include the fact that the

> diagnoses were made by staff who were not blinded to exposure status.

> The design also allowed for the potential misclassification because a

> woman implanted before moving to Olmstead County might have been

> missed. And, since there were no interviews, the study was unable to

> exclude cases with known etiologies and did not control for

> confounding.

>

> Also curious is the fact that the final paper failed to address

> changes that took place between the development of the protocol, the

> presentation of the abstract, and the publication of the final study.

> As the following table depicts, numerous, but unexplained, changes

> occurred:

>

> THE MAYO CLINIC STUDY;

>

> COMPARISON OF PROTOCOL, ABSTRACT AND PUBLICATION

>

> CONNECTIVE TISSUE DISEASES SELECTED FOR INCLUSION

>

> PROTOCOL

> (3/92)

> ABSTRACT

> (11/93)

> PUBLICATION

> (6/94)

>

> Systemic sclerosis (SSc) Yes Yes Yes

> Systemic lupus (SLE) Yes Yes Yes

> Rheumatoid arthritis (RA) Yes Yes Yes

> Sjogren's syndrome (SS) Yes Yes Yes

> Myositis (M) Yes Yes Yes

> Ankylosing spondylitis No No Yes

> Psoriatic arthritis No No Yes

> Polychondritis No No Yes

> Systemic vasculitis No No Yes

> Polymyalgia rheumatica No No Yes

> Keratoconjunctivitis sicca Yes No No (?)*

> Overlap syndromes(s) Yes No No

>

> * Listed as outcome but not included among Connective Tissue Disease

>

> c. Other Design Limitations Cast Doubt on the Reliability of the Data

> Despite the fact that the literature has long discussed the atypical

> presentation of disease and symptoms in women with breast implants,

> not one of the manufacturer funded studies has examined that question.

> Instead, each has focused solely on classical diseases. Moreover,

> despite increasing evidence that the rupture rate for silicone gel

> implants exceeds 50 percent of all implants, no study has ever

> examined the relationship between ruptured implants and disease and

> symptoms and most studies have overestimated the prevalence of breast

> implants.

>

> It is beyond dispute that none of the controlled epidemiology studies

> has examined the connection between breast implants and the symptoms

> that clinicians have reported. Instead, the studies have focused on

> scleroderma or on defined classical disease presentations. While some

> have argued that the -Guerrero paper(777) examined symptoms and

> symptom complexes, that is simply not the case. As Dr. Graham Colditz,

> one of the study's authors, has testified:

>

> Q. And when you designed this questionnaire and you designed the

> study, the Nurse's Study, you made a decision to only look for

> classical, traditional rheumatic disease; isn't that true?

>

> A. Yes.

>

> Q. And that's all the study concerns; covers?

>

> A. Yes.(778)

>

> Similarly, Dr. Speizer, one of the co-authors of the paper, has

> testified by affidavit(779) as follows:

>

> The sentence at page 1668 of the Article, "[w]e studies women with

> possible early, milder, or atypical forms of connective-tissue disease

> or with any sign or symptom of a connective-tissue disease who did not

> meet standard classification criteria" refers to categories a, b, and

> c above. There was no attempt to define "atypical disease," nor is the

> reference in the quoted sentence meant to define or refer to any

> specific set of conditions which may or may not have been called

> "atypical disease" by others. Rather, the word "atypical" was used to

> refer to women who reported classical connective tissue disease on

> their biennial questionnaires, but did not meet the criteria of

> category d.

>

> Indeed, as the Silverman Kessler review noted:

>

> . . . no study has specifically addressed atypical connective tissue

> diseases, and the few studies that have attempted to consider symptoms

> that may indicate atypical syndromes have had design flaws that render

> them inconclusive.

>

> The claim that the -Guerrero paper examined an association

> between silicone breast implant and atypical disease or non-defined

> symptoms is simply not supported by the facts. Thus, despite the

> manufacturers' attempts to the contrary, no study has examined

> atypicality.

>

> One very interesting issue that has never received tremendous

> attention is the fact that many of the controlled epidemiology studies

> use a breast implant prevalence rate that even the manufacturers agree

> is too high. Assume, simply by way of example, that 100,000 out of a

> population of 1,000,000 have a common problem. One could say that 10%

> of the population is affected. Assume, however, that 100,000 out of a

> population of 2,000,000 share a common problem. Then, only five

> percent of the population is affected, a far smaller number. Thus, the

> larger the size of the overall population, the smaller the

> complication rate. Early literature discussed a prevalence rate of

> between one and two million women with implants, but even Dow Corning

> has now published studies suggesting that the most likely estimate is

> 815,700 with a range of 715,757 to 924,729.(780)Many of the studies,

> and two in particular, used far higher prevalence rates for implants.

> For example, the paper,(781) began by announcing that

> "etween 1 and 2 million American women have undergone breast

> [surgery]." -Guererro(782) used even higher estimates: "ince

> 1962, approximately 1 million to 2.2 million women in the United

> States and Canada have received silicone breast implants . . . ." This

> over-estimate of the prevalence of breast implants is certainly a

> reason to question the reliability of the conclusions of these

> studies.

>

> Again, the manufacturers' control at the outset leads to questions

> about the reliability of the results. Thus, the epidemiologic evidence

> on which the manufacturers rely is inconclusive and certainly cannot

> be cited as "proof" of a lack of association.

>

> F. APPLICATION OF THE BRADFORD HILL CRITERIA, IN A MANNER CONSISTENT

> WITH SIR BRADFORD HILL'S ADMONITIONS, SUPPORTS AN ASSOCIATION BETWEEN

> IMPLANTS AND IMMUNE DYSFUNCTION

> The manufacturers recite as their mantra the Bradford Hill criteria.

> But as Dr. Sander Greenland told the Panel, those criteria should be

> used with a great deal of caution and with care to understand their

> limitations, limitations that Sir Bradford Hill included in his

> original article.(783)

>

> Sir Bradford Hill's criteria(784) were developed for use in the field

> of occupational medicine, but have been widely applied in other

> fields. As he acknowledged, "[t]he 'cause' of illness may be immediate

> and direct, it may be remote and indirect . . . ." To assist in

> assessing "cause," he developed nine criteria, criteria that the

> manufacturers claim "prove" there is no relationship between exposure

> to silicone gel breast implants and immune dysfunction. Contrary to

> their assertion, however, the criteria -- when applied consistent with

> Sir Bradford Hill's cautions -- support an association between

> silicone gel breast implants and immune system dysfunction and

> symptoms of disease.

>

> First, is the issue of strength. Sir Bradford Hill cautioned: "[w]e

> must not be too ready to dismiss a cause-and-effect hypothesis merely

> on the grounds that the observed association appears to be slight.

> There are many occasions in medicine when this is in truth so." Thus,

> it would be wrong, under Sir Bradford Hill's analysis, to discount the

> importance of the positive epidemiologic evidence supporting causation

> or to rely too heavily on that which fails to detect a relationship.

>

> Second, is the issue of consistency. He defined consistency as

> repeated observations by different persons under different

> circumstances. From the clinical reports of rheumatologists and

> immunologists, to the laboratory findings of research scientists, to

> the observations of pathologists, the findings are consistent, across

> different disciplines, in favor of a relationship.

>

> Sir Bradford Hill's third criteria is "specificity." His analysis is

> particularly telling in the context of evaluating an association

> between implants and immune dysfunction. He cautioned, "[w]e must not

> . . . over-emphasize the importance of this characteristic." In this

> context, no one denies that women develop autoimmune diseases,

> symptoms of disease or immune dysfunction when they do not have breast

> implants. Thus, the specificity analysis is not a relevant one in this

> context.

>

> The fourth criteria is a temporal relationship. The case reports, the

> case series, the literature on immunology, and, obviously, the

> pathology all focus on women whose symptoms appeared after their

> implants or women whose symptoms were exacerbated post implantation.

> Thus, the temporality requirement is met. Fifth, is what has become

> known as "dose-response." In instances in which there is something

> other than a linear relationship, he cautioned, "[w]e should then need

> to envisage some much more complex relationship to satisfy the

> cause-and-effect hypothesis." In the implant context, Dow Corning's

> chief epidemiologist has himself questioned whether a "classical

> toxicological" analysis of dose response was appropriate because of

> allegations of immune dysfunction. In this situation, particle size

> and total surface area may be as important as quantity.(785)

>

> The sixth issue is the one on which Section two of this submission

> focused -- bioplausibility. As detailed in that section, there is

> ample evidence of bioplausibility to support a conclusion in favor of

> causation. Seventh, the information must be coherent, an issue again

> tied to bioplausibility. Next, and an element that is compelling, is

> the issue of experiment. This inquiry focuses on what happens when the

> suspected offending agent is removed. Is there improvement? The

> evidence of remission - or even resolution of significant medical

> symptoms - following explantation is striking and lends further

> support to the causation model. Finally, Sir Bradford Hill urges the

> examination of analogies. The analogous evidence from the silica

> experience and its applicability based on the evidence of silica used

> in the silicone elastomer, and silicone degradation in vivo to silica

> is significant.

>

> In short, the Bradford Hill criteria when properly applied support the

> conclusion that there is a reasonable scientific basis for one to

> conclude that silicone gel breast implants cause immune system

> dysfunction and symptoms of disease.

>

>

>

> 697. Hennekens, C.H., Lee, I.M., Cook, N.R., "Self-Reported Breast

> Implants and Connective-Tissue Diseases in Female Health

> Professionals: A Retrospective Cohort Study," JAMA 275(8):616-621

> (Feb. 1996) [Record No. 0160].

>

> 698. Relative risk 1.24, p = .0015, 95% confidence interval, 1.08 -

> 1.41, two-sided.

>

> 699. Relative risk 1.30, p = .060, 95 % confidence interval, 1.05 -

> 1.62, two-sided.

>

> 700. Relative risk 1.18, p = .096, 95 % confidence interval, 0.97 -

> 1.43, two-sided.

>

> 701. Relative risk 1.49, p = .067, 95 % confidence interval, 0.97 -

> 2.28, two-sided.

>

> 702. Relative risk 1.52, p = .068, 95 % confidence interval, 0.07 -

> 2.37, two-sided.

>

> 703. Relative risk 1.84, p = .060, 95 % confidence interval, 0.98 -

> 3.46, two-sided.

>

> 704. Ory, Affidavit, p. 4, para. 10 (7/25/96) [Record No.

> 7184].

>

> 705. Sander Greenland, 7/23/97 Transcript of Panel Hearing, p. 465.

>

> 706. Id., p. 467.

>

> 707. Rothman, K., "A Show of Confidence," NEJM 299(24):1362-1363

> (correspondence) (1978) [Record No. 7185].

>

> 708. Id.

>

> 709. Id.

>

> 710. Record No. 7100 is the complete notice package from the original

> global settlement. The disease definitions are included as Exhibit D

> to that notice.

>

> 711. The Claims Office utilized trained nurses to review the claims

> and established a fraud unit to assure the integrity of the claims

> process.

>

> 712. In saying "the Court," the reference is actually to Judge Pointer

> working in conjunction with the claims office and the claims office's

> statistical and computer consultant.

>

> 713. Dr. Swan's entire report is contained in Record No. 7186,

> "Disease Occurrence in Silicone Breast Implanted Women: Estimates from

> the Sample of Domestic Claimants" (10/95).

>

> 714. It is likely that women with scleroderma, because of the severity

> of their illness, are over-represented in the opt-out group and

> underrepresented in the group that failed to register.

>

> 715. 2,679 represents the 2,994 included in the sample minus the major

> deficiencies (309) and the unknown (9).

>

> 716. 25,398 and 746 represent the anticipated number of major

> deficiencies and unknown, respectively.

>

> 717. This estimate can also be derived by multiplying the observed

> number of scleroderma cases (approved and minor) (5) by the ratio of

> valid claims in the sample to valid claims in the population: Total SS

> = 5 x (70,467/849) = 415.

>

> 718. The estimate of 75 cases is, admittedly, uncertain. It does,

> however, appear higher than that projected by -Guerrero who

> stated that a finding of 38 cases of scleroderma in implanted women

> "would indicate a possible association." -Guerrero, J., Schur,

> P., Sergent, J.S., et al., "Silicone Breast Implants and Rheumatic

> Disease: Clinical, Immunological, and Epidemiological Studies,"

> Arthritis & Rheumatism37(2):158-168 (1994) [Record No. 1747].

>

> 719. Kumagai, Y., Shiokawa, Y., Medsger, T., et al.,. "Clinical

> Spectrum of Connective Tissue Disease After Cosmetic Surgery:

> Observations on Eighteen Patients and a Review of the Japanese

> Literature," Arthritis & Rheumatism 27(1):1-12 (1984) [Record No.

> 1111].

>

> 720. Twelve had scleroderma (including 8 with progressive systemic

> sclerosis), 6 had rheumatoid arthritis, 5 had lupus, and 1 had

> polymyositis.

>

> 721. Giltay, E., Moens, H., Riley, A., et al., "Silicone Breast

> Prostheses and Rheumatic Symptoms: A Retrospective Follow-Up Study,"

> ls of Rheumatic Diseases 53:194-196 (1994) [Record No. 1731].

>

> 722. , S.E., O'Fallon, W.M., Kurland, L.T., et al., "Risk of

> Connective Tissue Diseases and Other Disorders after Breast

> Implantation," New England Journal of Medicine 330(24):1697-1702

> (6/16/94) [Record No. 1765].

>

> 723. Includes swelling of the wrist, swelling of three or more joints,

> symmetric joint swelling or any other documented arthritis or

> synovitis.

>

> 724. Includes dry eyes, dry mouth, or keratoconjunctivitis sicca.

>

> 725. Includes serosal inflammation such as pleuritis and pericarditis.

>

>

> 726. Press, R.I., Peebles, C.L., Kumagai, Y., et al., "Antinuclear

> Autoantibodies in Women with Silicone Breast Implants," The Lancet

> 340:1304-1307 (11/28/92) [Record No. 1636].

>

> 727. Claman, H.N., on, A.D., "Antinuclear Antibodies in

> Apparently Healthy Women with Breast Implants," Immunology of

> Silicones, pp. 265-268 (1996) [Record No. 0120].

>

> 728. Zazgornick, J., Piza, H., Kaiser, W., et al., "Autoimmune

> Reactions in Patients with Silicone Breast Implants," Wein Klin

> Wochenschur 108(24):781-787 (1996) [Record No. 7024].

>

> 729. Kayler, L.K., Goodman, P.H., "Breast Implants Increase The Risk

> of Arthralgia: An Epidemiological Meta-Analysis, J. Inves. Med.

> 43(1):129A (1995) [Record No. 3146].

>

> 730. Friis, S., Mellenkjier, L., McLaughlin, J.K., et al., "Connective

> Tissue Disease and Other Rheumatic Conditions Following Breast

> Implants in Denmark," ls of Plastic Surg. 39:1-8 (1977) [Record

> No. 7293].

>

> 731. Laing, T.J., Gillespie, B.W., Lacey, J.V., et al., "The

> Association Between Silicone Exposure and Undifferentiated Connective

> Tissue Disease Among Women in Michigan and Ohio," presented at

> American College of Rheumatism 60th Nat'l Scientific Meeting, October

> 18-22, 1996 [Record No. 0318].

>

> 732. Greenland, S., Finkle, W.F., "A Case Control Study of Prosthetic

> Implants and Selected Chronic Diseases," ls of Epidemiology

> 6:530-540 (1996) [Record No. 5154].

>

> 733. Mehihorn, J., Ziegler, V., "Epidemiological Analyses of the

> Relation Between Scleroderma, Exposure to Quartz and Silicosis in Men

> in East Germany," Abstract, Int. Epid. Assoc. (1997) [Record No.

> 7199].

>

> 734. Steenland, K., Goldsmith, D.F., "Silica Exposure and Autoimmune

> Disease," American Journal of Ind. Med.28:603-608 (1995) [Record No.

> 0239].

>

> 735. -Roman, J., Wichmann, Saliberri, J., et al., "Multiple

> Clinical and Biological Autoimmune Manifestations in 5 Workers After

> Occupational Exposure to Silica," Ann. Rheum. Dis. 52:534-538 (1993)

> [Record No. 0222].

>

> 736. Conrad, K., Mehlhorn, J., Luthke, K., et al., "Systemic Lupus

> Erythematosus After Heavy Exposure to Quartz Dust in Uranium Mines:

> Clinical and Serological Characteristics," Lupus 5:62-69 (1996)

> [Record No. 2300].

>

> 737. Steenland, K., Brown, D., "Mortality Study of Gold Miners Exposed

> to Silica and Nonasbestiform Amiphibole Minerals: An Update with 14

> More Years of Follow-Up," American Journal of Industrial Medicine

> 27(2):217-229 (1995) [Record No. 0354].

>

> 738. Strom, B., Reidenberg, M., Freundlich, B., et al. "Breast

> Silicone Implants and Risk of Systemic Lupus Erythematosus," Journal

> of Clinical Epidemiology 47(10):1211-1214 (1994) [Record No. 0654].

>

> 739. Burns, C., Gillespie, B., Heeringa, S., et al., "Response to

> Arthur D. Little comments of December 4, 1992 regarding the Women's

> Health Study protocol," DCC 269004712-2690047 (1/20/93) [Record No.

> 7187] at page 3, paragraph 3.

>

> 740. Colton, T., Statistics in Medicine, Little, Brown & Company, p.

> 124 (11/74) [Record No. 7253].

>

> 741. Horwitz, L., Ferleger, L., Statistics for Social Change, South

> End Press (1st Ed.), pp. 220-221 (1980) [Record No. 7254].

>

> 742. The PSEF is the non-profit education arm of the American Society

> for Plastic & Reconstructive Surgery. The manufacturers provided money

> to the PSEF to enable the PSEF to fund studies, including epidemiology

> studies. In addition, the manufacturers' attorneys were intimately

> involved in deciding which studies to fund. PSEF Notes re: Silicone

> Research Funding Summit Meeting, Hotel Nikko, Chicago, ASP 22922-22925

> (7/10/92) [Record No. 7043].

>

> 743. , S., Letter to H. Caffee attaching grant proposal

> (3/20/92) [Record No. 0517].

>

> 744. , S.E., O'Fallon, W.M., Kurland, L.T., et al., "Risk of

> Connective Tissue Diseases and Other Disorders After Breast

> Implantation," New England Journal of Medicine 330(24):1697-1702

> (6/16/94) [Record No. 1765].

>

> 745. Englert, H.J., Letter to Ralph Cook , DCC 269003456-269003459

> (10/6/92) [Record No. 7188].

>

> 746. Cook, R.R., Letter to H. Englert, DCC 279011631-279011637

> (11/4/92) [Record No. 7189].

>

> 747. Hennekens, C.H., Lee, I.M., Cook, N.R., et al., "Self-Reported

> Breast Implants and Connective-Tissue Diseases in Female Health

> Professionals: A Retrospective Cohort Study," JAMA 275(8):616-621

> (2/28/96) [Record No. 0160].

>

> 748. Hennekens, C.H., Buring, J., , C., et al., "A

> Retrospective Cohort Study of Breast Implants and Connective Tissue

> and Other Disorders," protocol, WHS 753-784 (6/18/92) [Record No.

> 2795] at p. WHS 759.

>

> 749. Hennekens, C.H., Lee, I.M., Cook, N.R., et al., "Self-Reported

> Breast Implants and Connective-Tissue Diseases in Female Health

> Professionals: A Retrospective Cohort Study," JAMA 275(8):616-621

> (2/28/96) [Record No. 0160] at p. 617.

>

> 750. Plaintiffs are attempting, through Dow Corning's bankruptcy

> proceedings, to obtain further information but have so far been unable

> to do so.

>

> 751. Cook, R., Letter to M. White, DCC 26000719-26000721

> (9/3/92) [Record No. 7255].

>

> 752. In another example of the role the manufacturers played in the

> design of the epidemiology studies, in a number of instances, women

> with only saline implants were included - not by the researcher, but

> at the insistence of the manufacturers. For example, in deciding

> whether to fund the proposal, Ralph Cook wrote, "if [sic] they

> develop of [sic] protocol and include all the appropriate caveats, and

> expand the study group to include those with saline implants, this

> could be a fundable project." Cook, R., Letter to Hollis Caffee,

> DCC0100001189 - 0109001194 (5/18/92) [Record No. 7256]; see also

> undated internal Dow Corning memo, DCC 269002729 [Record No.7296]

> ("Mention our interest in including saline implants.").

>

> 753. Silverman, B.G., Brown, S.L., Bright, R.A., et al., "Reported

> Complications of Silicone Gel Breast Implants: An Epidemiological

> Review, ls of Internal Medicine 124(8):744-56 (1996) [Record No.

> 0234].

>

> 754. The resulting study is Hochberg, M., Perlmutter, D., Medsger, T.,

> et al., "Lack of Association Between Augmentation Mammoplasty and

> Systemic Sclerosis (Scleroderma)," Arthritis & Rheumatism

> 39(7):1125-1131 (7/96) [Record No. 2925].

>

> 755. Cole, P., Review of Hochberg Proposal for Dow Corning, DCCKMM

> 401114-40115 (1/8/91) [Record No. 7190].

>

> 756. Id.

>

> 757. Cole, P., Review of Schottenfeld Proposal for Dow Corning, DCCKMM

> 401111-401113 (1/8/91) [Record No. 7191].

>

> 758. Id. at DCKMM 401112.

>

> 759. Weisman, M., Vecchione, T., Albert, D., et al.,

> "Connective-Tissue Disease Following Breast Augmentation: A

> Preliminary Test of the Human Adjuvant Disease Hypothesis," Plastic

> and Reconstructive Surgery 82(4):626-630 (10/88) [Record No. 1175].

>

> 760. Strom, B.L., Reidenberg, M.M., Freundlich, B., et al., "Breast

> Silicone Implants and Risk of Systemic Lupus Erythematosus, J. Clin.

> Epidem.47(10):1211-1214 (1994) [Record No. 0654].

>

> 761. , S.E., O'Fallon, W.M., Kurland, L.T., et al., "Risk of

> Connective-Tissue Diseases and Other Disorders After Breast Implants,"

> NEJM 330(24):1697-1702 (6/16/94) [Record No. 1765].

>

> 762. Id.

>

> 763. Id.

>

> 764. Woodbury, M.A., Memo to file re: Harvard Nurse's Health Study

> Proposal, DCC 279022399-279022404 (3/10/92) [Record No. 7192].

>

> 765. Schusterman, M.A., Kroll, S.A., Reece, G.P., et al., "Incidence

> of Autoimmune Disease in Patients After Breast Reconstruction with

> Silicone Gel Implants Versus Autogenous Tissue: A Preliminary Report,"

> ls of Plastic Surgery 31(1):16 (1993) [Record No. 1691].

>

> 766. Wells, K.E., Cruse, C., Baker, J., et al., "The Health Status of

> Women Following Cosmetic Surgery," Plastic and Reconstructive Surgery

> 93(5):907-912 (4/94) [Record No. 1752].

>

> 767. Id. at 909.

>

> 768. Hennekens, C.H., Lee, I.M., Cook, N.R., "Self-Reported Breast

> Implants and Connective Tissue Diseases in Female Health

> Professionals: A Retrospective Cohort Study," JAMA 275(8):616-621

> (2/28/96) [Record No. 0160].

>

> 769. Dow Corning, Implant Sales Figures 1962-1992 (6/11/93) [Record

> No. 7193].

>

> 770. Ory, 7/24/97 Transcript of Panel Hearing, p. 726.

>

> 771. Silverman, B.G., Brown, S.L., Bright, R.A., et al., "Reported

> Complications of Silicone Gel Breast Implants: An Epidemiologic

> Review," ls of Internal Medicine 124(8):744-756 (4/15/96) [Record

> No. 0234].

>

> 772. -Guerrero, J., Colditz, G., Karlson, E., et al., "Silicone

> Breast Implants and The Risk of Connective-Tissue Diseases and

> Symptoms," NEJM 332(25):1666-1670 (6/22/95) [Record No. 1819].

>

> 773. When the investigators presented their abstract at the American

> College of Rheumatology meeting in October, 1994, their data indicated

> 1209 nurses with implants in the cohort. The discrepancy was not

> explained in the paper. -Guerrero, J., Liang, M., Karlson, E.,

> et al., "Silicone Breast Implants and Connective Tissue Disease

> (CTD)," Abstract 736 (10/94) [Record No. 1786].

>

> 774. Again, the abstract reported different numbers -- 448 cases with

> definite connective tissue disease and five patients with connective

> tissue disease and implants. Id.

>

> 775. Silverman, B.G., Brown, S.L., Bright, R.A., et al., "Reported

> Complications of Silicone Gel Breast Implants: An Epidemiologic

> Review," ls of Internal Medicine 124(8):744-756 (4/15/96) [Record

> No. 0234].

>

> 776. , S.E., O'Fallon, W.M., Kurland, L.T., et al., "Risk of

> Connective-Tissue Diseases and Other Disorders After Breast Implants,"

> NEJM 330(24):1697-1702 (6/16/94) [Record No. 1765].

>

> 777. -Guerrero, J., Colditz, G., Karlson, E., et al., "Silicone

> Breast Implants and The Risk of Connective-Tissue Diseases and

> Symptoms," NEJM 332(25):1666-1670 (6/22/95) [Record No. 1819].

>

> 778. Deposition of Graham Colditz, p. 50 (1/21/95) [Record No. 7194].

>

> 779. Affidavit of E. Speizer, M.D., ¶ 4 (2/13/96) [Record No.

> 7195].

>

> 780. Cook, R.R., Perkins, L.L., "The Prevalence of Breast Implants

> Among Women in the United States," Immunology of Silicones 419-425

> [Record No. 7129]. Of additional interest on the prevalence rate issue

> is the fact that, in 1993, Dr. Roselie Bright from the FDA, published

> a report based on a survey in which she concluded that, as of 1988,

> 304,000 women had received implants. Bright, R.A., Jeng, L.L., .

> R.M., "National Survey of Self-Reported Breast Implants: 1988

> Estimates, J. Of Long-Term Effects of Medical Implants, 3(1):81-89

> (1993) [Record No. 2195].

>

> 781. , S.E., O'Fallon, W.M., Kurland, L.T., et al., "Risk of

> Connective Tissue Diseases and other Disorders after Breast

> Implantation," NEJM 330(24):1697-1702 (6/16/94) [Record No. 1765].

>

> 782. -Guerrero, J., Colditz, G., Karlson, E., et al., "Silicone

> Breast Implants and The Risk of Connective Tissue Diseases and

> Symptoms," NEJM, 332(25):1666-1670 (6/22/95) [Record No. 1819].

>

> 783. Sander Greenland, 7/23/97 Transcript of Panel Hearing, p. 486.

>

> 784. Bradford Hill, A., "The Environment and Disease: Association or

> Causation?" Proc. Royal Soc. Med. 58:295 (1966) [Record No. 7271].

>

> 785. Cook, R., Memo to File re: Conversation with Teel, DCC

> 279022491 (6/30/93) [Record No. 7288].

__________________________________________________