Breast Implants And Disease[1]
Dr. David S. Egilman, M.D., Clinical Associate
Professor, Brown University School of Medicine


I, Dr. David S. Egilman MD, MPH, am a medical doctor and Clinical Associate Professor of Community Medicine at Brown University. I am board certified in Internal Medicine and Preventive Occupational Medicine. My office address is 759 Granite Street, Braintree, Massachusetts, 02184.

I received a Bachelor of Science from Brown University in Molecular Biology in 1974. I received a medical degree from Brown University in 1978. I completed a three year medical residency in Internal Medicine at Strong Memorial Hospital in Rochester, New York, in 1981.

I completed a three year training program in epidemiology, called the National Institutes of Health Epidemiology Training Program in 1984. As part of this program, I completed a Master's in Public Health at the Harvard School of Public Health. At Harvard, I studied epidemiology, statistics, and occupational medicine.

I served two years at the National Institute for Occupational Safety & Health (NIOSH), designing and conducting small and large epidemiologic studies.

Since 1978, I have published a variety of letters and medical articles on the issues that relate to the manner in which cause-effect determinations are made in medicine (the epistemology of medicine). I have discussed the normal, accepted process of causal determination in medicine (described below) in several peer-reviewed articles.[2]  In addition, these ideas were accepted for presentation and were presented at the American Public Health Association meetings, in 1984. For the past five years, I have taught a course at Brown University that addresses the development of medical and scientific knowledge in the 20th century. This course deals specifically with the issues outlined in this report.

Finally, some of my opinions are based in part on my clinical experience and awareness of the ways that normal physicians in normal medical practice make decisions about etiologic relationships that affect patients' lives every day. Over 40 hours per week of my time is devoted to direct patient care.

As a practitioner, I utilize the epistemological method described below to make the following medical decisions: determining cause-effect relationships, judging drug side effects, judging therapeutic interventions, judging the presence of disease, providing forensic testimony and judging the causes of disability and death. The described methodology is consistent with the normal methods physicians use in the usual practice of medicine.

My opinions in this case are based on my training and experience in molecular biology, medicine, physics, chemistry, epidemiology, philosophy, ethics, public health, medical practice, review of corporate documents in this and other cases, review of depositions and deposition exhibits, discussions with researchers, and various medical literature.


I.  PRINCIPLES OF GENERAL CAUSATION

In making such judgments regarding medical causation, doctors utilize epidemiological data as one tool among many. Although doctors consider epidemiological research to be an important facet of medical epistemology, it is not regarded as being the sole or the most important contributor to decision-making. The branch of medical science concerned with the causes and origins of disease is called etiology. Epidemiology is not a synonym for etiology. Rather, etiologists use epidemiological evidence along with other evidence to draw conclusions about causal connections. Medical knowledge is based on empirical tests. Empirical tests are those gained from observation or experience.[3]  Epidemiology is but one form of empirical knowledge.

The evidence used to determine disease etiology may come from chemistry, physics, biochemistry, pathology, molecular biology, toxicology, clinical experience and other scientific disciplines. Types of data relied upon include observation, experiment, analogy, computer models, animal experiments, and bacteriologic experiments, as well as epidemiologic studies. Doctors consider the whole data set available to establish a causal nexus.


II.  THE EPISTEMOLOGICAL APPROACH OF SILICONE BREAST
     IMPLANT (SBI) MANUFACTURERS

Prior to the onset of litigation related to silicone breast implants (SBI), manufacturers never suggested that epidemiologic studies were necessary to determine the relationship of any health problem to SBI. SBI manufacturers never conducted, proposed nor sponsored any epidemiologic study of any known problem associated with SBI including contracture, migration, infection, hematoma, or rheumatologic complaint. The manufacturer's lack of interest in following the health effects of their products in women with epidemiological surveillance is further demonstrated by their failure to establish a registry of users or a systematic complaint mechanisms for users.

The SBI manufacturers evaluated the relationship between all of the above complaints and SBI based on case reports, clinical cases series, and animal studies. For example, Medical Engineering Corporation (MEC) Founder Wilfred Lynch wrote, "constrictive capsule formation is the complication of breast augmentation which is most troublesome at the moment. Its etiology is perplexing and most of the theories of contributing phenomena are quite subjective." [4]  Lynch did not base this statement of causation upon epidemiologic research. The fact that the specific mechanism was unknown did not prevent Lynch from concluding that the contracture was caused by the implant. Another comment by Wilfred Lynch, published in his 1978 Handbook of Silicone Rubber Fabrication, illustrates the type of testing the industry recommended:

"All basic materials which are to be used in the fabrication of implantable devices, or devices which will be in contact with fucous membrane for extended periods, must undergo long term implant tests in animals. In the course of these tests it must be demonstrated that there is no toxicity to the host, no undue inflammation of the tissues with which it has been in contact, and no migration of by-products to vital organs. The material itself must not show signs of deterioration." [p.217]

According to Lynch, animal testing is the sine qua non for determining product hazards.

In 1991, Micheal Syzcher, a consultant for Bristol-Myers[-Squibb] (BMS), published the following opinion on causation of SBI and disease:

"Autoimmune disease, with its attendant immunological sensitization may indeed be a serious risk associated with silicone gels. Human adjuvant disease or autoimmune disease after implantation of foreign material occurs subsequent to injection or implantation of paraffin/silicone, and possibly, silicone polymers. Patients develop signs, symptoms, and laboratory abnormalities suggestive but not diagnostic of a connective-tissue or autoimmune disease." [5]  [Emphasis added]

Syzcher goes on to reveal his reliance on case reports, mechanistic understanding, and animal studies to determine causation in humans:

"Several published reports have linked silicone fluids to various autoimmune diseases such as =93human adjuvant disease (110).  Based on animal models of adjuvant arthritis and experimental immunologic principles of adjuvant stimulation of the immune system, the term human adjuvant disease was applied to those diseases caused by exposure to silicone-containing materials (111).  Other syndromes were also implicated, such as scleroderma (112, 113), localized post traumatic musculo-skeletal conditions or interphalangeal osteoarthritis."

Heggers et. al., (115) conducted a study from which they concluded that although apparently inert, silicone is capable of eliciting a

". . . . cellular immune response demonstrated by the migration inhibition technique. This response is comparable to that elicited by purified protein derivative and may indicate that silicone acts as a hapten-like incomplete antigen."

Silicone-filled breast implants also contain silica (Si02) as a reinforcing filler in the device envelopes and gels (116).  Silica is known to be immunologically active, and may potentiate the development of autoimmune diseases induced by mycobacterial or other antigens present in resident microorganisms. An association of tuberculosis, silicone implants, and the subsequent development of connective tissue disease has been reported, and closely corresponds to an animal model of =93adjuvant disease (117).

Silicone fluids have been used in surgery for many years, notably in joint and breast augmentation prostheses. Silicones have generally been regarded as a relatively inert material that provokes little or no tissue reaction (118).  However, as the use of this material increases, reports of local and systematic reactions have emerged. This led Endo (119) to state that the notion that silicone is without important biologic effects may be na=EFve and require reexamination. [references in original text]

Syzcher based his opinion on case reports and case series and thus followed the normal medical practice used to determine the relationship between an exposure and a disease. This process, which does not require the sine qua non presense of a positive controlled epidemiolgic study.

Testimony and exhibits demonstrate that Medical Engineering Corporation (MEC) and/or Bristol Myers[-Squibb] (BMS) made claims about its products based on evidence other than statistically significant epidemiology studies. MEC did not base business decisions about sale of breast implants on epidemiology, as one of its documents notes:

"We don't have to PROVE validity to sell products, i.e., SCL's there's no proven validity there either . . . .  "   It continues, "Remind Lynn that we are not making any profit now, we have no alternative but to grasp at straws or anything else that might make a buck, we are desperate and we have to take risks or die.[6]

Labeling instructions were based on only the publication of particular case reports and in some cases were included without citation to any publication at all. For example the Bilumen Implant insert carried a protocol by John Munna, M.D.[7]  Not only did MEC not have any epidemiologic evidence supporting Munna's protocol, they entered into an arrangement with Munna where he received a percentage of each bilumen implant sale in exchange for his writing certain "scientific" papers.

Warnings included in the packaging are further examples of acceptance of association by manufacturers based only on case reports. The statement "closed capsulotomy (i.e., manual compression of the breast) is not recommended because it may cause rupture with potential gel extravasation" was added to the package insert after case reports appeared in the medical literature reporting rupture after closed capsulotomies.[8]

Based on a case report published by MEC consultant Dr. James Baker concerning closed capsulotomy as an alternative to open capsulotomy, MEC added this article to many other case reports cited in its Bibliography of its Biocompatibility Data Sheets.[9]  Company sales representatives used these sheets and they were intended to be left with physicians to encourage them to purchase MEC's implants. The bibliography contained many references to non-epidemiological studies.[10]

There are other MEC documents which demonstrate the company relied on studies, articles and case reports for a variety of different reasons. Several articles discussing surgical technique are included as part of the company's Training Manual for MEC Sales Representatives.[11]  In 1982 when the FDA wanted to classify breast implants as Class III devices requiring a Premarket Approval Application (PMA), MEC opposed the Class III classification. Betty Lock's letter to ASPRS, the society of plastic surgeons which joined in opposition with the implant manufacturers, cited case reports in support of the company's claims that implants add to the "quality of life," claims concerning capsular contracture, and claims concerning gel bleed.[12]  Similarly, the Health Industry Manufacturer's Association letter from Harold Buzzell to FDA attaching "Health Industry Manufacturers Association Comments on FDA's Proposed Classification of General and Plastic Surgery Devices" also cites non-epidemiological articles for support.[13]


III.  PROBABILISTIC CAUSAL ANALYSES

In determining the etiology of infectious diseases (caused by a microorganism), doctors have used Koch's postulates as a basis for deciding what experimental evidence is necessary to establish cause. Koch's postulates are: (1) the microorganism must be observed in every case of the disease; (2) it must be isolated and grown in pure culture; (3) the pure culture must, when inoculated into a susceptible animal, reproduce the disease; and (4) the microorganism must be observed in, and recovered from, the experimentally diseased animal.[14]

Although epidemiology plays absolutely no role in Koch's postulates, the medical community has completely accepted them as valid epistemological criteria for certain causal relationships.

However, doctors do not apply Koch's postulates directly or by analogy to determine causation for injuries and non-infectious diseases. Like infectious disease, the etiology of non-infectious diseases and injuries is multi-factorial, and most of the factors are unknown. However, in contrast to infectious diseases, no single factor is always a sine qua non of non-infectious disease. For example in the infectious disease tuberculosis, all patients with the disease have the TB bacillus in their bodies. But all non-infectious disease have multiple causes and have no sine qua non causal factor. For example, smoking and asbestos are both independent causes of lung disease and lung disease occurs in non-smoking, non-asbestos exposed people.

Both TB and lung cancer demonstrate the multi-factorial nature of disease etiology. Only a very small percentage of smokers, and a very small percentage of persons carrying the TB bacillus, ever get either lung cancer or tuberculosis. A combination of other known and unknown factors determine which smokers develop cancer or which individuals with the bacillus develop tuberculosis. Although doctors accept that asbestos and cigarette smoking cause cancer, cancer occurs in less than 20 percent of smokers and slightly over half of asbestos insulation workers.

Therefore, the reliance on microscopic evidence implicit in Koch's postulates is not a universal causal criterion for non-infectious diseases. Neither pathologists nor epidemiologists nor clinicians require the presence of microscopic evidence of previous smoking to attribute lung cancer to a person's smoking. Case reports and uncontrolled population data convinced most physicians that smoking was a cause of lung cancer, and that exposure to ionizing radiation caused certain cancers. Astute pathologists and readers of pathologic literature were the first to recognize that asbestos was a carcinogen; epidemiologists did not make this association until many years later.[15][16]  Pathologists and pathologic evidence is still the most important type of evidence for most legal proceedings and is usually the arbiter of choice for the determination of the cause of death. Epidemiologist are not called to the morgue to determine cause of death for death certificates.

In non-infectious diseases, probability analyses can be used to determine disease etiology. It is possible to integrate different data sets into one causal probability through a model influenced by Bayesian decision-making (see Figures 1, 2). This method integrates available information into one model for causation, representing the results of each study type on a continuous line from 0 to 100, with 0 representing no evidence for causation and 100 representing the certainty of evidence of causation. Then the study types are arranged hierarchically, and the overall probability of causality is sequentially modified by each study type in turn, until a final value between 0 and 100 is reached.

It is necessary to describe this model for causation because of misrepresentation of the scientific method in public discourse. Some researchers have attempted to elevate one entire class of data to the level of final arbiter of causal proof. For instance, scientists working for tobacco companies argued that cigarettes had not been proven to cause lung cancer because of the lack of animal studies establishing such a relationship, because the exact mechanism of cancer induction is unknown and the specific carcinogenic substance(s) is (are) unknown. Some scientists argue that the evidence linking silicone breast implants to disease is not sufficient to establish a cause-effect relationship because of the insufficiency of epidemiological data. The scientific community has never exclusively relied upon epidemiology as the accepted method of evaluating cause-effect relations for making normal medical decisions. It is reasonable for different people to assign a different priority to different classes of studies. However, it is not reasonable to ignore entire classes of data, or, in an ad hoc manner, to elevate one class of data as carrying the ultimate burden of proof, especially for making public policy decisions regarding potential health risks.


IV.  IMPORTANCE OF CLINICAL EXPERIENCE

The textbook titled Medical Decision Making identifies the following information available to physicians making medical decisions: personal experience, published experience, and attributes of the patient.[17]  A physician's opinion is guided by personal experience with similar events or by the experiences of colleagues. Thus, a surgeon's estimate of the probability that a patient will survive an appendectomy is guided by personal experience with this operation in similar patients.[18]  In addition, physicians rely upon published experience, in the form of reports quantifying the risk or success associated with a certain procedure. For instance, the above mentioned surgeon may rely on a report giving the rate of death after appendectomy in order to estimate the probability that a patient may die from the operation. Published experience is particularly influential when the physician has little personal experience.[19] [emphasis added]  Finally, attributes of the patient are important to alert the physician for unusual characteristics of the patient that put him at higher or lower risk than the average.[20]

These principles of making medical decisions are exactly analogous to making cause-effect evaluations. In the case of appendectomy and likelihood of dying, the appendectomy is a determinant and the outcome is risk of dying. As can be seen from the above, in making these decisions, physicians rely heavily on their prior experience. Clinicians use their prior experience with similar events to estimate probability. Experienced clinicians have seen so many patients that they have a good intuitive understanding of which events occur commonly and which events are unusual. Personal experience is, and will continue to be, the principal factor influencing a physician's probability estimates.[21]

Doctors relying on clinical experience depend on an implicit control or comparison group -- all the other patients they have seen. A case report is essentially a small epidemiologic study. A case report is made because the case is unusual or informative with respect to all other cases the doctor has seen.

Clinical experience is considered particularly important in the diagnosis of occupational and environmental diseases because of the importance of the history. The history is the most important piece of information in determining if an individual case of disease or injury is related to exposure. For example possible causes of asthma include, cold air, exercise, pollen, chemical exposures, and dust. Doctors determine general and specific individual causation and causal factors from the history of the time course of exposure and symptoms onset and relief of symptoms correlated with removal of exposure. Physical examination and laboratory tests are helpful, but ultimately it is information obtained from an occupational history that determines the likelihood that a medical problem is work-related.[22]  The history is also critical in determining whether or not SBI induced disease in an individual. The time course of exposure correlated with the onset and continuation of symptoms, the pattern of signs, symptoms and laboratory abnormalities and the effect of explantation on symptoms are all considered in the physician determination of general and specific causation. This is the usual method used by physicians in determining disease causation and in selecting treatments. This is particularly true in dealing with specific criteria for causation, such as temporality (discussed in more detail below). With a well-taken history, the course of a patient's complaint can provide evidence as to whether the complaint is caused by an exposure. Doctors ask questions like: Do the symptoms begin after the start of the exposure? Do they disappear if the exposure is removed? Do laboratory abnormalities follow the exposure? Are local effects present? Is there a pattern of symptoms or signs? A physician's clinical experience, and the quality of their history, is especially important in cases where an injury or disease has multiple known causes.


V.  EXAMPLES FROM MEDICAL PRACTICE

1)  Injuries

When someone is run over by a car and dies, their death can be attributed to the traumatic injury of being struck by an automobile, even in the absence of specific knowledge of the cause of death. The immediate cause of death may have been a heart attack, a broken neck, or suffocation. Furthermore, even if the exact mechanism whereby an injury is produced is unknown, causal conclusions may safely be drawn. An epidemiological study is not necessary to determine that the trauma from the car caused death. In addition, if an epidemiological study of Chevrolets indicated that Chevrolets can kill people hit in traffic it is not necessary to perform another study to establish that Fords can kill people. Normal medical practice is the application of inference, not the deductive process of mathematical proofs.

2)  Diseases

A further example is that of diabetes. There were no epidemiologic studies done to show that insulin was related to diabetes. The causal connection was established through animal studies only. Treatment with insulin began as soon as insulin was available, without controlled trials to demonstrate proof of its effectiveness. Clinical experience combined with animal experimentation was sufficient.

3)  Proposed Treatments

Penicillin was used to treat infections as soon as it was available. No epidemiology was cited in support of the use of penicillin for the treatment of patients sick with pneumonia. The first epidemiologic studies on the effectiveness of penicillin and pneumonia were not published until the 1970's. The epidemiologic studies failed to find any reduction in short term mortality from the use of penicillin. Doctors used penicillin because of laboratory studies and because they had a vague understanding of the mechanism of action (they knew it killed bacteria).


VI.  EXAMPLES FROM MEDICAL LITERATURE

The importance of case reports, or personal experience, in making medical decisions is reflected in assessments of causation relied upon every day by physicians. When physicians turn to textbooks for opinions on general causation, they find all sources of data presented, often with no distinction made between epidemiologic and other data.

To assess the importance of animal studies and human health, the US Congress' Office of Technology calculated the percentage of papers In journals that relied on animal, non-animal or human data. In all but one of the biomedical and behavioral journals reviewed, the majority of papers published were animals studies.[23]  The same publication contained a diagram of the steps in biomedical research that preceded successful coronary artery bypass graft surgery. The diagram visually illustrated the importance of and reliance upon various types of animal research for medical decision making. See Attachment 1.

The Physician's Desk Reference, which includes sections on side effects and precautions for prescribed medications, is one example.[24]  The information contained in this text is supported by case reports, animal studies, and epidemiologic literature. The entries are not separated according to the source of information. All of the sources are considered important in making decisions about whether a certain drug may cause a side effect in a specific patient. A physician will reduce dosage, or change medications, when a patient reports side effects consistent with those described in the PDR.

The textbook Principles of Surgery identifies the following causes of low back pain: Compression fracture, vertebral process fracture, sprain and strain, ruptured disc.[25]  This list is derived from an article published in 1960. At the time of publication, these relationships were not supported by epidemiologic studies. The relationships were then and are now understood to be causal by virtue of case reports, and physicians' general understanding of bio-mechanics.

In texts devoted exclusively to occupational medicine, causal associations are frequently made without reference to epidemiology. The following diseases are attributed to occupational causes in the absence of epidemiologic data:

Hand infections:  Mycobacterium marinum infections are unusual infections related to puncture wounds suffered from handling fish, crab, shrimp, or coral.[26]  =93SPorotrichosis is caused by the fungi Sporothix schenkii and usually presents with cutaneous cellulitis and lymphangitis after a puncture wound secondary to a contaminated thorn.[27]

Chronic Paronychia:  Repeated immersion of the finger tips with consequent paronychia affects, in particular, bar tenders, kitchen workers, and laundry workers. Prolonged immersion of the hands in water, particularly hot water that contains detergents, increases the risk of paronychia.[28]

Onycholysis:  Enzymes in laundry detergent have been reported to cause painful onycholysis with hemorrhage. Onycholysis has also been reported from hydrofluoric acid in a rust-removing agent and to sodium hypochlorite in an 18-year-old lifeguard who added 16% sodium hypochlorite to the pool daily.[29] [emphasis added]

Koilonychia:  Koilonychia has been reported in a woman who was employed as a wire coil winder. Koilonychia has been reported from organic solvents used to clean metal parts and accessories in an office furnishing factory.[30]  [emphasis added]

These examples show that physicians sometimes require only one case study in order to warn the medical community that certain occupations may be at risk of work-related disease. Furthermore, case reports serve to establish other important causal connections, especially for diseases or injuries that have many possible causes . Thus, case reports can make a new link between an occupation and an established disease or injury. This is an example of the use of the causal criterion analogy (see below for further detail), in which, because of similar types of exposure, a case report documents the existence of a risk within an occupation not known to be at risk. (Also see Appendix 1 for more examples of cause-effect relationships established with case series.)


    TABLE 1 - CAUSAL RELATIONSHIPS ESTABLISHED BY CASE REPORTS:

    • Thalidomide and malformation
    • Omniflox and liver failure
    • Isocyantes and asthma
    • Cold exposure and asthma
    • Allergen exposure and asthma
    • Vinyl chloride and cancer
    • Vinyl chloride and scleroderma
    • Adulterated rapeseed oil and scleroderma
    • Vaginal cancer and DES (early epidemiologic studies failed to confirm but did not impact on the acceptance of the causal relationship)
    • Silica dust and scleroderma
    • Solvents and scleroderma
    • Epoxy resins and scleroderma
    • L-Tryptophan and scleroderma
    • Bleomycin and scleroderma
    • Pentazocine and scleroderma
    • Specific side effects of hundreds of drugs, e.g., Chloramphenicol and aplastic anemia


Case reports have also provided a basis for many treatment modalities. TABLE 2 contains examples from the rheumatologic literature.


    TABLE 2 - TREATMENT MODALITIES IN RHEUMATOID DISEASE
                      BASED ON CASE

    • Hypertrophic osteoarthropathy:

      Steinfeld, A.D., and Munzenrider, J.E.: The response of hypertrophic pulmonary osteoarthropathy to radiotherapy. Radiology 113:709, 1974.

      Lopez-Enriquez, E. Morales, A. R., and Robert F.: Effect of atropine sulfate in pulmonary hypertrophic osteoarthropathy. Arthritis Rheum. 23:822, 1980.

      Lokich, J. J.: Pulmonary osteoarthropathy; association with mesenchymal tumor metastases to the lungs, J.A.M.A. 238:37, 1977.

      Leung, F.W., Williams, A.J., and Fan, P.: Indomethacin therapy for hypertrophic pulmonary osteoarthropathy in patients with bronchogenic carcinoma, West J. Med. 142: 345, 1985.

    • Primary and Secondary Hemochromatosis:

      Angevine, C.D., and Jacox, R.F.: Unusual connective tisue manifestations of hemochromatosis. Arthritis Rheum. 17: 477, 1974.

      McCarthy, J.T., Libertin, C.R., Mitchell, J.C., III, and Fairbands, V. F.: Hemosiderosis in a dialysis patient: Treatment with hemofiltration and deferoxamine chelation theraphy. Mayo Clin. Proc. 57:439, 1982.

      Cohen, A., Cohen, I.J., and Schwartz, E.: Scurvy and altered iron stores in thalassemia major. N. Engl. J. Med. 304:158, 1981.

    • Sarcoidosis:

      Kaplan, H.: Sarcoid arthritis with a response to colchicine. N. Engl. J. Med. 263:778, 1960.

      Kaplan, H.: Further experience with colchicine in the treatment of sarcoid arthritis. N. Engl. J. Med. 268:761, 1963.

      Harris, E. D., Jr., and Millis, M.: Treatment with colchicine of the periarticular inflammation associated with sarcoidosis: A need for continued appraisal. Arthritis Rhem. 14:130, 1971.

      Neville, E., Carstairs, L.S., and James, D.G.: Bone sarcoidosis. Ann. N.Y. Acad. Sci. 278:475, 1976.

      Franco-Saenz, R., Ludwig, G.D., and Henderson, L.W.: Sarcoidosis of the skull. Ann. Intern. Med. 72:929, 1970.

      Baldwin, D.M., Roberts, J.G., and Croft, H.W.: Vertebral sarcoidosis. J. Bone Joint Surg. 56:629, 1974.

      Perlman, S.G., Damergis, J., Witorsch, P., et al.: Vertebral sarcoidosis with paravertebral ossification, Arthritis Rheum. 21:271, 1978.

      Watson, R.C., and Cahen, I.: Pathological fracture in long bone sarcoidosis. J. Bone Joint Surg. 55:613, 1973.

      Marcove, R.C., Rooney, R. and Weis, L.D.: Osteosclerotic lesions in sarcoidosis. Clin. Orthrop. Rel. Res. 129:248, 1977.

      Schriber, R.A., and Firooznia, H.: Extensive phalangeal cystic lesions-limited sarcoidosis. Arthritis Rheum. 18:123, 1975.

    • Amyloidosis:

      Stone, M.J., and Frankel, E.P.: The clinical syndrome of light chain myelome. A study of 35 patients with special reference to the occurence of amyloidosis. Am. J. Med. 58:601, 1975.

      Smith, M.E., and Byuwaters, E.G.L.: Mortality and prognosis related to the amyloidosis of Still's disease. Ann. rheum. Dis. 27:137, 1968.

      Goldfinger, W.E.: Colchicine for familial Mediterranean fever (letter). N. Engl. J. Med. 287:1302, 1972.

    • Polychondritis:

      Svenson, K.L.G., Holmdahl, R., Klareskog, L., Wibell, L., Sjoberg, O., Klintmalm, G.B.G., and Bostrom, H.: Cyclosporin A treatment in a case of relapsing polychondritis. Scand. J. Rheumatol. 13:229, 1984.

    • Osteoarthritis (Drug therapies):

      Anderson, R.J., Potts, D.W., Gabow, P.A., et al.: Unrecognized adult salicylate intoxication. Ann. Intern. Med. 85:745, 1976.

      Hodgkinson, R., and Woolf, D.: A five-year clinical trial of indemethacin in osteoarthrosis of the hip. Practitioner 210:372, 1974.


Articles in the medical literature discuss the importance of evidence besides epidemiology.

Fleming et al. reviewed and evaluated whether the investigation of disease clusters continues to play an important role in establishing disease-toxin connections in the workplace.[31]  They identified 87 original disease cluster reports that established disease-toxin connections in occupational medicine (from 1775 to 1990). They identified four advantages in using cluster reports from the workplace to identify new hazards: natural denominator boundaries, shared exposures, the ability to form intermediate hypotheses, and the possibility of locating comparable populations in which to study these hypotheses.=94 They stated that, =93because new products, intermediate products, and procedures are introduced into working environments faster than epidemiologic and toxicological studies can be designed to evaluate their potential risks, disease cluster investigations will remain central to the understanding of disease, and to protecting workers.[32]


VII.  THE INTERPRETATION OF EPIDEMIOLOGY IS SUBJECTIVE

As illustrated above, epidemiology is not necessary when making traditional medical conclusions regarding causes and effects. It is one piece of evidence among many. Epidemiology is an aspect of various disciplines that contributes to the general knowledge present in the field. It does not constitute a field of knowledge in and of itself:

Instead, a multitude of medical sciences and areas of practice embody epidemiologic problems. Hence, cancer epidemiology is a specialty within oncology, malformation epidemiology within teratology, health-care epidemiology within health-care administration, and so on. In this regard epidemiology is akin to morphology, for example - an aspect of various sciences and other fields as opposed to a science or other subject-matter in itself.[33]

Because of this, the interpretation of epidemiologic evidence is inherently subjective. This is true of any piece of evidence used to evaluate cause-effect relationships. A cause-effect analysis that rests solely on a definitive evaluation of the sum of epidemiologic evidence regarding a certain determinant-disease relationship will not do justice to the universe of non-epidemiologic evidence that is also available. Thus, despite all the studies on, for example, the effects of smoking, physical exercise, or diet on the risk of myocardial infarction, or on the efficacy of anticoagulant medication in the prevention of its recurrence, the opinions of even experts on these topics remain quite diverse and, hence, subjective.[34]


VIII.  HILL'S CRITERIA FOR DETERMINING CAUSALITY

As an overall model for determining causality, Hill's criteria are well accepted.[35]  They are: temporality, biologic gradient (dose-response), consistency, biologic plausibility, strength of association, analogy, experimental evidence, coherence, and specificity. All of Hill's criteria are subject to criticism, and only temporality is individually necessary to prove causality. Of Hill's nine criteria, epidemiologic data is only required for determination of strength of association.

Strength of association is a reflection of the strength of effect of a study. The relevance of strength of association is limited because the strength of an effect measured in a study is related in large part to the prevalence of other co-factors and not to the factor studied. Strength of association is not a measure of the importance of a particular factor in causation. While studies with large rate ratios are less likely to suffer from errors due to bias or confounding, it is important to note that weak causal associations are as likely to be causal and as important as are strong associations.[36]  In addition, a rate ratio of two is not required to establish that a factor contributed to a disease in a particular individual. For example chronic smoking of less than a pack a day induces less than a two fold increase in the risk of heart disease, nonetheless all physician's would state that smoking contributed to an individual smokers heart disease if he/she smoked at this rate. Epidemiologic studies can, when evaluated together provide more confidence in an association even in the absence of a statistically significant finding in any individual study. Consider for example five different polls that all indicate that a particular candidate for office is ahead by between two and three points all of which are within the sampling era of each individual poll. It would be reasonable to conclude that the candidate was going to win on a more likely than not basis.

The temporality criterion requires that the cause precede the effect. While this is generally relevant, conformity with temporality does not mean every study must evaluate this issue. In addition, there will be some cases where strict temporality is not necessary in order to evaluate etiologic relationships. As Weed states, It is interesting to note that, in general terms, causality need not require antecedence. Counter examples include simultaneous cause-effect relationships.[37]  Temporality is usually established through non-epidemiologic evidence. Tobacco companies argue correctly that the current body of epidemiology literature cannot distinguish temporality from a genetic link between the tendency to smoke and risk factors for cancer.

Conclusive support for the temporal relation of smoking and cancer is derived from molecular and animal data.

Biologic gradient is the existence of a dose-response relationship for the proposed cause-effect combination. A dose-response relationship is not always necessary in order to establish causation. This is true for several reasons:

First, it is possible for sufficient evidence to be amassed for an association to be considered causal without any form of dose-response relation being observed. As two notable examples, most epidemiologists found persuasive the early evidence of association between vinyl chloride and angiosarcoma of the liver, and between diethylstilbestrol and adenocarcinoma of the vagina, even though no dose-response relations were demonstrated. The acceptance of these two cancer-exposure relations was firmly established on case reports only. The presence of a dose-response pattern in epidemiologic data is, after all, partly a function of the opportunity to study such a pattern. Second, the interpretation of an apparent dose-response relation in the data must include the possible non-causal reasons for its appearance, such as confounding and other sources of bias. Hence, one might expect to see a non-causal dose-response relation between alcohol consumption and lung cancer due to a correlation between alcohol and smoking. Third, a dose-response curve reflects complex biological mechanisms and may take any form. For instance, there may be a `threshold' dose below which there is no effect or a flat portion along which all doses produce the same magnitude of effect =85 Finally, estimates of effect made in relative terms (i.e., with the rate ratio) may obscure or present the misleading appearance of a relation of dose to the absolute magnitude of response, which is measured by the rate difference.[38]

Consistency requires that a proposed effect be observed repeatedly under different circumstances. This criterion is useful, and it can be met by many different conditions and types of study. However repetition of findings under similar conditions is not necessarily supportive of this criterion.

Specificity requires that each cause have a single effect. This is rarely a useful criterion because many causes have multiple effects. Asbestos causes asbestosis, lung cancer mesothelioma and other cancers. Smoking causes heart disease, lung cancer, oral cancer, etc. Trauma from a car accident can cause many different injuries.

Biological plausibility (mechanistic understanding) asks if the theory of causation (mechanistic) fits known mechanisms of injury causation. But usually, doctors do not require a specific understanding of the underlying mechanism of an injury or disease before assessing causation. For instance, doctors still do not understand exactly how smoking causes cancer, yet we routinely attribute cancers to smoking. In addition, when an exposure may result in many different intermediate causes of an outcome, it is not necessary to know which particular mechanism caused the outcome, within a reasonable degree of medical certainty. For example, a bullet wound in the chest may damage many different organs and cause the death of the person who is shot. Damage to the heart, lung or aorta may have caused the death. Despite the lack of knowledge of a specific causal mechanism, no doctor would hesitate to state that the bullet wound caused the death. This is true, even though the dead individual may have suffered a heart attack at the exact time that he/she was shot. This and other possible error is the reason doctors' opinions are given to a reasonable degree of medical certainty.

Coherence asks if the causal theory is not inconsistent with what is already known of the injury or disease.

Experimental evidence can consist of laboratory studies, animal studies, controlled clinical trials, or observational pathology studies. This includes articles discussing mechanisms of injury production, whose authors use experimental evidence and observation as the basis for their conclusions. Animal studies are relevant to human inference. Animal studies are performed for application to human health, not to animal health. Animal studies are not conducted to determine risks to mice, rats, dogs or cats. They are not conducted out of concern for mouse or rat health. They are done because it is generally felt that inferences can be drawn from animal studies about human risks. If inferences are not to be drawn, there should be a specific justification for failure to do so since many studies indicate that illnesses in humans are reproduced in animals. Animal studies comprise the major criteria for Koch's postulates and thus have been the fundamental basis for medical epistemology since the 19th century. If negative animals studies are relevant, positive studies must also be relevant. It is for this reason that regulatory agencies including the NIH, FDA, EPA, NIOSH and the National Toxicology Program rely on, conduct and fund animal studies. Eighteen of 21 agents found carcinogenic in animals have also been found to cause cancer in humans. Of 23 human carcinogens 21 have produced some indication of cancer in animals.[39]  Animal models are used to test causation and treatments for a variety of non-cancer effects including drug side effects, teratogenicity, asthma, heart disease, and medical device testing.

Analogy asks if epidemiological and other studies have established that an environmental exposure analogous to the exposure being considered may cause diseases similar to those reported for the exposure being considered.

Hill noted, "None of my nine view points can bring indisputable evidence for or against the cause and effect hypothesis, and none can be required as a sine qua non." [40]

Hill's final emphasis placed responsibility on scientists for making causal judgments with available known facts. He recognized that decisions have to be made in the absence of perfect data noting, All scientific work is incomplete -- whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the action that it appears to demand at a given time.[41]


IX.  SILICONE BREAST IMPLANTS (SBI) AND DISEASE [42]

Sufficient information is available in the medical literature to establish that the use of SBI can cause, aggravate, or exacerbate certain diseases, signs and symptoms.

These diseases are defined as follows:

    a.  Capsular contracture

    b.  Granuloma formation

    c.  Atypical connective tissue disease (ATCD) characterized by some or all of the following symptoms including:fatigue, myalgia, arthralgia, memory loss, paresthesias, dry eye, dry mouth, dysphagia, photosensitivity, Raynaud's phenomenon, pleurisy, stiffness, adenopathy, joint swelling and others, and or signs including:

    d.  Enlarged or tender thyroid, enlarged parotid, abnormal Schirmer, telangectasia, carpal tunnel syndrome, or erythema of the chest wall and others and or, associated laboratory abnormalities include: elevated cholesterol, ANA, esr, IgG, IgA, IgM, RF, anti-Ro (SSA), anti-La (SSB), and anti-microsomal antibody and/or others.

    e.  Myeloma and Monoclonal gamopathy of unknown significance (MGUS)

    f.  Scleroderma

Using all types of available data, it is possible to evaluate the relationship between SBI and disease. This can be accomplished through the application of the Hill model and Bayesian process. This framework asks two questions. The first is, Does the substance/factor cause the effect in general? In other words, does SBI use, in general, cause and disease? This can be answered, by considering Hill's criteria for causation, and using those criteria to place the proposed causal factor somewhere on the line between 0 and 100 in the proposed model. The second question is, Has the substance/factor contributed to the effect in a particular individual? No guidelines to approach this second question are proposed here, but it is the subtext of any discussion of general causation.

I will now apply Hill's criteria for causation to the question of the association of SBI and disease.

1. Specificity requires that each cause have a single effect. This criterion is not relevant because SBI causes disease through local or distant immune mechanisms. These mechanisms would be expected to result in different patterns of signs and symptoms in different individuals. This is the pattern of disease in other rheumatologic and immune based diseases.

2. Temporality requires that the cause precede the effect. temporality is a sine qua non for SBI and disease. In the case of breast implants the strongest evidence of temporality comes from the improvement in symptoms generally seen after explantation. This improvement is gradual and is therefore less likely to be due to a placebo effect. Of the 517 patients who underwent explantation, 188 improved, yielding a 40.65% improvement rate. In the explant studies that reported the ANA titer levels pre- and post-explantation, 27 of the 48 patients had reductions in post explant ANA levels. For a summary of the studies of breast implant explanations, see Appendix 2.

3. Consistency requires that a proposed effect be repeatedly observed under different circumstances. This criterion is satisfied by the observation that similar signs and symptoms of disease in women with SBI have been independently observed by different researchers in different locations. Biomarkers of immune system dysfunction are consistently found in symptomatic SBI patients, although it is expected that not all SBI implanted patients will be abnormal for all or even most abnormalities. Patient's with SLE, SS and RA have differing patterns of biomarkers positivity. The epidemiological evidence of lab abnormalities in women with silicone implants are presented in more detail in Appendix 3. Appendix 4 reviews the case series further describing the constellation of abnormalities in support of this criterion.

The studies that failed to find a difference in symptoms need to be examined closely. For example, Peters failed to find a difference in ANA percentage between controls and SBI patients.[43]  However his control group had a 28% positivity rate, a rate five times higher than that reported in the general medical literature. The prevalence of 26 % positivity in the implant group is comparable to the rate reported by other authors in SBI patients and is elevated based on most (if not all) other published sets of rates of ANA positivity in general populations. In the Peters study, it is of interest that four of the SBI patients had CTDs (three atypical), while none of the controls did.

Research in animals is consistent with the findings in human data. Comparable studies of animals and humans exposed to silicone gel implants have found identical effects.[44]  A list of animal studies with findings supportive of the conclusions from human research is in Appendix 5.

4. Biologic gradient is the existence of a dose-response relationship for the proposed etiologic factor. In the case of implants dose is related to latency, bleed rate, rupture, explantation, and type of implant (saline vs. gel vs. type of gel vs. size of implant). It has been shown that the rate of ANA induction increased with increasing duration of implantation.[45]  Pleiderer and Garrido found that the concentration of silicone found in human livers increased with decreased implant integrity.[46]  Wolf et al found that women with ruptured implants had higher anti-silicone anti-body titers in women with ruptured implants compared with intact and unimplanted women.[47]  Cuellar et al found that auto-immune abnormalities increased with duration of implantation.[48]  See Appendix 6 for a summary of biologic gradient data.

5. Biological plausibility (mechanistic understanding) asks if the theory of causation (mechanistic) fits known mechanisms of injury causation. The mechanism of SBI induced disease is supported by the known mechanism of other immune diseases, the known biologic action of silicone, the pattern of laboratory abnormalities in humans and animals, the distribution of silicone in the body, the in vivo distribution of reactive silica in the body.

Silicone is known to induce interferon production in animals and to induce antibody production in animals.[49] [50] [51] [52]  Interferon levels are elevated in humans in response to inflammation. Silicone has been shown to induce inflammation and act as an adjuvant in humans and animals. Silicone and has been shown to migrate throughout the body. Human data indicates that chronic elevation of interferon can result in auto-immune disease, and central nervous system abnormalities. Therefore there is a plausible mechanism of disease induction by silicone.

Although the exact mechanism(s) of silicone-induced inflammation remains under investigation, much is known about silicone-induced disease. Multiple studies indicate that there is a probable specific silicone antigen or anti-silicone immune cause. However, even if the inflammation is non-specific, that is, there is no specific understanding of the exact mechanism, non-specific inflammation in and of itself can produce the symptoms noted in silicone breast implants patients and is responsible for inflammation, scarring in silicate induced disease and silicosis.

6. Coherence asks if the theory is not inconsistent with what is already known of the injury. SBI as a cause for disease is not inconsistent with what is known about the immune diseases and SBI effects or current epidemiologic data. There is no data inconsistent with the causal induction of ACTD by SBI in patients.

7. Experimental evidence can consist of laboratory or animal studies, controlled clinical trials, or observational studies. This includes articles discussing mechanisms of injury production, whose authors use experimental evidence and observation as the basis for their conclusions. Animal and other studies provide valuable evidence that silicones can cause disease, on the mechanism of disease production. The biologic activity of silicones in animals has been reproduced in several different species and has human parallels and its importance has been recognized for silicone since 1964. Silicones have induced cancer in rats and on the basis of this data, in 1964, the head of the NCI, environmental cancer section stated:

"The conclusion therefore is sustained that an indiscriminate use of Silastic for parenteral implantation in man is potentially dangerous and thus inadvisable on the grounds of medical ethics and the exercise of good judgment." [53][54]

Animal studies, in particular with respect to silicone gel implants, provide important information on biologic plausibility and mechanistic understanding. TABLE 3 (below) includes the areas of knowledge demonstrated by animal studies.