by John H. Ferguson MD
From the Therapeutics and Technology Assessment Subcommittee, American Academy of Neurology, St. Paul, MN.
Taken from Neurology 1998;50:849-852. Permission granted by Lippincott Williams & Wilkins.

At a time when the only evidence for any association of silicone breast implants with disease consisted of case reports and case series, the weakest form of scientific evidence to prove an association,[1] large awards were won in court because of the supposed toxicity of these devices.[2] Because of a purported association of these implants with neurologic disease based primarily on case series without controls, this issue came to the attention of the American Academy of Neurology.[3]

Before the current issue of Neurology, there had been no published epidemiologic studies of neurologic illness related to breast implants. However, there have been several retrospective cohort studies testing the hypothesis that silicon breast implants are associated with various autoimmune disorders.[4-6]

These studies to date have been largely negative, showing no clear relation between breast implants and any disease. At last, two population-based studies addressing the issue of breast implants and neurologic disease have arrived from outside the United States and appear in this month's issue of Neurology.[7],[8] They should help to resolve the issue of the alleged association between breast implants and neurologic disease. These studies from two Scandinavian countries are remarkably similar in structure and results. Both used their national hospital discharge registry databases to find women who had had breast implants. Both used the same registries to find a very interesting and unique comparison group of women: those who had had breast reduction surgery. Both studies found very low rates of neurologic diagnoses in these cohorts (1%). However, both also found that the rates of neurologic diagnoses in the breast implant cohorts were similar to the rates in the breast reduction groups. In other words, in these two studies, one from Denmark and one from Sweden, the risk for developing a neurologic disease associated with having breast implants did not differ from the risk of developing a neurologic disease associated with having breast reduction surgery. Of interest and perhaps also of importance, in both countries the breast implant and breast reduction cohorts had a somewhat higher rate (although not statistically significant) of neurologic diagnoses than the calculated expected rates for hospitalized women in general. However, validation of observed neurologic diagnoses in the breast implant cohorts by chart review in both countries revealed misclassified cases or cases where the diagnosis was present at the time of surgery. Correcting for these brought the observed and the expected rates of neurologic diagnoses in the breast implant cohorts very close together.

There were some differences in the studies. Winther et al., of Denmark,8 reviewed records of breast implant cases that had ICD-8 codes of symptoms that could have been secondary to neurologic disorders, such as memory or visual complaints, balance disturbances, and back pain. Of the 12 patients found in this manner, only 1 chart indicated a primary neurologic disorder. It appears that this one case was not used in the overall analysis. Furthermore, Winther et al. did not attempt to verify all the neurologic diagnoses in the much larger breast reduction cohort, only those that showed an excess over expected numbers in the breast implant cohort. These data were unfortunately not shown. Winther et al. also found that women in both cohorts were hospitalized more often than Danish women in general both before and after the cosmetic surgery. In contrast, the cohort of women with breast implants studied by Nyren et al.7 of Sweden totaled 7433, 61/2 times the size of the Danish cohort. Additionally, this cohort was broken down into those having reconstruction secondary to breast cancer surgery and those having implants for other reasons, mainly cosmetic. Nyren et al. also chose the comparison breast reduction cohort (of 3,351) from a much larger group (33,716) that had breast reduction surgery, and matched them to the breast implant cohort for age within 5 years, hospital, and calendar year of operation. (Winther et al. did not match the two cohorts as closely, but state that the median age of surgery was 31 in both implant and breast reduction groups and that their reproductive patterns were similar.) Nyren et al. verified all the neurologic diagnoses in both cohorts by chart review and, especially in the multiple sclerosis diagnostic category, found cases that were misclassified or present at the time of surgery. However, in contrast to Winther et al., Nyren et al. made a direct analytic comparison of the breast implant and breast reduction cohorts after removal of the misclassified or preexisting cases.

How good are these studies? Some of the following concerns have been discussed by the authors. First, both of these studies were of women who were hospitalized, not outpatients. This is a strength in that diagnoses may be more specific and better documented. It could, however, represent a more restricted sample (based on selection bias) and thus miss patients with more subtle or more common but conceivably important neurologic symptoms such as headache, fatigue, muscle ache, and memory loss, all of which have been claimed in some reports as related to breast implants.[9] Such common symptoms are, unfortunately, components of other syndromes that have been difficult to categorize, such as those purported to be related to the Gulf War.[10],[11]

Second, although it seems intuitive that the chosen comparison group-women having breast reduction surgery-is an appropriate one in that both groups are seeking cosmetic improvements and may, therefore, be similar in other ways, there could be some characteristics of this retrospective choice that biases the results (again, selection bias).[12] A recent study has indicated that women seeking breast implants were different from other women in several interesting ways. Women with breast implants were more likely to drink more alcohol, use oral contraceptives, use hair dyes, be younger at first pregnancy and at first birth, have a history of terminated pregnancies, and have a greater lifetime number of sexual partners compared to women who did not have breast implants. They were also less likely to be overweight.[13] Because of the higher hospitalization rates in both cohorts, Winther et al. propose that both groups may seek medical care more frequently than other women. A more appropriate comparison group may be age- and otherwise related matched controls who have not had augmentation mammoplasty. This would obviously be a more difficult group to identify using a hospital database.

Third, as the authors themselves have pointed out, they were limited in their search for neurologic diagnoses by existing ICD-8 codes (some ICD-9 in the Swedish study) and might therefore miss more recent diagnostic categories such as chronic inflammatory demyelinating polyneuropathy. The completeness of the chart coding and the codes themselves are clearly factors that might limit the accuracy and comprehensiveness of the diagnoses. Although the comparison group was similarly limited, more subtle or uncoded illness might go undetected by these methods.

Fourth, chart review was only performed for the observed cases of neurologic diagnoses. Some random sample of record review of a portion of other charts in each cohort may have been valuable in assessing the completeness of documenting neurologic diagnoses.

Fifth, it is conceivable that the follow-up times of approximately 8 to 10 years in these studies were not long enough to show an effect. In addition, funding for both studies was in part provided by one of the implant manufacturers. Some may use this fact in an attempt to discredit such studies. However, in the studies that are the subject of this editorial, these data are derived from very reliable resources in countries not as affected by litigation mania as the United States, and from patients for whom the majority of surgical procedures and neurologic diagnoses predated the publicity created by the 1992 FDA [14] action and the ensuing flood of litigation.[15] Although one might debate interpretation, these studies and data can stand on their own.

These are the first epidemiologic studies with a "control" group as a comparison to test the hypothesis that breast implants are associated with neurologic disorders. The breast implant and breast reduction groups do not differ in their risks for neurologic diagnoses. Even though both groups in both studies seemed to have a slightly higher rate of neurologic diagnoses compared to women in general, they did not differ from each other. It seems reasonable to conclude that the neurologic disorders found in women with breast implants in these studies are not associated with the breast implants per se. Is this proof that breast implants are not associated with neurologic disorders? No, but it is the best data presently available, and the data do not support an association.

These studies would be much more difficult in the United States for a number of reasons. The populations Sweden and Denmark are likely more homogenous than those of the United States. In Sweden and Denmark there is a national health care system with relatively uniform health-related data-bases covering the majority of the populations described in these reports. Thus, it would be easier to retrieve data on all women or nearly all women who had breast implants or breast reduction surgeries in both countries than it would be in the United States with its hodgepodge of private, state, and federal health care insurance systems. In the United States we have no country-wide hospital data systems except for the Medicare population, which is not the most suitable group for a study of breast implants. Because of their national health database systems, these studies from Denmark and Sweden are more nearly population-based and the results more likely to be reliable despite some of the potential draw-backs outlined previously.

The silicone breast implant issue highlights a basic problem in our litigation-prone society. Injury, or presumed injury from industry-produced materials and products, will often be addressed first in the courts[16] before appropriate scientific (epidemiologic) data are available to prove or disprove an association, much less causation. Courts must make a decision for the plaintiff or the defense and cannot wait for "further studies."[17] Sometimes these class action lawsuits have presaged injuries or illnesses associated with exposure to a product that later was upheld by medical science, as was the case with asbestos and mesothelioma, and prenatal diethylstilbestrol (DES) and vaginal cancer.16 In many others, however, such as Agent Orange, Bendectin, and certainly silicone breast implants, the science has never supported the plaintiffs' claims, yet juries have continued to award large sums of money in class action settlements which in the case of breast implants has resulted in the declaration of bankruptcy of one of the manufacturers. [18] Although the toxic tort cases may be some of the most egregious, media-highlighted, and costly examples of the misuse and abuse of science in the courtroom, they are by no means the only examples. [19], [20]

Are there lessons here for neurologists in particular and for clinical medicine in general? We may disdain a legal and court system that uses adversarial expert witnesses, "hired guns" for either side, in the adjudication of a dispute with the result that good science and scientific principle are shredded, ignored, or distorted in the process. Yet often it is our case reports or case series presented in court that are the basis for these claims. To the extent that we as medical practitioners report our anecdotal experiences as primary witnesses, or report the medical literature comprised of cases or case series as expert witnesses, we may be part of the problem. As clinical neurologists, we have all seen impressive examples of patients whose illnesses seemed to result from some exposure or whose illnesses seemed to respond dramatically to a particular treatment. We are impressed with these cases; they are embedded in our practice experience, but they are still anecdotal. They be-and should be-used to generate hypotheses about association or causation but they are not enough to infer association, much less causation, for an entire similarly exposed population. Data is not the plural of anecdote.

The kinds of studies necessary to link an environmental exposure to disease were outlined over 30 years ago by Hill. [21] Unfortunately, these studies are seldom done prospectively, and in general plaintiffs appear in court before any appropriate observational (usually retrospective) studies have been performed. This problem was highlighted by the case of Bendectin, a drug used for nausea and vomiting during pregnancy, in which the outcome fortunately will enhance the appropriate use of science in the courtroom.

Numerous studies have shown no relationship of Bendectin to fetal limb malformations, a dispute which finally reached the Supreme Court in a landmark case, Daubert v Merrell Dow, concerning what is admissible as scientific evidence in court. [22], [23] The result of this decision has been, in effect, to strengthen the use of appropriate science in court proceedings by urging federal trial court judges to follow sound scientific principles on what is allowable scientific evidence in court. The Daubert decision has been strengthened by the recent Joiner decision by the Supreme Court to uphold the Federal Trial Judges' gatekeeping role regarding the admissibility of scientific evidence. [24] Impressive efforts are under way to educate judges in their gatekeeping role. [25]

Furthermore, with the more frequent use of court-appointed experts26-even expert panels-as opposed to adversarial expert witnesses, we are more likely to have cases adjudicated by the appropriate use of relevant science and scientific principle rather than "junk science." [26] In fact, in several recent breast implant cases, judges have appointed neutral expert panels to address the scientific base of these complaints, and the panels have concluded that there was no evidence to support the plaintiffs' charges. [27]

The courts and the legal system may dismay, baffle, irritate, and provoke us as physicians. On the other hand, they may provide us with an opportunity. We should learn to use not only our patients but also the courts' decisions involving medical and health science issues to generate hypotheses that may be answered by appropriately conducted scientific studies. The reports in this month's Neurology are examples of such studies-probably undertaken indirectly as a result of court decisions in the United States implicating breast implants in neurologic disorders-and offer the best epidemiologic data to date of the lack of association between breast implants and neurologic disease.

Acknowledgment
The author gratefully acknowledges the helpful critique of Douglas Weed, MD, PhD, Chief of Preventive Oncology, National Cancer Institute.

References
[1] Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology. The essentials. Baltimore: Williams and Wilkins, 1996.

[2] Angell M. Science on Trial. New York: W.W. Norton, 1996.

[3] "Ferguson JH. Silicone breast implants and neurologic disorders: report of the Practice Committee of the American Academy of Neurology." Neurology 1997;48:1504-1507.

[4] Gabriel SE, O'Fallon W, Kurland LT, Beard CM, Woods JE, Melton LJ. "Risk of connective-tissue diseases and other disorders after breast implantation." New England Journal of Medicine 1994;330:1697-1702.

[5] Hennekens CH, Lee I-M, Cook NR, et al. "Self-reported breast implants and connective-tissue diseases in female health professionals. A retrospective cohort study." JAMA 1996;275:616-621.

[6] Sanchez-Guerrero J, Colditz GA, Karlson EW, Hunter DJ, Speizer FE, Liang MH. "Silicone breast implants and the risk of connective-tissue diseases and symptoms." New England Journal of Medicine 1995;332:1666-1670.

[7] Nyren O, McLaughlin JK, Yin L, et al. "Breast implants and risk of neurologic disease: a population-based cohort study in Sweden." Neurology 1998;50:956-961.

[8] Winther JF, Bach FW, Friis S, et al. "Neurologic disease among women with breast implants." Neurology 1998;50:951-955.

[9] Ostermeyer-Shoaib B, Patten BM, Calkins DS. "Adjuvant breast disease: an evaluation of 100 symptomatic women with breast implants or silicone fluid injections." Keio J Med 1994;43:79-87.

[10] "The Persian Gulf experience and health. NIH Technology Assessment Workshop Panel." JAMA 1994;272:391-396.

[11] Hyams KC, Wignall FS, Roswell R. "War syndromes and their evaluation: from the U.S. Civil War to the Persian Gulf War." Ann Intern Med 1996;125:398-405.

[12] Gordis L, Epidemiology. Philadelphia: W.B. Saunders, 1996.

[13] Cook LS, Daling JR, Voigt LF, et al. "Characteristics of women with and without breast augmentation." JAMA 1997;277:1612-1617.

[14] Kessler DA. "The basis of the FDA's decision on breast implants." N Engl J Med 1992;326:1713-1715.

[15] Angell M. "Do breast implants cause systemic disease? Science in the courtroom." N Engl J Med 1994;330:1748-1749.

[16] Jasanoff S. Science at the Bar. Cambridge, MA: Harvard University Press, 1995.

[17] Ferguson JH. "Interpreting scientific evidence: comparing the National Institutes of Health Consensus Development Program and courts of law." The Judges Journal 1997;36:21-25.

[18] Angell M. Shattuck lecture. "Evaluating the health risks of breast implants: the interplay of medical science, the law, and public opinion." N Engl J Med 1996;334:1523-1518.

[19] Ferguson JH, Dubinsky M, Kirsch PJ. "Court-ordered reimbursment for unproven medical technology: circumventing technology assessment." JAMA 1993;269:2116-2121.

[20] Huber PW. Galileo's revenge: junk science in the courtroom. New York, NY: Basic Books, 1991.

[21] Hill AB. "The environment and disease. Association and causation." Proceedings of the Royal Society of Medicine 1965;58:295-300.

[22] Gold JA, Zaremski MJ, Lev ER, "Shefrin DH. Daubert v Merrell Dow: The Supreme Court tackles scientific evidence in the courtroom." JAMA 1993;270:2964-2967.

[23] Blackmum J. "Daubert v Merrell Dow Pharmaceuticals Inc." Supreme Court of the United States 1993;92-102:1-17.

[24] "Editorial. Junk science in the courtroom." Washington Post. Washington, DC, Dec. 18,1997:20.

[25] Zweig FM. "Judges causation science sourcebook. Science in the court: finding your way through mass toxic torts." Bethesda, MD: Einstein Institute for Science, Health and the Courts, 1995.

[26] Marshall E. "New York courts seek 'neutral' experts." Science 1996;272:189.

[27] Charrow RP. "Courting science: rational decision making in an irrational environment." The Journal of NIH Research 1997;9:56-58.

Received January 22, 1998. Accepted in final form January 22, 1998. Address correspondence and reprint requests to Dr. John H. Ferguson, Chairman, Therapeutics and Technology Assessment Subcommittee, American Academy of Neurology, 1080 Montreal Ave., St. Paul, MN 55116.
NEUROLOGY 50 April 1998
Copyright (c) 1998 by the American Academy of Neurology. All rights reserved Published by Lippincott Williams & Wilkins

---

FRONTLINE / WGBH Educational Foundation / www.wgbh.org

web site copyright WGBH educational foundation