Bradford Hill and breast implant illness: no evidence for causal association with breast implants

KEYWORDS:

• Breast implant illness
• silicone breast implants
• autoimmune disease
• breast reconstruction
• breast augmentation

Dear Editor

With great interest, we read a recent literature review on Breast Implant Illness by Cohen Tervaert et al. in Expert Review of Clinical Immunology [1]. Breast Implant Illness, previously called Autoimmune Syndrome Induced by Adjuvants (ASIA), is discussed by the authors in light of an extensive list of symptoms that have been reported in women with silicone breast implants. Breast implant illness to date is a diagnosis per exclusionem for which no official evidence-based definition exists. The authors use the Bradford Hill criteria, developed for the assessment of causality of associations in epidemiology, to conclude that Breast Implant Illness should be regarded by the scientific community as being scientifically proven [2]. In our opinion, neither the scientific literature nor the Bradford Hill criteria point toward a causal association between Breast Implant Illness and silicone breast implants in such an unequivocal way as the authors assert. Below, we discuss a number of concerns with regard to the authors’ line of reasoning and their application of Hill’s framework for assessing causality in epidemiology.

The Bradford Hill criteria were developed in the late 1960s to assess the causality of associations in epidemiology and were originally used to demonstrate a causal link between smoking and lung cancer [2]. In modern epidemiology, the criteria strength, consistency, dose-response, and biological plausibility are still deemed valid and remain commonly used [3]. Cohen Tervaert et al. start out by considering the Strength of Association between silicone breast implants and both autoimmune diseases and breast implant illness. According to Hill, a causal association is more likely if the association between exposure and disease is stronger: a stronger association is less likely explained by bias or chance. In epidemiology, strength of association is objectively expressed as the magnitude of the relative risk (RR) or odds ratio (OR). Confidence intervals around the risk estimates as well as statistical significance are commonly used to sort out whether an association is likely to be real or due to chance. For breast implant illness, the association with silicone breast implants has not been formally assessed, as this constellation of symptoms does not have an evidence-based definition. According to Cohen Tervaert et al., this formal risk calculation may be omitted since ‘seasoned clinicians allied to increasingly health literate patients are better placed than randomised controlled trials to determine causality’ [4]. The authors argue that an association between silicone breast implants and ‘autoimmune-like symptoms’ has been found by various ‘cohort studies.’ However, the referred studies lack the required design to justify such conclusions since these are mostly case series and case reports without an appropriate comparison group. Although important for hypothesis generation, this kind of research cannot support epidemiologic conclusions. Yet, there are various relevant studies to be considered. For autoimmune diseases, the study by Watad et al. is the only one presenting a formal risk calculation that shows a significant difference between women with silicone breast implants (n = 24,651) and a comparison group (98,604) [5]. When including fibromyalgia in the category of autoimmune diseases, the authors observed a weak association, with an OR of 1.22 (95% CI 1.18–1.26) for ‘any autoimmune disease.’ However, for the majority of women, it was unclear whether the diagnosis was made before or after insertion of silicone breast implants. This precludes any causal conclusions according to Hill’s criterion of temporality, which states that an outcome has to follow an exposure in time [2]. A subgroup comparison between 1,797 women whose date of implant placement was known (and who were free of autoimmune disease at that time) and 7,109 women without implants showed a statistically significant hazard ratio of 1.45 (95% CI 1.21–1.73) for ‘any autoimmune disease’ associated with silicone breast implants. Some caveats apply to this study, however. The identified increased risk is small and can be easily explained by residual confounding, for example, by smoking. The authors adjusted for ever versus never smoking, but the duration and intensity of smoking (pack-years) were unknown. As smoking is a risk factor for several autoimmune diseases and women with silicone breast implants in the study smoked more often than women without implants, residual confounding is likely. Moreover, it is possible that women with breast implants have other characteristics that may increase the risk of autoimmune diseases, as we know that women with cosmetic breast implants are not comparable to the general population [6,7]. Besides, in terms of Hill’s criterion of Strength of association, a 45% increased hazard ratio represents a weak association, not a strong one. For rare conditions like autoimmune diseases, the implications of weak associations for absolute risk tend to be small.

Another criterion of Bradford Hill that is discussed in the article under scrutiny is the criterion of consistency; has an association been repeatedly observed by different investigators, in different populations, under different circumstances and during different time periods? Whereas the authors purport that this criterion is met because different studies on the same constellation of symptoms in relation to silicone breast implants have found similar results, in our opinion, the scientific literature on breast implant illness has been anything but consistent. A major limitation of the article by Cohen Tervaert et al. is the lack of a systematic evaluation of all available studies. The authors ignored several studies which reported on the risk of breast implant illness complaints and/or autoimmune diseases, but which refuted an association with silicone breast implants. Barbosa et al. found no statistically significant difference in the prevalence of nonspecific constitutional symptoms, rheumatoid arthritis and systemic lupus erythematosus, connective tissue disease, and allergic reactions between unselected women with silicone breast implants and propensity score matched controls without silicone breast implants [8]. Neither was a difference found in the prevalence of breast implant illness manifestations in a report by Miseré et al., who compared unselected women with silicone breast implants with healthy controls without silicone breast implants selected from friends and family [9]. In contrast to the conclusions of Miseré et al., Cohen Tervaert et al. refer to the latter study to support their claim that an association between silicone breast implants and autoimmune-like symptoms was actually found. Multiple systematic reviews have refuted the association between silicone breast implants and autoimmune diseases [10,11]. Moreover, a number of investigators have declared that there is not even sufficient evidence to support the existence of a disease entity called ‘Breast Implant Illness’ [12]. The same goes for the concept of ASIA, which Cohen Tervaert et al. uncritically accepted, while Ameratunga et al. wrote in 2017 that its diagnostic criteria were so ambiguous that ‘it did not appear to constitute a definable medical condition’ [13]. We must unfortunately conclude that Cohen Tervaert et al. tend to handle the scientific literature quite arbitrarily. They have selectively cherry-picked studies to support their arguments, ignoring reports that contradict them, and doing so occasionally misrepresented the conclusions drawn in the articles they refer to. Furthermore, in a proper systematic review, articles should always be assessed in terms of their methodological quality. Cohen Tervaert et al. liberally refer to reports without discussing potential biases, for example, the way in which patients were selected. This testifies to a lack of concern for completeness and methodological rigor and makes their conclusions ultimately unconvincing.

Some women with silicone breast implants experience systemic complaints that are commonly associated with rheumatologic diseases. Explanatory immunological mechanisms have been suggested but are barely supported by experimental evidence and therefore remain at the level of hypothesis [14]. To date, none of these theories have found broad acceptance. In the absence of proportionate organic abnormalities, it cannot be a priori excluded that breast implant illness might also be approached from a behavioral science point of view. Cohen Tervaert et al. themselves point out the striking similarities between breast implant illness and entities like fibromyalgia and chronic fatigue syndrome. Possibly, in these entities psychological and social mechanisms play a key role in the amplification of functional somatic symptoms [15,16]. Whatever the answer to this etiological puzzle may be, women with silicone breast implants deserve to be informed based on solid scientific evidence. Large epidemiologic studies with proper comparison groups will be crucial to prove or refute a causal association between silicone breast implants and breast implant illness. Members of the Dutch Breast Implant Associated Anaplastic Large Cell Lymphoma (BIA-ALCL) Research-Consortium are currently collaborating with the Dutch National Institute for Public Health and the Environment to establish such cohorts in multiple centers in the Netherlands.

Finally, we do agree with Cohen Tervaert et al. on the fact that women considering silicone breast implants have a right to proper counseling to make decisions based on their own values and priorities. Women who develop systemic complaints after implantation of silicone breast implants should receive a proper medical evaluation. If no other possible cause can be found, these women should have the option of explantation. Nevertheless, choices should always be based on realistic considerations of risks and benefits. Silicone breast implants and their purported adverse outcomes are a high-profile topic that can ever count on attention from the lay-press. Unfortunately, media attention is often asymmetrically focused on negative rather than positive outcomes. This means that overzealous conclusions based on misleading representations of the evidence can easily lead to alarming and sensationalist reporting in the media. Obviously, this causes undue distress in women with silicone breast implants, while most women experience benefits from reconstructive as well as cosmetic breast procedures with silicone breast implants [17,18]. It is the responsibility of every scientist to translate scientific knowledge in such a way to provide evidence-based risk estimations as well as an honest account of uncertainties caused by bias and/or lack of studies with sufficient power. When scientists discuss the possible side effects of silicone breast implants, they should always carefully formulate their conclusions realizing that any unfounded conclusions may unnecessarily lead to substantial concerns among women with silicone breast implants.

Editor disclosures

One of the senior editors for this journal has served as a medical expert for the United States government, in the US vaccine court, in defence against autoimmune/autoinflammatory syndrome induced by adjuvants (ASIA). He thus recused himself and held no decision-making powers on this or related articles. The decision-making editor for this manuscript has no relevant financial relationships or otherwise to disclose.

Acknowledgments

The Dutch Breast Implant Associated Anaplastic Large Cell Lymphoma (BIA-ALCL) Research-Consortium includes Mintsje de Boer, MD, PhD, Department of plastic and reconstructive surgery, Maastricht University Medical Centre, Daphne de Jong, MD, PhD, Professor of Haematopathology, Cancer Centre Amsterdam, Amsterdam, Hinne Rakhorst, MD, PhD, Department of Plastic and reconstructive surgery, Medisch Spectrum Twente, Enschede, Flora E. van Leeuwen, PhD, Professor of Epidemiology of Cancer, Netherlands Cancer Institute Antoni van Leeuwenhoek, Amsterdam, Marc A.M. Mureau, MD, PhD, Professor of Oncological Reconstructive Surgery, Erasmus MC Cancer Institute, University Medical Center Rotterdam, René R.W.J. van der Hulst, MD, PhD, Professor of Plastic and Reconstructive Surgery, Maastricht University Medical Center, Maastricht, Jan Paul de Boer, MD, PhD, Department of Medical Oncology, Netherlands Cancer Institute Antoni van Leeuwenhoek, Amsterdam, and Jonathan Spoor, MD, Department of Epidemiology, Netherlands Cancer Institute Antoni van Leeuwenhoek, Amsterdam, The Netherlands

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.