Cancer incidence in a cohort of Ontario and Quebec women having bilateral breast augmentation
The possibility that women, who receive breast implants for cosmetic purposes, have increased long-term risks of developing cancer continues to be debated. The objective of our study was to prospectively examine cancer incidence among women who received breast implants. A cohort was assembled of 24,558 women, 18 years of age and older, who underwent bilateral cosmetic breast augmentation, and 15,893 women who underwent other cosmetic procedures in Ontario or Quebec between 1974 and 1989. These plastic surgery patients were selected from the same clinics as the implant population. Incident cancers were identified by linking to Canadian registry data up to December 31, 1997. In total, 676 cancers were identified among women who received breast implants compared to 899 expected based on general population rates (standardized incidence ratio (SIR) = 0.75; 95% confidence interval (CI) = 0.70–0.81). Overall cancer incidence rates among women who received breast implants were similar to that of the other plastic surgery patients (relative risk (RR) = 0.91, 95% CI = 0.81–1.02). However, women who received breast implants had lower breast cancer rates than the plastic surgery patients (RR = 0.64, 95% CI = 0.53–0.79). No increased risks were observed among the implant population for any of the other cancer sites examined. Comparisons involving only women who received breast implants found no association between long-term breast cancer incidence and implant site (submuscular vs. subglandular), fill (saline vs. silicone) or envelope (polyurethane-coated or not). In conclusion, women undergoing cosmetic breast augmentation do not appear to be at an increased long-term risk of developing cancer. © 2005 Wiley-Liss, Inc.
Millions of women worldwide have received silicone breast implants for cosmetic or medical reasons, but there remains considerable controversy about their long-term health effects, particularly for cancer. In the United States, it has been estimated that only 20% of breast implants have been used for reconstruction or other medical purposes, while the remaining 80% for breast augmentation.1 Silicone breast implants include those that are either filled with silicone gel, or inflatable saline implants with an outer shell made of a silicone elastomer. Reports of adverse reactions led the U.S. Food and Drug Administration (FDA) to remove silicone gel-filled implants (SGFI) from the market in 1992. The use of these implants has since been restricted to patients with breast cancer and certain other conditions enrolled in strictly controlled clinical trials. Inflatable saline implants remain in widespread use.
Although silicone in animals has generally not been associated with increased risks of cancer,2, 3, 4 silicone gel has been shown to be immunogenic,5 and to induce plasmacytomas in genetically predisposed mice.6 The site of implantation has led most cancer investigations to focus on breast cancer incidence or mortality. A number of epidemiological studies, including both cohort7, 8, 9, 10, 11, 12 and case-control designs,13, 14, 15 have consistently demonstrated that women who receive implants for augmentation purposes are at low or average risk of developing breast cancer. In 1999, the National Institute of Medicine and the International Agency for Research on Cancer concluded that there is epidemiologic evidence against a carcinogenic effect of silicone implants on the breast.1, 16
Relatively few studies have evaluated the relationship between breast implants and the incidence or mortality of non-breast cancers. The findings from these studies have been largely negative, though excesses have been inconsistently reported for some types of cancer including the stomach, brain, lung, vulva, liver and cervix.17, 18 There has also been some concern that breast implants may increase the risk of multiple myeloma,19 particularly, among younger women (<45 years of age).20 Taken as a whole, the epidemiological evidence for risk of cancer at other body sites has been deemed inconclusive.16
Previously conducted epidemiologic investigations of breast implants and cancer are subject to some important limitations. Most studies have used relatively small populations, and therefore, lacked the statistical power needed to characterize cancer risk at sites other than the breast. The small sample sizes have also resulted in a limited ability to conduct subgroup analyses, including an evaluation of variations in risk according to implant or surgery characteristics. Other identified shortcomings of past research include an inadequate control group, and an insufficient amount of follow-up time. Extended follow-up time is particularly relevant given that immunologic changes and deterioration of the implant capsule may only occur over an extended period of time,21, 22 and moreover, the existing literature suggests that the expected latency period between first exposure to carcinogenic agents and solid tumor development is lengthy.16
To provide further insight about long-term cancer risks associated with breast implants, we report herein results from a cohort study of women recruited in 2 Canadian provinces. This is the largest breast implant cohort study conducted to date with some women followed for up to 24 years. As such, the accumulated person–years of follow-up, and hence, the identified number of incident cancer cases provide the statistical power needed to evaluate variations in breast cancer risk for different breast implant characteristics, and to evaluate differences in rates for rarer forms of cancer.
Material and methods
Study design and subjects
The study population consisted of women, 18 years of age and older, who were residents of either Ontario or Quebec, and who underwent bilateral cosmetic augmentation mammoplasty in their province of residence, between January 1, 1974 and December 31, 1989. A similarly aged control cohort that consisted of women who received other common elective cosmetic surgeries was also assembled. These eligible cosmetic procedures, like breast augmentation, were not billable to the publicly funded health insurance plans of Ontario or Quebec. The cosmetic procedures included: chemical peel or dermabrasion, coronal brow lift (eyebrow and forehead lift), otoplasty (ear surgery), rhinoplasty (nose surgery), rhytidectomy (face lift) or blepharoplasty (eyelid surgery). Plastic surgery patients were frequency matched to the breast implant recipients by year of entry into the cohort, and by surgeon. While the design called for a two-to-one implant to other plastic surgery patient matching ratio, in practice, there was a slight departure from this ratio. This is due to the fact that recruitment was done based on a priori expected case numbers, and some women who received breast implants did not fulfill the requisite inclusion criteria. Because the study was undertaken primarily to evaluate the long-term cancer risks associated with SGFIs, we excluded from the other plastic surgery patient group those women who had received other types of silicone or artificial implants.
In Ontario, eligible subjects were identified by trained abstractors through the office records of plastic surgeons who had ever performed augmentation mammoplasty in the province between 1974 and 1989. Pilot studies revealed that the surgeons who performed augmentation mammoplasty also performed most of the control procedures, thus allowing other plastic surgery patients to be selected from the same clinics. Over the study period, it was estimated that 133 plastic surgeons had ever performed bilateral augmentation mammoplasty; nearly three-quarters of these procedures were performed by 20 plastic surgeons.
In Quebec, breast implant recipients and other plastic surgery patients were identified from the MED-ECHO files23 and from records of plastic surgeons in private plastic surgery clinics for the period between January 1, 1974 and December 31, 1989. The MED-ECHO database is a computerized system that describes all hospital discharges that occur in Quebec and includes both in-patient stays and day surgeries. The implant and other plastic surgery cohorts consisted of women who were operated on by approximately 100 surgeons who have practiced in Quebec since 1974.
The follow-up of women who received breast implants started from the calendar date of their index surgery. Excluded from both the implant and control cohorts were women who had undergone any previous major breast surgery, including reduction mammoplasty, breast lift and breast cancer surgery. We also excluded women who had a male genotype, or who had a history of cancer (excluding non-melanoma skin cancer) of any site before breast implant surgery. Finally, we excluded 65 women whose index bilateral breast augmentation surgery involved different implants for the left and right breasts.
In total, the cohort consisted of 40,451 eligible women, of whom, 24,558 received breast implants (7,153 women from Ontario and 17,405 from Quebec). The control group consisted of 15,893 women (4,418 from Ontario and 11,475 from Quebec). In both provinces, detailed information was collected about the surgical procedure and implant characteristics including: location of clinic, type of procedure, date of surgery, type of implant, site of implantation, fill volume, laterality and date of discharge. Personal identifying information was also collected to link patients to the mortality and cancer incidence data. This information included name (surname, given and maiden names), birth date, residential address and health insurance number. No data were collected from the women on individual-level risk factors for cancer (e.g., smoking, family history and reproductive history).
Ethics approval for the study was granted by the University of Toronto's Office on Research Studies, the ethics committee of Saint-Sacrement Hospital and the Ethics Committee for Clinical Research of Laval University.
The mortality experience of the cohort was ascertained by linking the cohort to the Canadian Vital Statistics Death database using a probabilistic procedure referred to as the Generalized Record Linkage System (GRLS).24 GRLS compares common fields in the 2 files to be linked, assigns weights to the resulting links and calculates a total weight. Links with a sufficiently high weight are accepted as a match. The Canadian Vital Statistics Death database, maintained by Statistics Canada, contains death data for all Canadian residents from 1950 onwards. The utility of linking epidemiological data to these databases has been demonstrated through a number of studies.25, 26, 27
After the mortality linkage was completed, GRLS was used to identify incident cases of cancer diagnosed among cohort members from 1969 until December 31, 1997. This was accomplished through record linkage of the personal identifying information of the cohort to records in the Canadian Cancer Registry. With the addition of Quebec registration data to the national registry in 1970, Canada is one of the few nations whose entire population has been covered by cancer registration.28 This computer-accessible, population-based national registry is managed by Statistics Canada through collaboration with provincial and territorial cancer registries. We linked the cohort to the cancer incidence data before the index date of surgery to exclude from analysis those individuals who had been diagnosed with cancer before their surgery was performed.
Probabilistic record linkage was done using personal identifying information that included surname, surname at birth, given names, birth date, social insurance numbers and place of residence at the time of the index surgery. Individuals with no match to the cancer incidence and mortality databases were assumed to be cancer-free and alive, respectively, as of December 31, 1997. The results of previous record linkage mortality studies suggest that the number of deaths that would be missed would be quite small given the personal identifying information available for the cohort.29 Given the high quality and completeness in the Canadian Cancer Registry, it is not unreasonable to assume the vast majority of incident cancer cases would be detected. Nonetheless, it is important to note that some incident cases may have been missed due to some underreporting of cancer cases in Quebec before 197828 as well as an inability to identify cases that might have occurred among women who have moved outside of Canada. Despite this, we know of no reason why incomplete ascertainment would be differential between the case and control population, a condition necessary to bias risk estimates generated from the internal cohort comparisons.
The person–time of follow-up was calculated for each woman from 1 year after the date of surgery until the earliest of: date of cancer diagnosis, date of death or December 31, 1997. The number of person–years and incident cases of cancer were tabulated across strata defined by: implant or other plastic surgery patient, province of residence at time of implant (Quebec or Ontario), attained age (18–24, 25–29, 30–34, …, 75–79, ≥80), calendar period of follow-up (1974–1977, 1978–1981, …, 1994–1997), period of surgery (1974–1977, 1978–1981, 1982–1985, 1986–1989), age at surgery (<35, 35–44 and ≥45) and time since surgery (1 – <5, 5 – <10, 10 – <15, and ≥15 years). Attained age, calendar period and time since surgery were time-dependent variables whose values changed over the course of the follow-up. The DATAB module of the Epicure software program30 was used to tabulate the person–years and incident number of cancers within each stratum.
We compared cancer incidence rates in the breast implant and surgical control cohorts to the general population by using provincial rates obtained from the Ontario and Quebec cancer registries. More specifically, the expected number of incident cancers in this cohort were obtained by multiplying the tabulated person–years of follow-up by the corresponding cancer incidence rates observed in the general population according to province (Ontario or Quebec), age (by 5-year age intervals) and calendar period (1974–1977, 1978–1981, …, 1990–1993 and 1994–1997). Differences in incidence rates between the implant and surgical control cohorts relative to the general population were formally evaluated by calculating the standardized incidence ratio (SIR), which is the ratio of the observed-to-expected number of incident cancers. The 95% confidence intervals (CIs) for each SIR were calculated by assuming that the observed cancers followed a Poisson distribution, using formulae detailed elsewhere.31
Comparisons of the implant recipients to the other plastic surgery patients, rather than the general population, has been recommended because women who undergo cosmetic surgery have been shown to differ in important ways from the general population.32 These internal cohort comparisons were performed using multivariable Poisson regression models. To evaluate whether these relative risks (RRs) differed by province, a test of homogeneity was conducted by including in the regression model a first-order interaction term of province and implant status. The 2 provinces were deemed to have different risk estimates if the interaction term was found to be statistically significant based on a two-tailed alpha of <5%.
We also performed more detailed analyses of breast cancer incidence rates by restricting analyses to include only those women who received breast implants. These analyses were conducted to identify possible differences in breast cancer risk according to the following implant characteristics: type of fill, type of envelope, site of implantation (subglandular or submuscular) and fill volume. For implant fill volume, women were categorized based on the quartiles of the observed frequency distribution of the mean value of the right and left implants (<175, 175–<200, 200–<225, and ≥225 cc).
All Poisson regression analyses were conducted using the SAS software package.33 The SAS procedure GENMOD was used to calculate the RRs of incident cancers and their corresponding 95% CIs. The RRs were adjusted for linear and quadratic age components, province of residence and calendar period. The potential confounding influence of the following factors were also examined: age at surgery, year of surgery and time since surgery. This was accomplished by examining the individual changes in the RR with and without these terms in the model.
A listing of reasons for excluding women from analyses, along with the accompanying frequencies, is presented in Table I. A total of 366,608 and 238,441 person–years of follow-up were accrued in the implant (n = 24,558) and the other plastic surgery (n = 15,893) cohorts, respectively (Table II). The median age at surgery for women who received breast implants was 31.3 years; the corresponding figure for the other plastic surgery patients was 31.9 years. Over half of the implants were received by women between the ages of 25 and 34 years. More than 70% of the women in each of the implants and other plastic surgery cohorts were followed for at least 10 years. On average, the length of follow-up was longer among Quebec implant recipients (15.4 years), relative to their Ontario counterparts (13.8 years). The most common procedures performed among the other plastic surgery patients were: rhinoplasty (48.8%), blepharoplasty (23.5%), otoplasty (10.9%), chemical peel (10.3%), rhytidectomy (5.4%) and coronal brow lift (1.1%).
Table I. Ineligible and Excluded Records, and Patients Included in The Analysis, Canadian Breast Implant Cohort Study
|Initial cohort size||24,897||16,591|
| Missing age at surgery||46||1|
| Age less than 18||84||625|
| Had prior major breast surgery||40||4|
| Had other silicone implants||20||1|
| Ever received doses of silicone||3||2|
| Diagnosis of cancer prior to index surgery||66||62|
| Had a male genotype||8||1|
| Reside outside of Ontario or Quebec at the time of surgery||7||2|
| Different types of implants were used by laterality||65||Not relevant|
|Patients included in analysis||24,558||15,893|
Table II. Frequency Distribution for Selected Characteristics of Women Who Received Breast Implants and Women Who Received Other Surgeries, by Province, Canadian Breast Implant Cohort Study
|Age at surgery|
| 18 to <25||1,160||16.2||952||21.6||2,505||14.4||2,529||22.0||3,665||14.9||3,481||21.9|
| 25 to <30||1,850||25.9||825||18.7||4,111||23.6||2,239||19.5||5,961||24.3||3,064||19.3|
| 30 to <35||1,934||27.0||759||17.2||4,934||28.4||2,069||18.0||6,868||28.0||2,828||17.8|
| 35 to <40||1,172||16.4||618||14.0||3,023||17.4||1,739||15.2||4,195||17.1||2,357||14.8|
| 40 to <45||588||8.2||494||11.2||1,480||8.5||1,068||9.3||2,068||8.4||1,562||9.8|
|Year of surgery|
|Length of follow-up (years)|
| 1 to <5||48||0.7||47||1.1||125||0.7||91||0.8||173||0.7||138||0.9|
| 5 to <10||1,767||24.7||1,190||26.9||2,419||13.9||1,122||9.8||4,186||17.1||2,312||14.6|
| 10 to <15||2,657||37.2||1,836||41.6||5,920||34.0||3,976||34.7||8,577||34.9||5,812||36.6|
|Person–years of follow-up1||98,455||–||57,676||–||268,152||–||180,765||–||366,607||–||238,441||–|
The majority of women (80%) received implants that were filled with silicone (Table III). Among the implant cohort, few Ontario women received breast implants that were coated in polyurethane (<1%), whereas 14% of women in Quebec received such implants. Virtually, all of these polyurethane-coated implants were received between 1985 and 1989. The site of implantation for 56% of women in both Quebec and Ontario was subglandular. The breast implant volumes were significantly higher in the Ontario series of women who received breast implants when compared with that of their Quebec counterparts. For 43% of Ontario women, the average implant fill volume was at least 225 cc, whereas, only 15% of Quebec women had average implant fill volumes of this magnitude. Among women who received breast implants, there were statistically significant differences between provinces, as measured by the χ2 test statistic, for type of implant, implant fill volume and site of implantation.
Table III. Frequency Distribution for Selected Breast Implant Characteristics (AT Index Surgery), by Province, Canadian Breast Implant Cohort Study
|Type of fill|
| Saline & silicone||2,647||37||949||5.5||3,596||14.6|
|Fill volume (in cc)|
| Less than 175||1,056||14.8||5,209||29.9||6,265||25.5|
| 175 to less than 200||1,340||18.7||4,661||26.8||6,001||24.4|
| 200 to less than 225||1,648||23.0||4,761||27.4||6,409||26.1|
| 225 and higher||3,109||43.5||2,613||15.0||5,722||23.3|
|Site of implantation|
In total, 676 and 553 incident cancers were identified among women in the implant and other plastic surgery cohorts, respectively (Table IV). External comparisons to the general population indicated reduced rates of cancer for all sites combined in both the implant (SIR = 0.75, 95% CI = 0.70–0.81) and other plastic surgery cohorts (SIR = 0.81, 95% CI = 0.74–0.88). Based on incidence rates in the general population, the observed number of breast cancers was significantly lower than the expected number in both the implant (SIR = 0.57, 95% CI = 0.49–0.65) and other plastic surgery patient (SIR = 0.87, 95% CI = 0.75–0.99) cohorts. Statistically significant lower-than-expected rates were also observed for endometrial cancer (SIR = 0.53, 95% CI = 0.33–0.79) in the implant cohort. As well, statistically significant lower than expected rates were observed for colorectal and thyroid cancers and cancers of all other sites combined among the other plastic surgery patients. Three incident cases of multiple myeloma were diagnosed among implant recipients; this was not significantly different from the expected number (SIR = 0.46, 95% CI = 0.09–1.33) [data not shown].
Table IV. Standardized Incidence Ratios (SIRs)1 for Selected Cancers Based on Ontario and Quebec Cancer Incidence Rates (1974–1997) Among Breast Implant and Control Patients, and Relative Risks (RRs)2 of Cancer Incidence for Breast Implant vs. Control Patients Based On Internal Cohort Comparisons
|All sites||140–208 (excl 173)||676||899.3||0.75*||553||683.7||0.81*||0.91||0.81–1.02|
|Lung||162.2–5, .8, .9||96||87.9||1.09||81||73.1||1.11||0.93||0.69–1.26|
|Nervous system||191, 192||11||16.9||0.65||11||12.5||0.88||0.66||0.28–1.54|
|Non Hodgkin's lymphoma||200, 202||25||33.5||0.75||20||25.6||0.78||0.97||0.53–1.76|
|Leukemia||204–206, 207.0, .2, .8, 208||12||17.6||0.68||9||13.7||0.66||0.94||0.39–2.25|
|Other cancer sites not listed above|| ||63||72.2||0.87||36||60.1||0.60*||1.35||0.89–2.04|
Internal cohort comparisons revealed reduced incidence rates of breast cancer among those who received breast implants (RR = 0.64, 95% CI = 0.53–0.79) relative to those with undergoing other plastic surgeries (Table IV). No statistically significant differences between the implant and the other plastic surgery group were observed for any of the other cancer sites examined.
The test of homogeneity for provincial differences in the RR was statistically significant for cancers of the breast and pancreas (p < 0.05), but not for any of the other cancer sites examined. Province-specific RRs for breast cancer for Ontario and Quebec were reduced and statistically significant at 0.40 (95% CI = 0.26–0.60) and 0.75 (95% CI = 0.60–0.94), respectively. The province-specific RRs between the breast implant and other plastic surgery cohorts for cancer of the pancreas were unstable because of a small number of identified cases, and were not significantly different from unity (Ontario RR = 6.95, 95% CI = 0.85–54.89; Quebec RR = 0.45, 95% CI = 0.17–1.19).
Table V presents the risks of breast cancer among implant recipients relative to the other plastic surgery patients for age at surgery, time since surgery and period of surgery. When compared to Ontario, the RRs calculated using the Quebec component of the cohort were nearer to 1 for all comparisons. No statistically significant excess rate of breast cancer relative to the surgical control cohort was observed in either province for stratified analyses performed across categories of age at surgery, time since surgery and period of surgery.
Table V. Relative Risks1 (RRs) and 95% Confidence Intervals (CIs) of Breast Cancer for Selected Implant Surgery Characteristics, Comparisons Made to Other Plastic Surgery Patients, by Province
|Age at surgery (y)|
|Time since surgery (y)|
| 1 to <5||24||25||0.84||0.47–1.50||5||15||0.27||0.10–0.77||29||40||0.63||0.39–1.03|
| 5 to <10||45||53||0.65||0.43–0.98||10||20||0.39||0.18–0.85||55||73||0.58||0.41–0.83|
| 10 to <15||43||39||0.79||0.51–1.23||19||20||0.67||0.35–1.27||62||59||0.74||0.51–1.07|
|Period of surgery|
Among women who received breast implants, there was no statistically significant difference in breast cancer rates between those whose implants were filled with saline relative to those filled with silicone gel (Table VI). Women who received polyurethane-coated breast implants had a RR of 1.51 (95% CI = 0.86–2.65) relative to women with no such coating. When compared with women who received subglandular breast implants, the RR for those who received submuscular implants was 1.23 (95% CI = 0.90–1.69). Fill volume was not associated with the risk of breast cancer.
Table VI. Relative Risks1 (RRs) and 95% Confidence Intervals (CIs) of Breast Cancer for Selected Breast Implant Characteristics
|Type of fill|
| Saline and silicone||45,087||23||1.12||0.70–1.80|
|Fill volume (cc)|
| 175 to <200||86,314||52||1.18||0.80–1.76|
| 200 to <225||87,767||54||1.22||0.82–1.80|
|Site of implantation|
|Total||339,714||188|| || |
Differences in breast cancer rates for selected breast implant characteristics, by site of implantation, were examined and the results are presented in Table VII. An elevated risk of breast cancer was observed among those women who received subglandular breast implants coated with polyurethane relative to those with no such coating (RR = 1.99, 95% CI = 1.07–3.71); however, this result was based on only 15 incident breast cancers. Analyses by follow-up interval revealed decreasing risks with increased follow-up (Table VIII). However, these risks were highly unstable as they were based on a small number of incident cases. Overall, the incidence of breast cancer among women who received subglandular polyurethane-coated implants was similar to that of the general population (SIR = 1.00, 95% CI = 0.56–1.65) [result not shown].
Table VII. Relative Risks1 (RRs) and 95% Confidence Intervals (CIs) of Breast Cancer for Selected Implant Characteristics, by Site of Implantation
|Type of fill|| || || || || || || || ||0.76|
| Silicone||128,790||65||1.0||–||64,280||42||1.0||–|| |
| Saline||9,304||4||1.40||0.48–4.03||869||0||0.00||N.E.|| |
| Saline and silicone||22,905||10||1.18||0.59–2.37||18,487||11||0.92||0.43–1.95|| |
| Unknown||27,147||15||0.81||0.46–1.43||24,616||14||0.80||0.43–1.47|| |
|Polyurethane coating|| || || || || || || || ||0.12|
| No||124,185||50||1.0||–||70,629||44||1.0||–|| |
| Yes||20,919||15||1.99||1.07–3.71||1,203||0||0.00||N.E.|| |
| Unknown||43,042||29||1.27||0.78–2.07||36,420||23||0.97||0.57–1.65|| |
|Fill volume (cc)|| || || || || || || || ||0.86|
| <175||57,064||27||1.0||–||22,214||13||1.0||–|| |
| 175 to <200||48,192||28||1.24||0.73–2.10||26,372||16||1.01||0.48–2.09|| |
| 200 to <225||44,265||22||1.06||0.60–1.87||33,018||23||1.19||0.60–2.34|| |
| ≥225||37,449||16||1.06||0.56–1.99||26,239||15||0.93||0.44–1.99|| |
| Unknown||1,176||1||1.27||0.17–9.35||409||0||–||–|| |
|Total||188,146||94|| || ||108,252||67|| || || |
Table VIII. Relative Risks1 (RRs) and 95% Confidence Intervals (CIs) of Breast Cancer for Women Who Received Subglandular Polyurethane Coated Breast Implants Relative to Other Women Who Received Subglandular Implants, by Time Since Surgery
|1 to <5 years|
|5 to <10 years|
|All follow-up time|
Comparisons between the implant and the other plastic surgery patient cohorts for all cancers combined, cancers of the genital organs, lung cancer and colorectal cancer were also conducted across strata defined by the surgery and implant characteristics listed in Tables V–VII. Among women who received breast implants that were placed subglandular, an increased risk of colorectal cancer was observed for unknown envelope relative to those with a silicone envelope (RR = 2.62, 95% CI = 1.19–5.75); however, no significantly elevated RRs were found for any of the other comparisons.
This cohort study provides further evidence that silicone gel-filled breast implants are unlikely to cause breast cancer or cancers occurring at other sites. The follow-up interval was as much as 24 years for some women, and no increased breast cancer rates were found among women having at least 10 years of follow-up. More than 70% of the cohort members were followed for at least 10 years. Further, while previous studies have demonstrated increased risks for other cancer sites,17, 34 no such associations were evident in our study.
The quality of the national cancer registry data used in our study is an important strength. In the cohort, the proportion of all cancers and breast cancers that were confirmed histologically was 82 and 93%, respectively.
Our finding of lower overall cancer incidence rates among women who received breast implants relative to the general population is consistent with previous work.8, 10, 12 These lower rates may be partly due to the fact that women who choose to undergo cosmetic breast augmentation are a highly selective group that have been shown, on average, to have a different risk factor profile for cancer than women in the general population.32, 35, 36 Specifically, on average, women who receive breast implants have lower body mass index, higher smoking rates, greater use of oral contraceptives, greater numbers of sexual partners and have children at a younger age when compared to women who have not. Therefore, observed difference in cancer rates between the breast implant and general population may easily be influenced by differences in the prevalence of these risk factors. Additional support for this argument can be found with the similar patterns of risk that we found for endometrial cancer, a cancer that has many of the same risk factors as breast cancer. Analyses performed among US women demonstrated that those who undergo cosmetic surgeries are more similar to those who receive breast implants for cosmetic purposes than women in the general population with respect to a large number of risk factors for cancer, including sociodemographic status.37 As a result, differences in cancer incidence rates among those who received breast implants and those undergoing other cosmetic procedures are less likely to be influenced by these risk factors. Our analyses suggest this is indeed the case, as cancer incidence rates among women who received breast implants were more similar to the rates observed among the cohort of women who received other plastic surgery procedures than to the general population.
For the reasons outlined above, comparisons between the breast implant and surgical control patients to the general population may be influenced by the “healthy woman” effect. To better understand the role of this possible effect, we performed additional analyses to evaluate whether lower cancer rates persisted with prolonged follow-up. This was done by repeating analyses by excluding the first 5, and first 10 years of follow-up. The exclusion of these follow-up intervals resulted in no appreciable change in the risk estimates for breast cancer and all cancer sites for both the internal and external cohort analyses. The health woman effect would be expected to diminish over time. We therefore feel that there are other selection factors at work that could contribute to lower cancer rates that were observed with extended follow-up.
Residential mobility and the resulting loss to follow-up may contribute to lower-than-expected incidence rates among the breast implant and other plastic surgery patients, relative to the general population. The inability to adequately capture mobility was cited as a limitation in a previous Canadian record linkage study that evaluated cancer incidence in women who received breast implants in Alberta.7, 9 In the present study, losses to follow-up were minimized with respect to mobility by linking the cohort to the national cancer incidence data collected by all provincial registries, and to the Canadian Vital Statistics Database. Cancer registration in Canada is near complete through the cooperation of the provincial cancer registries, and the reporting of deaths is mandatory; therefore, few such events occurring in Canada would have been missed. The ability to link the cohort to these databases was excellent, given the detailed personal identifying information available, particularly in the later period of the study where health insurance numbers were available and could be used. In the Ontario component of the cohort, 306 out of the 325 identified cancers were diagnosed in the province. Similarly, among the Quebec component, 899 of the 922 incident cancers were diagnosed in the same province. While it is possible that a small number of incident cancers were missed among women in the cohort whose diagnosis occurred while living outside Canada, given the large size of the cohort, the expected bias on the standardized incidence ratios is expected to be small. Perhaps more importantly, while incomplete case ascertainment through record linkage may have influenced external comparisons to the general population, it is unlikely to have affected comparisons made between the implant and surgical control cohorts. Specifically, we know of no reason why there would be differential case ascertainment between the implant and other plastic surgery cohorts.
With the exception of breast cancer, there were no statistically significant differences in cancer incidence for any of the other sites examined between women who received breast implants, and those who underwent other cosmetic surgical procedures. As with previous investigations,9, 10 we found lower breast cancer incidence rates among women who received breast implants. Several biological mechanisms have been suggested in which the implant procedure itself may confer a lower risk of breast cancer. These include: an enhanced immune system as a result of receiving a breast implant, whereby carcinogens and transformed cells are more easily destroyed; the weight and volume of the implant compresses the glandular tissue resulting in a decreased blood supply that may reduce the rate of cell proliferation; and a long-term reduction in the metabolic rate resulting from lower temperature of the breast tissue.38 However, we feel it is more likely that other characteristics contributed to the observed differences in breast cancer rates between women who received implants and the other surgical patients. For example, women with a family history of breast cancer may have elected not to receive breast implants, as these devices have been shown to interfere with the detection of breast cancer.39 Moreover, women who receive breast implants may be at a lower risk of developing breast cancer because they have smaller breasts. Although the epidemiological evidence linking breast size with cancer risk is inconsistent, arguments for an association include: breast cancer occurs more commonly in the left breast40 that tends to be the larger breast in many women41; breast size may serve as a proxy for the volume of epithelium at risk, and reflect fat deposits in the breast that influence estrogen levels and act as a repository for lipid-soluble carcinogens42; women who have had breast reduction surgery have lower breast cancer rates.43
Other studies have found that women undergoing cosmetic procedures have similar income levels to women who receive breast implants37; which are in turn higher than the general population.32, 44 Given that income has been positively associated with breast cancer incidence rates in Canada,45 our finding of lower breast cancer incidence rates in both the implant and plastic surgery populations is somewhat surprising. As previously mentioned, this difference may be due in part to the effects of other unmeasured confounders, or incomplete case ascertainment.
While family history and smaller breast size are plausible arguments to support a lower breast cancer risks among those who received breast implants, they are unlikely to explain the 13% lower risk of developing breast cancer among the other plastic surgery patients relative to the general population. To further explore differences in rates between these 2 populations, we compared breast cancer incidence rates across the different types of plastic surgery procedures. The SIRs for rhinoplasty, blepharoplasty, otoplasty, chemical peel and rhytidectomy were 0.79, 0.88, 0.93, 1.02 and 0.96, respectively; none of these values were statistically significant. Unfortunately, the relative lack of variability in these estimates does not provide additional information on possible selection factors at work.
We excluded the first year after surgery in the follow-up of this cohort, which is consistent with approaches undertaken by other investigators.8, 17 Before receiving breast implants, women typically undergo a detailed breast examination with or without mammography. Such examinations may lead to a reduction in the rates of breast cancer during the first year after breast implant surgery. Sensitivity analyses revealed that when analyses were repeated by including the first year after surgery, the RRs for breast cancer were largely unchanged.
Statistically significant reduced rates of breast cancer among women who received breast implants when compared to either the general population or other plastic surgery patients were observed in both the Ontario and Quebec components of the cohort. However, the magnitude of this reduction differed by province, namely, the RR among Quebec women who received a breast implant relative to the plastic surgery patients was closer to unity (RR = 0.75) than that in Ontario (RR = 0.40). The reason(s) for this difference is not clear, yet it cannot be fully explained by a greater proportion of women in Quebec who received polyurethane-coated implants, nor a longer follow-up period for the Quebec component of the cohort. Specifically, when women who received polyurethane implants were excluded from analyses, the estimated RRs derived by comparing breast cancer rates among women who received implants to other plastic surgery patients were 0.73 and 0.40 for Quebec and Ontario, respectively. Similarly, when analyses were restricted to exclude the first 10 years of follow-up the corresponding RR estimates for Quebec and Ontario were 0.86 and 0.44, respectively.
It should be noted that the number of patients recruited in Ontario was substantially lower than originally planned. Unlike in Quebec, no central list of plastic surgery procedures carried-out in hospitals existed in Ontario. It was only feasible to identify patients through careful review of all office records of plastic surgeons. Of 133 plastic surgeons potentially eligible, only 51 (38%) permitted access to all of their records. Reasons for non-participation included: uncertainty regarding the legal and ethical implications of permitting access without patient consent; substantial record destruction during the earlier years of the study and inability to locate many of the plastic surgeons or their records (moved outside Ontario; retired; died).
In spite of recruiting fewer Ontario plastic surgeons than expected, while this eroded our study power somewhat, we do not believe it seriously biased the findings. Where records had been destroyed in earlier years, we simply excluded all patients who had their index surgery (implants or otherwise) during these years. The potential for this selection bias to adversely affect our internal cohort analyses was minimized by matching breast implant patients to other plastic surgery patients, from the same practice, during the same year of index surgery.
The sample size in our study was sufficiently large to compare breast cancer incidence among women who received breast implants, according to the type of implant received. It is an advantage to perform these comparisons only among women who received implants, given that the risk factor profile within this group is likely to be more similar than comparisons involving the surgical control cohort or the general population. Our results indicate that there were no differences in breast cancer rates between women who received SGFI and those filled with saline.
The site of implantation may influence breast cancer incidence rates. The placement of the breast implants may be either subglandular, which is between the muscle and breast tissue, or submuscular, which is below the chest muscle. Subglandular placed breast implants are associated with higher contracture rates than those placed submuscular.46, 47, 48 In addition, subglandular placement of the breast implants has been shown to interfere with the imaging of the breast with mammography.49, 50 As a result, when compared to submuscular placed implants, subglandular placement could delay the diagnosis of incident breast cancer, leading to lower breast cancer incidence rates that would disappear with increasing follow-up. However, this phenomenon did not occur in our data. Specifically, the RR of breast cancer among those with submuscular implants compared to subglandular implants increased from 1.23 (95% CI = 0.90–1.69) to 1.47 (95% CI = 0.95–2.28) when the first 10 years of follow-up were excluded.
Analyses were also undertaken among women who received breast implants to determine whether polyurethane-coated implants were associated with higher breast cancer incidence rates. Polyurethane-coated implants were withdrawn from the market in 1991 as polyurethane was found to degrade into compounds that are, or could become, carcinogens in animals (i.e., 2,4-toluenediamine or precursors).51 To our knowledge, there has only been 1 study that has examined cancer incidence among women who received polyurethane-coated implants. However, this prospective study included only 213 women of whom there were 2 identified breast cancers; moreover, no reference group was included in the study, and therefore, risks could not be adequately assessed.52
In our study, a non-statistically significant elevated RR was obtained when comparing breast cancer rates among women with polyurethane-coated implants relative to breast implants with no such coating. When we restricted our analyses to women with subglandular implant placement, those with a polyurethane-coated implant had a statistically significant 2-fold increase in risk relative to women with no such coating. There was a progressive reduction in this risk as follow-up was extended, and no increased risks were observed with 10 or more years of follow-up. This has important implications given that polyurethane-coated implants were withdrawn from the Canadian market approximately 15 years ago. Specifically, it suggests that women who received such implants and are currently still alive, are not at an increased risk of developing breast cancer.
While we found increased breast cancer rates among women with subglandular placed polyurethane-coated implants within the 10-year period following surgery, the possibility that polyurethane directly contributes to this increased risk is tenuous. First, previous work suggests that the devices themselves do not produce sufficient TDA levels to cause breast cancer. Specifically, urine analyses of approximately 110,000 women who received polyurethane foam coated implants found that TDA levels were at such low levels that the associated risk of developing cancer from these implants was estimated to be about 1 in 1 million.53 It is unclear whether subglandular placement of the implant leads to higher levels of exposure of breast tissue to polyurethane foam degradation products. Second, this excess risk, if true, would be expected to increase with follow-up, given the potential for increased duration of exposure to polyurethane and allowing for a latency period between exposure and breast cancer development. Third, the finding of increased RR should also be interpreted cautiously as unmeasured individual level risk factors at the time of surgery may have confounded our comparisons and as the number of breast cancers on which the estimation is based is quite small (15 cases among women with polyurethane coated implants). Finally, the increase in RR was found in this subgroup of women after much exploration of the data, which increases the possibility that the observed association could be due to chance.
In summary, the study results suggest that silicone gel-filled breast implants do not directly increase the risk of developing breast cancer or cancers at other sites. Our finding of a positive association between breast cancer and subglandular polyurethane-coated implants is noteworthy, but is based on a small number of identified cases after much exploration of the data. Given that most of the women in our study population received breast implants at a young age, continued follow-up of the cohort is important to examine associations between breast implants and the incidence of breast cancer, particularly in postmenopausal women.
We are especially grateful to the participation of the plastic surgeons in Ontario and Quebec, and for the opportunity to access provincial cancer registry data. We thank the expert advice provided by Drs. Louise Duranceau, Pierre Langlois and Walter Peters, and comments provided on an earlier draft by Drs. Louise Brinton and Anthony Miller. We also thank the provincial study managers, Ms. Andrée Christen from Quebec and Ms. Gemma Lee from Ontario, for their help in the design and conduct of the study. Finally, we thank the Occupational Health and Research Division at Statistics Canada for linking the cohort to the national cancer registry and vital statistics databases.