The Breast Implant Surveillance Study was conducted during the period 1993–1994 in the United States. Patients were identified through the population-based SEER cancer registries in San Francisco–Oakland, CA, in Seattle–Puget Sound, WA, and in the state of Iowa. Eligible patients included the 5862 females diagnosed younger than age 65 with early-stage or unstaged first primary breast cancer in 1983, 1985, 1987, or 1989 and treated with mastectomy during their first course of therapy.

Participation involved completing a standardized questionnaire inquiring about implant status (right breast only, left breast only, both breasts, no implant, or unknown) and implant type (silicone gel, saline, double lumen [consisting of silicone gel and saline], other, and unknown type), and providing signed consent for release of medical records for validation of implant usage and implant type. Patients in the Seattle–Puget Sound region and the state of Iowa were mailed a self-administered questionnaire, and nonrespondents were contacted for a telephone interview. Women from the San Francisco–Oakland region were administered questionnaires by telephone.

Next-of-kin were asked about the deceased patient's implant status and for consent for medical record review. For women reporting breast implants, a medical record review of breast implant usage (including the date and type of implant received, and removal and replacement status) was conducted by trained abstractors for women who reported having a breast implant. Patient and tumor characteristics at diagnosis, including age, race/ethnicity, year, stage, histology, grade, radiation treatment, vital status, and cause of death, were obtained from the SEER database.

Although socioeconomic status (SES) is not routinely collected by SEER, we were able to assign census block group level measures of SES to a subset of subjects (1989 San Francisco Bay Area patients). Using data from the 1990 US Census, we examined the impact of living in census block groups characterized by low education, by poverty, and by occupation (median income, < 20% below poverty versus ≥ 20% below poverty; education, no high school diploma versus high school graduate; and occupation, blue collar versus nonblue collar 10).

Patients who reported a breast implant were classified as having a particular type of implant if they had a unilateral implant, or if they had bilateral implants of the same type. Nineteen women with bilateral implants and discordant information about implant type were excluded from analyses stratified by the type of implant. An additional 133 women reporting implants but lacking implant information were excluded from all analyses requiring detailed implant information. The vital status (obtained annually through patient contact, death records, motor vehicle departments, voters' registration records, and Social Security files) was determined from the December 1999 SEER Public Use Tape. The outcome variables were death due to breast cancer and death due to nonbreast cancer causes, as routinely ascertained by SEER and as defined by the International Classification of Disease, Ninth Revision, codes 174.0–174.9 for deaths occurring between 1983 through 1998, and by the International Classification of Disease, 10th Revision, code C509 for deaths occurring in 1999. Survival time was calculated from the date of diagnosis to the earliest date: death, last known to be alive, or 31 December 1999 (study cutoff date).

For descriptive analyses, women with breast implants were compared with those without implants on characteristics at diagnosis (SEER region of residence, age, year, race/ethnicity, marital status, stage, grade, and histology), using chi-square tests and Fisher's exact test to assess differences. Two-sided P < 0.05 was considered statistically significant.

Of the 5862 patients eligible for the study, 4968 (84.7% of those eligible) patients participated in the study. After restricting the sample to women with primary invasive breast cancer and known survival time, a final sample size of 4385 patients were used for all survival analyses. Survival estimates were computed using the Kaplan–Meier method, and differences in survival were compared using the two-sided log-rank test. To adjust for patient and tumor characteristics, the risks of nonbreast cancer mortality and death due to breast cancer were modeled using Cox proportional hazards regression after censoring deaths from other or unknown causes. The assumption of proportionality was tested and met for all covariates used in the Cox analysis. All analyses were performed using SAS Version 8.02 (SAS Institute, Inc., Cary, NC, USA).

Eighty-five percent of study-eligible women completed the interview. Responders were slightly older compared with nonresponders. They also were more often non-Hispanic white, were diagnosed with in situ cancers, were less likely to have more than one primary tumor and were more likely to have lived until the end of the study period (Table 1).

Among the responders, 20% received a breast implant (Table 1). Their mean age at diagnosis was 47 years, and they were younger, on average, than women without implants. Somewhat higher proportions of women with implants were of non-Hispanic white race/ethnicity, resided in the San Francisco–Oakland region, were diagnosed in the late 1980s, and had tumors of lobular histology than women without implants. The percentage of women with in situ breast cancer was twice as high in women with breast implants as in nonimplanted women. In addition, women with implants were less likely to receive radiation therapy or to be diagnosed with more than one primary tumor.

Implant information was obtained for 866 women; these women were slightly older and more likely to be living than women with unknown implant information. Among the 1143 breast implants received (Table 2), the most common types were silicone gel and double lumen. The majority of women (67%) received a unilateral implant a median of 9.6 months after breast cancer diagnosis. Approximately one-third of the women had an implant removed; the majority of these women chose to have the implant replaced. Saline-filled implants were removed for 48% of women, although saline-filled tissue expanders (temporary implants) may be incorrectly included in this category. Fifty-four percent of women with 'other' implants had them removed, and 49% had them replaced.

At the end of the follow-up period, 231 (5.3%) patients did not have complete follow-up. Twenty-eight percent of all patients died, with nearly two-thirds of these deaths due to breast cancer (Table 3). Women with implants had a similar distribution of causes of death to those without implants (Table 3), except for a significantly larger proportion of deaths due to suicide – although this was based on a small number of deaths (0.4% versus 0.03% of all patients, respectively; P = 0.02). Among the 4385 patients with invasive breast cancer and known survival time, the 817 (19%) receiving a breast implant experienced better survival than women without implants, after adjustment for age at diagnosis (Fig. 1).

The multivariate Cox proportional hazards model showed that implant status was a significant factor associated with improved survival for deaths due to breast cancer and for nonbreast cancer mortality (Table 4). Risk of breast cancer death in women with implants was approximately one-half of that for women without implants, after adjustment for multiple clinical and sociodemographic factors. Age at diagnosis, stage, grade, histology, and radiation therapy were significant predictors of breast cancer death in this cohort, as they were for women without implants. With the exception of age, results were similar when modeled for nonbreast cancer mortality, although hazard rate ratios for women with implants were slightly higher overall. Among women with implants, the type of breast implant did not significantly impact survival, although women receiving saline implants had marginally lower risks than women receiving silicone gel implants.