Breast implants and anaplastic large cell lymphomas among females in the California Teachers Study cohort

Clinical case series have reported an apparent coexistence of breast implants and primary T-cell lymphomas of the breast, specifically for the anaplastic large cell lymphoma (ALCL) subtype (Gidengil et al 2015; Jewell et al 2011; Thompson & Prince 2013). In 2011, the Food and Drug Administration (FDA) reported a “possible association between breast implants and the development of ALCL” and encouraged “additional evaluation from other studies” (United States Food and Drug Administration 2011), particularly as epidemiological data have not yielded consistent results. A review of five prospective cohort studies comprising >43,000 women with cosmetic breast implants followed for up to 37 years found no excess ALCL risk (Lipworth et al 2008) . Data from a Danish cohort of approximately 20,000 women with breast implants identified 31 lymphomas, but no ALCLs Vase et al 2013) . Another study of >150,000 women with breast implants identified 3 breast ALCL, which was lower than expected based on cancer registry data [Spear et al (2010)] . However, a case-control study nested within the population-based cancer registry in the Netherlands with case ascertainment from 1994-2006, reported an association between silicone breast implants and ALCL (Odds Ratio [OR]=18.2, 95% Confidence Interval [CI]=2.1-156.8, based on 11 ALCLs).

Here, we evaluated the association between breast implants and incident T-cell lymphomas in the California Teachers Study (CTS) cohort, a prospective cohort study comprising 133,479 female public school professionals who enrolled in 1995-1996 (www.calteachersstudy.org). This analysis included 123,392 women, after excluding those with a previous diagnosis of haematological cancer, those who were not California residents at baseline, and those who only wished to participate in breast cancer studies. The use of human subjects in this study was approved by each participating institution.

In 1995-96, CTS participants were asked whether they ever had breast implants, their age at first breast implant, and the type of implant inserted (e.g., silicone gel, saline). The CTS cohort is followed annually for cancer diagnosis through linkages with the California Cancer Registry, which receives information on >99% of all cancer diagnoses occurring in California residents. Non-Hodgkin lymphomas (NHLs) were defined using International Classification of Diseases for Oncology, Third Edition (ICD-O-3) morphology codes Fritz et al 2013). Between 1 January 1995 and 31 December 2012, 89 women were identified with incident T-cell lymphomas; the definition and distribution of the subtypes are shown in Table I.

Table I.

Classification of incident T-cell lymphomas in the California Teachers Study cohort (1995-2012).

T-cell lymphoma subtypes	N	ICD-O-3 code
T-cell lymphomas, total	89	9590-9591, 9596, 9675, 9700-9702, 9705, 9708-9709, 9714, 9716-9719, 9727, 9729, 9820, 9827, 9831, 9832, 9834, 9835, 9837, 9948, 9970
Mycosis fungoides/Sezary syndrome/PCLPD	30	9700, 9701, 9718
Peripheral T-Cell Lymphoma	44	9702, 9705, 9708, 9709, 9714, 9716, 9717,
Angioimmunoblastic lymphoma	8	9705
Anaplastic large cell lymphoma	10	9714
Hepatosplenic T-cell lymphoma	1	9716
T-cell large granular lymphocytic leukaemia	2	9831
Cutaneous T-cell lymphoma, NOS	9	9709
PTCL-NOS	16	9702
Adult T-cell leukaemia/lymphoma	1	9827
ENNKTL/ANKL	1	9719, 9948
Precursor T-cell	4	9727, 9729, 9835, 9837
Prolymphocyte leukaemia	2	9832, 9834
T-cell NOS	5	9590, 9591

ICD-O-3, International Classification of Diseases for Oncology, Third Edition; PCLPD, Primary cutaneous lymphoproliferative disorders; PTCL, peripheral T-cell lymphoma; ENNKTL, Extranodal NK/T-cell lymphoma, nasal type; ANKL, aggressive NK-cell leukaemia; NOS, not otherwise specified

We fitted multivariate Cox proportional hazards regression models to estimate the relative risk (hazard ratio, HR) of T-cell lymphoma and subtypes associated with breast implants and the 95% CIs for the relative risk. Age (in days) was the time scale used for the analysis. Follow-up began on the date the enrolment questionnaire was submitted and ended on the first of any of the following events: diagnosis of any haematological malignancy (NHL, Hodgkin lymphoma [HL], multiple myeloma, leukaemia), relocation outside of California, death, or 31 December 2012 (end of follow-up). All models were stratified by age (in years) at cohort entry, and initially adjusted for race (non-Hispanic, white, other), family history of HL or NHL, socioeconomic status, smoking status and body mass index (Supplemental Table 1). Because these variables modified the HR by <10%, we present the most parsimonious model which adjusts for family history. Other ALCL risk factors, including coeliac disease and psoriasis (Wang et al 2014), were extremely rare and had no influence on risk estimates. Statistical analyses were performed using SAS, version 9.3 (SAS Institute, Inc., Cary, NC).

Of the 123,392 eligible participants, 2,990 women reported having had a breast implant at the time of their baseline questionnaire (1,715 silicone, 712 saline, 361 both; type not known in 202). Age at first implant ranged from 23 years to 67 years with a median of 37 years. Eighty-nine incident T-cell lymphomas were diagnosed during follow-up, which averaged 14 years. Of the 10 women diagnosed with incident ALCL, two reported having had breast implants at study entry (Table II). Calculation of HRs yielded no association between breast implant and T-cell lymphomas overall or the subgroup of peripheral T-cell lymphomas; a statistically significant association was observed for ALCL (HR= 10.9, 95% CI 2.18-54.0).

Table II.

Hazard ratios (HR) and 95% confidence intervals (CI) for the association between breast implants and T-cell lymphoma overall and by subtype in the California Teachers Study cohort (1995-2012).

	N	Person-Years	Cases (n=89)	HR*	95% CI
T-cell lymphomas
No breast implants	118834	1761306	85	1.00	(Reference)
Had breast implants	2990	44394	2	0.96	(0.23-3.91)
Missing	1568	21829	2	1.64	(0.40-6.73)
Peripheral T-cell l ymphoma
No breast implants	118834	1761306	41	1.00	(Reference)
Had breast implants	2990	44394	2	2.06	(0.49-8.59)
Anaplastic large cell lymphoma
No breast implants	118834	1761306	8	1.00	(Reference)
Had breast implants	2990	44394	2	10.9	(2.18-54.0)

^*adjusted for family history of non-Hodgkin lymphoma or Hodgkin lymphoma

In this analysis, a history of breast implant reported at study baseline was not associated with subsequent T-cell lymphoma risk but with a 10.9-fold increase in risk specific for ALCL. Of the two pathology-confirmed ALCL patients who reported a history of breast implant, the time lag from age at first implant to age at ALCL diagnosis was ~20 years. Both women reported using both saline and silicone implants. Neither had family history of lymphoma and neither had an identified history of coeliac disease; two recognized risk factors for ALCL. One ALCL was identified as primary site at the breast and the other at multiple lymph nodes. No other study participant diagnosed with any other subtype of T-cell lymphoma reported using breast implants and no other T-cell lymphoma identified in the eligible cohort was coded as arising in the breast.

Our data support a positive association between breast implants and ALCL risk, but the occurrence of ALCL among women with breast implants remains extremely low. The large magnitude of risk observed in the CTS between breast implants and ALCL was consistent with that previous reported (de Jong et al 2008), the only other population-based study that has published on this topic. The average time from breast implant to ALCL diagnosis in published clinical cases is ~10 years, which is considerably shorter than the ~20 years between implant and ALCL diagnosis identified in the CTS. The rarity of ALCL in the general population means both that any increased relative risk would translate to a relatively small absolute risk. Based on the low incidence rates estimated from Table II, particularly among the general population, it is not surprising that there is discordance with previous studies that did not find evidence of an association. Reasons for the discordance may include differences in population sampling, insufficient follow-up period, insufficient sample size, historical under-ascertainment of ALCL, or chance. Nevertheless, continued monitoring for ALCL and evaluation in even larger datasets with updated information of implant status is warranted, particularly as breast augmentation remains the most common cosmetic procedure in the United States.