Breast Implant-Associated Anaplastic Large Cell Lymphoma Incidence: Determining an Accurate Risk

Abstract

Objective:

This study sought to estimate the incidence and incidence rate of breast implant-associated anaplastic large cell lymphoma (BIA-ALCL) at a single institution which enables vigorous long-term follow-up and implant tracking for more accurate estimates.

Summary Background Data:

The reported incidence of BIA-ALCL is highly variable, ranging from 1 in 355 to 1 in 30,000 patients, demonstrating a need for more accurate estimates.

Methods:

All patients who underwent implant-based breast reconstruction from 1991–2017 were retrospectively identified. The incidence and incidence rate of BIA-ALCL were estimated per patient and per implant. A time-to-event analysis was performed using the Kaplan-Meier estimator and life table.

Results:

During the 26-year study period, 9,373 patients underwent reconstruction with 16,065 implants, of which 9,589(59.7%) were textured. Eleven patients were diagnosed with BIA-ALCL, all of whom had a history of textured implants. The overall incidence of BIA-ALCL was 1.79 per 1,000 patients (1 in 559) with textured implants and 1.15 per 1,000 textured implants (1 in 871), with a median time to diagnosis of 10.3 years (range, 6.4–15.5 years). Time-to-event analysis demonstrated a BIA-ALCL cumulative incidence of 0 at up to 6 years, increasing to 4.4 per 1,000 patients at 10–12 years and 9.4 per 1,000 patients at 14–16 years, although a sensitivity analysis showed loss to follow-up may have skewed these estimates.

Conclusions:

BIA-ALCL incidence and incidence rates may be higher than previous epidemiological estimates, with incidence increasing over time, particularly in patients exposed to textured implants for longer than 10 years.

Mini-Abstract

This study examined 9,373 breast reconstruction patients over 26 years and determined that BIA-ALCL incidence was 1.79 per 1,000 textured implant patients (1 in 559) with a median time to diagnosis of 10.3 years. BIA-ALCL incidence may be higher than previous epidemiological estimates, with incidence increasing over time.

Introduction

Breast implant-associated anaplastic large cell lymphoma (BIA-ALCL) is a rare form of peripheral T-cell lymphoma that develops around implants used for breast augmentation or reconstruction.¹ The underlying etiology of the disease is not fully understood but possibly related to chronic inflammation caused by textured devices, leading to the malignant transformation of T cells.^2–9 These malignant cells are characteristically anaplastic lymphoma kinase (ALK)-negative and CD30-positive.^{1, 10, 11}

BIA-ALCL is an important public health concern, as the use of breast implants has increased globally over the past 2 decades.^12–15 The American Society of Plastic Surgeons estimates that more than 400,000 implant-based cosmetic and reconstructive procedures were performed in the United States in 2017.¹⁵ Although sales of textured breast implants have decreased since 2000, more than 3 million of these devices have been used in patients in the United States, putting a significant number of women at risk for the development of BIA-ALCL.⁷ The risk of developing BIA-ALCL appears to be associated with prolonged exposure to textured implants, with most case series reporting BIA-ALCL onset at a median exposure time of 7.5–11 years,^{1, 7, 8, 16–18} although shorter times (0.4–2 years) have also been reported.^{16, 18, 19}

The reported incidence of BIA-ALCL varies widely in the literature, ranging from as high as 1 in 355 patients to as low as 1 in 30,000 patients.^{7, 8, 20–22} Incidence rates also vary, from as high as 1 in 3,215 person-years to as low as 1 in 500,000 person-years.^{7, 8, 22, 23} This variability is likely multifactorial and related to the relative rarity of the disease, passive data reporting systems, lack of accurate implant utilization rates, and long implant exposure times prior to the onset of disease.²⁴ Many studies have relied on varying data sources, such as population-based registry data, case report forms, the medical literature, and manufacturers’ sales statistics.^{7, 8, 21, 23, 25} For example, although 186 unique reports of BIA-ALCL in the United States have been documented by PROFILE (Patient Registry and Outcomes for Breast Implants and Anaplastic Large Cell Lymphoma Etiology and Epidemiology), this information alone, without accurate knowledge of the total number of implants used in the United States, makes it impossible to accurately estimate the incidence of the disease.¹⁶ Thus, a study in which the type of implant used, the exposure time, and the development of the disease are accurately documented may provide a more accurate assessment of the incidence and prevalence of BIA-ALCL.

A recent single-surgeon experience from our institution suggests that the risk of BIA-ALCL development among women who undergo reconstruction with textured breast implants may be higher than previously estimated. In the study, Cordeiro et al. described a cohort of 3,546 women who underwent breast reconstruction with textured implants between 1992 and 2017 and were prospectively followed long term. Of these women, 10 developed BIA-ALCL.²² This calculated risk of 1/355 women is significantly higher than previous estimates. The limitations inherent in a single-surgeon experience, however, suggest that a more in-depth analysis at the institutional level, with an examination of all patients during the time period, is warranted.

The primary objective of our study was to determine both the incidence and incidence rate of BIA-ALCL by year of implant exposure within a single institution that performs a high volume of prosthetic implant placement for breast reconstruction. Due to the presentation and local nature of BIA-ALCL development (i.e., in the periprosthetic capsule), we calculated the incidence of this disease per implant and per patient. Our secondary objective was to analyze trends in implant usage over time (i.e., textured versus smooth devices) at our institution to understand the impact of these devices on BIA-ALCL development among breast cancer survivors and determine the burden of risk in our population.

Methods

After institutional review board approval, we retrospectively identified all patients who underwent implant-based breast reconstruction between July 1, 1991 and June 30, 2017 at Memorial Sloan Kettering Cancer Center (MSK). Surgical details, including laterality of the procedure, implant type used (smooth vs. textured; manufacturer), and subsequent procedures for breast reconstruction, were recorded from patient medical records. Subsequent procedures were classified as explantation alone, implant removal and replacement, or explantation with conversion to autologous reconstruction for each breast of each patient. Additional variables of interest, including past medical history and demographics (i.e., age, race/ethnicity, body mass index [BMI]), were collected. All patients in whom the type of implant used was indeterminate were excluded from analysis.

BIA-ALCL diagnosis was confirmed by a board-certified hematopathologist based on cytology, immunohistochemistry, and microscopic examination. All confirmed cases contained CD30-positive, ALK-negative T cells in the periprosthetic fluid, implant capsule, and/or lymph node biopsies.

We estimated exposure time to implants using two methods: “Follow-Up,” which defines exposure as the time difference from initial placement of the permanent implant to the most recent follow-up with an institutional physician or advanced practice practitioner, and “Vital Status,” which defines exposure time as the time difference from initial placement of the permanent implant to the date of data collection (March 1, 2019) or date of death. Sensitivity analysis to determine robustness was performed by comparing the results of both methods.

Given the association between BIA-ALCL and textured devices, we estimated the exposure time in implant years for each implant type (smooth or textured). For both methods, the date of implant exchange to a different shell type, explantation, or explantation with conversion to autologous reconstruction was factored into the calculation. Exchange to an implant of the same type was considered a continuation of exposure for this calculation. This generated a longitudinal exposure time to each implant type and enabled us to study each implant separately in patients who underwent bilateral reconstruction. Using this approach, the total time of exposure to textured or smooth implants was independently calculated for laterality of breast, accounting for patients who had metachronous bilateral implant placement and patients who underwent unilateral removal, conversion to autologous reconstruction, or exchange to smooth implants. Total exposure time to textured implants was used to calculate the incidence rate by implant-year and person-year.

We performed a time-to-event analysis using the Kaplan-Meier estimator with 95% confidence interval (CI) to determine the cumulative incidence of BIA-ALCL. The occurrence of BIA-ALCL was compared to time exposure to textured implants for any period of time in order to give the most accurate representation of disease incidence. Four curves were generated to show the estimated cumulative incidence of BIA-ALCL per patient and per implant for both methods of estimated exposure time. Per-patient exposure time was defined as the cumulative exposure time to textured implants regardless of laterality of implant. Per-implant exposure time was defined as the cumulative exposure to textured implants for each individual breast. In this analysis, reasons for censoring (removal from the denominator population) included explantation without further reconstruction, explantation with conversion to autologous reconstruction, exchange to smooth implants, loss to follow-up, death, or not reaching event of interest within a given exposure time. A 20-year life table was constructed to report incidence in 2-year intervals. Trends in implant use over the study period were examined for both smooth and textured devices.

All data were analyzed with R Statistical Software²⁶(packages: ggfortify, ggplot2, survival). Microsoft Excel was used to calculate implant-specific exposure time and to generate figures.

Results

Over the 26-year study period, 9,373 patients underwent implant-based breast reconstruction and had complete electronic medical records for inclusion in the final analysis. Table 1 describes the patient characteristics of the overall study cohort. Mean age at first reconstruction was 49.5 ±10.1 years, and patients were predominantly white (n=7,848; 83.7%).

Table 1:

Total Implant-Based Post-Mastectomy Reconstruction Patient Characteristics

Characteristic	Value
Total patients, n	9,373
Age at first reconstruction, mean (SD), years	49.5 (10.1)
BMI, mean (SD)	26.1 (20.7)
Race, n (%)
White	7,848 (83.7%)
Asian/Indian Subcontinent	481 (5.1%)
African-American	674 (7.2%)
Native Hawaiian or Pacific Islander	2 (0.02%)
Other	120 (1.3%)
Refused to answer	213 (2.3%)
Unknown	33 (0.3%)
Breast cancer diagnosis, n (%)
Infiltrating ductal carcinoma	5,035 (53.7%)
Ductal carcinoma in situ	1,089 (11.6%)
Infiltrating lobular carcinoma	627 (6.7%)
Lobular carcinoma in situ	214 (2.3%)
Invasive ductal and lobular carcinoma	651 (6.9%)
Both ductal and lobular carcinoma in situ	241 (2.6%)
other or unknown	1,516 (16.2%)
Diabetes, n (%)	650 (6.9%)
Hypertension, n (%)	2,377 (25.4%)
Chemotherapy, n (%)	4,438 (47.3%)
Radiation therapy, n (%)	1,797 (19.2%)

BMI: body mass index

Table 2 summarizes operative characteristics and BIA-ALCL incidence and incidence rates. Of 9,373 patients, 5,771 (61.6%) underwent bilateral reconstruction and 3,602 (38.4%) underwent unilateral reconstruction. A total of 16,065 implants were placed during the study period, contributing to a total of 94,667 implant-years, with an average per-implant exposure time of 5.63 years (range, 0–27.92 years) per the Follow-Up method and 8.11 years (range, 0–28.02 years) per the Vital Status method, as described above. When examining implant surface type, 6,149 patients underwent placement of 9,589 textured implants, and 3,918 patients underwent placement of 6,476 smooth implants, inclusive of all replacement/secondary procedures. Assuming implant exposure time per the Follow-Up method, patients with textured implants had a total exposure time of 67,323 implant-years, with a median exposure time of 6.53 years per implant (range, 0–27.9 years). Patients with smooth implants had a total exposure time of 27,344 implant-years, with a median exposure time of 3.69 years per implant (range, 0–18.17 years). Assuming implant exposure time per the Vital Status method, patients with textured implants had a total exposure time of 91,915 implant-years, with a median of 9.37 years per implant (range, 0–27.90 years). Patients with smooth implants had a total exposure time of 36,625 implant-years, with a median of 5.31 years per implant (range, 0–21.16 years).

Table 2:

Operative Characteristics and Estimated Incidence and Incidence Rates of BIA-ALCL

Total Patient Characteristics			Follow-Up Exposure		Vital Status Exposure
Reconstruction Laterality	N patients	N implants	Total Implant-Years	Median Overall Exposure, years (range)	Total Implant-Years	Median Exposure, years (range)
Unilateral	3,602 (38.4%)	3,855 (24.0%)	24,617.73 (26.0%)	6.05 (0–26.69)	34,341.25 (26.7%)	8.94 (0–27.14)
Left	1,799 (19.2%)	1,912 (11.9%)	12,254.34 (12.9%)	6.05 (0–25.57)	16,837.5 (13.1%)	8.77 (0–25.88)
Right	1,803 (19.2%)	1,943 (12.1%)	12,363.40 (13.1%)	6.06 (0–26.69)	17,503.75 (13.6%)	9.13 (0–27.14)
Bilateral	5,771 (61.6%)	12,210 (76.0%)	70,049.07 (74.0%)	5.41 (0–27.92)	94,198.87 (73.3%)	7.64 (0–28.02)
Total	9,373	16,065	94,666.80	5.63 (0–27.92)	12,8540.10	8.11 (0–28.02)
Total Implant- Type Characteristics*
	N patients	N implants	Total Implant-Years	Median Implant Follow-Up, years (range)	Total Implant-Years	Median Implant Follow-Up, years (range)
Textured	6,149 (61.1%)	9,589 (59.7%)	67,323.25 (71.12%)	6.53 (0–27.90)	91,914.96 (71.5%)	9.37 (0–27.90)
Smooth	3,918 (38.9%)	6,476 (40.3%)	27,343.55 (28.88%)	3.69 (0–18.17)	36,625.16 (28.5%)	5.31 (0–21.16)

BIA-ALCL = breast implant-associated anaplastic large cell lymphoma

^*These patients include replacement procedures.

Eleven cases of BIA-ALCL were diagnosed, all in patients treated with textured implants (Table 3). The median time to diagnosis was 10.26 years (range, 6.43–15.52 years). Ten of the 11 patients had undergone bilateral reconstruction; all cases of BIA-ALCL were diagnosed unilaterally. Although one patient had smooth implants at the time of diagnosis, with a smooth implant exposure time of 1 year, this patient was initially exposed to a textured implant for 10.26 years prior to exchange to a smooth implant due to the development of a spontaneous seroma. All BIA-ALCL cases had exposure to Allergan / Inamed / McGhan Biocell textured breast implants.

Table 3:

BIA-ALCL Patient Description

Patient	Age at Diagnosis (years)	Year of Diagnosis	Implants Used	Implant Type at Diagnosis	Exposure Time to Diagnosis (years)	Breast Cancer Laterality	ALCL Laterality	Presenting Symptom	BIA-ALCL Treatment
1	51.40	2015	Inamed 168	Textured	6.43	Left	Left	Seroma	Capsulectomy
2	77.33	2014	McGhan 153, Allergan 15*	Smooth*	10.26	Bilateral	Right	Mass	Capsulectomy
3	72.79	2016	McGhan 120	Textured	12.92	Bilateral	Right	Seroma	Capsulectomy
4	51.22	2010	Inamed 468	Textured	6.46	Right	Left	Seroma	Capsulectomy
5	61.74	2017	McGhan 468	Textured	9.53	Bilateral	Right	Seroma	Capsulectomy
6	50.61	2017	Inamed 410	Textured	10.09	Right	Left	Seroma	Capsulectomy
7	50.39	2015	McGhan 468	Textured	9.14	Right	Left	Seroma	Capsulectomy
8	54.57	2013	McGhan 468	Textured	11.37	Left	Right	Seroma	Capsulectomy
9	52.90	2018	McGhan 468	Textured	10.82	Right	Left	Chest Wall Pain	Capsulectomy
10	68.92	2018	McGhan 153, Inamed 410	Textured	14.58	Left	Left	Seroma	Capsulectomy
11	58.31	2018	McGhan 120	Textured	15.52	Left	Left	Mass	Capsulectomy
Median	54.57 (50.39–77.33)				10.26 (6.43 – 15.52)

BIA-ALCL = breast implant-associated anaplastic large cell lymphoma

^*Smooth device placed 1 year prior to diagnosis following 10 years of exposure to textured implant. Exchange to smooth device occurred due to recurrent seroma around textured implant.

The overall incidence of BIA-ALCL among patients who underwent implantation of textured implants (Table 4) was 1.79 per 1,000 patients (1 in 559) and 1.15 per 1,000 implants (1 in 871). When exposure time was estimated per the Follow-Up method, the estimated incidence rate of BIA-ALCL was 1 per 4,020 patient-years and 1 per 6,120 implant-years of textured implant exposure. When exposure time was estimated using the Vital Status method, these estimates decreased to 1 per 5,514 patient-years and 1 per 8,356 implant-years of textured implant exposure.

Table 4:

Estimated Incidences and Incidence Rates among Patients Exposed to Textured Implants

Units	Overall		Follow-Up Exposure		Vital Status Exposure
	Incidence Per Patient	Incidence Per Implant	Patient-Years Incidence Rate	Implant-Years Incidence Rate	Patient-Years Incidence Rate	Implant-Years Incidence Rate
1: X	1: 559	1: 871	1: 4,020	1: 6,120	1: 5,514	1: 8,356
X per 1000	1.79 per 1,000	1.15 per 1,000	0.25 per 1,000	0.16 per 1,000	0.18 per 1,000	0.12 per 1,000

Table 5 and Figure 1 demonstrate the cumulative incidence of BIA-ALCL. The first BIA-ALCL events occurred within 6–8 years of exposure. Most BIA-ALCL diagnoses occurred at 10–12 years after implantation, which includes the median time to diagnosis. In this time period, assuming exposure time per the Follow-Up method, the cumulative incidence was 4.35 (95% CI, 1.18–7.50) per 1,000 patients and 3.00 (95% Cl, 0.80–5.19) per 1,000 implants. However, when assuming exposure time per the Vital Status method, estimated cumulative incidences were notably lower: 2.65 (95% Cl, 0.77–4.35) per 1,000 patients and 1.77 (95% Cl, 0.51–3.03) per 1,000 implants. After 14–16 years of textured implant exposure, cumulative incidence plateaued per the Follow-Up exposure method at 9.40 (95% Cl, 2.70–16.06) per 1,000 patients and 6.66 (95% Cl, 1.85–11.46) per 1,000 implants. Per the Vital Status exposure method, these estimates were 4.82 (95% Cl, 1.70–7.93) and 3.31 (95% CI, 1.14–5.48), respectively.

Table 5:

Life Table of Per-Patient and Per-Implant BIA-ALCL Incidence in 2-Year Intervals and Median Exposure Time Among Cases

		Follow-Up Exposure						Vital Status Exposure
		Per-Patient Analysis			Per-Implant Analysis			Per-Patient Analysis			Per-Implant Analysis
2-Year Interval	Event (n)	Risk (n)	Censored (n)	Incidence Per Thousand (95% CI)	Risk (n)	Censored (n)	Incidence Per Thousand (95% CI)	Risk (n)	Censored (n)	Incidence Per Thousand (95% CI)	Risk (n)	Censored (n)	Incidence Per Thousand (95% CI)
0–2	0	5249	900	0	8051	1538	0	5825	324	0	9060	529	0
2–4	0	4312	938	0	6531	1522	0	5093	732	0	7871	1189	0
4–6	0	3371	940	0	5059	1470	0	4414	679	0	6777	1094	0
6–8	2	2508	865	0.63 (0–1.50)	3735	1328	0.42 (0–1.00)	3686	727	0.47 (0–1.12)	5628	1149	0.31 (0–0.73)
8–10	2	1660	842	1.55 (0–3.09)	2405	1322	1.04 (0–2.08)	2786	897	1.06 (0.01–2.11)	4148	1476	0.70 (0.01–1.39)
10–12	4	1050	606	4.35 (1.18–7.50)	1471	930	3.00 (0.80–5.19)	2016	766	2.65 (0.77–4.53)	2919	1225	1.77 (0.51–3.03)
12–14	1	682	367	5.49 (1.61–9.35)	945	525	3.82 (1.10–6.53)	1516	499	3.21 (1.03–5.39)	2142	776	2.17 (0.69–3.64)
14–16	2	395	286	9.40 (2.70–16.06)	537	408	6.66 (1.85–11.46)	1025	489	4.82 (1.70–7.93)	1418	722	3.31 (1.14–5.48)
16–18	0	176	218	9.40 (2.70–16.06)	228	307	6.66 (1.85–11.46)	593	432	4.82 (1.70–7.93)	785	633	3.31 (1.14–5.48)
18–20	0	54	122	9.40 (2.70–16.06)	63	165	6.66 (1.85–11.46)	239	356	4.82 (1.70–7.93)	295	492	3.31 (1.14–5.48)
Cumulative Incidence Per Thousand at Median Exposure Time of 10.26 years (95% CI)
		2.79 (0.48–5.11)			1.90 (0.32–3.49)			1.80 (0.34–3.26)			1.19 (0.22–2.17)

BIA-ALCL = breast implant-associated anaplastic large cell lymphoma

This life table describes the survival for all textured implants. Risk describes the number of implants at risk of developing BIA-ALCL at that interval in time. Event describes the number of implants that led to ALCL development. Censored refers to the number of patients who did not reach follow-up for that interval.

Figure 1:

Time-to-Event Analysis, BIA-ALCL diagnosis per patient or per implant by exposure to textured implants

* The Kaplan-Meier curve shows the gradual increase in percent incidence of the disease by patient exposure to textured implants. The 95% confidence intervals are shown in light, dashed lines.

Figure 2 demonstrates the trends in utilization of smooth versus textured implants from 1991 to 2017 at MSK. The number of implants used for breast reconstruction at our institution has steadily increased over time. Textured devices outnumbered smooth implants from 1991 to 2009. Beginning in 2009, however, there was a decrease in textured implant use with a concurrent increase in smooth implant utilization. Subsequently, the use of smooth implants outnumbered textured devices after 2011. This trend continues at our institution, and we rarely use textured devices now(data not shown).

Figure 2:

Utilization of breast implants by implant type at Memorial Sloan Kettering Cancer Center, 1991–2017

*This graph represents the gradual increase in volume of implant-based procedures from 1991 to 2017, also stratified by implant surface.

Discussion

This is the largest institutional series evaluating the incidence of BIA-ALCL using time-to-event analysis. Using 26 years of experience with alloplastic reconstruction at a single center, this long-term follow-up has created a setting to enable the appropriate determination of both the numerator and denominator for BIA-ALCL development. Our estimated incidences of 1.79 per 1,000 patients and 1.15 per 1,000 implants are higher than previous estimates.^{7, 8, 21, 23, 27} Our estimated incidence rates of 1 per 6,120 implant-years and 1 per 4,020 person-years of textured implant exposure are also higher than previously reported estimates.^{7, 8, 23} Furthermore, our estimated cumulative incidence at 14–20 years was 9.40 per 1,000 patients and 6.66 per 1,000 implants. Although the 95% CIs of these estimates is wide, these relatively high cumulative risk estimates are important when counseling women who have had textured implants for long periods of time.

As loss to follow-up was a major potential bias that would lead to an underestimate of exposure time, we performed a second calculation of exposure time as the time difference from initial placement of the permanent implant to the date of data collection or death, and we defined this interval as Vital Status exposure time. This estimation assumes that all patients were event free after their latest follow-up and up until death and allows us to perform a sensitivity analysis on our estimated incidence rate. This analysis also assumes that all patients who did not follow up were not diagnosed with BIA-ALCL at an outside institution. Our estimated incidence rates using the Vital Status exposure time analysis method were lower than our exposure time estimates at 1 in 5,514 patient-years and 1 in 8,356 implant-years. Similarly, the cumulative incidence up to and after 14–20-years was much lower using the Vital Status exposure time analysis method at 4.82 per 1,000 patients and 3.31 per 1,000 implants. Nevertheless, these estimates of incidence rate are higher than previous studies suggest, although additional work is necessary given the wide CIs.^{7, 8, 23}

Our findings are in contrast to those of several recent studies that reported lower incidence and incidence rates of BIA-ALCL.^{7, 8, 20, 21, 23, 27} Although these studies attempted to quantify BIA-ALCL incidence and incidence rates on an epidemiologic scale, their results are subject to bias and are likely underestimates of the true incidence due to a combination of the rarity of the condition, the long exposure time for disease development, and the passive data reporting systems. For example, de Jong et al. reported a country-wide incidence of 11 cases in 100,000–300,000 women and an incidence rate among women with breast prostheses of 0.1–0.3 per 100,000 women per year (1 per 1,000,000 to 1 per 333,000 person-years) in the Netherlands.²³ However, there appears to be no indication whether these rates relate specifically to textured implant exposure. Doren et al. reported an incidence of 1 per 30,000 women and an incidence rate of 2.03 per 1 million person-years (1 per 490,000 person-years), comparing the number of new diagnoses per year to the number of textured implants in the same year.⁷ However, their estimate is biased due to their data sources; one database was used to identify BIA-ALCL diagnoses while a separate database was used to identify the population at risk. For implant-specific incidence rate, the study by Loch-Wilkenson et al. showed an incidence rate of 1 per 25,640 implant-years among patients with Biocell textured products and 1 per 357,000 implant-years among patients with Siltex textured products.⁸ The study, however, also approximated exposure years using sales data, likely leading to an overestimate of exposure years.

Estimates of incidence rate that do not account for exposure time are likely to be inaccurate. Since BIA-ALCL takes time to develop, one cannot reasonably estimate the incidence of a given disease by comparing a given year’s annual diagnoses to that same year’s annual implants; doing so, perpetually inflates the “person-years” in the denominator. Previous estimates that used time-to-event analyses (i.e., Kaplan-Meier) did not adjust for censoring, a major limitation of registry or manufacturer records. Censoring occurs when there is incomplete follow-up and must be considered in any time-to-event analyses when examining a dynamic cohort. With these data sources, patients are not rigorously followed in terms of their implant exposure time or revision procedures, leading to incorrect estimates. In the context of BIA-ALCL, which is associated with implant-exposure time, the current estimates do not account for these considerations.

This is the largest study examining BIA-ALCL in a time-to-event fashion. Through almost 7 years of follow-up, the incidence of BIA-ALCL was nearly 0%. Following this time point, however, there was nearly an exponential rise in incidence. The cumulative incidence was relatively small at each year or event point, with a culmination of 6.66 per 1,000 implants beyond 14–16 years. This time interval is limited by a follow-up of 537 implants as a result of censorship due to loss to follow-up. Reasons for loss to follow-up could be transfer of care to another institution or voluntary lack of follow-up. Given the concern for possibly overestimating risk due to censoring, we also examined cumulative incidence using the Vital Status method, which assumed event-free exposure for all patients, with censoring only with death. This yielded a cumulative incidence of 3.31 per 1,000 implants at 14–16 years and beyond. Our study directly addressed loss to follow up through standard censoring in the Kaplan-Meier time to event analysis. Patients lost to follow up or who did not reach a time point in the analysis are presented in table 5 as “censored.” Such patients were then removed from the denominator of the risk calculation. This gives the most accurate understanding of the incidence rate, and is in accordance with standard practice for reporting patients lost to follow up in time to event analyses^{28, 29}.

In our study, incidence rate was analyzed by implant-year exposure and person-year exposure, in line with previous studies. Considering that textured devices may be associated with an increased risk of BIA-ALCL, patients who undergo bilateral breast reconstruction with textured devices as opposed to unilateral reconstruction could conceivably be at an increased risk for BIA-ALCL development. Stated differently, it is possible that each breast/implant is an independent event with its own inherent risk of BIA-ALCL development. Our study anecdotally supports this hypothesis, as 10 of the 11 patients in our cohort who developed BIA-ALCL had bilateral reconstructions, whereas only 61.6% of the overall cohort had bilateral reconstructions. Clearly, additional research is necessary to test this hypothesis.

Of note, textured implant sales data suggest that a large volume of procedures were performed from 1990 to 2005.⁷ Though rates of implant-based reconstruction have continued to increase since 2000,^{12, 15} sales of textured implants have decreased over that same period of time.⁷ Nevertheless, annual sales of textured implants were approximately 75,000 as recently as 2015.⁷ Our trend analysis of implant-based procedures shows that many of our patients are approaching 10 years of exposure to textured implants, thus necessitating patient education and follow-up care. The number of patients with BIA-ALCL will likely increase simply because more patients will have an exposure time similar to the median time of diagnosis found in this cohort (10.26 years) and other reports.^{16, 19, 30, 31}

Management options for patients with textured devices include continued implant monitoring, exchange for smooth devices, or conversion to autologous tissue reconstruction. It is unknown how the removal of a textured device impacts the likelihood of future BIA-ALCL development, although it is reasonable to hypothesize that the removal of the implant substantially decreases the risk. In our series, one patient had textured devices for 10 years, which were exchanged for smooth implants 1 year prior to diagnosis. It is possible that BIA-ALCL was present but undiagnosed at the time of exchange. Risk reduction has yet to be clearly quantified in such exchange procedures and may prove even more difficult than assessing the risk of developing BIA-ALCL itself. Additionally, estimates of incidence and incidence rate of BIA-ALCL appear to vary depending on the surface texturing of the device, suggesting that not all textured devices carry the same risk.^{8, 21}

At MSK, all patients with breast implants (both textured and smooth) have been contacted regarding the BIA-ALCL risk in textured implant devices and were encouraged to follow up with their plastic surgeon, particularly if any changes were noted in the reconstructed breast. Patients seen in clinic are carefully counseled regarding their estimated risk and options for management. Our reconstructive practice has shifted to utilizing primarily smooth surface implants. The institutional stance to contralateral prophylactic mastectomy and reconstruction has not significantly changed.

This study has a number of limitations. It is a single-institution experience with breast reconstruction, which may limit the generalizability of the results. The incidence of BIA-ALCL in this patient population may differ from that of patients who received textured implants for cosmetic reasons alone. The study is retrospective and limited by reviewer bias, medical record completeness, and selection bias that manifests in several ways. First, due to patients transferring care to other institutions, this study may underestimate the true incidence of BIA-ALCL, as patients who develop BIA-ALCL may be captured elsewhere. Second, the mean follow-up time for smooth compared with textured devices was significantly shorter in our study. This makes direct comparisons and conclusions difficult. Third, our results showed a significant loss to follow-up before the median time of BIA-ALCL diagnosis, which may have led to both an underestimation of risk if patients developed BIA-ALCL and sought care elsewhere or an overestimation if patients who did not follow up were healthy without adverse events. These opposing consequences of attrition bias both exemplify the challenges associated with analyzing this disease and underscore the importance of long-term follow-up. Additionally, this is a more in-depth analysis of the patients presented by Cordeiro et al., who comprised a majority of this study cohort²². The current findings should be viewed along with the initial report; however, the data collected, as well as the analyses, were completely independent to reduce risk of additional bias. Despite these limitations, this study attempts to better understand the risk associated with textured implants and the development of BIA-ALCL.

Conclusions

BIA-ALCL following textured implant placement may have a higher incidence rate than initially estimated. Given the number of patients with textured devices, the prevalence of BIA-ALCL will likely increase in the coming years. In addition to understanding the relationship between textured devices and BIA-ALCL development, an understanding of the incidence reduction imparted with implant exchange or autologous tissue conversion remains elusive and warrants close examination.