Systematic Under-Ascertainment of Anaplastic Large Cell Lymphoma Cases in Cancer Registries: Report from a Comprehensive Cancer Center

Summary

Introduction:

Breast implant-associated anaplastic large cell lymphoma (ALCL) has been rapidly rising in the US and around the world, leading to a mandated “black-box” label on all silicone- and saline-filled implants by the Food and Drug Administration (FDA). Because regulatory decisions in the US and around the world have been influenced primarily by risk estimates derived from cancer registries, it is important to determine their validity in identifying cases of ALCL.

Method:

We reviewed all cases of ALCL submitted to the New York State Cancer Registry from a large comprehensive cancer center in New York City from 2007 to 2019. To determine the possibility of misdiagnosis or under-diagnosis of ALCL cases reported to cancer registries, we accessed the sensitivity and specificity of the ICD-O-3 codes 9714 (ALCL) and 9702 (Mature T-cell lymphoma, not otherwise specified [T-NOS]) to identify pathologically-proven ALCL.

Results:

We reviewed 2,286,164 pathology reports from 47,466 unique patients with primary cancers. Twenty-eight cases of histologically-proven ALCL were identified. The sensitivity and specificity of the ICD-O-3 code 9714 (ALCL) were 82% and 100%, respectively. The sensitivity of the combined codes 9714/9702 (ALCL/T-NOS) was 96% and the specificity was 44%.

Conclusion:

Previous epidemiological studies that influenced regulatory decisions by the FDA may have systematically underestimated the risk of ALCL by at least 20%. We encourage updated global risk estimates of breast ALCL using methods that ensure adequate case ascertainment.

Lay Summary

The incidence of breast implant-associated anaplastic large cell lymphoma (ALCL) has been rapidly rising worldwide. We reviewed all cases of ALCL submitted to the New York State Cancer Registry from a large comprehensive cancer center, and found that previous studies may have underestimated the risk of ALCL by at least 20%.

To the Editor,

Breast implant-associated anaplastic large cell lymphoma (ALCL) has been rapidly rising in the US and around the world, leading to a mandated “black-box” label on all silicone- and saline-filled implants in 2022.¹ Much of the data that led to regulatory action by the FDA was based on epidemiological studies using national cancer registries.^2–4 The FDA’s initial warning in 2011 included estimated incidence rates calculated from the National Cancer Institute (NCI)’s Surveillance, Epidemiology, and End Results (SEER) Program. Estimates of breast ALCL in the US have tripled since that time.¹ Though pathological misclassification of rare lymphoma subtypes may be common in cancer registries, consequential regulatory decisions are made based on accurate counts of these rare histopathologies. Therefore, to determine the possibility of misdiagnosis or under-diagnosis of ALCL cases reported to cancer registries, we reviewed all cases of ALCL submitted to the New York State Cancer Registry from a large NCI-designated Comprehensive Cancer Center in New York City.

We assessed the sensitivity and specificity of the ICD-O-3 codes 9714 (Anaplastic large cell lymphoma, T cell and Null cell type [ALCL]), commonly used in the medical literature⁵, and 9702 (Mature T-cell lymphoma, not otherwise specified [T-NOS]). We collected all cases newly diagnosed from January 2007 to December 2019 in our institutional cancer registry (Supplementary File). We also performed a text search of all pathology reports for terms synonymous with the codes. All pathology reports were individually reviewed.

Of 47,466 unique patients with primary cancers reported to our state registry, we identified 23 and 61 unique primary tumors coded as ICD-O-3 9714 (ALCL) and 9714/9702 (ALCL/T-NOS), respectively. We identified additional cases of ALCL using a text search of our pathology reports. Out of 2,286,164 pathology reports, we identified 28 cases of histologically-proven ALCL. The sensitivity and specificity of ICD-O-3 code 9714 (ALCL) to detect ALCL were 82% and 100%, respectively (Table 1). The sensitivity of the combined codes 9714/9702 (ALCL/T-NOS) was 96% and the specificity was 44%.

Table 1.

Sensitivity and Specificity of ICD-O-3 Codes to Identify Histologically-Proven ALCL in an Institutional Cancer Registry

	n	Sensitivity	Specificity
9714 (ALCL)	23	82%	100%
9714/9702 (ALCL/T-NOS)	61	96%	44%

ALCL- Anaplastic large cell lymphoma; T-NOS- T-cell lymphoma, not otherwise specified

Our findings demonstrate that use of the ALCL code alone may lead to a gross underestimate of risk in outcomes research, as it had a specificity of 100% but a sensitivity of only 82% and, therefore, produced undercounting of approximately one fifth of ALCL cases. Using the combined ALCL/T-NOS codes decreased the specificity to 44% but increased the sensitivity to 96%, suggesting that most cases would be ascertained. We, therefore, recommend that estimates of ALCL in outcomes research report a range of risk estimates that include ICD-O-3 codes for both ALCL and ALCL/T-NOS. It should be noted that the code T-NOS in SEER and SEER/National Program of Cancer Registries (NPCR) data occurs less frequently within the breast than in other anatomical sites.¹ We, therefore, expect that the specificity of the combined codes for ALCL/T-NOS is higher when the research question is limited to breast ALCL. We previously externally validated these codes by comparing SEER and SEER/NPCR data to FDA data.¹ We estimated 353 and 310 cases of breast ALCL were diagnosed through 2017 using SEER and SEER/NPCR data, respectively, compared with 333 cases reported to the FDA over a similar period. Together, these studies provide external (FDA) and internal (histological review) validation of the ALCL and ALCL/T-NOS codes.

We recently reported an increase in breast ALCL incidence rate from initial FDA estimates of 3 cases for 2001–2007 to 14.5–19.6 cases for 2012–2018 per 100 million persons per year. The lower range estimate only included ALCL cases and could, therefore, be directly compared to previous FDA estimates, while the upper range included cases of ALCL/T-NOS.¹ Some well-cited and influential epidemiological studies that used registry data in the US and Denmark solely used the ICD-O-3 code 9714 to identify cases of breast ALCL for incidence rate calculations.^2–4 Based on our current and previous studies, incidence rates and risk calculations may have been systematically underestimated.¹ This systematic error may exacerbate greater discrepancies in reported incidence rates and risk calculations that are due to the rapidly increasing rate of ALCL in the US and around the world. We, therefore, encourage more contemporary studies of global risk estimates of breast ALCL to use methods that ensure adequate case ascertainment.

Limitations of this study include pathological review conducted at a single institution. It is unknown if practice patterns at our academic institution are generalizable to the US. However, we expect that under-ascertainment of cases may be greater at community and non-academic centers. This limitation of our study only further emphasizes the importance of providing a range of risk estimates for ALCL that includes T-NOS.

In conclusion, cancer registry studies of ALCL that exclude T-NOS may lead to a systematic underestimation of the risk of ALCL. We encourage health outcomes researchers who use cancer registries to provide a range of risk estimates that include ALCL and ALCL/T-NOS.