# Active bone marrow dose reconstruction in a large‐scale cohort study of cancer patients treated with photon radiotherapy

## TL;DR

This study shows how to use electronic medical records to calculate bone marrow radiation doses in cancer patients, which can help understand the link between radiotherapy and health risks like leukemia.

## Contribution

The study introduces a scalable method for reconstructing bone marrow dose and dose-volume metrics using electronic medical records in a large cohort of radiotherapy patients.

## Key findings

- Pelvic irradiation resulted in the highest mean active bone marrow dose per prescribed dose.
- Intracranial irradiation had the lowest mean active bone marrow dose per prescribed dose.
- Dose-volume analysis showed that mean dose estimates poorly represent the full dose distribution to bone marrow.

## Abstract

Previous studies have investigated the dose‐response relationship between external beam radiotherapy (EBRT) and leukemia by reconstructing the mean active bone marrow (ABM) dose as part of the exposure assessment. However, no prior study has leveraged electronic medical records (EMR) to reconstruct ABM dose and dose‐volume in an EBRT patient population.

To support future studies on the relationship between radiation exposure and adverse health effects among EBRT patients, we demonstrate methods to retrospectively calculate ABM dose and dose‐volume metrics in a large patient cohort using EMR.

We retrieved complete EMR for 639 individuals (five pediatric and 634 adult) from a retrospective cohort of cancer patients treated with photon radiotherapy between 2003 and 2011, covering a wide range of treatment sites and target volumes. Using a previously published methodology by our group, bone site contours were derived for each patient treatment session via three‐dimensional skeletal registration with a habitus‐matched computational phantom. Segmentation accuracy was assessed analytically and by visual inspection of the patient‐phantom skeletal overlap. ABM dose and dose‐volume metrics were computed using the derived skeletal segmentation and the dose distribution calculated by the treatment planning system.

Analytically and visually acceptable skeletal segmentation was achieved for 478 patients (three pediatric, 475 adult). Across all treatment sessions, pelvic irradiation resulted in the highest mean ABM dose per prescribed dose (distribution 25th–75th percentile: 0.1820–0.2870 Gy/Gy_Rx, n = 183), while intracranial irradiation produced the lowest among the categorized field types (0.0030–0.0391 Gy/Gy_Rx, n = 37). Dose‐volume analyses indicated that the mean ABM dose estimate provides a limited characterization of the ABM dose distribution, with most patients having a 99th percentile dose to marrow volume that was at least one order of magnitude higher than the mean dose estimate.

This study presents a scalable EMR‐based approach for reconstructing ABM dose and dose‐volume metrics in EBRT patients. By capturing clinically relevant dose heterogeneity, we have estimated a more informative exposure variable than mean ABM dose for studies on the role of radiotherapy in leukemogenesis and other adverse health effects. Future work will improve the generalizability of these methods for certain treatment regions and pediatric populations.