Head and Neck Radiation Therapy Patterns of Practice Variability Identified as a Challenge to Real-World Big Data: Results From the Learning from Analysis of Multicentre Big Data Aggregation (LAMBDA) Consortium

Abstract

Purpose: Outside of randomized clinical trials, it is difficult to develop clinically relevant evidence-based recommendations for radiation therapy (RT) practice guidelines owing to lack of comprehensive real-world data. To address this knowledge gap, we formed the Learning from Analysis of Multicenter Big Data Aggregation consortium to cooperatively implement RT data standardization, develop software solutions for data analysis, and recommend clinical practice change based on real-world data analyzed. The first phase of this "Big Data" study aimed at characterizing variability in clinical practice patterns of dosimetric data for organs at risk (OARs) that would undermine subsequent use of large-scale, electronically aggregated data to characterize associations with outcomes. Evidence from this study was used as the basis for practical recommendations to improve data quality.

Methods and materials: Dosimetric details of patients with head and neck cancer treated with radiation therapy between 2014 and 2019 were analyzed. Institutional patterns of practice were characterized, including structure nomenclature, volumes, and frequency of contouring. Dose volume histogram (DVH) distributions were characterized and compared with institutional constraints and literature values.

Results: Plans for 4664 patients treated to a mean plan dose of 64.4 ± 13.2 Gy in 32 ± 4 fractions were aggregated. Before implementation of TG-263 guidelines in each institution, there was variability in OAR nomenclature across institutions and structures. With evidence from this study, we identified a targeted and practical set of recommendations aimed at improving the quality of real-world data.

Conclusions: Quantifying similarities and differences among institutions for OAR structures and DVH metrics is the launching point for next steps to investigate potential relationships between DVH parameters and patient outcomes.

Figure 1

Interinstitutional variability in (a) naming and (b) use of the 13 structures used by 3 of 5 institutions for at least 50% of plans. Musc_Constricts includes Musc_Constrict_S, Musc_Constrict_I, Pharynx, and Pharynx-PTV. Each structure bar represents the range of values (number of name variants or plans with structure segmented) of the 5 institutions. Glnd_Submands and Parotids indicates contouring both left and right structures.

Figure 2

Variability in contoured structure volumes is evident in (a) the wide range of values relative to median. Median volume values are provided adjacent to structure names on the Y axis. (b) Three groupings were identified for low (green shading), moderate (gray shading), and high (no shading) variability of contoured volumes based on k-means clustering of volumes according to intrainstitutional variability (<[Q3 – Q1]/median>) and interinstitutional variability (<CHD>).

Figure 3

Statistical dose volume histogram (DVH) curves illustrating variation of doses for (a) Glnd Submand_Low: a structure with low interinstitutional volume variability (<CHD> = 0.25 ± 0.059) and (b) Larynx: a structure with moderate interinstitutional volume variability (<CHD> 0.41 ± 0.1). The median (dashed line) and ranges encompassing 25% to 75% (dark pink), 15% to 85% (medium pink), and 5% to 95% (light pink) of the DVH curves from each institution are shown.

Figure 4

Histograms illustrating intrainstitutional and interinstitutional variation of 1 dose volume histogram metric: the mean (Gy) for (a) Glnd_Submand_Low and (b) Larynx.

Figure 5

Bayesian network analysis of predictive relationships (strong: solid lines; moderate: dashed lines) among volume-based dose volume histogram metrics for all institutions.