How Low Can You Go: Exploring the Role of Low-Dose Computed Tomography in Interstitial Lung Disease Detection

Lung cancer screening using low-dose computed tomography (LDCT) is being implemented globally and has been shown to reduce lung cancer mortality (1). Despite clear recommendations (2, 3), standardization for the identification and management of incidental findings is lacking. However, the widespread adoption of LDCT presents a unique opportunity to detect early interstitial lung disease (ILD) in an at-risk population of current and former smokers. Interstitial lung abnormalities (ILA), which may represent early or subclinical ILD, identified on LDCT have been associated not only with increased lung cancer incidence and lung cancer–specific mortality (4) but also with adverse respiratory outcomes, including lung function decline, fibrotic progression, and mortality (5–7). Despite this, ILA and ILD are often inconsistently reported by radiologists, with varying follow-up recommendations provided (8). Although the clinical significance of ILD is well established, the implications of nonfibrotic ILAs remain less clearly defined. It may therefore be more impactful to focus research efforts on fibrotic features, particularly those suggestive of usual interstitial pneumonia (UIP), to improve the diagnostic utility of this lower-resolution imaging modality.

In this issue of AnnalsATS, Wang and colleagues (pp. 1314–1320) sought to achieve this by investigating whether quantitative ILD measures obtained through Computer-Aided Lung Informatics for Pathology Evaluation and Ratings (CALIPER) are associated with clinically meaningful outcomes in the National Lung Screening Trial. A deep learning–based classifier for UIP (DL-UIP) was also used (9). Their large 11,518-participant cohort analysis demonstrated increased all-cause mortality among individuals with CALIPER-derived ground-glass opacity and reticulation, and particularly among those with honeycombing (hazard ratio, 6.23; 95% confidence interval [CI], 4.23–9.16; P < 0.001).

Notably, reticular abnormalities were also linked to increased lung cancer–specific mortality. The prevalence of DL-UIP was low at 0.2%, but DL-UIP positivity was significantly associated with mortality (hazard ratio, 4.75; 95% CI, 2.50–9.04; P < 0.001).

Although LDCT can detect ILAs, high-resolution computed tomography (HRCT) remains the gold standard imaging modality for ILD evaluation (2, 10). Nevertheless, this study highlights the potential of LDCT as a population-based screening tool for ILD. Imagine a scenario in which asymptomatic at-risk individuals have ILA detected through LDCT, receive timely referral to pulmonary specialists, undergo HRCT confirmation of clinical ILD, and begin earlier intervention—ultimately leading to improved clinical outcomes.

There are several clear advantages of using LDCT for healthcare screening. For many individuals, lung cancer screening may represent the only opportunity to undergo chest imaging, making it critical for clinicians to extract all clinically relevant information, from coronary artery calcium scores to thyroid nodules to ILA. Compared with HRCT, LDCT delivers lower radiation exposure and does not require expiratory or prone imaging. Importantly, LDCT screening targets a high-risk population of current and former smokers who are more likely to develop ILD and other smoking-related comorbidities. Moreover, many lung cancer screening programs are implemented through academic centers, providing infrastructure that could support further research into early prevention trials in individuals with ILAs (11).

However, this approach is not without limitations. The exact false-positive and false-negative rates of ILD identification on LDCT remain unclear. Although Wang and colleagues have reported encouraging sensitivity and specificity estimates (77% and 84%, respectively), together with others (12), technique factors such as suboptimal inspiration can produce false interpretations. A careful balance between sensitivity and specificity must be reached, while considering the implications of overdiagnosis and the impact of lead time bias. Furthermore, the lower resolution of LDCT can result in imaging misclassification. Nonetheless, ILD findings can still be identified on dose settings as low as 20 mA and slice thickness up to 3–5 mm (13). Finally, although this study establishes a radiologic signal that correlates with prognostic outcomes, LDCT cannot yet be considered a diagnostic test for ILD. Further validation is required, but this represents a significant step forward in supporting the role of quantitative CT (qCT) and artificial intelligence–based classification tools in LDCT, paving the way for future research and clinical application.

To realize the potential of LDCT in ILD screening, several key steps must be taken (Figure 1). First, local lung cancer screening programs need to be established. Next, a global standardized approach to image interpretation, particularly incidental findings, must be developed, whether through radiologist scoring or qCT analysis. From a practical standpoint, incidental findings such as ILA require additional interpretation time and healthcare resources. If radiologists interpret these findings, there are costs and workload implications. Alternatively, using qCT software like CALIPER would necessitate access to centralized, validated, and affordable algorithms. Ideally, these would be either open source or financially attainable by public healthcare systems. If qCT is used, a single validated software and algorithm must reliably detect lung nodules, ILA, and other incidental findings such as calcium scores and osteoporosis. Description outputs must also be standardized, which may include morphological descriptors (i.e., ground-glass opacities, reticulation, honeycombing), patterns (i.e., UIP), and quantitative thresholds for total parenchymal involvement (2, 12). Local referral pathways will need to be established to ensure timely evaluation of symptoms and physiology and to distinguish ILA from ILD (14). Finally, as more patients with ILA will be identified, healthcare systems must plan for downstream resource implications, and future cost analysis needs to be performed to understand the resultant impact. Although these steps may seem ambitious, the study by Wang and colleagues reminds us that the future of ILD screening may already be here; what remains is the coordinated effort to harness this tool to its full potential.

Figure 1.

Key steps required for the implementation of low-dose computed tomography in the screening and early detection of interstitial lung abnormalities and interstitial lung disease. ILD = interstitial lung disease, LDCT = low-dose computed tomography; qCT = quantitative computed tomography.