Annual Brain MRI Surveillance for Detection of Brain Metastasis in Patients With Lung Cancer

Brain metastasis is a common and clinically important complication of lung cancer. Earlier detection may allow timely local therapy, including stereotactic radiosurgery, whole-brain radiotherapy, neurosurgical resection, or systemic treatment. However, the population-level benefit of routine brain MRI surveillance after lung cancer diagnosis remains uncertain. Randomized evidence is limited, and surveillance practices vary across clinical settings.

This study will use a sequential target trial emulation framework to evaluate the clinical utility of periodic brain MRI surveillance in patients with lung cancer. The target trial of interest is a hypothetical randomized strategy comparing active periodic brain MRI surveillance with inactive surveillance among patients with newly diagnosed lung cancer who have no prior brain metastasis and no clinical indication for diagnostic brain MRI at the time of trial entry.

A new emulated trial will be created at monthly intervals after lung cancer diagnosis. At each trial baseline, patients will be eligible if they are alive, have not developed brain metastasis, have not undergone brain MRI during the preceding surveillance interval, and have no recent neurologic symptoms or clinical circumstances suggesting diagnostic rather than surveillance MRI. Patients with symptoms or diagnostic indications may be excluded from that specific trial but may become eligible again in later trials if they meet eligibility criteria. Patients who develop brain metastasis or die will be permanently excluded from subsequent trials.

The main exposure will be active brain MRI surveillance. In the primary analysis, patients who undergo brain MRI during a given monthly trial will be classified as entering an active surveillance state for one year. After one year, they will return to the inactive surveillance state unless they undergo another surveillance brain MRI. A sensitivity analysis will define the active surveillance state as lasting two years after brain MRI.

Diagnostic brain MRI will be distinguished from surveillance brain MRI using claims-based clinical criteria. Brain MRI will be classified as diagnostic if relevant neurologic symptoms, emergency department use, hospital admission, neurologic consultation, or prescriptions for medications commonly used for intracranial edema or seizure management occur within predefined time windows around the MRI date. Symptoms and clinical indicators will include signs of increased intracranial pressure, focal neurologic deficits, seizures, and altered cognition or consciousness, based on ICD-10 diagnosis codes and related treatment patterns.

Propensity score matching will be performed separately within each monthly trial. Active and inactive surveillance groups will be matched 1:1 using a caliper of 0.2 standard deviations of the logit of the propensity score. The propensity score model will include age, sex, diabetes, hypertension, dyslipidemia, chronic obstructive pulmonary disease, chemotherapy, radiotherapy, surgery, and year of cancer diagnosis. This design will account for time-updated changes in treatment and clinical status at each trial baseline.

The primary outcome will be incident brain metastasis, defined using diagnostic codes for brain metastasis. Secondary outcomes will include asymptomatic brain metastasis, potentially treatable brain metastasis, receipt of brain metastasis-directed treatment, and all-cause mortality. Among patients who develop brain metastasis, outcomes will be compared according to whether brain metastasis was detected during an active surveillance period or outside an active surveillance period.