Indeterminate Pulmonary Nodules: Causes and Management

Most pulmonary nodules found on CT scans turn out to be harmless. More than 95% of all nodules picked up incidentally on chest imaging are benign, yet distinguishing the small fraction that harbor early-stage lung cancer from the overwhelming majority that do not is one of the trickiest judgment calls in medicine.¹ That uncertainty is what makes a nodule “indeterminate”: it sits in a gray zone where it cannot be confidently called benign or malignant based on a single scan alone. What happens next depends on the nodule’s size, shape, and density, on the patient’s personal risk profile, and increasingly on newer tools ranging from blood tests to artificial intelligence.

How Common Are Pulmonary Nodules

Pulmonary nodules are extraordinarily common, and the more imaging we do, the more we find. In a large retrospective cohort of over 120,000 CT scans, roughly half showed at least one nodule 4 mm or larger.² A review of 24 studies totaling nearly 31,000 participants confirmed that the rate of detection depends heavily on the imaging field of view and the minimum size threshold used for reporting.³ Many of these nodules are found by accident during scans ordered for entirely unrelated reasons, such as a chest CT after a car accident or imaging for abdominal pain. This “incidental” discovery is what most people experience: you went in for something else, and a spot showed up on your lung.

The cancer risk for any given nodule is low overall, but it rises sharply with size. Lung cancer is diagnosed in just under one in ten patients whose nodule measures more than 8 mm, while smaller nodules rarely turn out to be malignant.¹ This size relationship is one reason doctors do not rush to biopsy every tiny spot they see.

What Causes Benign Nodules

The vast majority of lung nodules that turn out to be harmless fall into a few familiar categories. The most common are granulomas, small clumps of inflammatory tissue that form in response to a past infection or an inhaled irritant. In parts of the United States where certain soil-dwelling fungi are endemic, old granulomas from histoplasmosis or coccidioidomycosis are an everyday finding on chest CTs. Calcified granulomas, hamartomas (benign tumors made up of a mix of cartilage, fat, and connective tissue), and scarring from healed infections are all well-recognized causes of nodules that show up on scans and stay put forever.⁴

Intrapulmonary lymph nodes, small collections of immune tissue within the lung itself, are another frequent source. They tend to be small, smooth, and located in predictable positions near the lung’s lower lobes. Rounded areas of atelectasis (collapsed lung tissue) and arteriovenous malformations round out the list of benign look-alikes. The challenge is that none of these can always be distinguished from early cancer on a single CT image, especially when the nodule is small.

When a Nodule Signals Cancer

Among nodules that do turn out to be malignant, the most common culprit by far is primary lung cancer, particularly adenocarcinoma. In one large cohort, among patients whose nodules were surgically removed or biopsied and found to be cancerous, adenocarcinoma accounted for about 94% of cases.² In patients who already have a history of cancer elsewhere in the body, a new lung nodule raises a second concern: metastasis. Whether a new spot is more likely to be a metastasis or a brand-new primary lung cancer depends on the original cancer type, smoking history, and the nodule’s appearance on imaging.⁵

People who smoke or formerly smoked are at higher risk for a primary lung cancer, while certain cancer types, such as melanoma, sarcoma, and colon cancer, are more prone to sending metastases to the lungs. This distinction matters because the treatment path is entirely different: a solitary new primary lung cancer may be curable with surgery, while scattered metastatic disease calls for systemic therapy.

What Imaging Features Tell Doctors

Radiologists look at several features on a CT scan to estimate whether a nodule is worrisome. Size is the starting point, but shape, border characteristics, and internal density all provide clues. A spiculated border, where the nodule has irregular, ray-like projections extending outward, is one of the strongest markers of malignancy. In one study comparing benign and malignant nodules, spiculation was present in about 87% of malignant nodules versus 55% of benign ones.⁶ Lobulation, where the nodule surface is scalloped rather than smooth, and uneven internal density were also significantly more common in cancerous nodules in the same study.

Calcification patterns can be reassuring. Dense, central, or “popcorn” calcification within a nodule usually points to a granuloma or hamartoma. Eccentric or stippled calcification is less reassuring and can occasionally be seen in cancers that have engulfed a pre-existing granuloma.⁴ Beyond shape and density, radiologists also look for “pleural tags,” thin lines connecting the nodule to the lung’s outer lining, which were observed in about 58% of malignant lesions compared with 27% of benign ones in the same pathology-correlated study.⁶

Ground-Glass Nodules and Their Slow Clock

Not all nodules are solid. Some appear hazy on CT, like frosted glass, and these “ground-glass nodules” (GGNs) behave very differently from their solid counterparts. Pure ground-glass nodules grow so slowly that their doubling time often exceeds three to four years. In a study tracking persistent pure GGNs over a median of more than six years, only about half had grown by the five-year mark, and their median volume doubling time was nearly 1,500 days.⁷

Even GGNs that have been stable for three full years are not guaranteed to stay that way forever. A separate study found that about 7% of patients with GGNs stable over the first three years experienced growth during longer follow-up. Risk factors for delayed growth included being 65 or older, a prior history of lung cancer, initial size of 8 mm or larger, and the presence of a solid component within the ground-glass area.⁸ This slow-motion biology is why guidelines often recommend extended surveillance for GGNs and why a “part-solid” nodule, one with both ground-glass and solid elements, is generally treated with more urgency than a purely ground-glass one.

Predicting Cancer Risk With Models

Because the decision about whether to biopsy, scan again later, or leave a nodule alone carries real consequences, doctors often use mathematical prediction models to estimate the probability of malignancy. The most widely studied include the Mayo Clinic model, the Brock University model, and the Herder model. These take into account variables like nodule size, patient age, smoking history, cancer history, and imaging characteristics to produce a probability estimate.

Validation studies have generally found that these models perform reasonably well, though none is perfect. In one large validation, the Mayo and Brock models performed similarly with area-under-the-curve (AUC) values near 0.90, both significantly outperforming the Veterans Association model.⁹ In another cohort, the Mayo Clinic model slightly outperformed the Brock model, though both tended to overestimate the probability of cancer.¹⁰ The Herder model, which adds PET/CT metabolic activity data to the clinical and imaging variables, has shown the highest accuracy in populations where that information is available, with an AUC exceeding 0.92.⁹

These models are useful as a starting framework, but they are not decision-makers. A model might peg a nodule’s cancer risk at 15%, which falls in a gray zone where watchful waiting, a PET scan, or a biopsy could all be reasonable. That is where clinical judgment and patient preferences enter the picture.

PET/CT and Metabolic Imaging

When a nodule’s appearance on CT is ambiguous, doctors often turn to PET/CT. This scan uses a radioactive sugar tracer that lights up areas of high metabolic activity. Because cancer cells burn through glucose faster than most normal tissue, a “hot” nodule on PET raises the suspicion of malignancy. Combined PET/CT has achieved accuracy as high as 93% in characterizing solitary lung nodules, outperforming CT and PET alone.¹¹

PET has important blind spots, though. Very small nodules and certain slow-growing cancers, including some adenocarcinomas and carcinoid tumors, may not take up enough tracer to light up. In a study of nodules found during lung cancer screening, PET/CT visual analysis achieved 84% sensitivity and 95% specificity, but when stricter metabolic thresholds were applied, sensitivity dropped sharply while specificity stayed high.¹² False positives can also occur with infections and inflammatory conditions, particularly granulomatous diseases and active fungal infections, which mimic the metabolic signature of cancer on PET.

Blood Tests and Biomarkers

A blood test that could reliably rule cancer in or out would transform nodule management, and researchers have made some progress in that direction. One approach measures autoantibodies, immune proteins the body produces in response to tumor-associated antigens sometimes years before cancer becomes visible. A validated autoantibody test has shown high specificity, around 90% to 96% depending on the positivity threshold, meaning a positive result is unlikely to be a false alarm.¹³ The trade-off is sensitivity: the test catches only a small fraction of cancers, roughly 9% to 16% depending on the threshold used.¹⁴

In practical terms, this means the test works best as a “rule-in” tool. A positive result can bump a patient’s estimated cancer probability up enough to justify moving more quickly toward biopsy or surgery, and cost-effectiveness modeling has found the approach compares favorably to CT surveillance alone.¹⁵ A negative result, however, does not reliably rule out cancer. Researchers are exploring other circulating biomarkers, including cell-free DNA and protein panels, but none has yet reached the same level of clinical validation.

Biopsy Options When Answers Are Needed

When imaging and risk models leave a nodule in the intermediate zone, tissue sampling becomes necessary. The two main approaches are transthoracic needle biopsy, where a needle is inserted through the chest wall under CT guidance, and navigational bronchoscopy, where a thin scope is threaded through the airways to reach the nodule from inside the lung.

A recent trial published in the New England Journal of Medicine directly compared the two. Navigational bronchoscopy achieved a diagnostic accuracy of about 79% versus roughly 74% for transthoracic biopsy, a difference that was small enough to confirm the bronchoscopic approach was not inferior.¹⁶ The more striking finding was the safety gap: complications occurred in about 5% of bronchoscopy cases versus 29% of transthoracic biopsies, driven almost entirely by pneumothorax, the collapse of part of the lung when air leaks into the chest cavity. Around 28% of transthoracic biopsy patients developed a pneumothorax compared to roughly 3% in the bronchoscopy group, and the need for a chest tube was about 12% versus under 1%.¹⁶

Nodule location matters in choosing between these procedures. Bronchoscopy works best for nodules near the airways, while transthoracic biopsy may be the only option for peripheral nodules that sit far from a bronchial path. For nodules that remain indeterminate after both imaging and biopsy attempts, minimally invasive surgical resection through video-assisted or robotic-assisted thoracic surgery can serve as both diagnostic and therapeutic, removing the nodule and sending it to pathology for a definitive answer.¹⁷

Artificial Intelligence in Nodule Assessment

AI tools trained on thousands of CT images are increasingly being tested as aids for nodule classification. Several algorithms have demonstrated strong performance in distinguishing benign from malignant nodules, and in one study, AI alone achieved about 90% accuracy, comparable to experienced physicians reading the same scans. When AI output was combined with multidisciplinary team review, accuracy climbed to about 97%, substantially above either alone.¹⁸ A multi-center clinical trial in China validated a model called DeepFAN, showing that it improved junior radiologists’ accuracy, confidence, and consistency when used as an assistant.¹⁹

AI’s practical promise lies less in replacing radiologists than in reducing the follow-up burden. If an algorithm can identify with high confidence which nodules are extremely unlikely to be cancer, fewer patients would need repeat scans and fewer would undergo unnecessary invasive procedures.²⁰ That said, regulatory approval is still catching up to the research, and most current guidelines have not yet incorporated AI into their formal decision algorithms.

The Overdiagnosis Problem

Finding nodules early sounds like an unqualified good, but more detection does not always translate into better outcomes. Overdiagnosis, where imaging identifies a cancer that would never have caused symptoms or shortened a person’s life, is a recognized concern in lung cancer screening. Slow-growing cancers, particularly some ground-glass-predominant adenocarcinomas, may never progress to the point of clinical relevance, yet once discovered they are almost always treated. Guidelines from the European Society of Thoracic Imaging have explicitly aimed to minimize unnecessary follow-up while still catching cancers early enough to prevent a stage shift that would change prognosis.²¹

False-positive scans carry their own costs. In the National Lung Screening Trial, about 2.7% of false-positive results in the low-dose CT arm led to an invasive procedure such as bronchoscopy or needle biopsy, and roughly one in ten of those invasive procedures resulted in at least one complication.²² These are small numbers in isolation, but when millions of scans are performed across a population, the downstream harm is not trivial. The updated 2017 Fleischner Society guidelines addressed this partly by raising the minimum nodule size threshold for recommending routine follow-up, giving clinicians more flexibility to tailor surveillance to individual risk.²³

The Emotional Weight of a Lung Nodule

The physical risks of finding a nodule are well quantified, but the psychological toll often gets overlooked. Being told you have a spot on your lung is frightening, and the period of watchful waiting can be genuinely distressing. In one study of patients with incidentally discovered nodules, nearly 60% reported some degree of anxiety.²⁴ The Watch the Spot trial found that roughly a third of patients reported mild distress, with another 9% experiencing moderate and about 7% experiencing severe distress.²⁵ Greater distress was linked to larger nodule size, lack of timely clinician notification, female sex, and a history of smoking.

This distress appears to influence treatment preferences in ways that can be medically counterproductive. A systematic review of shared decision-making in nodule management found that most patients surveyed preferred some type of biopsy over observation alone at all risk levels evaluated, including when the estimated chance of cancer was as low as 10%. Only about a quarter of patients estimated their cancer risk accurately; most grossly overestimated the likelihood that the nodule was malignant.²⁶ Clear communication about actual risk numbers, and what watchful waiting actually involves, can go a long way toward aligning patient preferences with evidence-based care.

Why Geography Matters More Than You’d Think

Where you live can dramatically change the probability that a lung nodule is an infection rather than cancer. In the United States, certain fungal infections are concentrated in specific regions: histoplasmosis in the Ohio and Mississippi River valleys, coccidioidomycosis in the Southwest, blastomycosis in the Upper Midwest, and cryptococcosis in the Pacific Northwest.²⁷ These fungal infections produce nodules that can look remarkably similar to cancer on CT and can even light up on PET scans, making geographic context an essential piece of the diagnostic puzzle.

Coccidioidomycosis nodules, for instance, were more likely to have smooth borders, cavitation, and satellite nodules on imaging, while lung cancer nodules tended to be spiculated and associated with underlying chronic lung disease.²⁸ But those features overlap enough that many nodules remain indeterminate even with an experienced radiologist reading them. Complicating matters further, traditional geographic boundaries for these fungi are shifting with climate change and human migration, meaning a patient in a historically non-endemic area can still present with a fungal nodule.²⁹ A thorough travel and exposure history remains one of the simplest and most underused tools for narrowing the differential.

The Economics of Nodule Surveillance

Tracking a nodule over time with repeated CT scans is far cheaper than surgery, but the costs are not zero, especially when multiplied across millions of patients. Cost-effectiveness analyses have started to weigh different management strategies against each other. One modeling study found that for subsolid nodules discovered during lung cancer screening, strategies that avoided treating pure ground-glass nodules entirely were cost-effective, and that setting a minimum solid-component size of 4 mm before triggering treatment was optimal under standard willingness-to-pay thresholds.³⁰ These findings reinforce the clinical impression that pure ground-glass nodules behave so indolently that aggressive early intervention often costs more than it saves, both financially and in terms of patient well-being.