CT scans to measure lung tumors can be unreliable, potentially leading patients and doctors to believe cancer is growing when it’s not, according to a study.

In principle, that could mean stopping a treatment that is actually keeping the tumor in check, researchers said in the study, which they said was the first to test how reliable lung cancer scans are, and appeared in the Journal of Clinical Oncology.

The patient and the doctor both need to understand that small changes don’t necessarily mean much, said Gregory Riely, at Memorial Sloan-Kettering Cancer Center in New York.

Changes of up to 10 percent can happen simply as a result of the inherent variability of CT imaging.

For the study, the Sloan-Kettering team asked patients with late-stage lung cancer if they’d be willing to have two chest CT scans within minutes. Thirty-three patients said ‘yes’.

Then the researchers gave the images to three radiologists who had no idea the scans had been repeated before the tumors could have grown or shrunk appreciably.

But the radiologists reported many changes, ranging from a 23 percent shrinkage to a 31 percent growth.

Overall, three percent of the tumors appeared to have grown so much that doctors would diagnose disease progression according to common criteria. And the smaller the tumor, the bigger the variation.

Riely said some doctors will make treatment decisions based on tiny changes seen on scans, although that might be a costly mistake, according to the study’s findings.