NEJM author admits to skewing patient protocol in CT lung cancer screening trial
In the correction, which appeared in Thursday's issue of the journal, lead author Claudia Henschke, MD, and colleagues from the Weill Medical College of Cornell University in New York City, submitted the clarification, regarding the selection process, which led to false survival rates in the initial trial.
In March, a New York Times investigation revealed that the study, which found that 80 percent of lung cancer deaths could be prevented through widespread use of CT scans, was largely funded by a cigarette company—Vector Group. In response, the NEJM editors said they would more actively seek out funding disclosures as of April.
At the time, NEJM editor, Robert Schwartz, MD, and colleagues wrote that “we and our readers were surprised to learn that the source of the funding of the charitable foundation was, in fact, a large corporation that could have an interest in the study results.”
On Thursday, Henschke wrote that at 37 of the 38 participating sites, people interested in participating in the trial were first interviewed and then given a precoded questionnaire by the study staff. Eligibility was assessed by means of a computer algorithm, and those who did not meet the eligibility criteria were excluded, she wrote. At the 38th site, the questionnaire was not administered, and therefore not recorded before enrollment.
Except for the 12 patients excluded after enrollment, no patients were excluded from the study after they had been enrolled on the basis of the computer algorithm. Inclusion of these 12 changes the 10-year survival rate for patients with lung cancer from the 80 percent reported previously for 484 patients to 81 percent for 496 patients. The other reported findings remain unchanged, according to Henschke.
The article reported that eight patients with clinical stage I lung cancer remained untreated and died within five years after diagnosis. However, only three had a pathological diagnosis of stage I lung cancer, she wrote. Another four had stage I disease confirmed on CT, but further workup was delayed despite repeated promptings, and pathological diagnosis was made only after the cancer had progressed to stage IV. The remaining patient had a solitary nodule on baseline CT that grew at a rate consistent with primary lung cancer, refused biopsy and treatment, and died of lung cancer six months after the last CT showing lung cancer. Thus, all eight patients died from lung cancer within five years after their actual or potential diagnosis during stage I.
“Since, however, pathological diagnosis of lung cancer was required by the International Early Lung Cancer Action Project (I-ELCAP) investigators, I should have classified four of the eight patients as having stage IV lung cancer and the remaining patient who had not received a pathological diagnosis during stage I as having an interim diagnosis. The remaining 483 patients received an antemortem pathological diagnosis of their lung cancer,” Henschke wrote. “Thus, the correct number of patients who were untreated and had a diagnosis of stage I lung cancer is three, not eight, and the total number of patients who had clinical stage I lung cancer is 407, not 412.”
These corrections increased the 10-year Kaplan–Meier survival rate for clinical stage I lung cancer from 88 percent to 90 percent. The overall Kaplan–Meier survival rate remained the same, since all patients with any stage of lung cancer were included in that analysis, according to the lead author.