Many dermatopathologists offering a second opinion about melanocytic skin lesions prefer to have access to the first diagnostic report, but a controlled trial demonstrates that this has a powerful influence on perception, diminishing the value and accuracy of an independent analysis.
In a novel effort to determine whether previous interpretations sway second opinions, 149 dermatopathologists were asked to read melanocytic skin biopsy specimens without access to the initial pathology report. A year or more later they read them again but now with access to the initial reading.
The study showed that the participants, independent of many variables, such as years of experience or frequency with which they offered second options, were more likely to upgrade or downgrade the severity of the specimens in accordance with the initial report even if their original reading was correct.
If the goal of a second dermatopathologist opinion is to obtain an independent diagnostic opinion, the message from this study is that they “should be blinded to first opinions,” according to the authors of this study, led by Joann G. Elmore, MD, professor of medicine, University of California, Los Angeles. The study was published online recently in JAMA Dermatology.
Two-Phase Study Has 1-Year Washout
The study was conducted in two phases. In phase 1, a nationally representative sample of volunteer dermatopathologists performed 878 interpretations. In phase 2, conducted after a washout period of 12 months or more, the dermatopathologists read a random subset of the same cases evaluated in phase 1, but this time, unlike the first, they were first exposed to prior pathology reports.
Ultimately, “the dermatologists provided more than 5000 interpretations of study cases, which was a big contribution of time,” Elmore said in an interview. Grateful to their critical contribution, she speculated that they were driven by the importance of the question being asked.
When categorized by the Melanocytic Pathology Assessment Tool (MPAT), which rates specimens from benign (class 1) to pT1b invasive melanoma (class V), the influence of the prior report went in both directions, so that the likelihood of upgrading or downgrading went in accordance with the grading in the original dermatopathology report.
As a result, the risk of a less severe interpretation on the second relative to the first reading was 38% greater if the initial dermatopathology report had a lower grade (relative risk, 1.38; 95% CI, 1.19 – 1.59). The risk of upgrading the second report if the initial pathology report had a higher grade was increased by more than 50% (RR, 1.52; 95% CI, 1.34 – 1.73).
The greater likelihood of upgrading than downgrading is “understandable,” Elmore commented. I think this is consistent with the concern about missing something,” she explained.
According to Elmore, one of the greatest concerns regarding the bias imposed by the original pathology report is that the switch of opinions often went from one that was accurate to one that was inaccurate.
If the phase 1 diagnosis was accurate but upgraded in the phase 2 diagnosis, the risk of inaccuracy was almost doubled (RR, 1.96; 95% CI, 1.31 – 2.93). If the phase 1 report was inaccurate, the relative risk of changing the phase 2 diagnosis was still high but lower than if it was accurate (RR, 1.46; 95% CI, 1.27 – 1.68).
“That is, even when the phase 1 diagnoses agreed with the consensus reference diagnosis, they were swayed away from the correct diagnosis in phase 2 [when the initial pathology report characterized the specimen as higher grade],” Elmore reported.
Conversely, the risk of downgrading was about the same whether the phase 1 evaluation was accurate (RR, 1.37; 95% CI, 1.14 – 1.64) or inaccurate (RR 1.32; 95% CI, 1.07 – 1.64).
Downward and upward shifts in severity from an accurate diagnosis are concerning because of the likelihood they will lead to overtreatment or undertreatment. The problem, according to data from this study, is that dermatologists making a second opinion cannot judge their own susceptibility to being swayed by the original report.
Pathologists Might Be Unaware of Bias
At baseline, the participants were asked whether they thought they were influenced by the first interpretation when providing a second opinion. Although 69% acknowledged that they might be “somewhat influenced,” 31% maintained that they do not take initial reports into consideration. When the two groups were compared, the risk of downgrading was nearly identical. The risk of upgrading was lower in those claiming to disregard initial reports (RR, 1.29) relative to those who said they were “somewhat influenced” by a previous diagnosis (RR, 1.64), but the difference was not significant.
The actual risk of bias incurred by prior pathology reports might be greater than that captured in this study for several reasons, according to the investigators. They pointed out that all participants were experienced and board-certified and might therefore be expected to be more confident in their interpretations than an unselected group of dermatopathologists. In addition, participants might have been more careful in their interpretations knowing they were participating in a study.
“There are a lot of data to support the value of second opinions [in dermatopathology and other areas], but we need to consider the process of how they are being obtained,”. Elmore said. “There needs to be a greater emphasis on providing an independent analysis.”
More than 60% of the dermatologists participating in this study reported that they agreed or strongly agreed with the premise that they prefer to have the original dermatopathology report when they offer a second opinion. Elmore said that the desire of those offering a second opinion to have as much information in front of them as possible is understandable, but the bias imposed by the original report weakens the value of the second opinion.
Blind Reading of Pathology Reports Needed
“These data suggest that seeing the original report sways opinions and that includes swaying opinions away from an accurate reading,” Elmore said. She thinks that for dermatopathologists to render a valuable and independent second opinion, the specimens should be examined “at least initially” without access to the first report.
The results of this study were not surprising to Vishal Anil Patel, MD, director of the Cutaneous Oncology Program, George Washington University Cancer Center, Washington, DC. He made the point that physicians “are human first and foremost and not perfect machines.” As a result, he suggested bias and error are inevitable.
Although strategies to avoid bias are likely to offer some protection against inaccuracy, he said that diagnostic support tools such as artificial intelligence might be the right direction for improving inter- and intra-rater reliability.
Ruifeng Guo, MD, PhD, a consultant in the Division of Anatomic Pathology, Mayo Clinic, Rochester, Minnesota, agreed with the basic premise of the study, but he cautioned that restricting access to the initial pathology report might not always be the right approach.
It is true that “dermatopathologists providing a second opinion in diagnosing cutaneous melanoma are mostly unaware of the risk of bias if they read the initial pathology report,” according to Guo, but restricting access also comes with risks.
“There are also times critical information may be contained in the initial pathology report that needs to be considered when providing a second opinion consultation,” he noted. Ultimately, the decision to read or not read the initial report should be decided “on an individual basis.”
The study was funded by grants from the National Cancer Institute. Elmore, Patel, and Guo reported no relevant financial relationships.
JAMA Dermatol. Published online August 10, 2022. Abstract
Ted Bosworth is a medical journalist based in New York City.