Six factors may predict invasive breast cancer recurrence among patients diagnosed with and treated for ductal carcinoma in situ, according to results of a systematic review and meta-analyses published in Cancer Epidemiology, Biomarkers & Prevention.
Moreover, the review revealed common biases in studies assessing invasive recurrence after DCIS, including inadequate handling of cofounding factors.
“As most DCIS lesions will never make it to breast cancer, it is urgent to find out how to distinguish harmless from hazardous DCIS to spare many women the burden of needless treatment,” Jelle Wesseling, MD, PhD, professor of breast pathology at the divisions of diagnostic oncology and molecular pathology at the Netherlands Cancer Institute and Leiden University Medical Center, said in an interview with HemOnc Today. “To do so, we wanted to explore what we have learned from the past thus far. Progress to identify such prognostic factors has been slow. Current guidelines dictate surgical excision of DCIS, yet the majority of cases do not progress nor become life-threatening.”
Wesseling and colleagues searched PubMed and identified 1,781 articles published between 1970 and June 2018 that evaluated the risk for invasive recurrence among women with primary DCIS. Eligible studies included a minimum of 10 ipsilateral-invasive breast cancer events and at least 1 year of follow-up.
Forty studies met the inclusion criteria, with patient sample sizes ranging from 52 to 37,692 and median follow-up periods of 3.2 years to 15.8 years.
Researchers conducted meta-analyses to approximate the average effect size of the prognostic factors.
The meta-analyses identified six statistically significant prognostic factors for invasive recurrence after DCIS diagnosis: African American race (pooled estimate [ES] = 1.43; 95% CI, 1.15-1.79); premenopausal status (ES = 1.59; 95% CI, 1.2-2.11); detection by palpation (ES = 1.84; 95% CI, 1.47-2.29), involved margins (ES = 1.63; 95% CI, 1.14-2.32), high histologic grade (ES = 1.36; 95% CI, 1.04-1.77) and high expression of p16 (ES = 1.51; 95% CI, 1.04-2.19).
The researchers analyzed the risk for bias in these studies using six domains of the Quality in Prognosis Studies tool: study participation, study attrition, endpoint definition, prognostic factor measurement, confounding measurement and handling, and statistical analysis and reporting.
In 39 of the 40 studies evaluated, a high or moderate bias occurred in at least one of the six Quality in Prognosis Study domains. Twenty-two studies demonstrated a high risk for bias in at least one domain. Inadequate handling of confounders and poorly characterized study groups conferred the highest risk for bias.
According to the researchers, several other prognostic factors need to be confirmed in high-quality studies with larger patient numbers.
The researchers acknowledged limitations to their study, noting that the Quality in Prognosis Study tool required subjective assessment in assigning scores for the six domains. The wide variation in prognostic factors among studies meant that the evidence relied upon a few available studies. Additionally, studies categorized as high quality were not permitted to have high risk for bias in any of the Quality in Prognostic Studies domains.
Future studies also will need to assess the etiology of the recurrent cancers, the researchers noted.
“We found that insufficient handling of confounding factors — especially treatment for DCIS, and poorly described study groups — were the two most frequently occurring biases,” Wesseling told HemOnc Today. “We were aware of the high frequency of biases among previous studies. Some biases are inevitable, as it can be difficult to establish fully annotated cohorts, but others are easy to prevent. Our goal was to increase awareness and help the research community avoid these biases in the future.” – by Jennifer Byrne
Ryser MD, et al. J Natl Cancer Inst. 2019;doi:10.1093/jnci/djy220.
For more information:
Jelle Wesseling, MD, PhD, can be reached at Netherlands Cancer Institute, Department of Pathology, Plesmanlaan 121, 1066CX Amsterdam, the Netherlands; email: email@example.com.
Disclosures: The authors report no relevant financial disclosures.