News

RCTs vs. observational studies: mortality ‘strikingly’ different following RT

Key clinical point: Analysis of randomized trial data compared with observational data on mortality after breast cancer radiotherapy shows strikingly different results.

Major finding: Radiotherapy after mastectomy and axillary dissection was associated with lower mortality by clinical trial data (rate ratio, 0.84; 95% CI, 0.76-0.94) but was associated with higher mortality by observational data (1.34; 95% CI, 1.31-1.37).

Data sources: Randomized evidence came from the Early Breast Cancer Trialists’ Collaborative Group, meta-analyses of 31 trials (n = 13,932); observational evidence came from the SEER data base (n = 393,840).

Disclosures: Dr. Henson reported having no disclosures. One of her coauthors reported ties to industry.

View on the News

Is there a battle between randomized clinical trials and observational studies?

The hierarchy of evidence establishes that meta-analyses and randomized clinical trials (RCTs) offer the highest quality of evidence, and RCTs clearly offer the best assessment of effects of therapy. Observational studies, on the other hand, have considerable limitations, yet they have helped establish key causal relationships. Large prospective cohort studies, such as Framingham Heart Study, the National Child Development Study, and the Nurse’s Health Study, among others, continue to provide important data.

The study by Henson et al. reopens the debate over the value of RCTs versus observational studies (J Clin Oncol. 2016 Jan 18. doi: 10.1200/JCO.2015.62.0294).

It is well known the nonrandomized studies can produce misleading results due to selection bias. Selection bias favoring treatment for healthier patients produces improved survival among treated patients. On the other hand, selection bias disfavoring treatment occurs in studies of patients with more aggressive tumors who receive more treatment, and despite interventions, have worse outcomes.

The observation that postmastectomy radiotherapy is associated with worse outcomes reflects a real phenomenon, likely explained by confounding by indication. Patients treated with radiotherapy likely have worse prognoses, and despite therapy, had worse outcomes.

In the Henson et al. study, detailed data on treatment, comorbid conditions, or other health determinants were not available. In recognizing the importance of observational data for comparative effectiveness research, we must also understand and account for the limitations.

There should be no battle between RCTs and observational data, as both can provide valid and important knowledge to help clinicians make decisions and deliver evidence-based compassionate care.

Dr. Mariana Chavez-MacGregor is assistant professor in the department of breast medical oncology at the University of Texas MD Anderson Cancer Center, Houston. Dr. Sharon Giordano is associate professor in the department of breast medical oncology at the university. These remarks were part of an editorial (J Clin Oncol. 2016 Jan. 18. doi: 10.1200/JCO.2015.64.7487). Dr. Chavez-MacGregor reported financial ties with Roche, Novarits, InVitae, Pfizer, Genomic Health, and Genentech/Roche. Dr. Giordano reported having no disclosures.


 

FROM THE JOURNAL OF CLINICAL ONCOLOGY

References

Analyses of mortality after breast cancer radiotherapy using randomized clinical trial data versus observational data produced strikingly different results, according to researchers.

Analyses of randomized data indicated radiotherapy reduced mortality after breast-conserving surgery and mastectomy in node-positive disease; by contrast, SEER data analyses showed radiotherapy was associated with a significantly larger reduction in breast cancer mortality after breast-conserving surgery but higher mortality after mastectomy. Among patients with node-positive breast cancer who underwent mastectomy and axillary dissection, radiotherapy was associated with lower mortality by clinical trial data (rate ratio, 0.84; 95% CI, 0.76-0.94) but was associated with higher mortality by observational data (1.34; 1.31-1.37).

Furthermore, analyses of randomized trial data indicated increased mortality from heart disease (1.27; 1.12-1.44) and lung cancer (1.78; 1.30-2.46) following radiotherapy, but analyses of SEER data indicated reduced mortality from heart disease (0.56; 0.53-0.60) and lung cancer (0.86; 0.75-0.99) associated with radiotherapy.

“It is not plausible that these negative associations in the SEER data are causal and, clearly, they are strongly influenced by factors other than the effect or radiotherapy,” wrote Dr. Katherine Henson of the University of Oxford (England), and colleagues (J Clin Oncol. 2016 Jan 18. doi: 10.1200/JCO.2015.62.0294).

“Randomized trials are needed wherever possible to investigate the effect of treatment on mortality from the original cancer,” according to the investigators, and, “selection biases can be problematic even when analyzing treatment-related toxicities using observational data.”

Analyses of randomized data demonstrated reduced breast cancer mortality associated with radiotherapy after breast-conserving surgery (0.82; 95% CI, 0.75-0.90). Compared with randomized data, analyses of observational data showed a much greater reduction in mortality associated with radiotherapy after breast-conserving surgery (0.64; 95% CI, 0.62-0.66).

Randomized evidence came from the Early Breast Cancer Trialists’ Collaborative Group, meta analyses of 17 trials (n = 10,801) of radiotherapy after breast-conserving surgery, 14 trials (n = 3,131) of radiotherapy after mastectomy, and 78 trials (n = 42,080) of mortality from causes other than breast cancer. Observational evidence came from the SEER data base (n = 393,840).

The researchers offered plausible explanations for selection bias that may have resulted in the divergent results calculated using observational data. Because radiotherapy following mastectomy is indicated only in the presence of adverse disease characteristics, patients who did not receive radiotherapy may have survived longer because of especially favorable characteristics, despite of lack of radiotherapy.

“When evaluating rare late effects for which sufficient randomized evidence cannot reasonably be obtained, analyses of observational data comparing treated and untreated patients may often be the only source of information, but they must always be interpreted with considerable caution,” wrote the investigators.

Next Article:

   Comments ()