In the opinion of this commentator, one of the more exciting developments in clinical investigation over the past several years is the increasing recognition of the importance of real-world data and its role for physicians, regulators, policy makers, patients, and society.
In the opinion of this commentator, one of the more exciting developments in clinical investigation over the past several years is the increasing recognition of the importance of real-world data and its role for physicians, regulators, policy makers, patients, and society. As critical as clinical trials have been to the establishment of a robust evidence-based medicine for the past half century, formal prospective studies can realistically address only a very small fraction of meaningful clinical questions and concerns.
Further, it is increasingly appreciated that in the oncology arena, clinical trials simply do not represent the overall large population of individuals who develop a malignant disease. This is particularly relevant in phase 3 randomized trials, in which the use of control and experimental groups is desired to be as homogeneous as possible to isolate the unique efficacy and toxicity profiles of the investigational strategy. As a result, patients with common comorbidities (ie, current and past cardiac, renal, and pulmonary dysfunction) are often excluded from these studies. Further, the risk of cancer increases with age, and older patients (> aged 70-75 years) are recognized to be relatively underrepresented in clinical cancer trials.
The advent of integrated electronic medical records and the development of robust health system, proprietary, and governmental databases has provided investigators with the opportunity to obtain meaningful answers to clinical questions. Leveraging these data is rapidly becoming the investigative world’s new reality. Clinical trials, including randomized studies will always play a critical role in medicine, but the future will almost certainly increasingly focus on obtaining reliable real-world data to inform decision-making or to raise issues and concerns about current clinical practice that will require future study.
Consider, for example, a recent issue of the high-impact general medical journal, Annals of Internal Medicine. Included in its contents were peer-reviewed articles that examined a national database from Denmark that suggested the use of the diabetic drug, metformin, by men preconception was strongly associated with major birth defects in their sons1; a population-based study from Norway and Sweden that examined the risk of the development of a myocardial infarction following the performance of outpatient ophthalmologic procedures (no increased risk demonstrated)2; a review of the large Kaiser Permanente Southern California database examining the subsequent development of shoulder conditions following vaccination—findings from which demonstrated a small but statistically significant risk3; and an evaluation of the Medicare database for the risk of anaphylaxis associated with differing formulations of intravenous iron, which confirmed an overall low risk and also revealed potential differences between various commercial products that may have clinical relevance.4
For each of these studies, a clinically relevant question was proposed. Some of the observations were reassuring, such as the findings for ophthalmologic procedures1,5 and intravenous iron use.4 Others, however, were concerning, such as the data on preconception use of metformin2,6 or shoulder conditions following vaccination.3 In each case, follow-up data, including examination of other large databases to confirm or refute the findings, are essential.5
The potential to use sources of real-world data in routine clinical decision-making is palpable.6 In the oncology sphere, it is not difficult to identify studies using real-world data that have major clinical implications. For example, clinical trials have revealed the importance of adjuvant checkpoint inhibitor therapy for high-risk malignant melanoma. But what happens when the therapy is employed in routine clinical practice outside the carefully controlled and monitored setting of a clinical trial?
A population-based registry in the Netherlands has collected outcomes from patients with melanoma who have received adjuvant therapy with a checkpoint inhibitor since 2019.7 In a recent report of 641 patients, favorable efficacy was observed (70.6% recurrence-free survival at 12 months), but a rather high rate of grade 3 or greater toxicity was noted (18%) and 61% of patients were required or elected to prematurely discontinue treatment. These data, which require confirmation in other real-world experiences, point to comparable efficacy but also the potential for greater risk of serious toxicity compared with clinical trial experience.
An examination of the Flatiron Health Record database including more than 2200 patients with ovarian cancers who had received a minimum of 2 lines of therapy provides a provocative response to these clinically relevant questions.8 Of the 222 patients who had completed second- or third-line platinum-based chemotherapy and had been followed for a minimum of 2 months, 20% had a documented BRCA mutation. Of the population with a BRCA mutation, 63% received maintenance therapy with a PARP inhibitor, 17% were treated with maintenance bevacizumab, and 20% were observed without any antineoplastic therapy. In the BRCA wildtype population the administration of a PARP inhibitor decreased to 40% and bevacizumab use increased modestly to 23%, and observation almost doubled at 36%.
Of note, during the study period the percentage of individuals receiving PARP inhibitor maintenance was found to increase every 3 months by 1.3%, with no change over time in bevacizumab administration. In addition to the presence of a BRCA mutation, younger patients were more likely to receive an oral PARP inhibitor compared with older women. This study provides insight into the use of particular antineoplastic strategies in ovarian cancer beyond their delivery within the confines of a clinical trial and suggests that patient populations in the real-world clinical setting are likely to be treated with different therapeutic approaches.