Technological Advances Come With a Cautionary Note

Maurie Markman, MD

The recent sentencing of the Golden State Killer, an individual responsible for more than a dozen murders and 4 times as many rapes in California over many years, reminds us once again of the enormous potential for innovative technology to favorably transform our world while also raising a legitimate concern over how such advances may be employed for goals that would not be so positively viewed.¹ The killer, whose wanton spree of violence began in 1975, was not captured until investigators using a DNA sample from a double murder in 1980 were ultimately able to find the suspect through a genealogy website where they identified a match to a distant relative.¹ Although arguing that the result of this specific effort was inappropriate would be irrational, concern exists about the potential for individual privacy to be compromised if such technology were to be used for questionable purposes.

Similarly, in the realm of medicine, we see enormous opportunities for truly novel technologies to revolutionize the health and welfare of all members of society. A recent report revealed the essential equivalence of multichannel electrocardiogram readings obtained on a commercially available smartphone to those of a standard electrocardiogram, including for individuals with ST elevation and non-ST elevation myocardial infarctions.² One can easily imagine the individual and societal impact of the potential for much earlier diagnosis of acute coronary syndromes as well as essential intervention following the development of a serious cardiac event.

Potential for Artificial Intelligence

In the oncology arena, it is not difficult to find examples where rapidly proliferating artificial intelligence technology has transformed, or is on the verge of transforming, current management paradigms, particularly in the domain of cancer diagnostics. For example, a recent report of a novel system for the reading of breast cancer screening images revealed the rather impressive clinical utility of this artificial intelligence process.³ The specific algorithm was shown to be superior to 6 experienced radiologists who participated in this study and revealed a reduction in both false-positive and false-negative evaluations. Further, when a double-reading process (2 reviewers) was employed, the addition of the artificial intelligence platform resulted in an 88% reduction of the required effort of the second radiologist.

Similarly, a recent report of a “deep learning model” employing MRI of intratumoral and peritumoral regions of women undergoing radical surgery for cervical cancer revealed the benefits of this noninvasive approach in documenting lymph node involvement compared with standard MRI alone.⁴ Of note, the results of this imaging approach were shown to be highly prognostic for disease-free survival in this setting.

As a final brief example, investigators recently reported the development of a machine learning strategy that examined the molecular profiles of 7791 tumors involving 22 cancer types to determine whether these data could be helpful in the determination of the site of origin of a given malignancy.5 The algorithm was able to correctly diagnose the cancer in 73.8% of cases in this training set and in 74.1% of an independent group of 11,644 malignancies.⁵ In addition, in a group of 141 patients with a presumptive diagnosis of cancer of unknown primary site, the algorithm predicted the site of origin in 67.4% of the cases,⁵ suggesting the potential utility of this strategy in routine clinical pathology practice.

Pitfalls of Early Adoption

However, we must acknowledge that we cannot always simply assume that the use of newer technology will be associated with a positive outcome. For example, dangers have recently been highlighted for the potential misuse of poorly constructed and considered race-adjusted algorithms designed to improve the quality of care that may result in the exact opposite outcome.⁶

Similarly, the excessively early adoption of what may appear to be innovative device technology developed to improve surgical outcomes may actually compromise survival, as recently revealed for the use of minimally invasive radical surgical approaches in the management of early-stage cervical cancer.^7,8 One wonders why the gynecologic oncology community so actively embraced the routine use of this strategy before phase 3 randomized trial data on the safety and benefits of this approach were available.

In this regard, it is somewhat ironic, and perhaps a bit disturbing, that the same issue of the New England Journal of Medicine that reported the results of such a phase 3 trial revealing the inferior survival of women randomized to minimally invasive radical surgery also included the results of a retrospective analysis of 1225 patients treated with this technique compared with a nonrandomized but carefully matched control group managed with an open surgical technique where reduced survival for the minimal invasive technique was also observed.^7,8

Although it would be pure speculation to estimate how many women might not have had their lives shortened if the results of a solidly evidence-based randomized trial had been reported before minimally invasive radical surgery for early-stage cervical cancer became routine in clinical practice, it would be quite appropriate to ask such a question with the goal of helping others avoid a serious technological misadventure in the cancer management arena in the future. Advances in technology have a tremendous potential to improve both survival and quality of life for individuals with malignant disease, but all new approaches—regardless of their innovative nature—require an objective, rigorous assessment of their safety and clinical utility.

References:

1. Murphy H, Arango T. Guilty pleas, again and again, end Golden State Killer’s reign of terror. New York Times. June 30, 2020:A19.
2. Spaccarotella CAM, Polimeni A, Migliarino S, et al. Multichannel electrocardiograms obtained by a smartwatch for the diagnosis of ST-segment changes. JAMA Cardiol. Published online August 31, 2020. doi:10.1001/jamacardio.2020.3994
3. McKinney SM, Sieniek M, Godbole V, et al. International evaluation of an AI system for breast cancer screening. Nature. 2020;577(7788):89-94. doi:10.1038/s41586-019-1799-6
4. Wu Q, Wang S, Zhang S, et al. Development of a deep learning model to identify lymph node me-tastasis on magnetic resonance imaging in patients with cervical cancer. JAMA Netw Open. 2020; 3(7):e220121625. doi:10.10110/jamanetworkopen.2020.11625
5. Penson A, Camacho N, Zheng Y, et al. Development of genome-derived tumor type prediction to inform clinical cancer care. JAMA Oncol. 2019;6(1):84-91. doi:10.1001/jamaoncol.2019.3985
6. Vyas DA, Eisenstein LG, Jones DS. Hidden in plain sight—reconsidering the use of race correction in clinical algorithms. N Engl Med. 2020;383(9):874-882. doi:10.1056/NEJMms2004740
7. Ramirez PT, Frumovitz M, Pareja R, et al. Minimally invasive versus abdominal radical hysterectomy for cervical cancer. N Engl J Med. 2018;379(20):1895-1904. doi:10.1056/NEJMoa1806395
8. Melamed A, Margul DJ, Chen L, et al. Survival after minimally invasive radical hysterectomy for early-stage cervical cancer. N Engl J Med. 2018;379(20):1905-1914. doi:10.1056/NEJMoa1804923