Routine Cancer Antigen 125 Surveillance—The Fatal Attraction of Testing

Editorial
  July 21, 2016
 

The article by Esselen et al¹ in this issue of JAMA Oncology belongs to the category of health services research that asks how the introduction of new evidence influences clinical practice. In the instance they examined—publication of a study that showed routine cancer antigen 125 (CA-125) surveillance was harmful—the answer seems to be “not at all.” In the 6 NCI-designated cancer centers they studied, the use and frequency of CA-125 testing of women in remission after initial treatment for ovarian cancer was similar in the years before and after the study. In both periods, almost all such patients received CA-125 testing approximately every 3 to 4 months. This is actually quite a remarkable observation. Why did practice not change?

It is instructive to reread the original study by Rustin et al^2,3 that questioned the use of CA-125 testing. It was a well-conducted multicenter randomized trial of 529 women who were in remission after initial chemotherapy for ovarian cancer. All patients received CA-125 testing every 3 months, but the treating physicians did not know the results. If and when a CA-125 test exceeded twice normal, patients were randomized into what they called early and delayed treatment groups, with randomization stratified by several prognostic factors. For patients in the early treatment group, their physicians were notified of the CA-125 results. In the delayed treatment group the results were withheld, and any decisions about further treatment were based on clinical recurrence. Most women in the early treatment group started chemotherapy soon after the notification of elevated CA-125 results. Those in the delayed treatment group started about 5 months later. Median survival was not different: 25.7 months in the early treatment group vs 27.1 months in the delayed treatment group. However, median time to deterioration in global health score or death was significantly longer in the delayed treatment group (5.8 vs 3.2 months, P =  .002). The delayed treatment group also underwent fewer cycles of chemotherapy, were less likely to receive third-line chemotherapy, and had 5 more months to live in remission after their initial treatment.^2,3

The question for discussion then becomes: why was there no change in practice in response to good evidence? It is not as if physicians are incapable of rapid change in practice in the face of new information. It happens all the time when the information favors new or more treatment.^4,5 For example, we assessed change in practice associated with the report showing a 2% absolute survival improvement at 18 months with taxane plus doxorubicin compared with doxorubicin alone in breast cancer patients with positive nodes. The oral report⁶ of these results in 1998, 5 years before the actual publication of the trial, was associated with a rapid shift to taxanes with a more than doubling in use within 3 months.⁵

Why did this not occur with the Rustin study? One could argue that there was no change because the clinical guidelines did not change. This is somewhat circular, because the 6 centers studied by Esselen et al¹ were leading academic medical centers, populated by clinicians who also sit on the committees that write the guidelines.

I will propose 2 sets of factors contributing to the lack of an obvious change in practice in response to the Rustin study. The first relates to the fundamental attractiveness of testing, and the second involves how the Rustin study was framed by the authors and subsequent commentators.

Physicians practice in a sea of uncertainty. Making life-altering decisions based on incomplete information is 1 of the major stresses of becoming a physician. There are good and bad ways of dealing with that stress. Over confidence is 1 bad way. We all have encountered physicians who practice a “my way or the highway” style. A more acceptable means to reduce uncertainty is to seek more information through testing. The past 2 decades have witnessed a proliferation of new cancer tests and improvements in older ones. Highly sensitive scans and molecular and genetic markers allow us to more accurately characterize risk for or extent of disease and response to treatment, reducing uncertainty in decision making.

As the tests proliferated, questions arose whether more information is always a good thing. Magnetic resonance imaging for uncomplicated low back pain can lead to unnecessary surgery.⁷ After routine fetal monitoring was introduced, childbirth became a surgical procedure for a quarter of women.⁸ The dissemination of prostate-specific antigen screening was followed by an iatrogenic epidemic of prostate cancer.^9,10

This is a difficult message to accept. How can more information possibly hurt? Avoiding information sounds antiscientific. Can we not still obtain the information but interpret it more wisely? A nice sentiment, but it lacks evidence of feasibility in real life. The analyses by Esselen et al¹ suggest that information on increased CA-125 drove initiation of second line chemotherapy as rapidly after the report^2,3 by Rustin as before.

Another difficulty involves telling physicians that they may be doing the wrong thing. Whether it involves tests or treatments, it is no easy task to convince physicians that they are causing more harm than good. There are a number of predictable responses to reports such as the 1 by Rustin et al^2,3 One is what I call the moving target. “Okay, so that surgery (or radiotherapy or whatever) had bad outcomes, but you don’t understand that the techniques have improved dramatically since the time of your data.” Such a response has the advantage of sometimes being true. Other responses include questioning the patient selection: “Those aren’t the people I see in my practice” or the physicians studied: “No wonder it didn’t work. I don’t do it that way!”

All 3 of these responses and more greeted the Rustin study.^11,12 What did not occur was any change in practice. Readers might want to conduct a thought experiment to ask themselves: “What if the results reported for delayed treatment were found instead for a new chemotherapy regimen? How would the report have had an effect on practice?”

The second set of reasons has to do with framing the results of the Rustin study. The 2 approaches studied by Ruskin et al^2,3 were characterized as early vs delayed treatment, but that was not what they studied, at least not in my opinion. They were in essence comparing CA-125 surveillance every 3 months vs no testing. The “no testing” group received CA-125 testing, but since the results were not released to the physicians, it was the same as not testing. The findings of the study are then best framed as routine surveillance with CA-125 testing is harmful in patients diagnosed with ovarian cancer in remission after initial treatment.

Another factor that impedes clear thinking about harmful tests and treatments is the injection of medical economics into the discussion. Discussions of cost-effectiveness can make clinicians uncomfortable. The results of cost-effectiveness analyses can vary widely depending on the assumptions. And why should physicians act as societal financial gatekeepers? It is difficult enough helping patients decide on what is best for them without introducing the concept of what is best for society. In addition, such discussions can be seen by our patients as attempts at rationing their care. The suspicion that other patients, perhaps those with better insurance or better connections, might be receiving better treatment can destroy trust in the physician.

But in the context of CA-125 testing, the issue of cost-effectiveness becomes absurd. How can a test or treatment be cost-effective if it is not effective —indeed, if it is harmful? Thus, I am uncomfortable with the focus on the economic impact of continued CA-125 testing by Esselen et al.¹ It is not by any means irrelevant, but it muddies the issue. Once again, the findings of the Ruskin study are that routine surveillance testing is harmful. That should be the focus for subsequent decisions by clinicians, payers, and policy makers—not the costs involved.

And, of course, the women with ovarian cancer contribute to the decisions to obtain C-125 testing. While physicians may be challenged dealing with uncertainty, many of our patients refuse absolutely to tolerate it. They want to know, even if that knowledge can lead to harm. Thus, in a shared decision making context, most patients will choose more information.¹³ But why would clinicians present the option of CA-125 testing? Shared decision making does not require that physicians present the patient with harmful options. We do not discuss heart transplants with patients who have mild congestive heart failure. Why would we discuss CA-125 testing with women who have ovarian cancer in remission?

Will the practice of obtaining routine CA-125 testing change? I doubt it. The moving target argument is too strong. There will be better tests, and more effective treatments. And studies like that of Rustin with uncomfortable results are usually not replicated.¹⁴ The issue is allowed to fade away. The fatal attraction of more information is too compelling.

References

1

Esselen  KM, Cronin  AM, Bixel  K,  et al.  Use of CA-125 tests and computed tomographic scans for surveillance in ovarian cancer [published online July 21, 2016]. JAMA Oncol. doi:10.1001/jamaoncol.2016.1842.

2

Rustin  GJ, Van Den Berg  ME. A randomized trial in ovarian cancer (OC) of early treatment of relapse based on CA125 level alone vs delayed treatment based on conventional clinical indicators (MRC OV05/EORTC 55955 trials). ASCO Ann Meet Proc. 2009;27(18s):1.

3

Rustin  GJ, van der Burg  ME, Griffin  CL,  et al; MRC OV05; EORTC 55955 investigators.  Early versus delayed treatment of relapsed ovarian cancer (MRC OV05/EORTC 55955): a randomised trial. Lancet. 2010;376(9747):1155-1163.
PubMed   |  Link to Article

4

Jackevicius  CA, Anderson  GM, Leiter  L, Tu  JV.  Use of the statins in patients after acute myocardial infarction: does evidence change practice? Arch Intern Med. 2001;161(2):183-188.
PubMed   |  Link to Article

5

Giordano  SH, Duan  Z, Kuo  YF, Hortobagyi  GN, Freeman  J, Goodwin  JS.  Impact of a scientific presentation on community treatment patterns for primary breast cancer. J Natl Cancer Inst. 2006;98(6):382-388.
PubMed   |  Link to Article

6

Henderson  IC, Berry  D, Demetri  GD,  et al. Improved disease-free and overall survival from the addition of sequential paclitaxel but not from the escalation of doxorubicin dose in the adjuvant chemotherapy of patients with node-positive primary breast cancer. Proc ASCO. 1998;17:101a.

7

Webster  BS, Cifuentes  M.  Relationship of early magnetic resonance imaging for work-related acute low back pain with disability and medical utilization outcomes. J Occup Environ Med. 2010;52(9):900-907.
PubMed   |  Link to Article

8

Placek  PJ, Taffel  SM, Moien  M.  Cesarean rate increases in 1985. Am J Public Health. 1987;77(2):241-242.
PubMed   |  Link to Article

9

Welch  HG, Albertsen  PC.  Prostate cancer diagnosis and treatment after the introduction of prostate-specific antigen screening: 1986-2005. J Natl Cancer Inst. 2009;101(19):1325-1329.
PubMed   |  Link to Article

10

Howrey  BT, Kuo  YF, Lin  YL, Goodwin  JS.  The impact of PSA screening on prostate cancer mortality and overdiagnosis of prostate cancer in the United States. J Gerontol A Biol Sci Med Sci. 2013;68(1):56-61.
PubMed   |  Link to Article

11

Morris  RT, Monk  BJ.  Ovarian cancer: relevant therapy, not timing, is paramount. Lancet. 2010;376(9747):1120-1122.
PubMed   |  Link to Article

12

Rustin  G, van der Burg  M, Griffin  C, Qian  W, Swart  AM.  Early versus delayed treatment of relapsed ovarian cancer. Lancet. 2011;377(9763):380-381.
PubMed   |  Link to Article

13

Markman  M, Petersen  J, Belland  A, Burg  K.  CA-125 monitoring in ovarian cancer: patient survey responses to the results of the MRC/EORTC CA-125 Surveillance Trial. Oncology. 2010;78(1):1-2.
PubMed   |  Link to Article

14

Goodwin  JS, Goodwin  JM.  The tomato effect. Rejection of highly efficacious therapies. JAMA. 1984;251(18):2387-2390.
PubMed   |  Link to Article