How Strong Is the Evidence Supporting Thromboprophylaxis in Surgical Oncology?

Introduction

Venous thromboembolism (VTE) has been described as the second leading cause of death in patients with cancer and the most common cause of death after surgery.^1,2 VTE can also cause morbidity including post-thrombotic syndrome (PTS), pulmonary hypertension, and chronic pain.³ On the basis of a high level of evidence from randomized controlled trials (RCTs), the ASCO Clinical Practice Guidelines strongly recommends pharmacologic thromboprophylaxis for patients undergoing surgery for cancer.⁴

Herein, we argue that critical weaknesses in the literature preclude meaningful interpretation of many of these RCTs. These weaknesses are the unexplored potential for heterogeneity in VTE end points, the unclear effect of thromboprophylaxis on survival, and a lack of supportive care outcomes in RCTs. Until these deficiencies can be rectified, evidence supporting the use of thromboprophylaxis should be considered weak and incomplete.

There Is Unexplored Heterogeneity in VTE End Points

The end points of RCTs evaluating thromboprophylaxis often contain heterogeneity of unknown significance. In many trials, especially surgical ones, patients are screened for asymptomatic deep vein thrombosis (DVT) using ultrasound or venography.^5-12 However, the clinical significance of screen-detected DVT has never been proven in a controlled trial. Indeed, at least one trial showed that screening does not affect clinical outcomes.¹³ The ASCO guidelines do not recommend screening for DVT in routine practice.⁴ Therefore, screen-detected DVT should be considered an unproven surrogate end point.

Another source of heterogeneity are incidental VTE and peripheral (ie, segmental or subsegmental) pulmonary embolisms (PEs). Their incidence increased since the advent of high-resolution computed tomography scanners.^14,15 Although they can be associated with a poor prognosis, their clinical significance has also never been proven, and some may be inconsequential.^16,17 This uncertainty is appropriately reflected in the guidelines.⁴ Only an informal consensus with a low quality of evidence supports treatment for incidental PE or DVT. An informal consensus on the basis of an insufficient quality of evidence supports treatment of subsegmental PE.

Despite their unknown significance, incidental VTE and peripheral PE are often included in the end point of RCTs. This was the case in the recent AVERT and CASSINI trials for oncology patients.^18,19 However, investigators frequently omit details such as the number, location, or clinical impact of these lesions. To understand why this might be important, consider two hypothetical patients: one with a large, central PE causing hemodynamic instability and another with shortness of breath, an isolated subsegmental PE, and pneumonia. In the CASSINI trial, both events might have been adjudicated as symptomatic nonfatal PEs.

This potential for heterogeneity in VTE events is a major blind spot in the literature. For example, the 2020 ASCO guidelines cited three relatively new RCTs in surgical patients.^6,11,20 Control patients experienced 28 VTEs, but 24 were screen-detected DVT. Although there were two PEs, it was not reported whether they were central or peripheral or there were coincident diagnoses. Because the clinical significance of most of these events is unclear, it is difficult to draw strong conclusions from RCTs showing they were prevented.

There Is Rationale and Equipoise for De-Escalation Trials for Thromboembolic Disease

It may be common practice in oncology to treat incidental VTE or peripheral PE, but there has never been an RCT showing that this benefits patients. A perceived lack of equipoise may be one reason why, but this perception is unfounded. The seminal trial by Barritt and Jordan²¹ suggested that untreated PE carries a high mortality risk but was published more than 60 years ago, has never been replicated, and would likely be unpublishable today because of its outdated and flawed methodology.

In stark contrast to the results reported by Barritt and Jordan are data from a more recent RCT that had both superior methodology and larger enrollment. Nielsen et al²² randomly assigned ambulatory patients with acute symptomatic DVT, half of whom had PEs detected by baseline ventilation perfusion scans, to anticoagulation versus a nonsteroidal anti-inflammatory drug. Remarkably, there was no difference in clot resolution or clot progression, and the only death in the study occurred in the anticoagulation group.

In addition to the surprising results reported by Nielsen et al, there is mounting evidence of overdiagnosis and overtreatment of VTE corresponding with the advent of PE protocol computed tomography scanners.^14,15,23 Moreover, the weight of contemporary prospective literature indicates that VTE is a marker of mortality risk, but not a common cause of death.²⁴ Therefore, it is possible that asymptomatic DVT, incidental VTE, or peripheral PE could be safely managed with observation in carefully selected patients. If this hypothesis is true, these lesions should probably not be included in the end points of RCTs evaluating thromboprophylaxis.

Thromboprophylaxis Does not Detectably Influence Survival

The risk of mortality posed by VTE is often given as rationale in studies evaluating thromboprophylaxis. A common claim is that VTE is the second leading cause of death in patients with cancer.²⁵ If this were true, effective strategies to prevent VTE should affect survival in an adequately powered study. However, although mortality is sometimes included in composite end points, RCTs are almost never powered to detect effects on mortality alone. Yet, even meta-analyses fail to demonstrate whether thromboprophylaxis imparts a mortality benefit or detriment.

The idea that thromboprophylaxis improves survival in surgical patients gained traction after the publication of one meta-analysis in 1988.²⁶ Although it analyzed over 70 RCTs, its overall effect estimates on the mortality end point depended entirely on two studies published over 40 years ago that had substantial risks of bias.²⁷ In any event, most of the patients enrolled in those two trials did not have cancer. Meta-analysis of dozens of other RCTs, including 31 with placebo controls, failed to show a survival benefit from thromboprophylaxis.²⁷

Contemporary meta-analyses have also consistently failed to detect an effect on survival. One cited in the ASCO guidelines to support recommendations for extended thromboprophylaxis in surgical oncology actually showed that all-cause mortality was slightly higher in the treatment group, although the effect was statistically insignificant.² In nonsurgical oncology patients, there is no evidence of a survival effect.²⁴ Since the possibility that thromboprophylaxis causes a small detrimental effect on mortality has not been excluded, there is a need for RCTs with adequate power to detect effects on survival.

Thromboprophylaxis Is a Promising but Unproven Supportive Care Strategy

Even if thromboprophylaxis does not detectably influence survival, it may be an important component of supportive care.³ Complications from VTE including PTS, chronic pain, and pulmonary hypertension can occur in as many as 25%-50% of patients. VTE can adversely affect quality of life, physical fitness, and mental fitness. Thromboprophylaxis would be valuable if it could safely mitigate these complications. However, although thromboprophylaxis is considered a supportive care strategy, RCTs rarely report supportive care end points. Instead, they report number of VTEs averted and weigh this against bleeding complications. However, this is not a valid comparison because it ignores the impact of salvage treatment.

Thromboprophylaxis is conceptually like adjuvant therapy for cancer. The strategy is to subject an entire at-risk population to an intervention that can reduce the probability of an undesired event. A downside of this approach is that although all patients are exposed to toxicity, only a fraction will benefit. Another caveat is that effective salvage therapy diminishes the rationale for adjuvant therapy. Consider adjuvant chemotherapy for stage III colon cancer: It improves survival because most distant recurrences cannot be salvaged.²⁸ If they could be, there would be no reason to subject all patients with stage III disease to the toxicity of chemotherapy; it would be better to simply wait and treat the recurrences.

The effectiveness of treatment for VTE also has major implications for thromboprophylaxis, but this possibility seems to have been overlooked. For example, extended thromboprophylaxis reduces proximal DVT by 1.4% implying a number needed to treat of 70 patients.² However, a placebo-controlled RCT showed that PTS only occurred in 14% of patients who developed a proximal DVT, indicating that most patients were salvaged with respect to the PTS end point.²⁹ Therefore, the number needed to treat for extended thromboprophylaxis to prevent one case of PTS would be 500, not 70. Exposing 500 surgical patients to increased bleeding risk is a high price to pay to prevent one case of PTS.

One could argue about the specifics, but the point of this hypothetical is that a single-minded focus on reducing the incidence of VTE is misguided. It is not safe to assume that thromboprophylaxis improves morbidity at the population level simply because it lowers VTE incidence. Thromboprophylaxis definitely increases bleeding, and patients with cancer are already at elevated risk of bleeding compared with those without cancer.^24,30 Moreover, the widespread use of screening for asymptomatic DVT may have compromised evaluation of safety end points. This is because overtreatment with full-strength anticoagulation for screen-detected DVT could have caused catchup bleeding complications in control patients.¹⁵

The relevant comparison, then, is not whether there are more VTEs avoided than bleeding episodes caused. It is whether the morbidity benefit accrued to the small fraction of patients who would have developed a symptomatic VTE and could not have been salvaged by treatment outweighs the toxicity experienced by the entire at-risk population. The fact is most patients will never develop a symptomatic VTE, and some of them will be harmed by thromboprophylaxis they did not need. We owe it to these patients to test whether thromboprophylaxis is a beneficial supportive care strategy, but this can only be accomplished through RCTs that report clinical and supportive care end points.

In conclusion, surrogate outcomes, such as short-term prevention in VTE, might be useful to mitigate competing risks in trials with low event rates, but they do not always translate into beneficial, patient-centered outcomes. Thromboprophylaxis definitely lowers the incidence of VTE in patients undergoing surgery for cancer. However, many VTEs have no proven clinical significance, thromboprophylaxis does not have a measurable effect on survival, and since RCTs do not usually report supportive care end points, there is little support for claims that it is a beneficial supportive care strategy.

Complete reporting of VTE events in clinical trials would be a low-cost way to enable investigators to explore their heterogeneity. Important details might include the location and number of VTEs, any associated clinical symptoms, and the presence of coincident diagnoses. Consistency in event reporting is also crucial; a recent meta-analysis provides a blueprint for this. Kraaijpoel and colleagues identified significant heterogeneity in the reporting of fatal PEs in clinical trials and have offered objective definitions for use in future studies.^31,32

Most importantly, new RCTs are needed. The significance of screen-detected DVT, incidental VTE, and peripheral PE should be directly tested in carefully selected patients. If these lesions can be safely managed with observation, they are probably not useful as end points in RCTs evaluating thromboprophylaxis. Finally, a reduction in VTE incidence cannot be extrapolated as proof of a beneficial effect on clinical outcomes. Consequently, future RCTs evaluating thromboprophylaxis should either be adequately powered to detect an effect on survival or must report supportive care end points.

AUTHOR CONTRIBUTIONS

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

How Strong Is the Evidence Supporting Thromboprophylaxis in Surgical Oncology?

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

No potential conflicts of interest were reported.