Demmy, Todd L.
From the Department of Thoracic Surgery, Roswell Park Cancer Institute, Buffalo, NY.
Address correspondence and reprint requests to Todd L. Demmy, Department of Thoracic Surgery, Roswell Park Cancer Institute, Buffalo, NY 14263; e-mail: email@example.com.
This exhaustive meta-analysis by Cheng and colleagues demonstrated advantages for thoracoscopic lobectomy (ie, without rib spreading) over traditional open operations and even video-assisted procedures allowing wound retraction. In their conclusion, the authors correctly emphasize that many benefits for this emerging technology are not proven by randomized controlled trials so they encourage more of such studies.
Randomization is a powerful tool that overcomes potentially biased patient selection with which the studies compiled in the review had to contend. It is therefore appropriate to focus this commentary on the challenges of selecting an appropriate study population and designing a randomized clinical trial. To be compelling, such a trial needs to achieve its accrual goals, successfully test its hypothesis, and have a study group that can be generalized to the broad population of lung cancer patients. Having practiced cardiac surgery before focusing on general thoracic surgery, I found it useful to categorize my comments thematically by ongoing controversies in coronary revascularization.
The Next Best Stent
This theme refers to the difficulty of comparing an existing standard to a treatment based on dynamic technology. Just as some percutaneous coronary interventions (PCI) evolved faster than a randomized trial comparing CABG could be initiated, VATS lobectomy techniques changed considerably over the span of this review. This resulted from improved technologies and accrued surgeon experiences at centers of excellence. High definition cameras, for instance, are recent innovations that provide superb exposures that facilitate dissections and reduce the needs for camera zooming by assistants. Surgeons choose from various posterior or anterior direction styles to dissect the hilum and often divide the bronchi or incomplete fissures in different orders than open resections.
Next, surgeons built their VATS lobectomy programs on different patient care philosophies that influenced their results. Many practitioners initially preferred patients with better pulmonary function, fewer comorbidities, and peripheral lung tumors. Accordingly, resections were easier by virtue of better lung collapse and physiologic tolerance of the “learning curve.” Adept thoracic surgeons rapidly built speed and skills for this low-risk group, but the in-patient outcomes were approachable by their experienced colleagues using fast-track pathways, epidural anesthesia, and limited thoracotomy incisions. Others started with frail patients for whom the anticipated morbidity of a thoracotomy was excessive. Because event rates are higher for these cases, fewer patients were needed to observe differences between groups. With thoracoscopic lobectomy, infirmed patients had more noticeable independence from nursing support or narcotic usage after discharge.
It is unclear which initial case mix builds skills for complex operations faster; however, most thoracoscopic surgeons expand their indications and some now exceed 70% of available referral cases. Combining surgeons with different thoracoscopic rates along with those doing only open cases yielded an average VATS lobectomy rate of about 25% within the STS General Thoracic Database. A high absolute VATS resection rate may be a good benchmark with which to credential surgeons for research protocols because it indicates less selection bias. Combined with good outcomes, this benchmark could also validate idiosyncratic surgical methods. Applied to an institution or a community, high rates indicate that a group of thoracic surgeons have similar skill levels. So with appropriate benchmarks, results of a randomized VATS versus open lobectomy study probably could be generalized to routine practice because the surgical techniques are relatively stable and applicable to most patients.
Off-pump Versus On-pump
This theme refers to issues more tangible to physicians than research subjects and that have the potential to adversely affect investigator equipoise. Although epicardial stabilizers and VATS lobectomy tools continue to improve, off-pump cardiac surgeons and thoracoscopic lung surgeons both adapt to a more challenging operating room environment. Some technical barriers are occasionally tedious like port-access limitations or blood in the anastomotic field. Surgeons persist in these challenges because they believe intuitively, by personal outcome studies, or by the published results of others that they are benefiting their patients. Surgeons may be encouraged by referring physicians. For instance, the cardiologist may note fewer embolic or inflammatory complications without cardiopulmonary bypass. Primary care doctors like the fewer chronic pain interventions and medical oncologists prefer the better chemotherapy tolerance associated with minimal access surgery. The interplay between specialists can become complex as PCI are mixed with less invasive surgical revascularization. Or medical oncologists use neoadjuvant chemotherapy to shrink large tumors to enable thoracoscopic lobectomy in high-risk patients.
Physician beliefs favoring less invasive techniques are formidable barriers to referral for clinical trials. And they prevent investigators' equipoise in their explanations to participants. They can be offset by finding investigators who are truly uncertain regarding the treatment options or by selecting certain patient groups for whom participating surgeons feel equipoise.
Percutaneous Intervention (PCI) Versus Coronary Bypass
The most significant barrier to randomized clinical surgical trials like that proposed is the natural tendency of human subjects to avoid pain and disruption to their lifestyles. Less established minimally invasive procedures are seductive not only for their initial promise of fewer problems, but because the traditional proven surgical therapy remains available later (usually) if needed. Of course, cardiac surgeons can cite circumstances where initial PCI affected the quality of their operation. The oncologic concerns with VATS lobectomy have been higher recurrence rates; yet, such problems have not been reported despite the common use of the operation (including within major cancer centers). Because lung cancer recurrences usually present within 2 years after resection, reports of problems should have emerged by this point. Another concern regarding the adequacy of lymph node dissections has been discounted by comparative, albeit mostly nonrandomized, trials showing similar lymphatic clearance by VATS. Also, there is uncertainty within the profession about the incremental benefit of lymph node dissection in the era of PET/CT imaging.
Two cooperative group protocols from the Cancer and Leukemia Group B (CALGB) attempted to randomize surgical therapies of different levels of invasion. CALGB 39804 compared thoracoscopic to open metastasectomy and had almost no accrual because of referral bias and patient concerns. CALGB 30102 compared two minimally invasive procedures (Pleurx catheter or Chest tube/Talc) for malignant pleural effusions but closed early for poor accrual largely because patients wanted to choose the technique that suited their desire for shorter inpatient or longer outpatient management. CALGB 39802 showed the feasibility of VATS lobectomy and established criteria that defined the procedure as having no rib spreading. Because most participating surgeons felt that randomization between open and thoracoscopic lobectomy was too difficult, CALGB 140301 was designed as a prospective multicenter registry study comparing both operations. Although the NCI initially approved it, after subsequent review the protocol was no longer of sufficient priority to compete with randomized trials.
It is interesting to note that the trials in this meta-analysis that yielded the 205 randomized cases occurred between 1995 and 2001 with no subsequent reports. This suggests that the interest or ability to randomize patients fell off. A similar comprehensive analysis of the world's literature by the Cochrane group for laparoscopic versus open cholecystectomy found 2338 patients were randomized before the procedure became standard of care.1 It is interesting that cholecystectomy which is at least 10 times more common than pulmonary lobectomy yielded a similar a ratio of randomized patients. Of the 38 randomized cholecystectomy studies, only four were published after 2000 and only five could muster over 50 patients per arm despite the high incidence of the operation. Furthermore, most of the randomized trials in that cholecystectomy analysis were evaluated to have high bias risk.
So How Can We Randomize Between Different Levels of Surgical Trauma?
For certain questions such as VATS versus open lobectomy, it may not be possible. If such a trial were designed, would the patients who are indifferent about the current engrained perceptions in outcomes really be a representative sample? One possibility is to offer minimally invasive surgery only to patients who take part in the clinical trial. This might be possible for more advanced thoracoscopic resections like pneumonectomy. Because complication and cancer recurrence events occur more frequently for complex thoracic cases, fewer patients may be needed for randomization. If results are equivalent for more complex cases, one might assume that the results will be at least as good for simple cases.
Yet this is an imperfect solution. Surgeons need to pilot clinical trial designs that accommodate the challenges listed above. One method is to use randomization schemes that reduce the odds for assignment to the less fashionable therapy—in this case open lobectomy. This may work when there is not as strong of an aversion to the less desirable treatment arm. Hybrid trial designs studying both randomized patients and those who refuse (ie, select their preferred therapy) have been performed but are more complex to implement and analyze. Such trials are attractive because more is learned about the broader group of patients and how it compares with the subjects willing to be randomized.
Unfortunately, for the question at hand, cooperative group scientists do not yet have a validated methodology to randomly assort those wanting to limit perturbations of their anatomy and physical function. Given its stage of acceptance, randomized VATS lobectomy trials would probably be like those in the laparoscopic cholecystectomy literature: small, occurring at single, less typical venues where sufficient investigator equipoise exists, or at high bias risk. Whether combining many flawed small randomized trials together in a meta-analysis is better than large multi-institution registry study that is flawed because of nonrandomization is a matter of debate. Some feel that there is sufficient evidence to pursue VATS lobectomy as the preferred approach to lung resection and further study is not necessary. Or that it won't be long until insurance claims databases, the STS General Thoracic Database or other large regional datasets provide supportive evidence. Yet, these data sources often include information from low-volume centers that have inconsistent results. Unlike general data sources, a prospective registry at sites with existing research programs in open and/or thoracoscopic lung cancer resection would provide our specialty with contemporary survival rates for these common operations at centers of excellence. Such useful information for North America has not been available since the Lung Cancer Study Group (LCSG) reports. Furthermore, high-volume centers have the surgeon experience recommended by Cheng and colleagues, and streamlined their business practices making cost-analyses easier.
Randomized trial models are powerful tools. But good surgeons and scientists pick the tools to fit the problem at hand. I hope that we will continue to study the issues summarized nicely in this meta-analysis and that we will be supported by government or other granting agencies because of the importance of the research even if we have to employ partially randomized, registry, or nontraditional methods to get the job done.
© 2007 Lippincott Williams & Wilkins, Inc.