Failure of Preclinical Studies to Accurately Predict the Effect of Regional Analgesia on Cancer Recurrence

A qualified medical librarian conducted a systematic search on our behalf. Medline (Ovid), Embase (Ovid), and PubMed were queried from 1970 through September 17, 2023, using both natural language and controlled vocabulary terms for regional anesthesia, cancer, surgery outcomes, and general anesthesia. The following terms were included in the search: ((((Neoplasms[Mesh] OR “MEDICAL ONCOLOGY”[Mesh]) OR ((neoplas*[tiab] OR cancer*[tiab] OR carcinoma*[tiab] OR malignan*[tiab] OR tumor*[tiab] OR tumor*[tiab] OR oncolog*[tiab] OR metasta*[tiab] OR osteosarcoma*[tiab] OR sarcoma*[tiab] OR melanoma*[tiab]))) AND (“anesthesia, conduction”[MeSH Terms] OR “anesthesia, epidural”[MeSH Terms] OR “anesthesia, spinal”[MeSH Terms] OR “nerve block”[MeSH Terms] OR (“regional”[Title/Abstract] AND “anesthesia*”[Title/Abstract]) OR (“spinal”[Title/Abstract] AND “anesthesia*”[Title/Abstract]) OR “nerve block*”[Title/Abstract] OR (“epidural”[Title/Abstract] AND “anesthesia*”[Title/Abstract]))) AND ((infect*[tiab] OR bacteria*[tiab] OR fungal*[tiab] OR mycos*[tiab] OR mycotic*[tiab] OR nosocomial*[tiab]) OR length-of-stay[tiab] OR ((readmit*[tiab] OR readmission*[tiab]) AND patient*[tiab]) OR (mortality[tiab] OR surviv*[tiab] OR fatal*[tiab] OR Kaplan-Meier[tiab]) OR (“progression free”[tiab] OR “disease free”[tiab]) OR (pulmonary[tiab] AND edema*[tiab]) OR “volume overload*”[tiab] OR ((vision*[tiab] OR visual*[tiab] OR eyesight*[tiab]) AND (low[tiab] OR loss*[tiab] OR lost[tiab] OR acuity[tiab] OR deficien*[tiab] OR diminish*[tiab] OR reduc*[tiab] OR decreas*[tiab])) OR (stroke*[tiab] OR ((cerebrovascular[tiab] OR vascular[tiab]) AND accident*[tiab])) OR ((brain*[tiab] OR cerebral*[tiab]) AND infarct*[tiab]) OR ((kidney*[tiab] OR renal*[tiab]) AND (fail*[tiab] OR insufficien*[tiab])) OR ((myocardi*[tiab] AND infarct*[tiab]) OR “heart attack*”[tiab]))) AND (random*[Title] AND (allocat*[Title] OR assign*[Title] OR blind*[Title] OR control*[Title] OR divided[Title] OR picked[Title] OR placebo*[Title] OR select*[Title] OR study[Title] OR trial[Title])). A total of 469 unique records were identified. The results were limited to randomized controlled trials published in English in which cancer recurrence was the primary outcome. We excluded publications reporting secondary analysis of randomized controlled trials. After examining abstracts, only four publications met our eligibility criteria.

An assumption behind the analysis to follow is that the major relevant trial results are correct. We thus briefly review the four robust trials on regional analgesia and cancer recurrence and their limitations. The first trial randomized 2,132 women having potentially curative breast cancer surgery to volatile general anesthesia alone versus T1 to T5 single-shot bupivacaine paravertebral blocks with propofol sedation and volatile anesthesia as needed. The trial started in 2007; the women assigned to paravertebral blocks were given about 4.5 times as much propofol, and only 17% received any volatile anesthesia. The median follow-up was 36 months. Baseline characteristics were well balanced, as would be expected in such a large trial. The adjusted hazard ratio (paravertebral vs. general anesthesia) for local or metastatic cancer recurrence was 0.97 (95% CI, 0.74 to 1.28) based on 213 outcome events (P = 0.84). The results in various subgroups were similar. Major limitations of the trial include relatively low pain scores and opioid use in both groups and relatively small operations that presumably provoked much less surgical stress than larger procedures.

A second trial also investigated the effect of regional anesthesia on breast cancer recurrence in 526 women who were randomized to a pectoral nerve block type II with ropivacaine (plus general anesthesia) or to general anesthesia alone.  Bispectral Index, propofol use, and remifentanil consumption were similar in each group. The adjusted hazard ratio for recurrence-free survival was 0.9 (95% CI, 0.76 to 1.32) based on 85 outcome events. Curiously, postoperative opioid use was not reported, and it thus remains unclear how effective the blocks were. It is also unclear whether volatile anesthetics were used and, if so, how much was given in each group.

The third trial randomized 1,712 patients with noncardiac thoracic and abdominal cancers (mostly colorectal) to general anesthesia alone or general anesthesia combined with thoracic epidural analgesia that was maintained postoperatively.  The duration of surgery averaged almost 4 h, and the baseline characteristics were well balanced. Postoperative opioid-equivalent use was only slightly reduced by epidural analgesia, although patients assigned to general anesthesia alone were mostly given morphine, whereas those assigned to neuraxial analgesia were mostly given epidural sufentanil. The adjusted hazard for recurrence-free survival was 0.97 (95% CI, 0.84 to 1.12) based on 790 outcome events (P = 0.69). Again, the results in various subgroups were similar. Major limitations of the trial include lack of detail about volatile anesthetic dosing and similar postoperative opioid use.

The fourth major trial of regional analgesia and cancer recurrence randomized 400 patients having lung cancer resections to general anesthesia alone or general anesthesia combined with thoracic epidural analgesia that was maintained postoperatively.  At a median of 32 months follow-up, the adjusted hazard ratio for general/epidural versus general anesthesia was 0.90 (95% CI, 0.60 to 1.35) based on 102 outcome events (P = 0.61). The results in various subgroups were similar. Patients assigned to general anesthesia alone used a median of 42 mg intravenous morphine versus 0 mg in those assigned to combined general/epidural anesthesia. A limitation is that the trial was done in a single center and had relatively wide CI.

Three of the four trials report CIs that do not preclude meaningful benefit or harm from regional analgesia. However, none of the point estimates suggests clinically meaningful benefits, which might have indicated that the results were underpowered rather than truly neutral. The pooled adjusted hazard ratio was 0.95 with tight 95% CI (0.85 to 1.06; P = 0.40; fig. 2). The trials included abdominal, breast, and lung cancers, which are the three most common types of nonskin cancer. It remains possible that regional analgesia reduces recurrence for other types of cancer, but that seems unlikely and would affect a small fraction of the overall cancer population. There is also a report that peritumor infiltration of breast cancer with local anesthesia reduces local recurrence and metastases, presumably via a novel local mechanism.

Combined, four robust trials (total N = 4,770) in patients with the most common types of cancer thus indicate that regional blocks provide little, if any, protection against cancer recurrence. These results were unexpected based on considerable in vitro and animal work suggesting that regional analgesia should reduce cancer recurrence after potentially curative cancer surgery. The question, then, is why there was such a disconnect between the preclinical results and clinical outcomes. In the following sections, we will thus evaluate preclinical work and discuss why laboratory and animal studies may have been misleading.

In vitro studies play a critical role in basic science research. These studies are foundational for deciphering the cellular mechanism of cancer dissemination. This section outlines the challenges of in vitro studies and how misinterpretation of their results may have misled clinical researchers to investigate the effect of regional anesthesia on cancer recurrence.

Billions of cells experience mutations and aberrancies that can lead to cancerous transitions daily in each of us. Fortunately, our immune systems are finely tuned to detect and eliminate cancerous cells.  The most important of the many protections against cancer are natural killer cells, a specialized type of T cell that is preprogramed to recognize and eliminate cancerous cells.  Natural killer cell function may be especially important during and after cancer surgery, when tumor disruption can release malignant cells into circulation.  For example, the count and function of human natural killer cells is reduced 30 to 40% by cancer surgery.  The effects of volatile anesthetics, opioids, and surgical stress on natural killer cell function are thus of paramount importance.

Tumor-killing assays are well accepted and have been used for nearly a half-century to investigate and develop immune therapies. However, data from natural killer cell activity assays have shortcomings, and the reproducibility of the experimental results is poor. In vitro killing assays have been used to evaluate the effect of mediators of the surgical stress response, opioids, and anesthetics on the number and function of natural killer cells. For example, since the mid-1970s, investigators have reported dose-dependent and reversible inhibition of human natural killer cell cytotoxic activity after in vitro exposure to volatile anesthetics such as halothane and nitrous oxide.  Others suggested a similar phenomenon using ex vivo natural killer cells from rodents exposed to halothane or isoflurane.

Bar-Yosef et al.  and Wada et al.  investigated ex vivo the effect of spinal anesthesia with bupivacaine on the natural killer cell function of rats having surgery and reported that spinal anesthesia restored the suppressive effects of surgery and volatile anesthetics on natural killer cell activity and numbers. These studies clearly suggested that regional anesthesia might preserve perioperative natural killer cell function.

Opioids also suppress natural killer cell activity in humans and rodents.  However, the effects are type-, dose-, and time-dependent.  For instance, high doses of morphine suppress the function of natural killer cells, whereas tramadol apparently improves function. Hydromorphone and oxycodone do not exert meaningful activity on the cytotoxic activity of natural killer cells.  Spinal morphine also suppresses ex vivo killing activity of natural killer cells, although only transiently and only by 20 to 25%.  The effect of opioids on natural killer cells thus appears to be highly variable, of modest magnitude, and short lived.

Surgical stress is characterized by release of catecholamines and other stress hormones that potentially impair natural killer cell function. For example, studies using ex vivo natural killer cells from rats report that surgery stress suppresses the killing of cancer cells by natural killer cells, an effect that is reversed by β blockers. However, surgical and nonsurgical studies in humans, using ex vivo natural killer cells, demonstrated that epidural or stellate ganglion blocks only slightly improve natural killer cell function and that the effect is not correlated with changes in systemic catecholamine concentrations.  There is thus only marginal evidence that surgical stress meaningfully impairs natural killer cell function or that neuraxial or regional blocks reverse the effect.

The studies in the last few paragraphs were considered critical evidence supporting the need for randomized trials. However, results from cytotoxicity assays can be misleading.  Perhaps they should have been interpreted with greater nuance and a better understanding of methodologic limitations. First, there is a clear discrepancy between in vitro experimental results using natural killer cells from animals and humans and even among various animal strains. For instance, MADB106 breast cancer cells were initially derived from breast adenocarcinoma in rats.  Furthermore, EL4 T-cell lymphoma cells (mouse origin) were used to investigate the impact of spinal anesthesia in immunocompetent mice.  However, it is well known that murine natural killer cells do not necessarily replicate the human counterpart’s function or activity under resting conditions.  Furthermore, batch-to-batch variation in natural killer cell acticity isolated from peripheral blood mononuclear cells can affect interpretation of the results. Second, in the presence of surgical tissue injury, the proposed protective effect of spinal or epidural anesthesia in humans is generally overwhelmed by the immunosuppressive effects of tissue damage. Third, the tumor microenvironment is rich in immune mediators (such as prostaglandins, interleukins, and catecholamines) that modulate the activity of natural killer cells—but are lost during cell manipulation or used in fixed concentrations.  Fourth, in vitro killing assays use immortalized cancer lines as targets for natural killer cells. An additional factor is that heterogeneity and genomic diversity are acquired during long-term cell culture which can promote sensitization or resistance to immune mediators (immune escape) and anticancer therapies.  Mutations may explain why treatment effects are divergent within cell lines even when experiments are adequately controlled. 

In summary, cytotoxicity assays were generally well performed and contributed to our understanding of how volatile anesthetics, opioids, and spinal anesthesia influence the function of natural killer cells. However, the results of these experiments were often overinterpreted, given the generally modest differences in killing activity, or misinterpreted due to important kinetics aspects (i.e., onset, peak effect, or duration of effect) inherent to cytotoxicity killing assays and species-related differences. Future research investigating the effect of perioperative pharmacologic intervention on rodents’ natural killer cell–mediated tumor killing should be assessed with newer and more comprehensive methods such as degranulation and maturation assays and a three-dimensional spheroid format using either whole peripheral blood mononuclear cells or purified natural killer cells. Three-dimensional formats can better model the in vivo tumor structure. If successful, testing the proposed interventions could progress to in vivo patient-derived xenograft or syngeneic mouse models.

In vitro assays are routinely used to investigate mechanisms of metastasis. In vitro assays examining cancer cell behaviors such as proliferation, migration, and invasion offer advantages, including being relatively inexpensive, easy to conduct, and highly reproducible.  Still, interpretation of their results needs to be carefully considered.

In 1968, Fink et al.  recognized the in vitro effect of halothane on sarcoma cells. Between then and the early 2000s, various studies reported that volatile anesthetics had no substantial effects on migration and invasion assays. However, more recent studies conducted with modern volatile anesthetics, including sevoflurane and desflurane, demonstrated in vitro effects on cell behaviors that might promote metastasis. In vitro investigations similarly demonstrated that opioids promote tumor proliferation, invasion, migration, and angiogenesis via the µ-opioid receptor in various lung or breast cancer cell lines. In vitro studies also suggested a critical role of stress hormones in the metastatic process. Sood et al.  and Yang et al.  showed that epinephrine and norepinephrine significantly increased invasion in ovarian and nasopharyngeal cancer cell lines. There is thus considerable evidence from tumor cell behavior assays that reducing volatile anesthetic use, giving less opioids, and decreasing surgical stress might reduce cancer recurrence.

Cell behavior assays nonetheless misdirected clinical investigators for various reasons. First, most studies used immortalized cancer cell lines propagated in two-dimensional cultures without a structured component (i.e., lymphatic vessel or collagen structure). It is now well understood that in vitro assays do not resemble the architecture and microenvironment of the tumor in patients. Consequently, malignant cell lines cultured in vitro poorly replicate human cancers with respect to phenotype, cancer cell composition, and treatment response.  Second, there is little information about receptor conformational changes or isoform types among cancer cell lines. For instance, there are more than 10 isoforms of the µ-opioid receptor, but which isoforms are in various cancer cell lines remains unknown, as is the extent of tumor heterogeneity in human specimens.  Third, in vitro experiments have commonly overlooked variables that influence responses, including seeding cell number, high drug doses, long incubation duration, type of culture media, and cells’ supernatant removal after incubation. For instance, a study reporting that sevoflurane promotes proliferation and aggressive cell behaviors in breast cancer cells in vitro was based on dissolving volatile anesthetic in culture media instead of using an airtight chamber at clinically meaningful concentrations. 

In summary, it is generally accepted that cancer cells must interact with other cells, degrade the extracellular matrix, and proliferate before establishing thriving metastases. Many studies reported that anesthetics, opioids, and surgical stress facilitate those mechanisms in vitro. However, proliferation, migration, and invasion assays poorly predict long-term consequences of anesthetics, opioids, and surgical stress. This led investigators to overestimate the generalizability of results from in vitro experiments based on a highly reductionist approach to the study of cancer.

In fairness, the limitations of classical in vitro models were poorly appreciated when the key in vitro cancer-anesthesia studies were done. Future in vitro investigations should consider stronger models, including three-dimensional ones, that presumably better predict in vivo cancer proliferation, migration, and invasion. It would also be helpful if investigators better conveyed the limitations of the models they use and the extent to which findings are likely (or not) to translate to the clinical scenario. Similarly, clinical investigators need to be more nuanced about the interpretation of basic science results and recognize that all in vitro models are abstractions, sometimes to the point of providing little guidance for trial design.

Animal models are used as a bridge between in vitro results and clinical tests. The underlying assumption in preclinical animal research is that results will predict human responses to a reasonable degree. Animal models were therefore used to study the impact of regional anesthesia on the metastatic process after surgery. For instance, two studies showed that surgery with halothane significantly increased lung tumor retention in rat and mouse models of metastatic breast cancer and fibrosarcoma. Similarly, adding intrathecal anesthesia with bupivacaine significantly attenuated the metastasis-promoting effect of surgery (fig. 3).  The obvious question is why these and related studies in animals failed to translate to humans.

There are two broad types of metastatic mouse and rat models. In orthotopic models, metastatic cells originate from the orthotopically growing tumors, followed by the migration of tumor cells to the metastasis site. In disseminated models, a tail vein injection of tumor cells promotes metastases in lung, bone marrow, or brain. Many studies investigating the effects of anesthetics, opioids, or stress used the latter model in which immortalized cancer cells (MADB106 and EL4 T cells) were injected via the tail vein.  The success of the tail vein model varies substantially depending on the metastatic potentials of the cells. Therefore, highly metastatic cells such as MADB106 are commonly used to gain efficacy in metastasis formation. While this method of generating distant lung metastasis is well accepted, it may only partially mimic metastasis formation in humans because: (1) it does not follow the early biologic steps of the metastatic process; (2) injecting a large bolus of tumor cells into the circulatory system is not the most common mechanism of metastatic disease after surgery; and (3) tumors grow faster in animal models than in humans.

It could be argued that the orthotopic and carcinogen-induced spontaneous tumor models are considered better paradigms of clinical cancer—and thus have more robust external validity—than the tail vein inoculation of cancer cell lines.  A major advantage of orthotopic and carcinogen-induced spontaneous tumor models is that cancers closely mimic early metastatic steps. However, when the major trials of regional anesthesia and cancer were designed, no experimental data from orthotopic and carcinogen-induced spontaneous tumor models were available—and they are still lacking.

In animals, postsurgical pain per se may facilitate the metastatic process. Conversely, a strong nociceptive block reduces metastasis.  For example, studies report a lower metastatic burden in rodents with profound blocks from intrathecal bupivacaine than those without.  Similarly, metastasis burden was higher in animals given intrathecal morphine.  Conversely, animals given systemic morphine showed a reduction in tumor burden. These studies suggested that robust pain control is needed to reduce cancer progression. Regional blocks generally provide excellent pain relief but perhaps not enough to influence metastases. Furthermore, in some trials the amount of pain in the reference groups was modest or unreported. 

Circulating tumor cells are common in patients with cancer and are the seeds of metastasis.  Most circulating cancer cells are eliminated by natural killer cells, but any that remain can become metastases (after many steps, each of which can prevent progression).  Animal studies report that spinal anesthesia with bupivacaine prevents the suppressive effects of surgery on natural killer cell activity ex vivo  whereas halothane inhibits the function of natural killer cells in the context of surgery.  Importantly, however, recent evidence from randomized trials and a meta-analysis indicates that counts and function of natural killer cells do not correlate with the number of circulating tumor cells. Hence, the observed “significant” differences in ex vivo cytotoxic assays between animals with and without regional blocks may not correspond to clinically meaningful immunological protection based on circulating tumor cell analysis.

Most animal studies only evaluated the effects of one or two drugs (i.e., volatile anesthetics or opioids) in combination with spinal anesthesia or spinal anesthesia alone. The rationale behind such an experimental design was to gain a mechanistic approach without confounding or confusing factors. However, multiple drugs are given during and after surgery including antibiotics, muscle relaxants, and steroids. Many such drugs directly influence the function of natural killer cells and may therefore have attenuated the effect of regional anesthesia on the in vitro killing activity of the cells.  In addition, animal studies overestimated by about 30% or more the effect of regional anesthesia on cancer recurrence, which is consistent with published data in other fields. 

A final concern is that animal research often omits formal randomization and investigator blinding, compromising internal validity, which is likely related to the lack of best-practice standards for animal testing.  Sample sizes are often unjustified and inadequate, and statistical analysis is sometimes unsophisticated. A survey of 271 articles about rodent and nonhuman primate studies found that 87% did not report random treatment allocation and 86% did not report investigator blinding.  For example, blinding and random allocation were not reported in studies by Bar-Yosef et al., Wada et al. and Page et al.  in their studies of regional analgesia and cancer.

Hypothesis-driven clinical trials are usually designed to control the type I and type II errors related to the primary outcome. However, laboratory research is often conducted with 6 to 12 animals/group and unjustified assumptions of large treatment effects—a practice with little scientific and statistical basis. For example, influential studies reporting the effect of spinal anesthesia on cancer recurrence did not provide sample size estimates.  In other words, unlike clinical research, there are few well established standards for the conduct of basic science. Too often, the number of animals is arbitrary, species selection is based on convenience and cost, and formal randomization and blinding are lacking. Furthermore, animal research often stops within rodents without follow-up in larger and more relevant species. In contrast, interesting observational analyses in patients are usually followed by small clinical trials and when appropriate, by large, robust trials.

In summary, clinical trials were based on animal data suggesting that regional blocks, avoiding volatile anesthetics, and reducing opioid administration might reduce cancer recurrence after surgery. However, animal models of cancer seem to be especially bad at predicting human responses perhaps explaining why clinical trialists were misled into believing that regional anesthesia might reduce cancer recurrence. It could be argued that better experimental paradigms such as the orthotopic and carcinogen-induced spontaneous tumor models should be considered in future investigations. Indeed, such models better reflect clinical cancer than the tail vein inoculation of cancer cell lines because they closely mimic early metastatic steps.  For example, Doornebal et al. used an orthotopic mouse model of breast cancer in which they compared the impact of morphine (versus saline) after mastectomy. In that model, morphine did not promote lymphatic or distant metastatic formation, consistent with clinical studies in breast Last, laboratory animal investigations should use adequate sample sizes and rigorous statistical methods, designs resembling those used routinely for clinical research to avoid type 1 and type 2.

Preclinical research involves experiments on cells, tissues, animals, and other biologic materials and seeks to better understand diseases’ biologic mechanisms and develop potential therapies. Disciplines include molecular biology, pharmacology, toxicology, and animal behavioral sciences. Preclinical investigations have been the foundation of major discoveries and advances in anesthesiology, including the mechanism of action of local anesthetics and hypnotic agents such as propofol which the World Health Organization declared as an “essential medicine.”

Translational science refers to the bench-to-bedside process that transforms knowledge from basic scientific observations into therapeutic advances.  The process is designed to select basic discoveries most likely to succeed in human trials.  For instance, studies in frogs provided the basis for much of what we currently know about the mechanisms of action of local anesthetics and how to optimize various drug characteristics.  A considerable amount of translational research was foundational for the success of propofol in clinical practice.

Experimental paradigms of human cancer have been used for decades to study tumor biology and assess the efficacy of therapeutic agents. However, cancers are highly complex diseases with variable courses and responses to therapy.  Consequently, experimental models only approximately replicate the conditions under which cancer progression or dissemination occurs in humans. 

Many preclinical studies suggested that regional analgesia might reduce the risk of cancer recurrence after potentially curative surgery—although it apparently does not. Limitations of the relevant in vitro studies included using established cell lines instead of patient-derived cancer cells, poor replication of tumor microenvironments, lack of three-dimensional tumor organization and extracellular matrix, and excessive drug concentrations and duration. The animal models were based on a reductionist approach to how tumorigenesis and metastasis occur in humans and used tumors with genomic makeup that differ from naturally occurring tumors. Furthermore, some of the animal models—such as tail vein injection of cancer cells—poorly replicate natural tumors.

While we now know that many of the cancer models used in preclinical research were flawed, they were often the best available at the time, and the studies were generally well conducted.  We thus do not criticize investigators who conducted preclinical studies of regional analgesia and cancer progression, some of which date back more than two decades. Nor were their conclusions unreasonable given their understanding at the time. Nonetheless, it appears that important limitations of in vitro and animal studies contributed to the (apparently incorrect) theory that regional analgesia prevents cancer recurrence after potentially curative cancer surgery.

It would be easy to look back and say that preclinical research about regional analgesia and cancer was flawed by use of suboptimal models. That assessment is accurate, but based on current understanding of tumor models that was unavailable when the first major trial started in 2007.  It is also true that clinical investigators had a poor understanding of the limitations of existing models—a poor understanding possibly shared by the preclinical investigators. It would also be easy to blame poor communication across the translational continuum, but in fact there was considerable discussion between basic scientists and clinical investigators, and some participated in both kinds of research. 

The value of a clinical trial is proportional to the number of people affected by their findings and their impact on decision-making. In our case, metastasis is a leading cause of death in patients with cancer and the second leading cause of death worldwide.  Therefore, investigating whether perioperative interventions might moderate cancer recurrence is clearly important.

Translational research is an iterative cycle of designing robust experimental studies, correctly interpreting the findings, and using the resulting understanding to design clinical trials that provide strong causal information about improving clinical practice. In the case of regional anesthesia and cancer recurrence, the discordance between preclinical studies and clinical trials apparently resulted at least in part from the use of models that we now know were suboptimal, although they were well accepted at the time. Consequently, the reliability of experimental in vitro and in vivo models was overestimated by basic scientists who failed to recognize how poorly reductionist models characterize the complexities of human cancers. Clinical investigators similarly misjudged the reliability of preclinical evidence, although basic scientists are presumably far better positioned to judge the reliability of their work and the extent to which it might reasonably extrapolate to humans. The result was a huge effort and the expense of four major trials, none of which support the preclinical evidence suggesting that regional analgesia might reduce cancer recurrence.

The question, of course, is whether the disconnect between preclinical research and trials might have been prevented. There is no simple answer to this question. Basic science methodology improves rapidly, as does understanding about weaknesses of previous methods. In contrast, it can easily take a decade or longer to organize and conduct a suite of clinical trials exploring a theory from various angles.  Clinical trials will thus always be based on science that will be at least somewhat out of date by the time the trials are completed. However, a major purpose of preclinical science is to guide trials, which constitutes the first step toward rational clinical implementation.  The quality of preclinical evidence results depends upon planning and conducting—even when this could involve considering experiments in higher animal species. It is thus reasonable to ask basic scientists to help with that transition since they are far better positioned than clinical investigators to judge which findings are likely to translate.

Although there have been occasional efforts, there is not currently a formal process for setting priorities for clinical trials or for basic scientists to guide clinical investigators.  We need such a process, because trials are expensive and time-consuming.  There will never be enough of them to test every interesting question. A multidisciplinary process that guides topic selection should make for better trials. 

Importantly, guidance groups should cross many areas, at least within a clinical specialty—a process that would force preclinical scientists to defend the strength of their recommendations to their colleagues who are presumably positioned to judge the merits of various proposals. Some sort of forced ranking process will help identify which of many areas of investigation are likely to provide clinical benefit. Clinical investigators can then focus on the top choices and consider their side of the issues, including the number of people at risk, likely treatment effect, availability of patients, regulatory issues, and the many other practical considerations that go into trial design. Done well, such a multidisciplinary process would guide the selection of topics for trials, some of which will meaningfully advance care.