Controlled Trials of “Palliative Care”—Where Do We Go from Here?

A Paradigm Shift from Early SPC Studies

These two studies have the potential to transform palliative care delivery and to meet the challenges posed by a growing demand for palliative care and a shrinking or stagnant palliative care workforce ⁵ to those with serious illness. They also highlight important new challenges in conducting research and interpreting our findings.

First, the primary research question has pivoted from “Does early SPC help patients?” to “How do we achieve similar outcomes with less resource-intensive care delivery models?” These fundamentally different questions require a shift in design and analysis from a focus on superiority to noninferiority, with attendant shifts in sample size, complexity, and budget. For example, the original Temel study required only 151 patients to show the superiority of early SPC, which took three years to recruit at a single center. The stepped-care trial needed 507 patients from three sites to demonstrate noninferiority versus early SPC. The virtual palliative care trial enrolled 1250 patients from 22 sites to demonstrate noninferiority versus early SPC. More complex trials are likely disproportionately susceptible to systemic disruptions, such as a pandemic; these trials took five years to complete. Even with more efficient data collection and multi-center collaborations, we will not be able to answer noninferiority questions at the same rate as we have answered superiority questions.

Second, a shift toward noninferiority studies side-steps the question of whether these interventions are actually superior to “usual care.” These studies assume that early SPC (the control group) is better than usual care, an assumption with face validity but potential threats to generalizability due to selection bias—the centers where these studies are conducted don’t look like most centers. While the original Temel study demonstrated a moderate effect size (Cohen’s d = 0.42), ¹ more recent meta-analyses of SPC RTCs have shown smaller effect sizes for quality of life (d = 0.12–0.33). ⁶ This could represent a regression toward the mean, or it could also mean that usual care is improving over time. Is it plausible that “usual care” in the last five years is as effective (or ineffective) as it was 15–20 years ago? After all, palliative care educators have made important gains in advancing the “primary palliative care” competencies of oncologists and other specialists who care for patients with serious illnesses. Studies of specific primary palliative care interventions have generally not shown significant improvements in outcomes, ⁷ but this does not rule out the possibility of a slow but substantial improvement over 15 years across the field. Furthermore, cancer-directed treatments have advanced, with implications for prognosis, symptom burden, and management of side effects. Immunotherapy and targeted therapy for metastatic cancer can reduce tumor burden with fewer side effects than systemic chemotherapy. Ultimately, early SPC and “usual care” are not medications—they are complex interventions with an effectiveness that is likely to change over time and space. This raises difficult-to-answer questions about the meaning of noninferiority in 2024.

A third challenge is the timing of the primary outcome. Johnson et al. recently published a meta-regression study looking at the effect size of multi-component SPC compared with “usual care” in 39 RCTs. They found that for quality of life, using the most robust estimates of minimally important differences in outcome, the greatest effect size was seen at 13–36 weeks. Notably, both the stepped-care and virtual care RCTs used a primary endpoint measured at 24 weeks, but the findings of Johnson et al. may suggest that we are committed to using this later endpoint from now on, which increases the difficulty and costs of research. It also increases the chances that studies will be abandoned due to poor accrual ⁸ or biased due to attrition.

The Most Important Unanswered Questions?

The complexity of our research questions drives the complexity of our research. Unfortunately, questions about early SPC have become dramatically more complex in recent years. Where do we go from here?

First and foremost, we still have a limited notion of what component(s) of SPC is decisive for improving patient-centered outcomes. SPC is sometimes a “kitchen sink” intervention—a whole team approach to addressing the components of total pain, for example. But it is ideally a robust and adaptive intervention (as outlined in these recent RCTs), identifying patient needs in a holistic manner and responding accordingly. SPC appears to be more effective for endpoints such as quality of life or physical symptoms than for mood symptoms. ⁸ If we understood why, we might be better able to scale this effect to reach more people—or deploy specific team members to address specific forms of distress.

Second, we understand little about the effectiveness of early (or stepped, or virtual) SPC care delivery models across different illnesses. Even in oncology, an RCT of early SPC found different outcomes for people with lung and gastrointestinal cancers; ⁹ we should not assume that the effect seen in these latest RCTs would be found in noncancer populations. Ultimately, the many (and growing) populations of interest to palliative care mean that is not realistic to expect that we can test all potential approaches in each population. It will also be difficult to justify the use of so much of our limited research resources to study a single research question, to the detriment of other clinical studies of medications or nonpharmacologic interventions.

Third, we must identify predictors of good or poor response. For example, Greer et al. found in an RCT that the coping strategies used by patients were a strong predictor of outcome. Patients who used an “approach-oriented” coping mechanism (featuring active coping, positive reframing and acceptance) had much better quality-of-life and depression scores after 24 weeks than patients who used an avoidant coping mechanism. ¹⁰ Methodologists are increasingly focused on heterogeneous treatment effects in a large RCT, ¹¹ and how we might use RCT data to better identify those who would benefit most, and least, from an intervention. Is it possible that a modified SPC intervention focused on illness understanding and reinforcing more beneficial coping strategies might be more effective than our current approaches?

Moving Forward: The Importance of Modest Expectations and Implementation

Recommendations regarding the provision of early SPC are here to stay. However, our current research focus must shift toward identifying the most effective components of our intervention, which patients can benefit the most, and how we can increase our impact without dramatic increases in resources. We must also be realistic in what we can expect from RCTs as we stop trying to prove that SPC is better than usual care, and start trying to identify lower-resource, scalable forms of early SPC that are as good as the standard, high-resource version. We must also continue to study implementation considerations, to help ensure that effective interventions are effectively applied in practice as they were in the RCTs.

As demand grows, we face a dual challenge: optimizing limited SPC resources to expand clinical impact while strategically allocating limited research resources to address increasingly complex questions. There are no easy answers.