Bolus-Only Hydromorphone Pump Fails Noninferiority Test in Refractory Cancer Pain Trial

For the roughly 10 to 20 percent of cancer patients whose pain remains uncontrolled despite guideline-based therapy, patient-controlled intravenous analgesia has become an increasingly common rescue and titration tool. Yet a basic question has gone largely unanswered: when a pump delivers hydromorphone on demand, does adding a continuous background infusion actually help? A prospective, multicenter, randomized noninferiority trial published in Supportive Care in Cancer now provides one of the most direct tests to date, and its answer is deliberately cautious. Bolus-only delivery did not prove noninferior to bolus plus continuous infusion on the trial’s primary endpoint, although both approaches produced rapid, clinically meaningful pain relief with comparable safety.

The trial enrolled 226 adults with refractory cancer pain between December 1, 2018, and August 31, 2022, across seven centers in a study registered with the Chinese Clinical Trial Registry (ChiCTR1800020014). Participants were 18 to 80 years old, had pathology- or cytology-confirmed solid tumors, and met a strict definition of refractory pain: persistent inadequate analgesia or intolerable opioid side effects despite optimized conventional therapy. Before enrollment, every patient had received standardized cancer-pain treatment for one to two weeks, including fentanyl transdermal patches, extended-release oxycodone tablets, and adjuvant analgesics, and had been on a stable extended-release opioid dose for at least seven days. Patients with paralytic ileus, brain metastases, uncontrolled nausea or vomiting, severe hepatic or renal failure, or recent PCA or hydromorphone exposure were excluded.

Randomization was 1:1, concealed in sequentially numbered sealed opaque envelopes and prepared by an independent nurse at each center. One hundred four patients were assigned to the bolus-only group and 111 to the bolus-plus-continuous-infusion group. In the bolus-only arm, each on-demand hydromorphone bolus was set at 10 plus or minus 5 percent of the total equianalgesic opioid dose taken in the preceding 24 hours. In the combined arm, a continuous infusion delivered the preceding 24-hour equianalgesic dose divided by 24 per hour, with the same bolus size available for breakthrough episodes. The lockout interval was fixed at 8 minutes in both groups, and pump parameters stayed unchanged for the first 24 hours before protocol-driven titration every 24 hours thereafter.

The primary endpoint was the pain relief rate at 72 hours after PCIA initiation, defined as the proportion of patients achieving at least a 50 percent reduction in resting pain intensity on the 11-point Numerical Rating Scale from baseline. With a prespecified noninferiority margin of minus 10 percent, one-sided alpha of 0.05, and 80 percent power, the trial required at least 103 patients per group. At 72 hours, the pain relief rate was 89.4 percent in the bolus-only group (93 of 104) and 94.6 percent in the combined group (105 of 111). The between-group difference yielded a 90 percent confidence interval of minus 11.2 percent to 1.1 percent. Because the lower bound crossed the minus 10 percent margin, noninferiority of bolus-only delivery was not demonstrated.

The authors stress what this result does and does not mean. The point estimate and upper confidence bound suggest only a modest absolute difference between the strategies, and a small but potentially meaningful benefit of background infusion cannot be excluded, particularly in patients with high baseline symptom burden. Formally, equivalence cannot be concluded either. A sensitivity analysis adjusted for a significant baseline imbalance in breakthrough pain intensity told a more supportive story: after adjustment, the between-group difference in pain relief rate was 2.28 percentage points in favor of bolus-only delivery, with a 90 percent confidence interval of minus 9.34 to 13.90 percentage points and a P value of 0.747. That adjusted lower bound sat above the noninferiority margin, providing supportive, though not definitive, evidence once the baseline imbalance was accounted for.

Secondary outcomes added texture to that ambiguity. Resting NRS scores fell significantly from baseline at every assessment from 1 hour through 72 hours in both groups, with all comparisons yielding P values below 0.0001, and no significant between-group differences at any time point. Breakthrough cancer pain intensity also declined over time in both arms, though the bolus-only group’s reduction was not statistically significant at the 1-hour mark before becoming significant from 4 hours onward. Because baseline breakthrough pain NRS scores differed significantly between groups, the investigators included baseline values as covariates in a mixed model for repeated measures; after adjustment, no significant between-group differences in breakthrough pain intensity remained at any post-baseline time point.

Where the modes diverged was in the frequency of breakthrough episodes. In the repeated-measures analysis, daily breakthrough pain episode frequency fell significantly from baseline through 24 hours in the bolus-only group but rose above baseline at 72 hours, while the combined-infusion group showed significant reductions through 48 hours with no significant reduction at 72 hours. At the final assessment, the estimated daily episode frequency was significantly higher in the bolus-only group, by 1.72 episodes per day (P = 0.029). The authors interpret this as possibly reflecting more stable analgesic coverage under continuous background infusion in some patients, consistent with earlier work linking steadier plasma opioid concentrations to fewer pain fluctuations and less rescue analgesia, though they frame it as an interpretation rather than a proven mechanism.

Opioid consumption told a complementary story. The increase in equivalent morphine consumption, which captures the additional opioid delivered by the pump on top of each patient’s unchanged baseline regimen, rose significantly from baseline at the 24-, 48-, and 72-hour marks in both groups, with no significant between-group differences at those time points. Within the first 24 hours, however, the combined-infusion group showed higher early opioid exposure, reflecting accumulation from the continuous basal rate. Notably, breakthrough episode counts were comparable across groups in that early window, suggesting the extra consumption was not driven by greater rescue need. The authors raise, but cannot resolve, the question of whether sustained opioid exposure might accelerate tolerance over longer periods, noting that dose escalation can also stem from tumor progression, pain sensitization, and other confounders.

Function and symptom burden improved in both arms. Brief Pain Inventory scores showed significant improvements at 72 hours across domains covering general activity, mood, walking ability, interpersonal relationships, enjoyment of life, and sleep, with all comparisons at P below 0.005 and no significant between-group differences in total or subscale scores. Edmonton Symptom Assessment System total scores, measured on a 0-to-90 scale across nine symptoms, fell significantly from baseline to 72 hours in both groups (P below 0.001), with no significant between-group difference at either time point despite a baseline difference in anxiety scores. Safety findings were similarly reassuring: nausea, vomiting, constipation, somnolence, dry mouth, and pruritus occurred at comparable rates, no unexpected or serious adverse events were recorded, and there were no cases of opioid abuse, catheter-related complications, or pump malfunctions.

The trial’s limitations frame how far its conclusions travel. The 72-hour observation window precluded assessment of opioid tolerance, endocrine effects, or durable quality-of-life changes. The open-label design may have introduced operational and reporting bias, partially mitigated by objective endpoints such as opioid consumption. Baseline imbalance in breakthrough pain intensity may have influenced downstream pain dynamics and rescue analgesic use, and the fixed dose-adjustment algorithm, while standardizing care, may not reflect individualized clinical practice. Eleven patients withdrew during the observation period, nine from the bolus-only group and two from the combined group, mostly due to tumor-related complications and systemic infections. The authors conclude that PCIA mode selection should not follow a one-size-fits-all approach: patients with frequent breakthrough pain or marked diurnal variability may benefit from background infusion, while those with more predictable pain patterns may be adequately managed with bolus-only dosing, potentially minimizing opioid exposure. Future work, they suggest, should extend follow-up, add pharmacokinetic analyses, and stratify by pain phenotype, circadian variation, and psychological comorbidity.

Subject of Research: Comparing bolus-only versus bolus plus continuous infusion hydromorphone patient-controlled intravenous analgesia in patients with refractory cancer pain

Article Title: Bolus-only versus bolus plus continuous infusion hydromorphone for refractory cancer pain: a prospective, multicenter, randomized noninferiority trial

Article References: Yan, Z., Fu, Y., Jin, Y., Zhao, W., Gong, L., Xie, G., Zhou, L., Tian, D., & Feng, Z. (2026). Bolus-only versus bolus plus continuous infusion hydromorphone for refractory cancer pain: a prospective, multicenter, randomized noninferiority trial. Supportive Care in Cancer, 34(11), Article 1139. https://doi.org/10.1007/s00520-026-11323-4

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11323-4

Keywords: refractory cancer pain, hydromorphone, patient-controlled analgesia, noninferiority trial, breakthrough cancer pain, opioid titration, continuous infusion, pain relief rate, Supportive Care in Cancer, randomized controlled trial, opioid consumption, palliative care