Establishing an efficacy hierarchy for breakthrough cancer pain management: a network meta-analysis of rapid-onset fentanyl formulations and morphine

Abstract

Background

Breakthrough cancer pain (BTP) is a common and debilitating condition affecting most patients with advanced malignancy. While multiple rapid-onset opioid formulations are available, a lack of comprehensive comparative evidence has hindered evidence-based treatment selection. We conducted a network meta-analysis (NMA) to evaluate the relative efficacy of all available formulations and characterize their temporal analgesic profiles for BTP management.

Methods

We performed a systematic search of PubMed, Embase, and the Cochrane Library for randomized controlled trials (RCTs) evaluating rapid-onset fentanyl formulations (intranasal fentanyl spray [INFS], fentanyl buccal tablet [FBT], fentanyl pectin nasal spray [FPNS], fentanyl sublingual tablet [FST], fentanyl buccal soluble film [FBSF], and oral transmucosal fentanyl citrate [OTFC]) and morphine preparations (oral morphine sulfate immediate release [MSIR] and subcutaneous morphine [SCM]) for BTP in adults. The primary outcome was the standardized mean difference (SMD) in pain intensity difference (PID) at 10, 15, 30, and 60 min post-administration. A frequentist random-effects NMA was conducted for each time point, and treatments were ranked using P-scores, a frequentist metric ranging from 0 to 1 that reflects the probability of a treatment being superior to competitors (higher values indicate better ranking; distinct from p-values for statistical significance).

Results

Seventeen RCTs involving 1,463 patients were included. Intranasal fentanyl spray (INFS) was consistently the most effective treatment across all time points, demonstrating statistically significant pain relief within 10 min (SMD: 0.85; 95% CI: 0.59–1.10) and showing its peak effect at 15 min (SMD: 1.05; 95% CI: 0.67–1.42). Fentanyl buccal tablet (FBT) consistently ranked as the second most effective agent. In contrast, immediate-release oral morphine (MSIR) did not achieve statistical superiority over placebo until the 30-minute mark (SMD: 0.20; 95% CI: 0.01–0.39; p = 0.036), confirming a comparatively delayed onset of action.

Conclusions

This analysis provides evidence for a relative efficacy hierarchy among rapid-onset opioids for BTP. INFS demonstrated the most rapid and sustained analgesia across all evaluated time points, while FBT consistently ranked second. The delayed onset of MSIR warrants consideration when selecting agents for rapid pain relief. Treatment selection should integrate these efficacy data with safety profiles, route feasibility, and patient-specific factors.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-04993-8.

Background

Cancer-related pain affects approximately 70–90% of patients with advanced malignancy and represents one of the most feared and debilitating symptoms of disease progression [1]. Despite significant advances in oncologic pain management, achieving optimal analgesia remains a substantial clinical challenge. The current standard of care involves around-the-clock (ATC) baseline analgesics, primarily long-acting opioids, to provide sustained control for persistent background pain [2, 3]. However, even when background (baseline) pain is well controlled through ATC opioid therapy, up to 95% of these patients experience breakthrough pain (BTP), transient, severe exacerbations of pain that “break through” the ATC analgesic regimen [4].

BTP is characterized by a rapid onset, often reaching peak intensity within 3–5 min, and moderate-to-severe intensity (typically 7–10 on a 0–10 numeric rating scale [NRS]). The duration of untreated episodes varies considerably: the seminal study by Portenoy and Hagen reported a median duration of 30 min (range: 1 s to 240 min) [5], while a large European multicenter study by Davies et al. found a median untreated episode duration of 60 min, implying that approximately half of episodes persist beyond one hour [6]. A systematic review by Haugen et al. noted typical durations of 15–30 min, with substantial variability across studies [7]. These temporal characteristics have important therapeutic implications: the rapid onset necessitates fast-acting analgesics, while the variable and sometimes prolonged duration underscores the need to evaluate sustained efficacy beyond the initial minutes. These episodes significantly impair patients’ quality of life, functional capacity, and psychological well-being, thereby compromising the outcomes of their overall cancer care [8, 9].

Given the rapid temporal evolution of BTP, the timing of analgesic assessment is critical. Evaluating the pain intensity difference (PID) at 10, 15, 30, and 60 min post-administration provides clinically relevant endpoints that correspond to the natural history of BTP episodes and align with patient expectations for relief. Consequently, an ideal medication for BTP must possess pharmacokinetic properties that match these temporal characteristics. Such an agent should exhibit a rapid onset of action (ideally within 10–15 min), deliver significant analgesia within 30 min, and maintain its effect for approximately 1–2 h to cover the typical duration of a BTP episode. Furthermore, it must be easily administered by patients or caregivers in ambulatory and home-care settings [10, 11].

Traditional immediate-release oral opioids, such as morphine sulfate immediate release (MSIR), have historically been the standard of care for BTP management [12]. However, these agents demonstrate a significant temporal mismatch with the typical BTP episode. Their relatively slow onset of action (30–60 min to peak effect) and prolonged duration are poorly aligned with the rapid-onset, shorter-duration profile of BTP [13]. This mismatch often leads to inadequate analgesia during the most intense phase of the pain episode and may result in subsequent oversedation or adverse effects that outlast the pain itself [14, 15].

Various rapid-onset fentanyl formulations have been developed that utilize transmucosal delivery routes to bypass first-pass hepatic metabolism [16, 17]. These include oral transmucosal (OTFC), buccal (FBT, FBSF), sublingual (FST), and nasal (INFS, FPNS) delivery systems, each with distinct pharmacokinetic properties including different rates of absorption, bioavailability, and time to peak plasma concentration. Whether these pharmacokinetic differences translate to clinically meaningful differences in analgesic efficacy has not been systematically evaluated across all available formulations.

These transmucosal formulations achieve measurable plasma concentrations within 5–15 min for nasal delivery systems and 10–25 min for buccal formulations, offering a clinically significant advantage over the 30–60 min onset of oral morphine [18]. While transmucosal fentanyl formulations offer theoretical advantages over oral opioids in terms of onset of action, their comparative efficacy and the clinical significance of pharmacokinetic differences between formulations remain uncertain. Furthermore, the rapid systemic absorption that may confer therapeutic benefit also carries important safety implications. All rapid-onset fentanyl products in the United States are available only through the Transmucosal Immediate-Release Fentanyl (TIRF) Risk Evaluation and Mitigation Strategy (REMS) Access program, a restricted distribution system mandated by the FDA due to the risks of respiratory depression, misuse, abuse, and accidental exposure, particularly in opioid-non-tolerant individuals [19]. Common class-wide adverse effects include nausea (7–23%), dizziness (5–17%), and somnolence (3–15%), with route-specific effects such as nasal irritation and epistaxis for intranasal formulations and oral mucosal ulceration for buccal preparations [20].

Despite the availability of these multiple agents, considerable uncertainty persists among clinicians in selecting the optimal formulation for individual patients [21]. The existing evidence is fragmented; previous clinical trials have typically evaluated these agents against a placebo or in limited head-to-head comparisons [22–24]. This absence of direct, comprehensive comparative evidence means that treatment decisions are often guided by institutional preference, cost, or patient-specific factors rather than robust efficacy data [25].

A previous network meta-analysis by Zeppetella et al. (2014) compared the efficacy of rapid-onset fentanyl formulations for BTP at the 15-minute time point and identified INFS as the most effective agent [26]. However, that analysis was limited to a single time point, included 10 RCTs, and could not characterize the temporal evolution of analgesic effect. Since then, seven additional RCTs have been published, substantially expanding the available evidence base. The present study builds upon this prior work by incorporating all currently available trials and, crucially, by evaluating efficacy at four distinct time points (10, 15, 30, and 60 min), thereby providing a comprehensive temporal profile of analgesic onset and duration for all available formulations.

To fill this critical knowledge gap and provide evidence-based guidance for clinical practice, we conducted a comprehensive network meta-analysis. This approach allows for the simultaneous comparison of multiple interventions—even those not directly compared in trials—and provides a probabilistic ranking of their relative efficacy at clinically relevant time points. This analysis aims to establish a clear hierarchy of effectiveness among available BTP medications, elucidate their optimal timing of analgesic effect, and ultimately inform clinical decision-making to improve pain management for patients with cancer.

Methods

Search strategy and study selection

This systematic review was prospectively registered with The Open Science Framework (https://osf.io/r9yfg). The study selection process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Additional file 1: Table S1). A comprehensive systematic literature search was conducted across three major electronic databases: PubMed, Embase, and the Cochrane Library from inception until August 1, 2025. The search strategy combined terms related to breakthrough cancer pain, fentanyl formulations, morphine preparations, and randomized controlled trials, with no language restrictions applied during the initial search phase (Additional file 1: Table S2). Two independent reviewers (PHC and JHC) screened titles and abstracts for eligibility, followed by full-text assessment of potentially relevant articles. Disagreements were resolved through discussion or consultation with a third reviewer (CLH).

Eligibility criteria

We included studies that met the following Population, Intervention, Comparator, Outcome, and Study Design (PICOS) criteria: Population—adult patients (≥ 18 years) with a diagnosis of cancer who were experiencing BTP; Interventions—any formulation of rapid-onset fentanyl (INFS, FPNS, FST, FBT, FBSF, OTFC) or immediate-release morphine (oral or subcutaneous); Comparators—placebo or any of the other active interventions listed; Outcomes—studies must have reported PID at one or more of the following time points: 10, 15, 30, or 60 min post-administration, with sufficient data for quantitative analysis; Study design—randomized controlled trials (RCTs).

Studies were excluded if they focused solely on dose-titration without a comparative arm, or investigated intravenous routes of administration. Both crossover and parallel-group RCT designs were eligible for inclusion. Our analysis focused on rapid-onset fentanyl formulations and morphine preparations (oral and subcutaneous) as these represent the agents most commonly evaluated in RCTs and recommended in current clinical guidelines for BTP management (ESMO, NCCN). Other opioids (e.g., oxycodone, hydromorphone, sufentanil) were not included because the available RCT evidence was insufficient to form a connected treatment network.

Data extraction and quality assessment

Two reviewers (PHC and JHC) independently extracted data from all included studies using a standardized form. Key data points included study characteristics, patient demographics, intervention details, and PID outcome data at 10, 15, 30, and 60 min. Subsequently, the same two reviewers assessed the risk of bias for each study using the Cochrane Risk of Bias tool 2 (RoB 2) [27]. This assessment covered five key domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Based on these domains, each study was assigned an overall risk of bias judgment of “low risk,” “some concerns,” or “high risk.” The detailed assessment is available in Additional file 1: Table S3.

Outcome measures

The primary outcome for this network meta-analysis was the pain intensity difference (PID) from baseline, assessed at four discrete time points: 10, 15, 30, and 60 min post-administration. To harmonize data across trials that used various pain assessment instruments (e.g., visual analog scales, numeric rating scales), the treatment effect for each comparison was quantified as the standardized mean difference (SMD) with its 95% confidence interval (CI). This approach ensures all outcome data are comparable for quantitative synthesis [28, 29].

Statistical analysis

We performed a frequentist network meta-analysis (NMA) for each time point (PID10, PID15, PID30, and PID60) using the netmeta package (version 2.8-0) in R (version 4.3.0) [30]. To account for anticipated between-study heterogeneity, a random-effects model was used for all analyses, assuming a common heterogeneity variance (τ²) across all treatment comparisons [31, 32].

The validity of the NMA relies on the underlying assumptions of transitivity and consistency. We assessed transitivity by clinically evaluating the distribution of potential effect modifiers (e.g., patient populations, baseline pain characteristics) across the different treatment comparisons. Consistency, the agreement between direct and indirect evidence, was statistically evaluated as described in the following sections.

Network meta-analysis (NMA) extends traditional pairwise meta-analysis by simultaneously synthesizing both direct evidence (from head-to-head trials comparing two treatments) and indirect evidence (inferred through a common comparator) within a single coherent statistical framework [33]. For example, if Treatment A is compared with placebo in one trial and Treatment B is also compared with placebo in another, NMA can estimate the relative effect of A versus B indirectly through the shared placebo comparator. The validity of indirect estimates relies on the assumption of transitivity and consistency. The contribution of direct evidence to each pairwise estimate within the network is presented in Additional file 1: Fig. S1.

Network geometry and treatment ranking

The geometry of the evidence for each time point was visualized using network plots, where nodes represent the interventions and edges represent the available direct comparisons. The size of each node is proportional to the total number of patients randomized, and the thickness of each edge is proportional to the number of direct comparisons [33].

To rank the relative efficacy of all interventions, we calculated P-scores for each treatment. The P-score is a frequentist ranking metric ranging from 0 to 1 that reflects the certainty of a treatment’s rank relative to all other treatments in the network. A higher P-score indicates a more favorable efficacy ranking [34].

Assessment of heterogeneity, inconsistency, and publication bias

Between-study heterogeneity was quantified using the I² statistic. We evaluated network inconsistency, the disagreement between direct and indirect evidence, using two methods: a global assessment with the design-by-treatment interaction model and a local assessment for specific comparisons using the netsplit approach [35]. Potential publication bias and small-study effects were assessed by visual inspection of comparison-adjusted funnel plots and formally with Egger’s regression test, where appropriate. A p-value < 0.05 was considered statistically significant for all tests.

Of the 17 included trials, 16 employed crossover designs; only Zecca 2017 used a parallel-group design. As individual crossover trials reported only arm-level summary statistics (mean, standard deviation, and sample size per treatment arm) rather than within-patient paired differences or correlation coefficients, we calculated standardized mean differences using the parallel-group formula. This approach is recognized as conservative: it overestimates the standard error by not accounting for the within-subject correlation inherent in crossover designs, resulting in wider confidence intervals and potential underestimation of treatment effect precision. Consequently, statistically significant findings from this analysis can be interpreted with confidence, while non-significant results may reflect insufficient precision rather than true absence of effect [36–38]. Additionally, as included studies used heterogeneous assessment schedules (e.g., 10, 15, 20, 30, 40, 45, 60 min), data were mapped to standardized time points (PID10, PID15, PID30, PID60) using the nearest available measurement when the exact target time point was not reported by a given study (e.g., 20-minute data mapped to PID15; 40-minute data mapped to PID30). This approach maximized data utilization and network connectivity but may have introduced additional heterogeneity at time points where multiple mapping distances were combined.

Results

Study selection

The literature search identified 279 records, from which 129 duplicates were removed. After screening the titles and abstracts of the remaining 150 articles, 69 were selected for full-text review. Of these, 17 randomized controlled trials met the eligibility criteria and were included in the qualitative and quantitative synthesis. The detailed PRISMA flow diagram is presented in Fig. 1.

Fig. 1.

PRISMA flowchart of the study selection process. The diagram illustrates the flow of information through the different phases of the systematic review, mapping out the number of records identified, included and excluded, and the reasons for exclusions

Study characteristics

The 17 included studies enrolled a total of 1,463 patients with cancer-related BTP. Key characteristics of these trials are summarized in Table 1. The studies were conducted internationally, primarily in the United States (8 studies), various European countries (7 studies), and Japan (2 studies). Of the 17 trials, 16 utilized a randomized, double-blind, crossover design and one (Zecca 2017) used a parallel-group design, with the majority comparing an active intervention against a placebo. The patient populations were diverse, with lung, breast, and colorectal cancers being the most common diagnoses. The evaluated interventions included six distinct rapid-onset fentanyl formulations (OTFC, FBT, FPNS, INFS, FBSF, and FST) and two forms of morphine (MSIR and subcutaneous), which were compared against each other or a placebo. Detailed information for each included study (population, intervention dosing, outcome measure, and risk-of-bias judgement) is summarised in Additional file 1: Table S4.

Table 1.

Characteristics of included studies

Study	Country	Design	N	Age (years)	Intervention	Dose	Primary outcome	Timepoints (min)
Farrar 1998	USA	DB, PC, crossover	89	54 ± 12	OTFC vs. placebo	200–1600 µg	SPID, TOTPAR	15, 30, 45, 60
Coluzzi 2001	USA	DB, DD, crossover	89	55 ± 11	OTFC vs. MSIR	200–1600 µg vs. 15–60 mg	PID15	15, 30, 45, 60
Portenoy 2006	USA	DB, PC, crossover	77	58 ± 13	FBT vs. placebo	100–800 µg	SPID30	15, 30, 45, 60
Slatkin 2007	USA	DB, PC, crossover	87	NR	FBT vs. placebo	Titrated	SPID60	10, 15, 30, 60
Rauck 2009	USA	DB, PC, crossover	61	NR	FST vs. placebo	100–800 µg	SPID30	10, 15, 30, 60
Kress 2009	Europe	DB, PC, crossover	111	61 ± 10	INFS vs. placebo	50–200 µg	PID10	10, 20, 40, 60
Mercadante 2009	Europe	OL, crossover	139	NR	INFS vs. OTFC	50–200 µg vs. 200–1600 µg	TMPR	5, 10, 15, 20, 30, 60
Rauck 2010	USA	DB, PC, crossover	80	57 ± 12	FBSF vs. placebo	200–1200 µg	SPID30	5, 10, 15, 30, 45, 60
Portenoy 2010	Americas	DB, PC, crossover	73	54 ± 11	FPNS vs. placebo	100–800 µg	SPID30	5, 10, 15, 30, 45, 60
Rauck 2012	USA	DB, PC, crossover	96	56 ± 12	FSS* vs. placebo	100–1600 µg	SPID30	5, 10, 15, 30, 45, 60
Kosugi 2014	Japan	DB, PC, crossover	73	61	FBT vs. placebo	50–800 µg	PID30	15, 30, 60
Novotna 2014	Czech	DB, PC, crossover	73	65 ± 12	FST vs. placebo	133–800 µg	SPID30	3, 6, 10, 15, 30, 60
Mercadante 2015	Italy	OL, crossover	72	62 ± 12	FBT vs. MSIR	100–800 µg vs. 10–80 mg	PID	15, 30
Shimoyama 2015	Japan	DB, PC, AC, crossover	47	59 ± 11	FST vs. placebo vs. MSIR	100–800 µg	PID30	30, 60
Thronæs 2015	Europe	DB, PC, crossover	15	61 ± 10	INFS vs. placebo	400 µg	PID10	5, 10, 30, 60
Mercadante 2016	Italy	OL, crossover	45	63 ± 13	FPNS vs. MSIR	100–800 µg vs. 10–80 mg	PID	15, 30
Zecca 2017	Italy	DB, DD, parallel	113	58–62	FST vs. SCM	100 µg vs. 5 mg	AVP30	10, 20, 30, 60

Note: Age presented as mean ± SD where available. All studies enrolled opioid-tolerant adult patients with cancer with breakthrough pain

Abbreviations: AC active-controlled, AVP average pain, DB double-blind, DD double-dummy, FBSF fentanyl buccal soluble film, FBT fentanyl buccal tablet, FPNS fentanyl pectin nasal spray, FSS fentanyl sublingual spray, FST fentanyl sublingual tablet; INFS intranasal fentanyl spray, MSIR morphine sulfate immediate release, N number of patients in efficacy analysis, NR not reported, OL open-label; OTFC oral transmucosal fentanyl citrate, PC placebo-controlled, PID pain intensity difference, SCM subcutaneous morphine, SPID sum of pain intensity differences, TMPR time to meaningful pain relief, TOTPAR total pain relief

*Fentanyl sublingual spray (FSS; Rauck 2012) was classified under the fentanyl sublingual tablet (FST) node in the network meta-analysis, as both formulations share the same sublingual delivery route and active ingredient

Risk of bias and network geometry

The overall risk of bias across the 17 studies was judged to be low to moderate. While most trials demonstrated adequate randomization processes, some studies raised concerns related to missing outcome data and the potential for selective reporting (Additional file 1: Table S3).

The network geometry of the available evidence for each time point is illustrated in Fig. 2. The complexity of the network varied by the outcome timing. For PID10, the network comprised 8 treatments from 10 studies. The network expanded at PID15 to include 9 treatments from 15 studies, and was most extensive at PID30, which included 9 treatments from 16 studies. At PID60, 14 studies contributed data for 9 treatments. In all networks, placebo was the most common comparator, serving as a central hub connecting most active treatments.

Fig. 2.

Network geometry of treatment comparisons for pain intensity difference (PID) at different time points. Each node represents an intervention, with node size proportional to the total number of patients randomized to that treatment across all contributing studies. Edge thickness reflects the number of studies providing direct evidence for that comparison, with the number displayed on each edge. Placebo serves as the central comparator hub connecting most active treatments. A PID at 10 min (PID10). B PID at 15 min (PID15). C PID at 30 min (PID30). D PID at 60 min (PID60). Abbreviations: INFS, intranasal fentanyl spray; FBT, fentanyl buccal tablet; FPNS, fentanyl pectin nasal spray; FST, fentanyl sublingual tablets; FBSF, fentanyl buccal soluble film; OTFC, oral transmucosal fentanyl citrate; MSIR, morphine sulfate immediate release; SCM, subcutaneous morphine

Network meta-analysis of efficacy

Early analgesic effect: 10 and 15 min

In the initial 15 min post-administration, a clear hierarchy of efficacy emerged among the treatments (Fig. 3A, B). At the 10-minute mark (10 studies, 8 treatments), intranasal fentanyl spray (INFS) demonstrated the most substantial analgesic effect (SMD: 0.85; 95% CI: 0.59–1.10), ranking highest with a P-score of 0.97. Subcutaneous morphine (SCM; SMD: 0.62; 95% CI: 0.20–1.03), fentanyl buccal tablet (FBT; SMD: 0.47; 95% CI: 0.17–0.77), and fentanyl sublingual tablets (FST; SMD: 0.30; 95% CI: 0.11–0.48) showed statistically significant efficacy compared to placebo. Other fentanyl formulations showed trends towards improvement but did not achieve statistical significance.

Fig. 3.

Forest plots of the network meta-analysis of treatment efficacy compared to placebo. The plots show the standardized mean difference (SMD) and 95% confidence intervals (CI) for each intervention at four time points. An SMD greater than zero favors the active intervention over placebo. The P-score for each treatment is also provided as a ranking metric. A PID at 10 min (PID10). B PID at 15 min (PID15). C PID at 30 min (PID30). D PID at 60 min (PID60)

At 15 min, the network expanded substantially to include 15 studies evaluating all 9 treatments. INFS demonstrated the largest treatment effect observed at any time point (SMD: 1.05; 95% CI: 0.67–1.42; P-score: 0.98). SCM was included in the PID15 network through data from Zecca 2017, which reported outcomes at the nearest available time point (20 min, mapped to PID15). SCM ranked second among all treatments at this time point (P-score = 0.80). The estimate reached statistical significance (SMD 0.70, 95% CI 0.10 to 1.30, p = 0.022), though the wide confidence interval reflects limited precision from a single contributing study. FBT showed significant efficacy (SMD: 0.48; 95% CI: 0.22–0.74; P-score: 0.67), as did FST (SMD: 0.37; 95% CI: 0.08–0.66; p = 0.012). FPNS showed a trend toward significance (SMD: 0.34; 95% CI: −0.06 to 0.75; p = 0.097). OTFC did not reach significance at this time point (SMD: 0.26; 95% CI: −0.08 to 0.59; p = 0.14). MSIR did not show a meaningful effect (SMD: −0.12; 95% CI: −0.48 to 0.24).

Peak and sustained analgesic effect: 30 and 60 min

The analgesic effects of most formulations continued to evolve at the 30- and 60-minute assessments, representing the period of peak and sustained relief for BTP (Fig. 3C, D). At 30 min (16 studies, 9 treatments), INFS maintained its top-ranked position with a large effect size (SMD: 1.04; 95% CI: 0.82–1.26; P-score: 1.00). Several other fentanyl formulations demonstrated statistically significant efficacy: FBT (SMD: 0.62; 95% CI: 0.46–0.78), OTFC (SMD: 0.58; 95% CI: 0.40–0.75), FPNS (SMD: 0.57; 95% CI: 0.30–0.84), FBSF (SMD: 0.54; 95% CI: 0.23–0.86), SCM (SMD: 0.53; 95% CI: 0.11–0.95), and FST (SMD: 0.38; 95% CI: 0.19–0.57). Importantly, MSIR achieved statistical significance at this time point (SMD: 0.20; 95% CI: 0.01–0.39; p = 0.036), marking the earliest time point at which oral morphine demonstrated a statistically significant effect.

By the 60-minute mark, all active treatments were statistically superior to placebo (14 studies, 9 treatments). INFS (SMD: 0.93; 95% CI: 0.66–1.20) and FBT (SMD: 0.83; 95% CI: 0.61–1.04) continued to show robust and sustained efficacy, remaining the top two ranked interventions. FPNS (SMD: 0.75; 95% CI: 0.37–1.13), OTFC (SMD: 0.69; 95% CI: 0.43–0.94), FBSF (SMD: 0.63; 95% CI: 0.27–1.00), subcutaneous morphine (SMD: 0.60; 95% CI: 0.15–1.06), and FST (SMD: 0.42; 95% CI: 0.24–0.61) all demonstrated statistically significant efficacy. MSIR showed a small-to-medium effect (SMD: 0.34; 95% CI: 0.04–0.64), confirming its delayed but ultimately effective analgesic profile.

Summary of treatment rankings and model diagnostics

Across all evaluated time points, INFS was consistently the highest-ranked intervention, with P-scores ranging from 0.91 to 1.00. FBT ranked among the top three formulations at all time points (P-scores: 0.67–0.79). The relative rankings of other fentanyl formulations varied across time points, reflecting their different pharmacokinetic profiles, such as the delayed but sustained efficacy of OTFC.

No significant inconsistency was detected at PID10, PID30, or PID60 (design-by-treatment interaction p = 0.43, 0.75, and 0.25, respectively; Additional file 1: Figs. S2, S3). At PID15, significant design-by-treatment interaction was observed (p = 0.018), likely attributable to the use of nearest-available timepoint data for studies lacking an exact 15-minute assessment; this inconsistency was no longer statistically significant (p = 0.060) in the sensitivity analysis excluding high-risk-of-bias studies. The comparison-adjusted funnel plots did not suggest significant publication bias (Additional file 1: Fig. S4).

A sensitivity analysis excluding four studies at high risk of bias (Rauck 2009, Mercadante 2009, Mercadante 2015, and Mercadante 2016) confirmed the robustness of the primary findings (Additional file 1: Fig. S5). The treatment ranking hierarchy was preserved at all four time points, with INFS maintaining the highest P-score at each assessment (PID10: 0.94; PID15: 0.95; PID30: 0.99; PID60: 0.94). All nine treatments remained in the sensitivity networks, and the direction and magnitude of treatment effects were consistent with the primary analysis.

The temporal evolution of treatment efficacy across all four time points is illustrated in Fig. 4. INFS demonstrated consistently superior performance, with peak SMD at PID15 (1.05, 95% CI 0.67–1.42), followed by a plateau at later time points. FBT showed a steady accumulation of analgesic effect. OTFC gained statistical significance from PID30 onward, while MSIR did not reach significance until PID30. Most fentanyl formulations converged toward similar efficacy levels by 60 min, while their differentiation was most pronounced in the critical first 15–30 min. A complementary heatmap visualisation of SMD versus placebo across all evaluated time points is provided in Additional file 1: Fig. S6.

Fig. 4.

Temporal profile of treatment efficacy for breakthrough cancer pain. The figure plots time after administration (10, 15, 30, 60 min) on the X-axis against SMD vs. placebo (with 95% CIs) on the Y-axis, with each treatment represented as a distinct colored line with confidence bands. INFS, intranasal fentanyl spray; FBT, fentanyl buccal tablet; FPNS, fentanyl pectin nasal spray; FST, fentanyl sublingual tablets; FBSF, fentanyl buccal soluble film; OTFC, oral transmucosal fentanyl citrate; MSIR, morphine sulfate immediate release; SCM, subcutaneous morphine

Discussion

In this comprehensive network meta-analysis comparing all available rapid-onset opioid formulations for BTP, our analysis of 17 RCTs including 1,463 patients provides evidence for a relative efficacy hierarchy based on temporal analgesic profiles. Our primary finding is that INFS provides the most rapid and potent analgesia, demonstrating superiority within 10 min and maintaining this effect through 60 min. FBT consistently ranked as a strong second-line alternative. Both formulations provide clinically meaningful pain relief during the critical initial phase of a BTP episode. Conversely, our findings confirm the significant limitations of immediate-release oral morphine, which did not achieve statistical significance until 30 min post-administration (PID30: SMD 0.20, p = 0.036), highlighting its suboptimal profile for this indication.

This analysis addresses a critical evidence gap in BTP management. Previously, the lack of direct, multi-treatment comparisons forced clinicians to select agents based on limited head-to-head data, institutional formularies, or anecdotal experience. By synthesizing both direct and indirect evidence, our findings provide additional evidence to help guide treatment selection. The consistent superiority of INFS across all time points, with FBT as a strong second-ranked formulation, may inform clinical decision-making when rapid onset is prioritized.

The observed efficacy hierarchy between INFS and FBT is well-supported by their distinct pharmacokinetic profiles. INFS achieves rapid systemic absorption via the highly vascular nasal mucosa, reaching peak plasma concentrations within 10–15 min, which aligns perfectly with the rapid escalation of a typical BTP episode [39–41]. In contrast, FBT utilizes a dual absorption mechanism (buccal and gastrointestinal), resulting in a slightly delayed but more sustained effect, with peak concentrations occurring around 45–50 min [22, 42, 43]. This profile may be advantageous for patients experiencing BTP episodes of longer duration.

These pharmacological differences directly inform patient-centered treatment selection. For instance, INFS is a particularly compelling option for the large proportion of patients with advanced cancer who suffer from oral complications, such as xerostomia or mucositis, which can impair buccal absorption [41, 44–46]. Conversely, FBT is a more suitable choice for patients with compromised nasal function due to congestion, rhinitis, or concurrent use of nasal decongestants [44, 45, 47]. To aid clinical interpretation, Cohen’s benchmarks classify SMD values of 0.2, 0.5, and 0.8 as small, medium, and large effects, respectively [48]. By this standard, INFS (SMD 0.93) and FBT (SMD 0.83) demonstrated large effects at 60 min, while MSIR showed a small-to-medium effect (SMD 0.34). These effect sizes are broadly consistent with clinically meaningful pain reduction, as the IMMPACT consensus recommendations define a moderately important improvement as approximately 30% or 2 points on a 0–10 NRS [49, 50]. The nature of the BTP episode itself is another key determinant; the ultra-rapid onset of INFS is ideal for pain that peaks within minutes, whereas the sustained profile of FBT may be preferable for more gradually developing or prolonged episodes.

Based on the efficacy data from this analysis, INFS may be considered for patients requiring the fastest possible onset of relief, particularly those with oral complications that limit buccal absorption. FBT represents an effective alternative for patients with nasal contraindications or preference for oral administration. However, these efficacy-based considerations should be weighed alongside the safety profile of each formulation and individual patient risk factors.

While our analysis focused on short-term analgesic efficacy, clinical decision-making must also integrate the safety profile of these formulations. All rapid-onset fentanyl products carry FDA black box warnings for life-threatening respiratory depression, accidental exposure, CYP3A4 drug interactions, risks from concomitant benzodiazepine use, medication errors between non-interchangeable formulations, and addiction/abuse/misuse potential. The TIRF REMS Access program requires enrollment of prescribers, pharmacies, and patients to mitigate these risks [19]. Nasal formulations, while demonstrating the fastest onset in our analysis, achieve peak plasma concentrations comparable to intravenous administration (within 12–15 min) [39], which creates a narrow therapeutic window for detecting and managing respiratory depression, particularly in home or unsupervised settings [42]. These safety considerations are particularly relevant for the growing number of patients managing BTP in outpatient and home-care settings, where immediate medical intervention for adverse events may not be available.

It is important to note that increased precision from indirect comparisons does not guarantee reduced bias. If the transitivity assumption is violated, indirect estimates may amplify rather than correct for bias. The direct evidence proportion for each comparison is presented in Additional file 1: Fig. S1, and comparisons driven predominantly by indirect evidence should therefore be interpreted with caution.

The variability in assessment time points across included studies necessitated a standardized approach to timepoint mapping. When a study did not report an outcome at the exact target time point, we used the nearest available measurement (e.g., 20-minute data mapped to PID15 for studies assessing at 10, 20, 40, and 60 min). This approach maximized data utilization and network connectivity but introduced a source of heterogeneity, as measurements from slightly different time points were pooled. This may partly explain the significant design-by-treatment interaction observed at PID15 (p = 0.018). Future trials adopting standardized assessment schedules (e.g., 5, 10, 15, 30, and 60 min) would facilitate more precise network analyses.

Our findings are consistent with, and substantially extend, the previous network meta-analysis by Zeppetella et al. (2014), which also identified the superiority of INFS at 15 min [26]. Our study provides a more granular temporal analysis across four key time points (10, 15, 30, and 60 min), offering a comprehensive view of the onset and duration of action for all available agents. Furthermore, our results align with the broader consensus from numerous clinical trials demonstrating that rapid-onset fentanyl formulations are significantly more effective than MSIR for BTP, a distinction that is critical for managing this acute pain condition [51].

Furthermore, our results provide the quantitative evidence needed to potentially refine current clinical guidelines, such as those from the European Society for Medical Oncology (ESMO). The 2018 ESMO guidelines recommend selecting a rapid-onset opioid based on individual pain patterns but treat the various formulations as largely interchangeable [52]. Our analysis challenges this perspective by establishing a clear and statistically robust efficacy hierarchy. The demonstration of substantial differences in effectiveness—particularly the consistent superiority of INFS and FBT and the delayed onset of MSIR—suggests that an evidence-based ranking of agents, in addition to patient preference, should be a key component of treatment selection.

An important consideration when interpreting these rankings is dose comparability. Most included trials employed individually titrated dosing protocols (e.g., INFS 50–200 µg, FBT 100–800 µg, OTFC 200–1600 µg), whereby each patient received a dose optimized for their individual analgesic response during an open-label titration phase. While this pragmatic approach mirrors real-world clinical practice and maximizes the generalizability of findings, it means that the NMA effectively compares each formulation at its individually optimized dose rather than at pharmacologically equivalent doses. Consequently, the observed efficacy differences may reflect a combination of intrinsic pharmacological properties and the effectiveness of each formulation’s dose-titration protocol.

Strengths and limitations

This study has several notable strengths. To our knowledge, it is the most comprehensive NMA to date on this topic, synthesizing data from 17 trials and over 1,400 patients to provide sufficient statistical power for detecting clinically meaningful differences. A key strength is our granular analysis at four distinct time points, which captures the dynamic temporal profile of analgesic response and offers practical insights into the onset and duration of action for different agents. Finally, the NMA methodology itself is a major advantage, allowing us to generate a unified evidence hierarchy by integrating both direct and indirect comparisons, thereby maximizing the use of the available evidence.

Nevertheless, certain limitations must be acknowledged. First, the network’s reliance on indirect evidence is substantial, as few head-to-head trials exist among the various active formulations. This may reduce the precision of some relative effect estimates. Second, our analysis is restricted to the first 60 min post-administration and cannot inform decisions regarding longer-term efficacy or safety. Third, the predominance of crossover trial designs (16 of 17 studies) necessitated the use of parallel-group formulae for SMD calculation, as paired statistics were not reported in the primary studies. This conservative approach may have resulted in wider confidence intervals and reduced statistical power compared to a proper paired analysis [36–38]. Fourth, and importantly, this analysis exclusively evaluated short-term analgesic efficacy and did not assess adverse events, tolerability, or safety outcomes. This is a significant limitation, as the safety profile is a critical determinant of clinical decision-making. Rapid-onset fentanyl formulations carry FDA black box warnings for life-threatening respiratory depression, and the risk is particularly relevant for nasal formulations used in unsupervised home settings [19]. Future comprehensive network meta-analyses that incorporate both efficacy and safety endpoints are needed. Fifth, as included studies used heterogeneous assessment schedules, data were mapped to standardized time points using the nearest available measurement when exact target time points were not reported. This may have introduced additional heterogeneity, particularly at PID15 where significant design-by-treatment interaction was detected (p = 0.018); notably, this inconsistency was no longer statistically significant (p = 0.060) in the sensitivity analysis excluding high-risk-of-bias studies, suggesting it was driven by the open-label studies rather than structural network bias. Sixth, our analysis was restricted to fentanyl formulations and morphine preparations; other opioids used in clinical practice for BTP (e.g., oxycodone, hydromorphone) were excluded due to insufficient RCT data, which may limit the completeness of the treatment hierarchy. Finally, although several studies reported pain assessments as early as 5 min post-administration, the available data at this time point were insufficient to construct a connected treatment network. Despite these limitations, the consistent treatment hierarchy across multiple time points, analytical approaches, and sensitivity analyses supports the overall robustness of these findings.

Conclusions

This network meta-analysis provides evidence for a relative efficacy hierarchy among rapid-onset opioids for BTP based on short-term analgesic outcomes. Intranasal fentanyl spray demonstrated the most rapid and sustained pain reduction across all evaluated time points, while fentanyl buccal tablet consistently ranked as the second most effective formulation. Immediate-release morphine showed a comparatively delayed onset, reaching statistically significant analgesia only from 30 min onward. However, treatment selection must integrate these efficacy data with considerations of adverse effect profiles, risk of respiratory depression, potential for misuse, route of administration feasibility, patient-specific factors (e.g., oral mucositis, nasal pathology), and the clinical context of administration (supervised vs. home setting). Our findings represent one component of the evidence base needed for individualized BTP management.

Future research should prioritize direct head-to-head trials between top-performing agents, investigate long-term safety profiles with particular attention to respiratory depression risk in home settings, and identify predictors of treatment response to further refine personalized BTP management.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AC

active-controlled

ATC

around-the-clock

AVP

average pain

BTP

breakthrough cancer pain

CI

confidence interval

DB

double-blind

DD

double-dummy

FBSF

fentanyl buccal soluble film

FBT

fentanyl buccal tablet

FPNS

fentanyl pectin nasal spray

FSS

fentanyl sublingual spray

FST

fentanyl sublingual tablet

INFS

intranasal fentanyl spray

MSIR

morphine sulfate immediate release

NMA

network meta-analysis

NRS

numeric rating scale

OTFC

oral transmucosal fentanyl citrate

PICOS

Population, Intervention, Comparator, Outcome, Study Design

PID

pain intensity difference

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

RCT

randomized controlled trial

REMS

Risk Evaluation and Mitigation Strategy

ROO

rapid-onset opioid

SCM

subcutaneous morphine

SD

standard deviation

SMD

standardized mean difference

TIRF

Transmucosal Immediate-Release Fentanyl

VAS

visual analog scale

Author contributions

PHT: Methodology, Formal analysis, Investigation, Data Curation, Writing—Original Draft, Visualization. CHL: Methodology, Validation, Investigation, Data Curation, Writing—Review & Editing. JHC: Investigation, Data Curation, Clinical interpretation of included trials, Writing—Review & Editing. CLH: Writing—Review & Editing. HJJ: Conceptualization, Supervision, Writing—Review & Editing. PHC: Conceptualization, Methodology, Supervision, Project administration, Funding acquisition, Writing—Review & Editing. All authors read and approved the final manuscript.

Authors’ social media handles

Po-Huang Chen – X: @chenpohuang.

Funding

This study was supported by research grants from the Tri-Service General Hospital (grant numbers: TSGH-D-114168, TSGH-D-115176 to PHC and TSGH-D-114074 to CHL). The funding bodies had no role in the design of the study, collection, analysis, and interpretation of data, or in writing the manuscript.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The data are derived from previously published studies which are cited in the reference list.

Declarations

Ethics approval and consent to participate

This systematic review and meta-analysis was conducted using previously published data from randomized controlled trials. As such, it did not require direct patient involvement. The study protocol was granted an exemption from formal review by the Institutional Review Board of the Tri-Service General Hospital, National Defense Medical University, on July 24, 2025 (TSGHIRB No. E202516045).

Consent for publication

Not applicable.

Use of Artificial Intelligence

AI-assisted tools were used for language editing during the manuscript revision process. All outputs were reviewed and approved by the authors.

Competing interests

The authors declare no competing interests.