Matching-Adjusted Indirect Comparison of Oral Paclitaxel (Liporaxel^®) Versus Weekly Intravenous Paclitaxel in Advanced Gastric Cancer

Abstract

Purpose

Intravenous (IV) paclitaxel is widely used in the treatment of advanced gastric cancer. Liporaxel^® is the first orally administered paclitaxel formulation and demonstrated non-inferiority to paclitaxel administered every 3 weeks in the phase III DREAM trial. As no head-to-head trial has compared Liporaxel^® with the widely used weekly paclitaxel regimen in gastric cancer, this study aimed to compare their relative efficacy.

Materials and Methods

This study utilized individual patient data from the DREAM trial and aggregate data for the weekly paclitaxel regimen from a systematic literature review. Given the absence of a common comparator, an unanchored matching-adjusted indirect comparison (MAIC) was performed for progression-free survival (PFS) and overall survival (OS) with weighting based on common and extended variables. Primary and sensitivity analyses were performed according to the effective sample size (ESS).

Results

Matching was performed across 5 studies. Despite a reduction in the ESS, baseline patient characteristics remained similarly distributed after matching. In the two studies included in the primary analysis, the adjusted hazard ratios from the unanchored MAIC, based on common variables, were 0.95 (95% confidence interval [CI], 0.72–1.26) and 1.03 (95% CI, 0.81–1.29) for PFS, and 0.94 (95% CI, 0.69–1.28) and 1.06 (95% CI, 0.84–1.34) for OS. The results obtained using the extended variables and sensitivity analyses were consistent with the primary analysis.

Conclusions

The efficacy of Liporaxel^® and the weekly IV paclitaxel regimen were comparable. This finding supports the previously established non-inferiority evidence and suggests Liporaxel^® as a valid treatment option for advanced gastric cancer.

INTRODUCTION

Gastric cancer remains a major global health burden, ranking as the fifth most commonly diagnosed malignancy and leading cause of cancer-related mortality [1]. Despite substantial advances in first-line treatment [2], most patients eventually experience disease progression and require subsequent systemic treatments. Since pivotal phase 3 trials have established the role of later-line systemic chemotherapy in patients with disease progression after first-line systemic therapy [3,4,5], these treatments play a significant role in managing patients with gastric cancer. Currently, the combination of ramucirumab and paclitaxel has become the standard second-line treatment, based on the survival benefit demonstrated in the RAINBOW trial [6]. However, because ramucirumab is an antiangiogenic agent, concerns about treatment-related complications [6] may preclude its use in a substantial proportion of patients.

Intravenous (IV) paclitaxel, weekly or every 3 weeks, is widely used for its efficacy and safety, as shown in previous phase II and III clinical trials [7,8,9,10]. However, the Cremophor EL–based injectable formulation of paclitaxel is associated with frequent hypersensitivity reactions ranging in severity from mild pruritus to anaphylaxis [11]. DHP107 (Liporaxel^®) was developed to overcome challenges associated with low solubility and drug efflux in the gastrointestinal tract. It is an oral formulation of paclitaxel designed to facilitate direct and efficient intestinal absorption without requiring P-glycoprotein inhibitors [12]. Among patients with advanced gastric cancer in whom first-line chemotherapy failed, the phase III DREAM study demonstrated that DHP107 was non-inferior to IV paclitaxel (175 mg/m² on day 1 every 3 weeks) as the primary endpoint of progression-free survival (PFS). In that study, the median PFS in the per-protocol population was 3.0 months with DHP107 and 2.6 months with IV paclitaxel, which is consistent with non-inferiority. The median PFS calculated from the full analysis of individual patient data (IPD) used in the present matching-adjusted indirect comparison (MAIC) was 2.9 months with DHP107. In the full analysis set, overall survival (OS) was 9.7 and 8.9 months, respectively, suggesting comparable efficacy [13]. More recently, the multinational phase III OPTIMAL study extended these findings to human epidermal growth factor receptor 2-negative recurrent or metastatic breast cancer, showing that DHP107 was non-inferior to weekly IV paclitaxel, with comparable efficacy outcomes (median PFS, 10.0 vs. 8.5 months; median OS, 32.6 vs. 31.8 months) [14], further supporting its role as a convenient oral alternative to IV paclitaxel.

In routine practice, the weekly paclitaxel regimen is preferentially used for managing patients with gastric cancer because it achieves efficacy comparable to the regimen administered every 3 weeks while demonstrating a more favorable toxicity profile [15]. However, Liporaxel^® has only been directly compared with the regimen administered every 3 weeks in a phase III clinical trial, and head-to-head evidence with the weekly regimen is lacking. In this MAIC study, by applying weights to align baseline characteristics across study populations [16,17,18], we aimed to evaluate the efficacy of Liporaxel^® relative to that of the widely used weekly paclitaxel regimen in gastric cancer using IPD.

MATERIALS AND METHODS

Study design

To assess the feasibility of an anchored MAIC, we searched for clinical trials directly comparing weekly and every-3-week paclitaxel regimens; however, no such studies were identified. Therefore, to enable a more robust unanchored indirect comparison, we conducted an unanchored MAIC using IPD for Liporaxel^® and aggregate data from clinical trials of the weekly paclitaxel regimen (Fig. 1). As this study involved a secondary analysis of de-identified individual patient data from the previously conducted DREAM trial and aggregate data from published studies, it was determined to be exempt from review by the Institutional Review Board of Asan Medical Center (IRB No. 2026-1346).

Fig. 1. Study design and analytical framework for the unanchored MAIC. An SLR identified weekly paclitaxel studies in advanced gastric cancer. Survival data were reconstructed as pseudo-IPD via the Guyot algorithm, and MAIC was performed by reweighting DREAM trial patients to match each comparator population for pairwise treatment effect estimation.

MAIC = matching-adjusted indirect comparison; SLR = systematic literature review; IPD = individual patient data; ESS = effective sample size.

Systematic literature review (SLR) for weekly paclitaxel

Aggregated data on the weekly paclitaxel regimen for advanced gastric cancer were identified through an SLR. A PubMed search was conducted on April 4, 2026, in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines [19]. The search terms were selected using the PICO framework (stands for: Patient or problem; Intervention or exposure; Comparison or control; Outcomes), and the inclusion and exclusion criteria are listed in Supplementary Table 1. A total of 118 studies were identified through the search; after title and abstract screening, 93 studies were excluded. The remaining 25 studies underwent full-text review, and ultimately, seven studies evaluating the weekly paclitaxel regimen were included (Supplementary Table 1).

Overview of the MAIC and statistical analysis

The primary outcomes were PFS and OS. Accordingly, for studies reporting outcomes of the weekly paclitaxel regimen, Kaplan–Meier curves for PFS and OS were digitized using WebPlotDigitizer to extract survival coordinates from published figures. The extracted coordinates, along with the published numbers at risk, were then used to reconstruct pseudo-IPD according to Guyot et al. [20]. Two studies were excluded because the numbers of patients at risk were not reported, precluding the application of this method (Supplementary Fig. 1).

Using aggregate data from 5 studies of the weekly paclitaxel regimen with available PFS and OS data, pairwise unanchored MAICs were conducted against Liporaxel^® IPD. The Liporaxel^® IPD were reweighted to match the baseline characteristics of each aggregate dataset, and PFS and OS were then compared. The matching variables used for weight estimation included common and extended variables.

Common variables were selected from known prognostic factors and potential effect modifiers in gastric cancer, including age, sex, Eastern Cooperative Oncology Group performance status (ECOG PS), prior gastrectomy, presence of peritoneal metastasis, metastatic disease burden, and prior treatment [21,22,23,24,25]. Among these, variables commonly available across the DREAM study and 5 weekly paclitaxel studies were included. Extended variables were included to address the higher risk of confounding inherent to unanchored MAIC and to improve the alignment of baseline characteristics across trial populations [16,17]. Among the variables evaluated, those that could not be applied consistently across all studies but could be incorporated into individual pairwise comparisons were selectively included as extended variables to reduce residual confounding [18]. Common variables were applied across all pairwise comparisons, whereas extended variables were incorporated when available for each comparison.

To assess covariate balance before and after weighting, standardized mean differences (SMDs) were calculated for all matching variables separately for each pairwise comparison using unweighted and weighted DHP107 IPD.

For each matching scenario, the effective sample size (ESS) following MAIC weighting was calculated. Comparisons with an ESS ≥50% of the original sample size were classified as the primary analysis, reflecting relatively less loss of precision due to weighting, whereas comparisons with an ESS <50% were treated as sensitivity analyses. This cutoff was a prespecified, author-defined threshold adopted for pragmatic purposes and did not, by itself, establish the absence of heterogeneity or the adequacy of matching.

To visualize and descriptively compare PFS and OS distributions across individual studies, we generated unadjusted Kaplan–Meier plots using the DREAM IPD before weighting and reconstructed pseudo-IPD from the five weekly paclitaxel studies using the Guyot algorithm. Using an unanchored MAIC approach, hazard ratios (HRs) and 95% confidence intervals (CIs) for PFS and OS were estimated for each comparison between the DREAM study and the five individual studies using Cox proportional hazards models. The HRs based on common and extended variables are presented. Formal comparative analyses of safety outcomes were not performed because the available safety data were not sufficiently comparable across studies, owing to differences in adverse event definitions, grading methods, reporting conventions, and follow-up. These differences limit the validity of formal cross-trial safety comparisons. Accordingly, safety outcomes are presented descriptively without statistical adjustment or formal between-study comparisons. Statistical significance was set at a 2-sided P-value <0.05. All statistical analyses were conducted using R statistical software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria) and Python (version 3.12.3; Python Software Foundation, Beaverton, OR, USA).

RESULTS

Study populations

Seven studies on the weekly paclitaxel regimen were identified [6,7,26,27,28,29,30], of which 5 studies allowing reconstruction of IPD from Kaplan–Meier curves were ultimately included in the unanchored MAIC analysis (Table 1). Across the 5 included studies [6,27,28,29,30], paclitaxel-based regimens were used as comparators, with sample sizes ranging from 45 to 357 patients.

Table 1. Key characteristics of clinical trials included in the MAIC.

Characteristic	DREAM [13]	WJOG4007 [30]	Shitara et al., 2014 [27]	RAINBOW [6]	BRIGHTER [28]	FRUTIGA [29]
Trial design	Phase III, RCT	Phase III, RCT	Phase II	Phase III, RCT	Phase III, RCT	Phase III, RCT
wPTX regimen	Not applicable (DHP107 IPD used for MAIC)	IV PTX 80 mg/m2 weekly (paclitaxel monotherapy comparator arm)	IV PTX 80 mg/m2 weekly (paclitaxel monotherapy comparator arm)	IV PTX 80 mg/m2 weekly (placebo + paclitaxel control arm)	IV PTX 80 mg/m2 weekly (placebo + paclitaxel control arm)	IV PTX 80 mg/m2 weekly (placebo + paclitaxel control arm)
Line of therapy	2nd line	2nd line	2nd line	2nd line	2nd line	2nd line
Number of patients	118	108	45	335	357	352
Primary endpoint	PFS	OS	PFS	OS	OS	OS
Tumor type	Gastric	Gastric	Gastric	Gastric/GEJ	Gastric/GEJ	Gastric/GEJ
Male (%)	77.1	77.8	64.4	72.5	71.1	68.2
ECOG PS 1 (%)	89.0	78.7	71.1	75.5	79.8	85.2
Peritoneal metastasis (%)	29.7	26.9	42.2	45.4	47.9	29.3
Prior gastrectomy (%)	48.3	34.3	44.4	37.9	28.3	42.3
MAIC role	IPD trial	Primary	Sensitivity	Sensitivity	Sensitivity	Primary

MAIC = matching-adjusted indirect comparison; RCT = randomized controlled trial; wPTX = weekly paclitaxel monotherapy; IV = intravenous; PTX = paclitaxel; PFS = progression-free survival; OS = overall survival; GEJ = gastroesophageal junction; ECOG PS = Eastern Cooperative Oncology Group performance status; IPD = individual patient data.

While PFS was the primary endpoint in the DREAM study [13], OS was used as the primary endpoint in 4 of the comparator studies. Baseline characteristics, such as sex, ECOG PS, presence of peritoneal metastasis, and prior gastrectomy, showed similar or partially different distributions compared with those in the DREAM study [13]. Of the 5 studies, WJOG4007 [30] and FRUTIGA [29], which were more similar to the DREAM study [13], were included in the primary analysis, whereas the remaining three were included in the sensitivity analyses.

Matching and ESS

The common variables used for matching across all pairwise unanchored MAICs included sex, ECOG PS, prior gastrectomy, and peritoneal metastasis. Extended variables included age group (≥65 years), presence of liver metastasis, and prior platinum-based therapy, and were incorporated according to the individual study (Table 2).

Table 2. Adjustment variable sets and effective sample sizes by comparator trial in the MAIC (ESS vs. DREAM trial, n=118).

Trial		No. of patients	Common variables				Extended additional variables			Core ESS		Extended ESS
			Male	ECOG PS 1	Peritoneal metastasis	Prior gastrectomy	Age ≥65 yr	Liver metastasis	Prior platinum	n	%	n	%
Primary analysis
	WJOG4007 [30]	108	✓	✓	✓	✓	✓	—	—	104.7	88.8	88.0	74.5
	FRUTIGA [29]	352	✓	✓	✓	✓	✓	✓	—	109.7	93.0	104.0	88.1
Sensitivity analysis
	RAINBOW [6]	335	✓	✓	✓	✓	✓	—	—	50.3	42.6	50.2	42.5
	BRIGHTER* [28]	357	✓	✓	✓	✓	—	—	—	52.0	44.1	52.0	44.1
	Shitara et al., 2014 [27]	45	✓	✓	✓	✓	✓	✓	✓	32.5	27.5	22.6	19.2

Common variables were uniformly applied across all comparisons. Extended variables were added per trial based on the availability of aggregate data in published reports. Primary analysis was defined using a prespecified author-defined ESS threshold of ≥50%.

MAIC = matching-adjusted indirect comparison; ESS = effective sample size; ECOG PS = Eastern Cooperative Oncology Group performance status; = included in the adjustment set; — = not available/not applied.

^*For BRIGHTER [28], no additional variables beyond the common set were available; therefore, the extended set was equivalent to the common set for this comparison.

As the heterogeneity between the study populations increased, ESS decreased. The WJOG4007 [30] and FRUTIGA [29] studies retained ESS values of 88.8% and 93.0%, respectively, whereas the remaining three studies showed lower ESS values ranging from 27.5% to 44.1%, reflecting a greater loss of precision after weighting (Table 2).

The covariate balance before and after weighting is presented in Table 3 for the primary analyses and in Supplementary Table 2 for the sensitivity analyses. Before weighting, several baseline characteristics differed between the DHP107 population and the comparator studies. After weighting, the SMDs for all matching variables were reduced to zero, indicating an exact balance for the covariate moments included in the MAIC weighting procedure.

Table 3. Covariate balance before and after MAIC weighting in the primary analyses.

Trial	Variable	Before (%)	Target (%)	SMD before	SMD after
WJOG4007 [30]	Male	77.1	77.8	−0.02	0.00
	ECOG PS 1	89.0	96.2	−0.28	0.00
	Peritoneal metastasis	29.7	25.9	0.09	0.00
	Prior gastrectomy	48.3	34.3	0.29	0.00
	Age ≥65 yr*	39.0	59.3	−0.42	0.00
FRUTIGA [29]	Male	77.1	68.2	0.20	0.00
	ECOG PS 1	89.0	85.2	0.11	0.00
	Peritoneal metastasis	29.7	29.5	0.00	0.00
	Prior gastrectomy	48.3	42.0	0.13	0.00
	Age ≥65 yr*	39.0	29.8	0.20	0.00
	Liver metastasis*	39.0	44.6	−0.11	0.00

After weighting, SMDs of zero reflect the exact balance in the matching covariate moments by construction and do not necessarily indicate a balance in the overall covariate distributions.

MAIC = matching-adjusted indirect comparison; SMD = standardized mean difference; ECOG PS = Eastern Cooperative Oncology Group performance status.

^*Extended matching variable.

PFS

The unadjusted Kaplan–Meier curves for PFS are presented in Fig. 2A. The Kaplan–Meier curves for Liporaxel^® and the 5 comparator studies showed crossing patterns. When matched using common variables, HRs for PFS were 0.95 (95% CI, 0.72–1.26; P=0.724) in the primary analysis for WJOG4007 and 1.03 (95% CI, 0.81–1.29; P=0.828) for FRUTIGA, with no statistically significant differences observed. When extended variables were applied, the ESS decreased; however, the difference in PFS between Liporaxel^® and the weekly paclitaxel regimen remained not statistically significant. The results of sensitivity analyses were consistent with those of the primary analyses (Table 4).

Fig. 2. Unadjusted Kaplan–Meier curves for survivals across the DREAM trial and comparator studies. (A) PFS and (B) OS. Survival data for the comparator studies were reconstructed from published Kaplan–Meier curves using the Guyot algorithm. For the DREAM trial [13], median PFS and OS were calculated from the raw full analysis set used in the present MAIC (PFS, 2.9 months; OS, 9.7 months). The published DREAM median PFS was 3.0 months in the per-protocol population, whereas the published median OS (full analysis set) was identical (9.7 months). The originally reported median PFS and OS values for the comparator arms were 3.6 and 9.5 months, respectively, in WJOG4007 [30]; 2.5 and 9.6 months, respectively, in Shitara et al. [27]; 2.9 and 7.4 months, respectively, in RAINBOW [6]; 3.7 and 7.4 months, respectively, in BRIGHTER [28]; and 2.7 and 8.4 months, respectively, in FRUTIGA [29]. These values are provided to facilitate comparison with the medians estimated from the reconstructed Kaplan–Meier curves. These curves reflect unadjusted, study-level data and are presented for descriptive purposes only. Differences in baseline characteristics across studies were subsequently addressed using MAIC.

PFS = progression-free survival; OS = overall survival; MAIC = matching-adjusted indirect comparison; wPTX = weekly paclitaxel monotherapy; mPFS = median progression-free survival; mOS = median overall survival.

Table 4. MAIC-adjusted hazard ratios for PFS and OS: Liporaxel^® versus weekly paclitaxel.

Trials		MAIC variable set	ESS	PFS		OS
				HR (95% CI)	P-value	HR (95% CI)	P-value
Primary analysis
	WJOG4007 [30]	Common	88.8%	0.95 (0.72–1.26)	0.724	0.94 (0.69–1.28)	0.698
		Extended	74.5%	0.96 (0.71–1.29)	0.786	1.03 (0.76–1.39)	0.868
	FRUTIGA [29]	Common	93.0%	1.03 (0.81–1.29)	0.828	1.06 (0.84–1.34)	0.623
		Extended	88.1%	1.06 (0.84–1.34)	0.630	1.05 (0.83–1.34)	0.674
Sensitivity analysis
	Shitara et al., 2014 [27]	Common	27.5%	0.95 (0.63–1.44)	0.813	1.13 (0.69–1.86)	0.621
		Extended	19.2%	0.86 (0.55–1.36)	0.524	1.22 (0.72–2.08)	0.457
	RAINBOW [6]	Common	42.6%	1.17 (0.90–1.52)	0.251	0.95 (0.73–1.25)	0.723
		Extended	42.5%	1.16 (0.90–1.52)	0.258	0.95 (0.73–1.25)	0.733
	BRIGHTER [28]	Common	44.1%	1.13 (0.87–1.48)	0.350	0.95 (0.71–1.26)	0.702
		Extended	44.1%	1.13 (0.87–1.48)	0.350	0.95 (0.71–1.26)	0.702

ESS expressed as percentage of Liporaxel^® (n=118). HR <1 favors Liporaxel^®.

MAIC = matching-adjusted indirect comparison; PFS = progression-free survival; OS = overall survival; ESS = effective sample size; HR = hazard ratio; CI = confidence interval.

OS

Unadjusted Kaplan–Meier curves for OS also showed crossing patterns between Liporaxel^® and the 5 comparator studies, similar to those observed for PFS (Fig. 2B). Based on adjusted HRs estimated using MAIC, the primary analysis with common variables yielded HRs of 0.94 (95% CI, 0.69–1.28; P=0.698) for WJOG4007 and 1.06 (95% CI, 0.84–1.34; P=0.623) for FRUTIGA, with no statistically significant differences observed. Analyses using extended variables and sensitivity analyses showed consistent results, supporting the primary findings (Table 4).

Safety

A descriptive summary of selected grade ≥3 adverse events reported in the DREAM trial [13] and the weekly IV paclitaxel comparator studies [6,27,28,29,30] is presented in Table 5. Hematologic toxicities were the most frequently reported grade ≥3 adverse events across all studies. In the DREAM trial, grade ≥3 neutropenia occurred in 42.4% of patients treated with Liporaxel^® and 53.4% of those receiving IV paclitaxel. Across the weekly IV paclitaxel comparator studies, the reported incidence of grade ≥3 neutropenia ranged from 6.3% to 36.4%. Grade ≥3 leukopenia (including decreased white blood cell counts), anemia, and febrile neutropenia were also reported at generally similar incidences across studies.

Table 5. Descriptive summary of selected grade ≥3 adverse events.

Adverse event	Liporaxel® (DREAM)	IV paclitaxel control (DREAM, 175 mg/m2, q3w)	Range across weekly IV paclitaxel comparator studies
Grade ≥3 adverse events	68.6%	83.9%	59.7%–63.3%*
Neutropenia	42.4%	53.4%	6.3%–36.4%
Leukopenia (including decreased WBC)	18.6%	16.9%	4.3%–23.5%
Anemia	14.4%	15.3%	8.9%–21.3%
Febrile neutropenia	5.9%	2.5%	0.3%–4.4%
Peripheral (sensory) neuropathy	2.5%	8.5%	0.0%–7.4%
Vomiting	2.5%	1.7%	0.0%–4.0%
Nausea	1.7%	0.0%	0.0%–2.9%
Diarrhea	0.8%	1.7%	0.0%–1.5%

IV = intravenous; q3w = every 3 weeks; WBC = white blood cell.

^*Includes only studies that explicitly reported an overall grade ≥3 adverse event rate: RAINBOW (62.6%) [6], BRIGHTER (59.7%) [28], FRUTIGA (63.3%) [29]. WJOG4007 [30] and Shitara et al. [27] did not report an overall grade ≥3 adverse event rate and were therefore excluded from the reported range.

Among non-hematologic toxicities, grade ≥3 peripheral (sensory) neuropathy was reported in 2.5% of patients treated with Liporaxel^®, compared with 8.5% of patients receiving IV paclitaxel in the DREAM trial [13]; the 2.5% rate was within the range reported in the weekly IV paclitaxel comparator studies (0%–7.4%) [6,27,28,29,30]. Grade ≥3 gastrointestinal adverse events, including nausea, vomiting, and diarrhea, were infrequent across studies, although the reported incidence of individual events varied across studies. No treatment-related deaths were reported among patients treated with Liporaxel^® in the DREAM trial, whereas treatment-related deaths were reported in some weekly IV paclitaxel comparator studies.

DISCUSSION

In this unanchored MAIC study, we sought to address the lack of direct comparative evidence between Liporaxel^® and weekly IV paclitaxel in previously treated advanced gastric cancer. Although the phase III DREAM study demonstrated the non-inferiority of Liporaxel^® to every-3-week paclitaxel, weekly paclitaxel remains the more widely used regimen in clinical practice because of its favorable tolerability profile. By reweighting IPD from the DREAM study to match aggregate baseline characteristics from published weekly paclitaxel trials, we found that Liporaxel^® was associated with survival outcomes comparable to those of weekly paclitaxel. These results were consistent in the primary analyses, in which the ESSs were well preserved, and further supported by the sensitivity analyses. Moreover, these results are consistent with those of the OPTIMAL study, which demonstrated the non-inferiority of Liporaxel^® to weekly IV paclitaxel in a different tumor type (breast cancer), further supporting the validity of the current analysis [14]. Collectively, our findings provide indirect evidence supporting Liporaxel^® as a potential oral alternative to weekly IV paclitaxel in the second-line or later treatment of advanced gastric cancer.

Ramucirumab plus paclitaxel is the current standard second-line treatment. However, ramucirumab is associated with anti-angiogenic agent–related complications, such as bleeding, thromboembolic events, hypertension, proteinuria, and gastrointestinal perforation [6]. These complications can occasionally be life-threatening and may preclude the use of ramucirumab in some patients with gastric cancer. Therefore, IV paclitaxel alone remains an important treatment option in second- or later-line settings. In this context, our findings further support the role of Liporaxel^® as a second-line or later treatment option by demonstrating outcomes comparable to those of weekly paclitaxel, which is often preferred over every-3-week paclitaxel because of its more favorable safety profile. Moreover, weekly IV paclitaxel administration requires repeated hospital visits and prolonged infusion times, imposing substantial practical and clinical burdens. In addition, Cremophor EL–based injectable formulations require premedication to reduce the risk of hypersensitivity reactions, further increasing treatment complexity. By avoiding IV infusion and formulation-related premedication, Liporaxel^® may offer a more feasible and patient-convenient treatment option while maintaining efficacy comparable to weekly paclitaxel. Given that ramucirumab plus paclitaxel remains the standard second-line treatment, these findings also suggest that Liporaxel^® could be explored in combination with ramucirumab as a more convenient taxane backbone while maintaining the therapeutic rationale for combining ramucirumab with paclitaxel. Further prospective studies are warranted to evaluate the efficacy, safety, and feasibility of this combination.

Because no common comparator was identified through the SLR, an unanchored MAIC was conducted, and no statistically significant differences in PFS or OS were observed between Liporaxel^® and the weekly paclitaxel regimen. Consistent findings were observed across analyses using common variables selected from known prognostic factors and potential effect modifiers in gastric cancer that were available across the Liporaxel^® IPD from the DREAM and all 5 comparator studies. Similar results were also observed in analyses that incorporated extended variables where feasible. Furthermore, consistent results were observed across ESS-based primary and sensitivity analyses, demonstrating comparable PFS and OS outcomes between Liporaxel^® and the weekly paclitaxel regimen.

The descriptive safety findings provide an additional clinical context for the efficacy results. Liporaxel^® demonstrated a manageable safety profile with no reported hypersensitivity reactions in the DREAM trial. However, these observations should not be interpreted as evidence of comparative safety because no formally adjusted comparison was performed, and substantial heterogeneity existed in definitions of adverse events, grading conventions, follow-up duration, and concomitant treatments across studies. Accordingly, safety findings should be interpreted as descriptive rather than comparative.

In this study, a non-inferiority margin was not prespecified for unanchored MAIC. However, when applying the sequential margin of 1.48 used in the DREAM study to establish non-inferiority for the primary endpoint of PFS [13], the upper bounds of the 95% CIs from the primary PFS analyses using common variables in this study were all within this margin. Moreover, when applying a more stringent non-inferiority margin of 1.30, as used in other phase III clinical trials on gastric cancer [31], the results met this criterion. Furthermore, according to a systematic review of oncology clinical trials [32], the median non-inferiority margin was 1.29, and the results of this study satisfied this threshold. However, as these non-inferiority margins were applied post hoc rather than prespecified, caution is warranted when interpreting these findings.

This study had several limitations. First, the unanchored MAIC has inherent limitations in adjusting for confounding factors between the groups; therefore, the results should be interpreted with caution. Although variables beyond the common set were selectively incorporated as extended variables where feasible, the risk of bias remained owing to methodological constraints [16,17,18]. Second, PFS and OS data for the weekly paclitaxel regimen were reconstructed using the Guyot algorithm [20] and, therefore, may differ from the original IPD. Nevertheless, the reconstructed Kaplan–Meier curves showed patterns similar to those reported in the original clinical studies. Third, unlike the DREAM study, some studies enrolled patients with gastroesophageal junction (GEJ) cancer. Although most patients had gastric cancer, a subset of patients with GEJ cancer was included, and the potential impact of this could not be further adjusted. Nevertheless, gastric and GEJ cancers have generally been grouped together in most clinical trials, and the clinical outcomes have not differed substantially between these populations. Therefore, this difference is unlikely to be a major limitation when interpreting the present findings. Fourth, the ESS decreased to as low as 27.5% after matching for common variables, indicating potential residual heterogeneity across the studies. Accordingly, ESS-based primary and sensitivity analyses were conducted, and the main findings were interpreted in populations with relatively low heterogeneity. Consistent results were obtained from these analyses. Finally, because Liporaxel^®, as an oral formulation of paclitaxel, is expected to have a different adverse event profile from IV formulations, safety outcomes were not compared. Nevertheless, the DREAM study [13] demonstrated that Liporaxel^® was generally well tolerated and associated with manageable adverse events. Collectively, these findings should be interpreted as exploratory and hypothesis-generating rather than definitive evidence, given the inherent limitations of unanchored MAIC. They should be regarded as supportive of, rather than substitutes for, direct comparative evidence.

In conclusion, this unanchored MAIC analysis found comparable survival outcomes between Liporaxel^® and the widely used weekly IV paclitaxel regimen in advanced gastric cancer patients. These findings are consistent with the results of phase 3 trials [13,14] and support the concept that Liporaxel^® may be considered a valid oral treatment option for patients with advanced gastric cancer.

SUPPLEMENTARY MATERIALS

Supplementary Table 1

Search strategy, PICO framework, and inclusion/exclusion criteria for the systematic literature review

Supplementary Table 2

Covariate balance before and after MAIC weighting in the sensitivity analyses

Supplementary Fig. 1

PRISMA flow diagram for the systematic literature review.