Response rates, survival status and adverse events of placebo in randomized control trials for hepatocellular carcinoma: A meta-analysis

Abstract

BACKGROUND

In randomized controlled trials (RCTs), the placebo arm has often been ignored as the attention tends to be focused on the treatment arm. We undertook a meta-analysis based on the data from the placebo arm in RCTs of hepatocellular carcinoma (HCC), the response rates and survival status, and adverse events (AEs) were summarized and evaluated.

AIM

To systematically evaluate the response rates, survival status and AEs in the placebo arms of RCTs for HCC.

METHODS

A systematic search was performed on PubMed, Ovid MEDLINE, Embase and Cochrane Library to identify relevant trials evaluating the efficacy of drugs for the treatment of HCC, published until December 31, 2023. Statistical analysis was performed using R statistical software (version 4.3.2).

RESULTS

A total of 18 RCTs, involving 2390 patients, met the criteria for inclusion in the meta-analysis. The pooled overall disease control rate and objective response rate in the placebo group were 38% [95% confidence interval (CI): 33%-42%] and 1% (95%CI: 1%-2%), respectively. Overall survival and progression-free survival in the placebo group were 7.9 months (95%CI: 7.6-8.31 months) and 1.9 months (95%CI: 1.6-2.1 months), respectively. The incidence of grade 3 or 4 AEs was 37% (95%CI: 30%-43%). Additionally, the incidence of interruptions or dose reductions due to AEs was 20% (95%CI: 13%-27%), while the incidence of treatment discontinuation due to AEs was 9% (95%CI: 6%-12%).

CONCLUSION

Over one-third of advanced HCC patients exhibit therapy-free disease control, with placebo-arm AEs observed. These findings guide single-arm trials design and enhance patient acceptance of anticancer therapies.

Key Words: Hepatocellular carcinoma; Placebo; Randomized controlled trials; Response rates; Survival status; Adverse events

Core Tip: While the therapeutic arm dominates hepatocellular carcinoma clinical trials, the placebo comparator serves as a critical but underutilized window into tumor biology. The natural course of hepatocellular carcinoma under placebo conditions remains poorly quantified, masking potential spontaneous stabilization mechanisms. Furthermore, placebo-associated adverse events in randomized controlled trials remain poorly characterized. Our meta-analysis addresses these gaps through systematically evaluated response rates, survival status and adverse events across placebo arms of hepatocellular carcinoma randomized controlled trials.

INTRODUCTION

Primary liver cancer was the sixth most commonly diagnosed cancer and the third leading cause of cancer death worldwide in 2022[1], with China accounting for 45.3% of new cases worldwide[2]. China bears a substantial burden of hepatocellular carcinoma (HCC) due to high prevalence of chronic hepatitis B infection, making it a significant public health concern nationwide[3]. Over the past decade, several clinical trials on new drugs for HCC have been conducted, including several single-arm trials (SATs) that have received approval from the United States Food and Drug Administration (FDA).

In the study of HCC, people are always concerned about the therapeutic effect of anticancer drugs, but ignore the results of the placebo arm in randomized controlled trials (RCTs). Is tumor regression solely attributable to the effect of medication? Or rather, can placebos also lead to stabilization or even regression of HCC? Previous studies have shown that solid tumors may also occasionally regress even without treatment, either spontaneously or due to placebo effects[4,5].

Therefore, we performed this analysis to reveal the response rates, survival status and adverse events (AEs) of placebo in RCTs of anticancer medicine in HCC. Spontaneous regression, stable disease and AEs may represent natural process during the disease course rather than resulting from drugs. Additionally, data on disease control rate (DCR), objective response rate (ORR), overall survival (OS), progression-free survival (PFS) and AEs in the placebo-controlled group can provide insights into future treatment strategies and serve as references for designing clinical trials, especially for SATs seeking accelerated approval for market launch.

MATERIALS AND METHODS

Study selection

This systematic review and meta-analysis adhered to the guidelines outlined in PRISMA[6] (Supplementary material 1) and AMSTAR[7] (Supplementary material 2). We systematically searched PubMed, Ovid MEDLINE, Embase and Cochrane Library to identify all RCTs evaluating anticancer drugs in advanced HCC patients, with publication cutoff date of December 31, 2023 and language restriction to English. The detailed search strategies for each database are presented in Supplementary material 3. Eventually 18 studies were included (Figure 1). The inclusion criteria were as follows: (1) Assessed advanced HCC in adult only; (2) Exclusion of (neo)adjuvant therapy; (3) Exclusion of local therapy such as surgery, radiation or transarterial chemoembolization; (4) RCTs with a sample size > 20 participants in each arm; (5) Inclusion of a placebo arm either placebo monotherapy or placebo combined with best supportive care without anticancer agents; (6) Exclusion of duplicate trial or subgroup analysis; and (7) Assessment of tumor response according to the Response Evaluation Criteria in Solid Tumors (RECIST or RECIST 1.1)[8,9]. Both Phase 2 and 3 RCTs in either first-line or later lines of therapy were included, if eligible. Nonrandomized trials, trials combing placebo with an active treatment, and trials lacking response rates data were excluded.

Figure 1 PRISMA flow chart.

Data extraction

After independently screening titles and abstracts by the two authors, the full text of potentially relevant studies was downloaded, and relevant data were independently extracted from the published reports by the same authors. Subsequently, this process was validated separately by two other authors, and any discrepancy was resolved through mutual negotiation.

We collected the following relevant trial characteristics: Study name, year of publication, lines of treatment, treatment type, number of patients randomized into each trial arm, tumor response [including DCR, ORR, complete response (CR), partial response (PR) and stable disease (SD)], OS and PFS, and AEs. ORR was defined as the sum of CR (complete disappearance of lesions) and PR (reduction in the sum of diameters of target lesions by ≥ 30%). DCR was defined as the sum of CR, PR and SD (< 30% reduction and < 20% increase in the sum of diameters of target lesions). This study did not involve any individual-level data; therefore, ethical approval was not applicable.

Data analysis

The statistical analysis was performed using R statistical software (version 4.3.2). The meta-analysis of rates was performed using a random effects model based on the inverse-variance approach to accommodate clinical heterogeneity. Rates following a normal distribution were directly calculated using the PRAW method, and studies that did not conform to a normal distribution and reported zero patient values used the Freeman–Tukey double arcsine transformation[10]. Confidence intervals (CIs) for each study were determined using the Clopper–Pearson method[11]. Statistical heterogeneity among the included studies was assessed using the DerSimonian–Laird method estimate for tau-squared (τ², which quantifies the variance of effect sizes for the data using the Cochrane’s Q statistic[12]). CIs for τ and τ² were derived using the Jackson method[13]. Statistical heterogeneity was considered statistically significant if P < 0.10, and quantified using the I² statistic, where I² < 25% indicated no heterogeneity, I² = 25%-50% represented moderate heterogeneity, and I² > 50% indicated considerable heterogeneity[14]. The median meta-analysis was performed using the Weighted Median of Medians to estimate the median. CIs for each study were determined using the sign test[15].

Forest plots for DCR, ORR, CR, PR, SD, median OS and median PFS were generated to assess the efficacy in the placebo arm in HCC RCTs. Forest plots of AEs were generated to assess the incidence of AEs in the placebo arm in HCC RCTs. Prespecified subgroup analyses were conducted to explore potential sources of statistical heterogeneity. Potential publication bias was assessed by Egger test, with statistical significance set at P < 0.05[16].

RESULTS

Characteristics of included studies

Out of the initially identified 1287 studies, 18 RCTs were found to meet the inclusion criteria, involving 2390 patients in the placebo arm (Figure 1). Targeted therapy was the predominant treatment modality (83%). Only three trials (17%) were performed in naive patients, while the remaining trials focused on second-line or later therapies. Approximately 22% of the trials were classified as Phase 2, with the remainder being Phase 3. The placebo arm comprised 2390 patients, whose tumor response rates (DCR and ORR) and median OS were evaluable. However, the number of evaluable patients for median PFS and AEs varied, with some missing. The details of the risk of bias assessment can be obtained for all studies in Supplementary material 4.

Response rates

In the RCT placebo arm, 917 of the evaluable patients attained disease control, resulting in an overall DCR of 38.4%. Using the random-effects model, the pooled DCR for placebo was 38% (95%CI: 33%-42%) (Figure 2). Additionally, 34 patients achieved an objective response, leading to a pooled ORR of 1.2%. Using the random-effects model, the pooled ORR for the placebo was 1% (95%CI: 1%-2%) (Figure 2). Please refer to the Supplementary material 5 for other tumor response data.

Figure 2 Forest plot of response rates in the placebo arm across the randomized controlled trials of hepatocellular carcinoma. A: Disease control rate in the placebo arm; B: Objective response rate in the placebo arm.

Survival status

Among the evaluable patients in the RCT placebo arm, the pooled median OS was 7.9 months (95%CI: 7.6-8.31 months) (Figure 3). Some studies did not report median PFS data. The pooled median PFS among evaluable patients in the RCT placebo arm was 1.9 months (95%CI: 1.6-2.1 months) (Figure 3). Given that median OS and PFS were continuous variables analyzed using the weighted median of medians, this analytical approach did not generate conventional heterogeneity metrics (e.g., I² statistics or Cochran's Q).

Figure 3 Forest plot of survival status in the placebo arm across the randomized controlled trials of hepatocellular carcinoma. A: Overall survival in the placebo arm; B: Progression-free survival in the placebo arm.

AEs

AE data were not detailed in some studies. Among the evaluable patients in the RCT placebo arm, AEs were reported in 1735 patients. Using a random-effects model, the pooled incidence of Grade 3 or 4 AEs was 37% (95%CI: 30%-43%) (Figure 4). A total of 458 patients experienced interruptions or dose reductions due to AEs, with a pooled incidence of 20% (95%CI: 13%-27%) (Figure 4). AEs requiring treatment discontinuation occurred in 180 patients, with a pooled incidence of 9% (95%CI: 6%-12%) (Figure 4). Please refer to the Supplementary material 5 for other AE data.

Figure 4 Forest plot of the incidence of adverse events in the placebo arm across the randomized controlled trials of hepatocellular carcinoma. A: Grade 3 or 4 adverse events (AEs) in the placebo arm; B: Interruption or reductions due to AEs in the placebo arm; C: Treatment discontinuation due to AEs in the placebo arm.

Subgroup analysis

Significant heterogeneity was observed in placebo DCR (I² = 83%, Q = 100.01, P < 0.0001); therefore, subgroup analyses were conducted based on the treatment lines (Figure 5). These analyses revealed a significant difference in placebo DCR between the first-line treatment (27%) and the second-line or later therapies (39%) (between groups χ² = 2.77, P = 0.096). Baseline functional status heterogeneity may further contribute to outcome variability. Given the unavailability of patient-level data across included studies, we conducted surrogate subgroup analyses by stratifying trials into categories based on aggregate functional status metrics: (1) Favorable vs unfavorable prognostic profiles; and (2) High-burden vs low-burden cohorts. No significant heterogeneity in placebo DCR was observed across subgroups stratified by Child–Pugh classification, Barcelona Clinic Liver Cancer staging, Eastern Cooperative Oncology Group performance status, macrovascular invasion status, extrahepatic disease, or HBV infection (Supplementary material 5). Consistently, subgroup analyses of ORR and AEs also demonstrated no significant variations.

Figure 5 Subgroup analyses based on the treatment lines in placebo disease control rate.

Publication bias

Potential publication bias was estimated with Egger's test. All studies yielded a score of P = 0.82, indicating a symmetric distribution across all studies and suggesting no publication bias. However, the funnel plot indicates that some of the studies were in significant regions, indicating possible heterogeneity. Please refer to the Supplementary material 5 for other data.

DISCUSSION

In cancer RCTs, oncologists and scientists focus more attention on the treatment arm, whereas the placebo arm is often ignored as a "stepping stone" for the therapeutic drugs. In contrast to previous studies, our study shifted this focus to the placebo arm, which has been neglected in recent decades. This systematic review and meta-analysis included 18 RCTs involving 2390 HCC patients. It marks the first study dedicated to exploring the survival status and AEs of placebo in RCTs for HCC.

This meta-analysis revealed that 38% of patients who were not treated with anticancer drugs achieved disease control, with median OS and median PFS of 7.9 and 1.9 months, respectively. Whether a high DCR equates to clinical benefit warrants careful consideration. Implementation of high-quality patient-reported outcome studies may provide critical insights by evaluating treatment benefits through the dimension of patient quality of life[17]. The median OS of 7.9 months in the placebo arm is comparable to the 6.8-months median OS reported by Cabibbo et al[18] in their natural course study of HCC. The observed difference reflects how stringent eligibility criteria in modern clinical trials (e.g., preserved liver function) limit the generalization of placebo arm survival data to unselected real-world cohorts. Nevertheless, these data remain clinically relevant for patients with adequate hepatic reserve, particularly those classified as Child–Pugh class A. Subgroup analysis indicated a higher DCR with second-line or later therapies compared to first-line treatment. This may stem from only three small-sample first-line treatment trials being included in our study. There is a possibility that some patients are sensitive to anticancer therapies, leading to intolerance rather than disease progression with first-line treatments. The observed efficacy in RCTs involving second-line or later therapies may partially stem from prior therapeutic interventions. Discrepancies may also stem from heterogeneous baseline functional status across study cohorts. However, the lack of individual patient data in source publications precluded stratified analysis to verify this hypothesis.

In our study, a large portion of patients exhibited stable disease during the follow-up, leading to a high DCR. Furthermore, we found an objective response in 34 of 2390 patients in the placebo arm, resulting in an ORR of 1%. The observed disparity in DCR and ORR appears to be significant, which echoes findings from prior clinical studies. Specifically, in the SHARP clinical trial, patients treated with sorafenib exhibited a DCR that was substantially higher than the ORR. In the sorafenib group, seven patients (2%) had a PR, while 211 patients (71%) had SD. Conversely, in the placebo group, only two patients (1%) had a PR, and 204 patients (67%) had SD[19]. This marked difference in DCR and ORR appears to be significant, suggesting a potential biological explanation rooted in the propensity of HCC to stabilize rather than regress.

This finding also suggests that even among patients not receiving any antitumor therapy (including chemotherapy, targeted agents, or immunotherapy) there may be a 1% chance of PR. This cannot be solely attributable to an incidental anecdote, as studies conducted by Bruix et al[20] and Finn et al[21] have reported ORRs of up to 4% in the placebo arm, which is higher than our study. Spontaneous tumor regression has been documented in various types of tumors as well. For example, approximately 20% of patients with desmoid tumors have experienced spontaneous remission[4]. Additionally, hormone-sensitive tumors like prostate cancer have shown a 15% rate of achieving PR after undergoing prednisone treatment[5]. Despite rare occurrence of spontaneous regression in patients with advanced HCC compared to the cancers mentioned above, it still cannot be ignored.

Our study demonstrates that AEs can occur in the placebo arm, as mentioned in some previous studies[22-24]. Fatigue, abdominal pain and decreased appetite were the most common AEs reported in the placebo arm, primarily consisting of subjective symptoms. In contrast, the treatment group also experienced rashes and objective abnormalities such as increased alanine aminotransferase or aspartate aminotransferase and hypertension. This indicates that AEs manifest not only in advanced HCC patients undergoing antitumor therapies, but also in those abstaining from conventional anticancer treatments. These AEs are related to disease progression, baseline functional status and individual biological behavior. Patients who experience specific symptoms at baseline are more likely to report these symptoms as drug side effects[25]. However, it is also possible that these reports are associated with the nocebo effect, which may be influenced by information about drug safety[23,26].

In advanced cancer treatment, the combined approach of conventional anticancer treatments and complementary and alternative medicine (CAM) is commonly applied[27,28]. Certainly, the concern over potential side effects of traditional treatments may prompt some patients to exclusively use CAM in search of a more comfortable treatment experience, but this approach often results in a worse prognosis. The study by Greenlee et al[29] suggests that patients using dietary supplements such as herbs were more likely to refuse chemotherapy. Also, a retrospective observational study suggests that patients receiving CAM are more likely to refuse conventional anticancer therapies and have a higher risk of death[30]. Our findings can raise patients’ consciousness that both conventional anticancer therapies and placebo treatment can incur AEs, and they should not solely avoid conventional anticancer treatment due to concerns about their potential side effects. In addition, it can also remind patients that the remission of cancer is not necessarily the result of anticancer drugs, but also the natural course of the disease. Clinical responses observed with CAM monotherapy may reflect nonspecific effects rather than direct antitumor activity. Current evidence does not support substituting CAM for conventional anticancer therapies.

In recent years, a few anti-HCC drugs have received accelerated approval from the FDA through SATs[31,32], primarily relying on ORR; a reasonable predictor of survival improvement[33]. However, whether a drug can be approved for market access is not only determined by ORR, but also by factors such as OS, PFS and whether the AEs are within acceptable ranges. Simultaneously, SATs may be prone to biases due to various factors, such as lack of randomization, potential confounding variables, or the absence of a control group for comparison. Here, we provide various indicators in the placebo-controlled group, including OS, as well as data on DCR, ORR and AEs, as reference for future SAT design. The rapidly changing drug development landscapes and the high failure rates in Phase 3 trials both underscore the need for more efficient and robust Phase 2 trial design. For trials that fail to demonstrate outcomes superior to those achieved in the placebo arm, proceeding to Phase 3 should be cautious, thereby avoiding unnecessary effort in labor, time cost and financial resources. The DCR, ORR, OS and PFS in the placebo-controlled group can also provide insights into the natural course of advanced HCC patients with adequate hepatic reserve, which is obscure to outline in the real-world data, leading to better understanding and providing references for designing future treatment strategies. By understanding the survival and efficacy data in the placebo arm, improvements can be made in sample size calculation, endpoint selection, follow-up time and other aspects of clinical trials, thus ensuring that the studies have sufficient statistical power and scientific reliability.

Possible publication bias and heterogeneity among included studies were potential limitations of our research. Publication bias towards negative studies may have affected our results by excluding studies with higher placebo response rates or longer survival durations. The lack of available supplemental data in some trials may limit our interpretation of the results. Therefore, it is uncertain whether all AEs are related to the placebo, and the correlation between clinicopathological variables such as age, gender, stage, Child–Pugh score, and indicators including DCR, ORR, OS, PFS and AEs could not be derived. Additionally, the inclusion of a relatively high proportion of second-line or later therapies in studies may lead to residual effects or AEs from previous treatments that can potentially influence the assessment of efficacy and AEs in subsequent research.

CONCLUSION

The findings of this meta-analysis suggest that 38% of patients with advanced HCC can experience a period of disease control without conventional anticancer treatments. In RCTs, AEs may also occur in the placebo group. These findings can serve as a valuable reference for future SATs, while also preventing patients from being misled by erroneous information, thereby contributing to a higher acceptance of conventional anticancer treatments.

ACKNOWLEDGEMENTS

Special thanks to Jin-Ming Yu for statistical support. Lastly, we appreciate all participating colleagues for making this research possible.