Introduction

For cancer patients undergoing chemotherapy, chemotherapy-induced nausea and vomiting (CINV) are two prevalent and potentially treatment-limiting side effects [1]. Female patients and younger patients have been reported to be at greater risk [2,3,4]. Patients who have experienced vomiting during previous chemotherapy and those with high expectations of severe nausea prior to chemotherapy are at greater risk as well [5].

CINV is classified according to its time of incidence as either acute (0–24 h post-chemotherapy) or delayed (24–120 h post-chemotherapy). CINV that occurs during the course of chemotherapy despite a prophylactic regimen is termed as breakthrough CINV [6, 7].

Only two groups of antiemetics have been developed to target specific biochemical CINV pathways. These include neurokinin (NK)1-receptor antagonists (e.g., aprepitant, rolapitant, and netupitant), and serotonergic (5-HT)3-receptor antagonists (e.g., ondansetron, palonosetron), whereas dopamine (D)2-receptor antagonists (e.g., prochlorperazine, metoclopramide) initially were developed for different indications [8,9,10,11]. Olanzapine was approved by the US Food and Drug Administration as an antipsychotic [12], but has been used off-label as an antiemetic due to its potential to bind to multiple receptors in the CINV pathway, specifically serotonergic 5-HT2a, 5-HT2c, 5-HT3, 5-HT6, and dopamine D1, D2, D3, and D4 receptors [13].

Several phase I–II trials first investigated the efficacy and safety of olanzapine [14,15,16,17,18,19]. A systematic review and meta-analysis of early phase trials reported that 97.2% and 83.1% of patients achieved complete response (defined as no emesis and no use of rescue antiemetics) in the acute and delayed phase, respectively [20].

A number of phase III randomized controlled trials were subsequently undertaken and published, and multiple systematic reviews and meta-analyses have been conducted [21,22,23,24,25,26,27]. However, no review has separately analyzed antiemetics for highly emetogenic chemotherapy (HEC) and moderately emetogenic chemotherapy (MEC) patients, an important distinction that leads to different clinical guideline recommendations. Notably, the American Society of Clinical Oncology (ASCO) [28] currently recommends olanzapine as part of a four-drug regimen for HEC patients, while the National Comprehensive Cancer Network (NCCN) [29] and the Multinational Association of Supportive Care in Cancer (MASCC)/the European Society for Medical Oncology (ESMO) [30] recommend the four-drug regimen as an option in HEC patients. None of these guidelines, however, recommend olanzapine for MEC patients [28, 30]. Furthermore, each of the published reviews has methodological limitations when appraised using AMSTAR-2, a critical appraisal tool for systematic reviews [31] (Appendix 1 Electronic Supplementary Material).

Given the growing interest in olanzapine and the need for a more rigorous review, the aim of this study is to review the efficacy and safety of olanzapine for the prophylaxis and rescue of CINV through a systematic review and meta-analysis. Furthermore, given the large body of existing data, the aim of this review will be to determine the shortfalls of existing literature to provide future direction for olanzapine research in the CINV setting through a cumulative meta-analysis and fragility assessment.

Methods

The protocol for this review has been included in Appendix 2 Electronic Supplementary Material. The reporting of this review is conducted in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses(PRISMA) checklist [32].

Search strategy

In the interest of conducting a rigorous and comprehensive review, a de novo search strategy was developed to search databases from their beginnings. Ovid MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials were searched from their beginning to April 24, 2020. Search restrictions were placed, so only English-language clinical trials were identified (Appendix 3 Electronic Supplementary Material).

Study selection

Two independent in-duplicate screenings were conducted. Where disagreements occurred, discussion of discrepancies occurred and consensus achieved, with the input of a senior author if required. Cohen’s kappa coefficient was calculated, to report the concordance.

Studies were first screened by title and abstract (level 1 screening). Studies were included after level 1 if they reported on olanzapine in a clinical trial for the setting of CINV. These abstracts then underwent full-text screening (level 2 screening) and were eligible for assessment of quantitative synthesis if they compared an olanzapine-containing regimen in one trial arm to a non-olanzapine-containing regimen in the other trial arm(s). Reference lists of included articles after level 2 screening were also assessed, to identify other potentially relevant randomized controlled trials. Studies with less than 5 patients per arm and non-randomized trials were excluded.

Data extraction

As with study selection, data extraction was conducted in duplicate and independently. Disagreements were resolved via discussion, to achieve consensus.

Study demographics of age range, percentage male, chemotherapy emetogenicity, and the difference between the olanzapine regimen and the comparative regimen were noted. The primary efficacy endpoints were the rate of complete response, in the acute (0–24 h post-chemotherapy), delayed (24–120 h post-chemotherapy), and overall (0–120 h post-chemotherapy) phases. The secondary efficacy endpoints were the rates of no nausea and no emesis, for each phase. Safety endpoints were the rate of no serious adverse events (i.e., no grade 3 or 4 toxicities, as assessed by Common Terminology Criteria for Adverse Events (CTCAE) criteria), as reported by authors. Grades 1 and 2 toxicities were not extracted for analysis, due to the paucity of data.

When a trial had two olanzapine-containing arms, the data across the two olanzapine arms were summed for analysis and compared to the non-olanzapine-containing arm.

Meta-analysis

The Mantel-Haenszel, random-effects analysis model was used to compute risk ratios (RRs) and accompanying 95% confidence intervals for each endpoint. A p value of less than 0.05 was considered statistically significant in the test for overall effect.

Studies were first separately meta-analyzed by regimen intent—whether olanzapine was administered for prophylaxis or for management of breakthrough nausea. It was then analyzed by age, separating studies reporting on adult and children. Adult studies were further meta-analyzed according to chemotherapy emetogenicity, olanzapine dosage, comparative regimens, and study quality; meta-analyses were conducted for the following subgroups of adult studies:

  1. 1.

    HEC studies, as determined by the MASCC/ESMO classification [33]

  2. 2.

    MEC studies, as determined by the MASCC/ESMO classification [33]

  3. 3.

    Olanzapine, administered as 10 mg daily PO

  4. 4.

    Olanzapine, administered as 5 mg daily PO

  5. 5.

    Studies with a double-blind, placebo-controlled design, where the control arm includes a placebo and all antiemetics of the olanzapine-containing arm except for olanzapine itself

  6. 6.

    Studies with an open-controlled design, where the control arm includes all the antiemetics of the olanzapine-containing arm except for olanzapine itself

  7. 7.

    Studies with an open-controlled design, where the control arm includes antiemetics not included in the olanzapine-containing arm

For endpoints that statistically favored one arm and had more than 3 included trials, absolute risk differences (RD) were computed to assess whether there is a 10% or greater difference, deemed to be the threshold for clinical significance by MASCC/ESMO [34]. These analyses were performed using Review Manager (RevMan 5.4) by Cochrane IMS.

Fragility assessment

Fragility indices were calculated for each statistically significant endpoint by subgroup, to quantitatively assess the robustness of the summary estimate. Determination of the index involves a series of iterative calculations, until the simulated study results change from statistically significant to statistically insignificant according to the Fisher’s exact test. Essentially, the index is the number of control patients that would need to change from a nonevent to an event outcome, to change the statistical conclusion of a trial [35]. These analyses were conducted using Stata 16.

Cumulative meta-analysis

A cumulative meta-analysis was conducted for each efficacy meta-analysis with more than 5 studies, also using the Mantel-Haenszel random-effects analysis model. These analyses will allow for the assessment of the impact of each trial on the meta-analysis summary effect size and 95% CI. These analyses were conducted using Comprehensive Meta-Analysis (Version 3) by Biostat.

Assessment of bias

The Cochrane Risk of Bias tool was used to assess the quality of included randomized controlled trials. Four reviewers (RC, LC, ML, CD) independently assessed bias, after which discussion and consensus was used to resolve any discrepancies. Funnel plots were generated to visually assess for publication bias, for each phase of the three efficacy endpoints where there are 5 or more trials; these were generated using Comprehensive Meta-Analysis (Version 3) by Biostat.

Results

Included studies

From the search strategies, 312 records were identified. After removing duplicate records and adding records identified from included trials, 178 records underwent level 1 screening. A total of 34 full-text articles were assessed for eligibility through level 2 screening, at which points 6 were excluded with reason—three were not a randomized controlled trial [36,37,38], one did not investigate olanzapine in the CINV setting [39], and two did not have an appropriate treatment regimen for inclusion in our review [40, 41]. Of the remaining 28 articles, 25 randomized controlled trials had extractable data and were included in this systematic review and meta-analysis(Appendix 4 Electronic Supplementary Material). Concordance, as measured by Cohen’s Kappa, for level 1 screening was 0.86, and 0.84 for level 2 (Appendix 5 Electronic Supplementary Material).

Three studies reported on olanzapine for the rescue of breakthrough CINV [42,43,44]; 22 studies reported on olanzapine in the prophylactic setting [45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66]. Only seven studies (one reporting on rescue of breakthrough CINV, and six reporting on prophylactic CINV) had no corresponding full-text articles [42, 48, 53, 55, 56, 60, 61]. One study reported on olanzapine for children [48]. Among the adult prophylactic studies, 15 reported exclusively on HEC patients [46, 49,50,51,52, 55, 56, 58, 60,61,62,63,64,65,66], three exclusively on MEC patients [53, 57, 59], and three on a patient population that consists of both HEC and MEC patients [45, 47, 54]. Eight studies compared olanzapine to a double-blind placebo-controlled regimen [47, 52, 59, 62,63,64,65,66] and thirteen used an opened controlled study design—nine studies used a control arm with antiemetics different from the antiemetics in the investigational (olanzapine-containing) arm [46, 49, 51, 53, 55, 56, 58, 60, 61] and four used a control arm with the same antiemetics as in the investigational (olanzapine-containing) arm except for olanzapine [45, 50, 54, 57]. 17 adult prophylactic studies used 10 mg doses of olanzapine [45, 46, 49,50,51,52,53,54,55,56,57,58,59, 61, 62, 65, 66], and 3 studies used 5 mg [47, 60, 63]; 1 used a mix of 5 mg and 10 mg [64] (Table 1).

Table 1 Study demographics
Full size table

Quality of included studies

The risk of bias assessment for each included study is reported in Appendix 6 Electronic Supplementary Material. Over half of all included studies had high risk of bias, due to concerns around lack of blinding.

Assessment for publication bias of olanzapine for the prophylaxis of CINV

Funnel plots are presented in Appendix 7, 8 Electronic Supplementary Material. There are no obvious asymmetries, suggesting no obvious concerns of publication bias in this body of literature.

Efficacy of olanzapine for the prophylaxis of CINV in children

In children, olanzapine was not statistically superior in the acute and overall phases, according to the one study by Long et al.

Efficacy of olanzapine for the prophylaxis of CINV in adults

Complete response

Acute phase

Olanzapine was statistically better than comparative regimens in the acute phase. Among HEC studies, studies using 10 mg olanzapine dosages, studies using a double-blind placebo-controlled design, and open-design studies comparing olanzapine to control regimens of antiemetics not included in the investigational arm, olanzapine was still statistically superior (Fig. 1.1). Olanzapine was clinically superior (risk difference greater than 10%) overall, in HEC studies, studies using 10 mg olanzapine doses, and for studies comparing olanzapine in a double-blind placebo-controlled design (Table 2).

Fig. 1
figure 1figure 1figure 1

Efficacy of olanzapine regimens compared to others for the prophylaxis of chemotherapy-induced nausea and vomiting (CINV)— Complete response. 1.1 Acute phase. 1.2 Delayed phase. 1.3 Overall phase

Full size image
Table 2 Absolute risk difference between olanzapine and other regimens for statistically significant differences
Full size table

Delayed phase

Olanzapine was also statistically and clinically superior in the delayed phase. This statistical and clinical superiority prevails in analyses of HEC studies, studies using 10-mg olanzapine doses, studies administering 5-mg of olanzapine, and studies assessing olanzapine in double-blind placebo control studies (Fig. 1.2; Table 2).

Overall phase

Olanzapine was statistically and clinically superior in the overall phase among all studies, HEC studies, 10-mg olanzapine studies, 5-mg olanzapine studies, and double-blind placebo-controlled studies (Fig. 1.3; Table 2).

Nausea control

For the acute, delayed, and overall phases, olanzapine was statistically superior to comparative regimens. This observation was similarly noted among HEC studies, studies where 10 mg of olanzapine was administered, and double-blind placebo-controlled trials. Olanzapine was also statistically and clinically superior to open-design studies using a control arm with different antiemetics than used in the investigational (olanzapine-containing) arm in the delayed and overall phases (Fig. 2; Table 2).

Fig. 2
figure 2figure 2figure 2

Efficacy of olanzapine regimens compared to others for the prophylaxis of chemotherapy-induced nausea and vomiting (CINV)—no nausea. 2.1 Acute phase. 2.2 Delayed phase. 2.3 Overall phase

Full size image

Emesis control

Neither olanzapine nor control arms were statistically superior to the comparator arm in the acute phase. Olanzapine was both statistically and clinically superior in the delayed and overall phases. Olanzapine was statistically and clinically better in the delayed phase among HEC trials and 10-mg olanzapine trials (Fig. 3; Table 2).

Fig. 3
figure 3figure 3figure 3

Efficacy of olanzapine regimens compared to others for the prophylaxis of chemotherapy-induced nausea and vomiting (CINV)—no emesis. 3.1 Acute phase. 3.2 Delayed phase. 3.3 Overall phase

Full size image

Cumulative meta-analysis and fragility assessment of olanzapine for the prophylaxis of CINV

Across all three time phases, the meta-analysis results for complete response are the most robust; results reporting on emetic control are the least robust of the three efficacy endpoints (Appendix 9 Electronic Supplementary Material). The most recent trials did not lead to a noticeable effect on the meta-analysis’ summary estimate for the endpoints of complete response and nausea control (Appendix 10, 11 Electronic Supplementary Material).

Olanzapine for the rescue of breakthrough CINV

Olanzapine was statistically superior to comparative regimens with respect to complete control, nausea control, and emetic control, according to the one study reporting on each outcome (Fig. 4). Olanzapine was also clinically superior in all these aforementioned endpoints—RD = 0.33 (95% CI: 0.10–0.56) for complete response in the acute phase, RD = 0.38 (95% CI: 0.18–0.57) for complete response in the overall phase, RD = 0.45 (95% CI: 0.28–0.62) for nausea control in the overall phase, and RD = 0.39 (95% CI: 0.21–0.56) for emetic control in the overall phase.

Fig. 4
figure 4

Efficacy of olanzapine regimens compared to others for the rescue of breakthrough chemotherapy-induced nausea and vomiting (CINV). 4.1 Complete response - acute phase. 4.2 Complete response - overall phase. 4.3 No nausea - overall phase. 4.4 No Emesis - overall phase

Full size image

Safety of olanzapine for the prophylaxis of CINV

Olanzapine is as safe as comparative regimens; the risk of serious adverse events is not statistically significant for olanzapine relative to other regimens (Appendix 12, 13 Electronic Supplementary Material).

Discussion

This review is the most rigorous systematic review to date investigating olanzapine in the CINV setting. A protocol was developed prior to the commencement, risk of bias for studies were assessed, and publication bias was assessed; some or all of these three methodological elements were omitted in prior reviews [21,22,23,24,25,26,27].

This review also has the highest statistical power and appraises all the clinically important endpoints. The most recent reviews by Zhou et al. in 2020 included 11 studies with 1107 patients [21]; other reviews by Bahbah et al. in 2019 and Sutherland et al. in 2018 included 9 RCTs with 1572 patients, and 14 trials with 1917 participants, respectively [22, 23]. This review summarizes the results across 25 studies, which reported on 4275 patients. One study reported the effect of olanzapine on children, and three studies reported on olanzapine for the rescue of breakthrough CINV; the remaining 23 studies reported on olanzapine for the prophylaxis of CINV in adults, across 4217 patients. Zhou et al. reported on acute and delayed emetic control with or without nausea control, Bahbah et al. meta-analyzed complete response and nausea control rates, and Sutherland et al. summarized instances where patients successfully experienced no nausea and no emesis; our review reports on complete response, nausea control, and emetic control.

For studies reporting on HEC patients, olanzapine is statistically and clinically superior in seven of nine efficacy endpoints in the prophylaxis setting; only complete emetic control in the acute and overall phases were not statistically different from comparative regimens. Meta-analysis results among studies employing 10-mg doses and among studies comparing olanzapine to placebo-controlled regimens indicated olanzapine as statistically and clinically superior in eight of nine efficacy endpoints for prophylaxis of CINV, with the exception of complete emetic control in the acute phase. These results support the international clinical guidelines [28,29,30] in their recommendation of 10-mg olanzapine in addition to standard antiemetic regimens for the prophylaxis of CINV among HEC patients.

Furthermore, this review includes important subgroup analyses not previously conducted among prophylactic studies, namely meta-analyzing studies reporting on MEC patients and 5-mg olanzapine dosing. Olanzapine is both statistically and clinically superior in only three of six efficacy endpoints where a 5-mg dosage is employed—complete response in the acute, delayed, and overall phases. However, it is important to note that over 800 patients across 4 studies were meta-analyzed for the efficacy endpoints of complete response; there was much less statistical power relative to meta-analyses looking at HEC patients alone. Furthermore, even for the efficacy endpoints of complete response, these meta-analysis results are much more fragile and less certain than those pertaining to olanzapine administered at 10-mg dose studies. Olanzapine may potentially be superior to comparative regimens when administered in 5-mg doses as indicated by point estimates, but the paucity of data results in low statistical power to find these differences statistically significant. Olanzapine has also recently been reported to be effective at 5-mg doses in controlling nausea and vomiting, unrelated to chemotherapy, for patients with advanced cancer [67]. More RCTs are needed in the CINV setting, to evaluate the efficacy of 5-mg olanzapine doses compared to non-olanzapine-containing regimens. Studies comparing 5-mg doses to 10-mg doses are also encouraged; an abstract recently presented by Mukhopadhyay et al. suggests that 5-mg and 10-mg doses may have similar efficacy, although it has no description of drop out patients or chemotherapy regimens in either arm, and no statistical calculations were published [39].

In the MEC setting, olanzapine is reported to be statistically and clinically superior in two of nine efficacy endpoints only—no nausea in the delayed phase, and no emesis in the overall phase. However, as with the results from the meta-analysis of 5-mg doses, there is a paucity of data in this setting. The results are less robust compared to those in the HEC setting, with the recent clinical trials having noticeable impacts on the summary effect size. More RCTs in this setting would allow for a better understanding of olanzapine’s true efficacy for MEC patients.

Olanzapine is reported to be clinically and statistically superior than other regimens for the rescue of breakthrough CINV. However, this review’s results are only supported by one included study for each efficacy endpoint. Results should be interpreted with caution. In both the prophylactic and rescue setting, olanzapine is reported to be equally as safe as other regimens. However, this too should be interpreted with caution, as the key adverse event of sedation is not routinely reported—many studies commonly reported only on serious (i.e., grade 3 or greater toxicity) adverse events, an observation also noted by our group several years ago [26]. It has been well-documented that olanzapine is a strong sedative, and patients commonly experience fatigue, drowsiness, and reduced general activity [20]. In the interest of reducing adverse events, further exploring the reduction of the dosage of olanzapine (i.e. more RCTs reporting on 5-mg olanzapine doses) is encouraged.

This review was not without limitations. Ideally, the protocol would have been registered on PROSPERO; given the COVID 19 pandemic, this was not a feasible option—protocol registration would have required several months, while in hindsight our review was already completed. There were numerous instances where there were high levels of heterogeneity; a random-effects model was applied in all circumstances to try to appropriately account for this. As well, as is the nature of meta-analyses, the results suffer from any intrinsic biases from included RCTs; over half of the studies have notable concerns of bias due to lack of blinding.

In conclusion, olanzapine is effective and safe for the prophylaxis and rescue of CINV. It has been well-documented in the HEC setting and when administered at 10-mg doses; it is statistically and clinically superior to comparative regimens, but its sedative properties can make it difficult to use in outpatient settings. It is unclear if olanzapine is effective in the MEC setting and when administered at a lower 5-mg dose, and further RCTs are needed for a more definitive conclusion. The sedative effect associated with 10 mg of olanzapine further corroborates the need for more investigations into using olanzapine at lower doses.