The searches retrieved a total of 5993 records (post deduplication) that were of possible relevance to the review (). These were screened and 104 full references were ordered. Of these, five had to be cancelled because they were unobtainable. Of the full references evaluated, 28 papers describing 16 studies were included in the review.
Study characteristics
Pegylated liposomal doxorubicin hydrochloride plus carboplatin compared with paclitaxel plus carboplatin
Two RCTs29,31 were identified for this comparison. The RCTs were of similar design but one was a Phase II RCT29 and one was a Phase III RCT.31 In addition, the dose of PLDH evaluated differed between the trials, with 45 mg/m2 and 30 mg/m2 used by Bafaloukos et al.29 and Pujade-Lauraine et al.,31 respectively. The licence for PLDH does not recommend a dose of PLDH for use in combination with platinum-based chemotherapy. Bafaloukos et al.29 note in the discussion that, at the time of initiation of the trial, limited information was available on the optimal dose for PLDH in combination with carboplatin. As highlighted by Bafaloukos et al.,29 retrospective analyses suggest that lower dose intensities of PLDH (30–40 mg/m2) are as clinically effective but with improved tolerance. Clinical experts have fed back that, in UK clinical practice, PLDH would most likely be used at a dose of 30 mg/m2 when combined with carboplatin.
Bafaloukos et al.29 report the results of a randomised study in which 204 patients with histologically confirmed recurrent ovarian cancer were randomised to either PLDH (45 mg/m2) plus carboplatin (AUC 5) every 28 days or paclitaxel (175 mg/m2) plus carboplatin (AUC 5) every 21 days. Patients recruited had disease that had recurred at least 6 months after platinum-based chemotherapy, i.e. women with platinum-sensitive disease. Women with only elevated CA125 levels (twice the ULN or more) as an indicator of disease were also included.
The primary aim of the study was to evaluate the comparative clinical effectiveness of the two treatment regimens in terms of response rate and toxicity in women with platinum-sensitive ovarian cancer relapsing after first-line platinum-based therapy. OS and TTP were analysed as secondary outcomes. Subsequent to randomisation, 15 patients were found to be ineligible (reasons provided). Therefore, analyses presented are based on data from 189 eligible patients (96 in the paclitaxel plus carboplatin group vs. 93 in the PLDH plus carboplatin group). The reported power calculation indicates that 201 patients were needed to identify a 20% difference in response rate between the groups. The study might have been underpowered to detect a difference between groups in response rate.
Randomisation (1 : 1) was performed at the central Hellenic Cooperative Oncology Group (HeCOG) Data Office in Athens but details on the method of randomisation were not reported. Stratification criteria were not applied at randomisation. Tumour response was evaluated using World Health Organization (WHO) criteria for patients with measurable disease and CA125 based on Rustin’s criteria for patients without measurable disease. Median duration of follow-up was reported as 43.6 months (95% CI 0.1 to 74.8 months), but the range of follow-up was not reported either for the full trial population or for the individual treatment groups.
All patients received standard premedication of dexamethasone, diphenhydramine and ranitidine (Zantac®, GSK) prior to paclitaxel. In the group receiving paclitaxel, premedication was administered twice, orally 12 hours before and again intravenously 30 minutes before paclitaxel infusion. In the group receiving PLDH, premedication was administered only intravenously prior to PLDH infusion. Six cycles of chemotherapy were administered, unless disease progression or unacceptable toxicity occurred. A maximum of 2 weeks’ delay was allowed for toxicity and treatment was discontinued if longer toxicity-related delays occurred. For grade 3 and grade 4 thrombocytopenia, a 25% and a 50% dose reduction, respectively, was recommended for all drugs.
A median of six cycles of paclitaxel plus carboplatin (range 1–9) and six cycles of PLDH plus carboplatin (range 1–8) were administered. Most patients in each group completed the planned treatment (68% in paclitaxel plus carboplatin and 70% in PLDH plus carboplatin).
In the second RCT identified for this comparison, Pujade-Lauraine et al.31 report the results of a randomised international, multicentre, open-label, Phase III non-inferiority trial (CALYPSO) in which 976 patients with platinum-sensitive (disease progression > 6 months after prior treatment) relapsed/recurrent ovarian cancer received a combination of PLDH plus carboplatin (n = 467) or carboplatin plus paclitaxel (n = 509). Prior treatment must have included a taxane and no more than two previous platinum-based regimens (i.e. patients had failed first- or second-line treatment). Patients with measurable [according to Response Evaluation Criteria in Solid Tumours (RECIST)] and CA125 assessable [according to Gynecologic Cancer Intergroup (GCIG)] criteria were eligible.
The primary publication presents results on PFS. Accompanying publications were identified that present results on mature OS data,56 clinical effectiveness results in the subgroup of patients with PPS ovarian cancer (relapse at between 6 and 12 months since receipt of last cycle of chemotherapy)57 and QoL.59
The trial was of a non-inferiority design with the aim of determining whether PLDH (30 mg/m2) plus carboplatin (AUC 5) every 4 weeks was non-inferior to the standard treatment of paclitaxel (175 mg/m2) plus carboplatin every 3 weeks.31 The goal was to evaluate the comparative effectiveness of the treatments in terms of efficacy and toxicity. The primary outcome of the trial was PFS, with OS, QoL and toxicity as prespecified secondary outcome measures. Determination of disease progression was based on RECIST and GCIG criteria modifications, and included any of the following: occurrence (clinically or by imaging) of any new lesion; increase in measurable and/or non-measurable tumour defined by RECIST; CA125 elevation defined by GCIG criteria; health status deterioration attributable to disease; and death from any cause before progression was diagnosed. Assessments were independently reviewed. All patients were observed for at least 5 years from random assignment to assess OS.
Randomisation was in permuted blocks of six in a 1 : 1 ratio, and patients were stratified based on therapy-free interval from last chemotherapy (6–12 months vs. 12 months), measurable disease (yes vs. no) and centre. Despite randomisation, an imbalance in treatment allocation was noted (467 randomised to PLDH plus carboplatin vs. 509 randomised to paclitaxel plus carboplatin).
All patients received antiemetics, including a serotonin antagonist and corticosteroid. Patients randomly assigned to paclitaxel plus carboplatin received premedication to prevent hypersensitivity reactions. Dose delay and dose reduction were allowed for haematological and non-haematological toxicity. In the absence of unacceptable toxicity or disease progression, patients were treated for a total of six courses of therapy; if SD or PR was achieved after six courses of therapy, patients were allowed to remain on therapy until progression.
To assert non-inferiority of PLDH plus carboplatin, it was estimated that a sample size of 898 evaluable patients (estimate of 745 progression) would be required.31 The calculation was based on non-inferiority margin with a HR of 1.23 at 15 months or a 7.9% absolute difference at 12 months (90% power and a one-sided CI of 95%).
Median follow-up was 22 months; median follow-up in the individual treatment groups not reported.31 The median number of cycles was six in each treatment group, with a range of cycles from 1 to 14 in the PLDH plus carboplatin group and 1 to 12 in the paclitaxel plus carboplatin group. A significantly larger proportion of patients in the PLDH plus carboplatin group completed at least six cycles of treatment (85% vs. 77%; p < 0.001).
Pegylated liposomal doxorubicin hydrochloride plus carboplatin compared with carboplatin alone
Alberts et al.28 reported the results of a randomised study in which 61 patients from the USA with recurrent stage III or i.v. epithelial or peritoneal ovarian carcinoma were randomised to PLDH (i.v. infusion of 30 mg/m2) plus carboplatin (AUC 5) once every 4 weeks (31 patients) or carboplatin (AUC 5) alone once every 4 weeks (30 patients). A follow-up study reporting final OS results was also identified.55
To be eligible for enrolment, patients had to have histologically diagnosed stage III or IV disease that was determined to be progressive based on RECIST or GCIG CA125 criteria. Patients also had to have a progression-free interval and a PFI of 6–24 months after first-line platinum-based chemotherapy, which indicates that the study focused on women with platinum-sensitive disease. Patients were excluded if Zubrod performance status was > 1. Prior treatment with up to 12 courses of a non-platinum containing consolidation treatment during the 6- to 24-month PFI was allowed on the proviso that treatment had been completed at least 28 days prior to registration.
The primary aim of the study was to evaluate the comparative clinical effectiveness of the two treatment regimens in terms of OS in women with platinum-sensitive ovarian cancer. PFS, confirmed CR rate and time to treatment failure were analysed as secondary outcomes. Objective response and disease progression were defined according to standard RECIST criteria.69 GCIG CA125 progression criteria were also implemented in defining disease progression.70
Details on the method of randomisation were not reported, but randomisation was 1 : 1 to each group and was reported to be equal between the groups. Randomisation was stratified by disease measurability, number of disease sites and serous histology. The power calculation reported indicates that the study had initially planned to recruit 900 patients over a period of 4.5 years. However, as a result of slow patient accrual, the study closed early with only 61 patients enrolled. Initially designed as a Phase III RCT, results were reported as for a Phase II RCT. Median duration of follow-up was reported as 22.4 months, but the range of follow-up was not reported either for the full trial population or for the individual treatment groups. Markman et al.55 reported a longer follow-up of the same trial. However, the duration of follow-up in this study is unclear.
Each treatment was given until progression, intolerable toxicity or a request from either the clinician or the patient to be removed from the study. Dose modifications were allowed, based on toxicity to PLDH. The maximum cumulative dose of PLDH was 600 mg/m2. Any patient with a compromised left ventricular ejection fraction (LVEF) (< 45% or decreases by a relative 20% from baseline) was removed from PLDH and continued on the carboplatin treatment. Carboplatin dose modifications were allowed for gastrointestinal and neurological toxicity. Patients with persistently greater than or equal to grade 2 peripheral neuropathy, despite dose reduction, were permanently taken off carboplatin treatments. The median number of treatment cycles given was 7 (range 1–18) for patients in the PLDH plus carboplatin group and 6 (range 2–16) for those in the carboplatin alone group. No major protocol violations were reported.
Trabectedin plus pegylated liposomal doxorubicin hydrochloride compared with pegylated liposomal doxorubicin hydrochloride alone
Monk et al.30 report the results of an open-label, randomised, multicentre (124 centres in 21 countries), Phase III trial involving 672 women with recurrent ovarian cancer after failure of first-line platinum-based chemotherapy (OVA-301). Patients with platinum-resistant ovarian cancer (PFI < 6 months) or platinum-sensitive ovarian cancer (PFI ≥ 6 months) were eligible, but those who experienced progression during first-line therapy (platinum refractory) were excluded. Measurable disease by RECIST criteria69 was also an inclusion criterion. Related publications identified were a follow-up study reporting mature OS analysis,64 clinical efficacy results for the subgroup of patients with PPS65 and full results for QoL.67
The aim of OVA-30130 was to compare the efficacy and safety of PLDH (30 mg/m2) plus trabectedin (1.1 mg/m2) every 21 days (n = 337) compared with PLDH (50 mg/m2) alone every 28 days (n = 335). The primary outcome was PFS, which was defined as time from randomisation to disease progression or death. Primary analysis of PFS was based on independent radiology review (radiological evaluation alone) by radiologists who were masked to treatment allocation. Secondary end points included OS, ORR (response maintained ≥ 4 weeks by RECIST69), and duration of response (calculated from date of first documentation of response to date of PD or death from PD). QoL was a tertiary outcome and was evaluated using the European Organisation for Research and Treatment of Cancer (EORTC) quality of life questionnaire C30 (QLQ-C30) and ovarian cancer-specific QLQ-OV28.71 All efficacy analyses were based on the intention-to-treat (ITT) principle.
Randomisation was by a permuted block method (1 : 1 ratio) and patients were stratified by performance status [Eastern Cooperative Oncology Group (ECOG) score of 0 or 1 vs. 2] and platinum sensitivity (sensitive vs. resistant). After enrolment of 440 patients, and before central radiology review, the study was amended, changing the two primary efficacy end points, OS and PFS, to a single primary end point, PFS. OS became a secondary end point; the sample size remained unchanged. The sample size calculation indicated that 415 PFS events were required to test statistical difference between treatment groups with at least 90% power; it is reported that approximately 650 patients were to be randomised over 2 years.
Treatment was continued until disease progression or confirmation of CR and could be continued for two or more cycles beyond confirmed CR. A maximum of two dose reductions for each drug was allowed (in the trabectedin plus PLDH group, trabectedin could be reduced to 0.9 mg/m2 and subsequently to 0.75 mg/m2, and PLDH to 25 mg/m2, then to 20 mg/m2; in the PLDH group, PLDH could be reduced to 37.5 mg/m2 and then to 28 mg/m2). Median cumulative trabectedin dose was 5.6 mg/m2 (range 1–23 mg/m2). For PLDH, median cumulative PLDH dose was 154.4 mg/m2 (range 15–630 mg/m2) and 216 mg/m2 (range 3–1061 mg/m2) when administered in combination with trabectedin and as a monotherapy, respectively. Incidence of dose reductions was similar between groups, whereas cycle delays were less frequent with PLDH alone than trabectedin plus PLDH.
Median duration of follow-up in the initial publication was not reported30 but median follow-up in the longer-term study was 47 months.64
The authors report that, despite stratification before randomisation, there was an imbalance between groups in mean baseline PFI that favoured PLDH alone (13.3 months with PLDH alone vs. 10.6 months with trabectedin plus PLDH; p = 0.009). Post hoc hypothesis-generating analyses on the influence of PFI on OS were carried out (discussed in Assessment of effectiveness).
It should be noted that use of trabectedin plus PLDH as an intervention in patients with PRR is not covered by the scope of this review. Clinical effectiveness data for only platinum-sensitive patients are presented.
Pegylated liposomal doxorubicin hydrochloride compared with topotecan
Gordon et al.49 report the results of a Phase III randomised study comparing PLDH compared with topotecan in 474 women with histologically proven recurrent epithelial ovarian carcinoma that recurred after or did not respond to first-line platinum-based chemotherapy. The RCT was open label in design and was carried out at multiple centres (104 sites) in the USA and Europe. Patients with either measurable or assessable disease were included, where measurable disease was defined as presence of bidimensionally measurable lesions with clearly defined margins based on imaging scans and assessable disease was defined as unidimensionally measurable lesions by imaging scan in conjunction with serum CA125 levels of > 100 U/ml. A follow-up publication reported data on more mature OS, together with subgroup analyses based on platinum sensitivity.54
Patients were randomised to receive either PLDH 50 mg/m2 as a 1-hour infusion every 28 days (239 patients) or topotecan 1.5 mg/m2 daily for five consecutive days every 21 days (235 patients).49 In the absence of disease progression, treatment in each group could be continued for up to 1 year. Treatment could also continue if the patient demonstrated sustained clinical benefit. Patients who discontinued treatment after 6 months (six cycles of PLDH or eight cycles of topotecan) were considered protocol completed.
The study was described as randomised, but details on the method of randomisation were not reported. Patients were stratified for platinum sensitivity and for the presence or absence of bulky disease (tumour mass > 5 cm). Patients were classified as platinum sensitive if they had a PFI of > 6 months after first-line platinum-based chemotherapy and platinum refractory if they had SD, progressed during initial platinum-based therapy or relapsed within 6 months after completion of therapy. In the subsequent publication,54 analyses for OS and PFS for the subgroups of patients with PPS disease (PFI > 6 to ≤ 12 months) and FPS disease (PFI > 12 months) are presented. The authors report that the main outcome measures of efficacy were PFS and OS. Overall response rate (confirmed CR plus PR), time to response, duration of response, QoL, and safety and toxicity were also assessed. The study was designed with 80% power to demonstrate statistical equivalence between the two treatment groups. The initial sample size calculation found that a total of 350 assessable patients, 175 patients in each treatment group, would need to be randomised. To accommodate two interim analyses (necessitating 5% more patients) and anticipated loss of 20% of randomised patients who might not be assessable for efficacy end points, the sample size was increased to 460.
Protocol deviations included: (1) failure to meet entry criteria (seven patients receiving PLDH, two patients receiving topotecan); (2) patients who continued on study after first clinically significant change in LVEF (13 patients receiving PLDH); (3) patients who continued treatment after documented disease progression (40 patients receiving PLDH, 42 patients receiving topotecan); and (4) patients who completed fewer than eight cycles of treatment but were deemed protocol completed by the investigator (20 patients receiving topotecan).
Dose modifications were permitted. Reasons for reduction in PLDH dose included PPE, haematological toxicity, elevated bilirubin, stomatitis, or all other grade 3 and grade 4 events until resolution to grade 2 or lower. In the event of severe neutropenia during any cycle with topotecan, the dose was reduced by 0.25 mg/m2 for subsequent courses. Treatment with either drug was temporarily suspended or discontinued in cases of: disease progression; serious or intolerable AEs precluding further treatment; inability to tolerate study drug despite dose modification; LVEF of < 45% or a 20% decrease from baseline; and patient’s decision to withdraw participation or patients requiring radiation.
Median duration of follow-up was not reported in either publication.49,54 In addition, information on mean or median number of cycles received in each treatment group was not provided. However, the mean cycle dose and cycle length for each treatment group were reported to be close to those specified in the protocol, indicating good compliance in following the dosing guidelines.
Pegylated liposomal doxorubicin hydrochloride compared with paclitaxel
In a publication available as only a conference abstract, O’Byrne et al.47 provided a brief overview of a trial comparing PLDH compared with paclitaxel. The search did not retrieve a full publication of this study. However, as part of TA91,13 the manufacturer of PLDH (Schering-Plough) provided a full trial report as part of the industry submission.72 The description of trial methodology and results for OS and adverse effects have been adapted from TA91.13
The trial by Schering-Plough was a Phase III, randomised, open-label study involving 216 women with epithelial ovarian carcinoma after failure of first-line platinum-based chemotherapy. Additionally, to be eligible, women had to have measurable disease and be taxane naive. The trial was designed to compare the clinical effectiveness and safety of PLDH (50 mg/m2) every 28 days compared with paclitaxel (175 mg/m2) every 21 days.
Randomisation was carried out in a 1 : 1 ratio, with patient stratification by platinum sensitivity and bulky disease. No details on the method of randomisation are reported.
TA9113 reports that the planned enrolment was for 438 patients but only 216 were randomised (108 in each treatment arm), with the trial closing early due to poor accrual. It is thought that poor accrual was associated with the approval of Taxol for use in combination with platinum-based therapy for the first-line treatment of ovarian cancer by the European Agency for the Evaluation of Medicinal Products.
Patients were assessed weekly for haematological toxicities, and radiological imaging was repeated every 7–8 weeks to assess disease status. Patients achieving either a CR or PR were re-evaluated 4 weeks later to confirm the initial observation of response. All participants were to have been followed for a minimum of 1 year for survival and disease progression.
At baseline, the two treatment groups were balanced in terms of age, treatment-free interval (TFI), disease bulk, the number of previous chemotherapy regimens, the type of previous chemotherapy agents received, histology and performance status.
As a result of the low recruitment rate, efficacy analysis in TA9113 was limited to OS. AEs were also described.
Topotecan compared with paclitaxel
ten Bokkel Huinink et al.21 report the results of an open-label, Phase III randomised study involving 235 patients with stage III/IV ovarian cancer, who had progressed during or after treatment with one platinum-based chemotherapy. The study was designed to compare the effectiveness and toxicity of topotecan (1.5 mg/m2) for five consecutive days every 21 days compared with paclitaxel (175 mg/m2) every 21 days. Enrolled patients had at least one bidimensionally measurable lesion as evidenced by CT or magnetic resonance imaging (MRI) scan, ultrasound or physical examination. Patients who had received more than one prior chemotherapy, or who had been previously treated with topotecan or paclitaxel, were ineligible. A second publication reporting more mature OS data was also identified.52 A related study reports results from an analysis of patients who received third-line treatment during the trial, and specifically crossover therapy with the treatment received in the other group.53
The primary outcome measures were response rate, duration of response and TTP. Response rate included CR or PR as a best response as determined by WHO criteria, with all responses independently reviewed by a radiologist who was masked to treatment allocation. Secondary outcome measures were time to response and OS. Of the 235 patients randomised, nine patients did not receive treatment and were excluded from analyses. An additional 24 patients were not evaluated for response, but were included in the calculation of response rate.
Randomisation was reported to be carried out by telephone, but details on the method of randomisation were not available. Patients were stratified by age (< 65 vs. ≥ 65 years), ascites (present vs. absent) and prior response to platinum-based therapy (resistant vs. early vs. interim vs. late response). Resistant disease was defined as no response to initial chemotherapy or having an initial PR or CR with subsequent progression while still receiving treatment. Early, interim and late response were defined as initial CR or PR with subsequent relapse within 3 months (early), 3–6 months (interim) or > 6 months (late) after cessation of chemotherapy.
Patients with a CR or PR continued treatment until either progression or 6 months past the maximal response; those who progressed were removed from the study. Those whose best response was SD after six cycles could be removed from the study or switched to the alternative regimen.
Patients on paclitaxel received premedication with dexamethasone and H1- and H2-receptor antagonists to prevent hypersensitivity. No premedication was initially given to those on topotecan but was allowed in subsequent cycles if nausea or vomiting occurred. Dose reductions in each group were permitted for toxicity. The minimum dose allowed was 1.0 mg/m2 per day for topotecan and 135 mg/m2 for paclitaxel; the dose of topotecan could also be escalated to a maximum of 2 mg/m2 per day. Patients were withdrawn from treatment if there was a > 2-week delay in treatment at the minimum dose of either medication because of toxicity. The target dose was achieved in 90% of cycles of topotecan and 98% of cycles of paclitaxel. Median number of cycles received was five in each group, with patients treated with topotecan receiving between 1 and 17 cycles compared with between 1 and 12 cycles for patients treated with paclitaxel.
A sample size calculation was not reported. Median duration of follow-up at the time of the first publication was unclear.21 Median follow-up at the time of the publication reporting more mature OS data was reported in TA91 to be 58.5 weeks in the topotecan group (range 0–86 weeks) and 52.6 weeks in the paclitaxel group (range 0–117 weeks).13
Gemcitabine plus carboplatin compared with carboplatin alone
Pfisterer et al.50 report the results of a Phase III international, open-label randomised study assessing the comparative clinical effectiveness of gemcitabine (1000 mg/m2) plus carboplatin (AUC 4) (n = 178) compared with carboplatin alone (AUC 5; n = 178) in patients with platinum-sensitive recurrent ovarian cancer, with recurrence occurring at least 6 months after completion of first-line platinum-based therapy. Patients were enrolled with measurable or assessable lesions according to Southwest Oncology Group (SWOG) criteria. Exclusion criteria included an ECOG score of > 2, inadequate bone marrow or kidney function or serious concomitant conditions, or life expectancy of < 12 weeks.
The primary outcome of the trial was PFS, with OS, response rate, duration of response, QoL and toxicity measured as secondary outcomes. It should be noted that the study was not powered to detect a difference between treatments in OS. Randomisation was carried out through the central Arbeitsgemeinschaft Gynaekologische Onkologie (AGO)-Ovarian Cancer Study Group (OVAR) office (method of randomisation not reported), with patients randomised at a 1 : 1 ratio. Patients were stratified by PFI (6–12 months vs. ≥ 12 months), first-line therapy (platinum plus paclitaxel vs. other platinum-based therapy) and bidimensionally measurable disease (yes vs. no).
Median duration of follow-up was reported as 17 months, but the range of follow-up was not reported either for the full trial population or the individual treatment groups. Treatment cycles in each group were repeated every 21 days for six cycles, in the absence of PD or unacceptable toxicity. At the investigator’s discretion, benefiting patients could receive a maximum of 10 cycles of therapy. The median number of cycles administered was six cycles in each group. Cycles could be postponed up to 2 weeks owing to toxicity, and longer toxicity-related delays led to treatment discontinuation. For grade 3 non-haematological toxicities (excluding nausea/vomiting), dose modifications and/or study discontinuation were at the investigator’s discretion. Patients in the gemcitabine plus carboplatin arm received 75.6% of the planned mean dose of gemcitabine (92.8% on day 1 and 63.4% on day 8) and 96.2% of the planned dose of carboplatin. Patients in the carboplatin arm received 98.2% of the planned dose.
Paclitaxel plus carboplatin compared with platinum-based therapy alone
Two RCTs were identified for this comparison.48,61 One RCT was a collaboration between the International Collaborative Ovarian Neoplasm (ICON) group and the AGO group and hereafter is referred to as ICON4/AGO-OVAR 2.2 (Parmar et al.61). The RCTs identified were of similar design, but one was a Phase II RCT (Gonzalez-Martin et al.48) and the other was a Phase III RCT (ICON4/AGO-OVAR 2.261).
The ICON4/AGO-OVAR 2.2 trial61 comprised results from two randomised trials that were run in parallel. ICON4/AGO-OVAR 2.261 was an international multicentre trial enrolling 802 patients in 119 hospitals across five countries. ICON4 was co-ordinated by the Istituto di Recerche farmacologiche Mario Negri (IRFMN), and the Medical Research Council’s Clinical Trials Unit (MRC CTU), and AGO-OVAR 2.2 was co-ordinated by AGO. Each co-ordinating unit had its own protocol, with minor differences in eligibility criteria.
All centres enrolled patients with relapsed epithelial ovarian cancer who had previously received platinum-based chemotherapy and had been treatment free for at least 6 months; patients in IRFMN were required to have been treatment free for a minimum of 12 months. The IRFMN and AGO-OVAR 2.2 protocols specified that women were to have received only one prior chemotherapy treatment to be eligible for enrolment, whereas the MRC-CTU protocol permitted women to have received more than one previous line of chemotherapy. Measurable disease at baseline was an entry criteria for patients randomised in centres co-ordinated by the IRFMN, but not MRC CTU or AGO co-ordinated centres. The IRFMN and MRC CTU protocols required that patients have had previous platinum-based chemotherapy, with or without paclitaxel. By contrast, the AGO protocol specified that patients must have previously received cisplatin plus paclitaxel or carboplatin plus paclitaxel. Patients with concomitant or previous malignant disease were ineligible.
The trial compared the clinical effectiveness of paclitaxel plus platinum-based chemotherapy with platinum-based chemotherapy alone. Patients were randomised to receive paclitaxel [175 (ICON4) or 185 (AGO-OVAR 2.2) mg/m2 as a 3-hour infusion] plus platinum chemotherapy (392 patients) or conventional platinum-based therapy (410 patients). Platinum-based therapy comprised carboplatin (AUC 5) or cisplatin (minimum 75 mg/m2 as monotherapy or 50 mg/m2 in combination therapy). In all protocols, cycles were administered every 21 days.
The aim of the study was to evaluate whether paclitaxel should be given in addition to platinum-based chemotherapy in patients with platinum-sensitive disease, who would otherwise be treated with conventional platinum-containing regimens. Randomisation used a computer minimisation method (1 : 1 ratio) and patients were stratified by multiple factors that were determined by the protocol of the assigned centre. In ICON4 protocols, patients were stratified by age, centre, last chemotherapy received, time since last chemotherapy completed and intended platinum treatment. In AGO-OVAR 2.2, patients were stratified by whether the patient had undergone secondary debulking surgery and time since completion of last chemotherapy.
The primary outcome measure of all protocols was OS; secondary outcomes were PFS and QoL. Progression required clinical or radiological evidence of disease (not only raised CA125 level). The sample size calculation found that 800 patients would be sufficient to detect an 11% difference between the groups if the control group survival was 50% at a power of 90% and a 5% significance level.
Median follow-up was 42 months. Of the full trial population, 72% of patients received a minimum of six cycles of assigned chemotherapy; reasons for not completing six cycles included disease progression or death, toxicity or patient preference.
Gonzalez-Martin et al.48 reported the results of a Phase II study, in which 81 patients with platinum-sensitive recurrent ovarian cancer, who had received no more than two previous chemotherapy lines, were randomised to receive carboplatin alone (AUC 5; 40 patients) every 21 days or paclitaxel (125 mg/m2 of > 3 hours) plus carboplatin (AUC 5; 41 patients) every 21 days. Patients had to have measurable disease as measured by CT or clinically evident but non-measurable disease that was evaluable by CA125 level, based on Rustin’s criteria. Patients who had an ECOG performance status of > 2, life expectancy of < 12 weeks or inadequate bone marrow, liver or kidney function were ineligible.
The primary outcome measure was ORR (CR or PR), which was evaluated using WHO criteria in those with measurable disease, or by CA125 level, according to Rustin’s criteria, in those without measurable disease. OS, TTP and QoL were reported as secondary outcome measures.
Both treatments were administered for a minimum of six cycles unless there was progression, unacceptable toxicity or a patient refused treatment. After six cycles, patients could continue for three further cycles if clinical benefit could be expected. All patients randomised to receive paclitaxel were treated with standard premedication 30 minutes before infusion, which comprised dexamethasone, diphenhydramine and ranitidine. In cases of grade 4 neutropenia or thrombocytopenia, doses were reduced to carboplatin AUC 4 (both groups) and paclitaxel 150 mg/m2.
Randomisation was reported to have been carried out by a central data centre (no further details reported). Patients were stratified by PFI [6–12 months (PPS) compared with > 12 months (FPS)] and number of previous chemotherapy lines (one vs. two). The trial was an unusual ‘pick up the winner’ design. The authors of the trial comment that this type of design has a ‘90% chance of selecting the better treatment if the difference is at least 15% and the smaller response rate is assumed to be 30%’. A sample size calculation is not presented, but the authors state that the trial was not designed or powered to detect differences in survival. The authors go on to comment that ‘no formal statistical comparison between the two arms was planned’, but a selection of statistical comparisons are reported ‘for exploratory purposes only’.
Patients in both treatment groups received a median of six cycles of treatment, with between two and nine cycles of carboplatin alone administered and between one and eight cycles of paclitaxel plus carboplatin administered. Three patients in the paclitaxel plus carboplatin arm did not receive one cycle of treatment. The proportion of patients requiring a dose reduction was small and was similar between the groups (4.7% with carboplatin alone vs. 6.6% with paclitaxel plus carboplatin). By contrast, a significantly larger proportion of patients required a dose delay in the carboplatin alone group (34.4%) compared with the paclitaxel plus carboplatin group (21%; p-value for difference < 0.006). The difference was attributed to the absence of haematological recovery by day 21 in the group receiving carboplatin alone.
The three patients who received no treatment in the paclitaxel plus carboplatin group were included in the ITT analysis of overall response but were excluded from other analyses. Median duration of follow-up was 67.7 weeks. At this time point, 32 patients had died and median OS has not been reached in the paclitaxel plus carboplatin group. The range of follow-up was not reported either for the full trial population or the individual treatment groups.
Paclitaxel plus carboplatin compared with paclitaxel alone
Lortholary et al.62 reported the results of a Phase II, multicentre, open-label, three-armed randomised trial (CARTAXHY) in patients with PRR recurrent ovarian cancer. Eligible patients were those who had received at least one prior therapy, with the most recent regimen combining platinum with a taxane agent. In addition, patients were required to have either measurable (according to RECIST criteria) or CA125 assessable disease, an ECOG performance status of ≤ 2 and a life expectancy of > 12 weeks. Patients with measurable disease (according to RECIST criteria) or evaluable disease (CA125) were enrolled. Patients who had previously been treated with weekly paclitaxel were excluded.
In total, 165 patients were randomised (1 : 1 : 1 ratio) to treatment with weekly paclitaxel (80 mg/m2 administered on days 1, 8 and 15 of a 4-week cycle; 57 patients), weekly paclitaxel plus carboplatin (AUC 5 administered on day 1 of a 4-week cycle; 51 patients) or weekly paclitaxel plus weekly topotecan (3 mg/m2 administered on days 1, 8 and 15 of a 4-week cycle; 57 patients). The combination of paclitaxel plus topotecan in the treatment of patients with PRR ovarian cancer is not covered by the scope of this review, and the efficacy results for this group are not presented.
The primary outcome was PFS. Secondary end points were response rate, OS, QoL and toxicity. QoL was assessed using the EORTC Quality of Life Questionnaire (EORTC-QLQ) and toxicity was assessed according to NCI-CTC. The efficacy analyses were based on the ITT principle.
Randomisation was carried out at the Group d‘Investigateurs Nationaux pour I’Etude des Cancers Ovariens (GINECO) data centre but details on the method of randomisation are not available. Patients were stratified according to centre, TFIs (progression during treatment vs. relapse between 0 and 3 months vs. relapse at > 3 months and ≤ 6 months), and presence of a measurable lesion at baseline.
Treatments were administered for six to nine cycles or until progression or unacceptable toxicity. On progression, patients treated with weekly paclitaxel or weekly paclitaxel plus weekly topotecan received carboplatin (AUC 5) and patients treated with weekly paclitaxel plus carboplatin went on to receive treatment of physician’s choice. One patient in the weekly paclitaxel group did not receive any treatment. Patients received a median three cycles in each group.
Dose reductions for toxicity of one level were to paclitaxel 65 mg/m2, carboplatin AUC 4 mg/ml/minute, and topotecan 2.4 mg/m2. Dose reductions of two levels were to paclitaxel 5 mg/m2, carboplatin AUC 3.5 mg/ml/minute, and topotecan 2 mg/m2. In cases in which there was a treatment delay of > 2 weeks, patients were discontinued from the study.
The sample size calculation indicated that 165 patients would be required for adequate power to detect a difference among groups with 80% power. Median duration of follow-up was 15 months.
Paclitaxel compared with oxaliplatin
Piccart et al.63 report the results of a multicentre (17 European centres across six countries), open-label, randomised, Phase II study. Patients were enrolled who had histologically or cytologically proven advanced ovarian cancer that had progressed or stabilised after prior treatment, with relapse occurring within 12 months of the last platinum-based chemotherapy regimen. No more than two prior cisplatin- and/or carboplatin-containing chemotherapy regimens were permitted. Patients were also ineligible if they had prior treatment with platinum derivatives other than cisplatin and carboplatin or with paclitaxel, docetaxel, or high-dose chemotherapy with haematopoietic stem cell support.
The primary aim of the trial was to evaluate the clinical effectiveness of oxaliplatin (Eloxatin®, Sanofi) (130 mg/m2 over 2 hours) every 21 days (n = 45) compared with paclitaxel (175 mg/m2 over 3 hours) every 21 days (n = 41).
Patients were assigned to their study group by the EORTC. No details on the method of randomisation are reported in the full publication. Patients were stratified by centre, performance status (0 vs. 1 vs. 2), PFI (0–6 months vs. 6–12 months) and number of prior platinum-based regimens (1 vs. 2). The primary outcome measure was the objective confirmed response rate, which was assigned as per WHO criteria and verified by two independent radiologists. Secondary outcome measures were TTP, OS, time to treatment failure and QoL.
For patients randomised to receive paclitaxel infusion, premedication included oral dexamethasone (20 mg) 12 and 6 hours before infusion, and diphenhydramine (50 mg intravenously) plus cimetidine (300 mg; Tagamet®, GSK) or ranitidine (50 mg intravenously) 30 minutes before the infusion. Antiemetic therapy before oxaliplatin infusion was a serotonin antagonist (5-HT3), with a single dose of corticosteroid (e.g. dexamethasone 20 mg).
Treatment in each group was continued until disease progression, unacceptable toxicity or patient refusal. The initial paclitaxel and oxaliplatin doses could be reduced in subsequent cycles, or the cycles could be delayed by 1 or 2 weeks, depending on toxicity. Dose reduction of below 90 mg/m2 for paclitaxel and 75 mg/m2 oxaliplatin per cycle was not permitted, and patients requiring these or lower doses went off study. The median number of cycles of treatment was six (range 1–8) in the paclitaxel group and four (range 1–8) in the oxaliplatin group. Most patients had a delivered relative dose intensity of at least 95%.
Median duration of follow-up was not reported. A total of five patients were not assessable for response (two in the paclitaxel arm and three in the oxaliplatin arm): four were ineligible because of eligibility deviations and one died 6 days after the first oxaliplatin cycle, as a result of a massive pulmonary thromboembolism (unrelated to treatment).
A sample size calculation was not reported. The authors comment that, despite the use of several centres, as a result of wider use of paclitaxel as a first-line treatment at the time the trial was initiated, accrual of paclitaxel-naive patients became slow in the later stages of the trial. It is unclear whether the trial was adequately powered to detect a difference between treatments.
Topotecan oral compared with topotecan intravenous
Gore et al.24 report the results of a multicentre, international (Europe, South Africa and North America) randomised trial of open-label design that compared topotecan administered orally (2.3 mg/m2 daily for five consecutive days; n = 135) with intravenously (1.5 mg/m2 daily for five consecutive days; n = 131). Both treatment regimens were given on a 21-day cycle. Patients were enrolled who had relapsed epithelial ovarian cancer (histological diagnosis) that was measurable at baseline and was of FIGO stage III or IV (266 patients randomised). To be eligible, patients were also required to have an ECOG score of ≤ 2. Patients had either progressed during or relapsed within 12 months of completing first-line chemotherapy, and only one prior chemotherapy regimen was permitted. Initial treatment must have been platinum based and could have been given in conjunction with a taxane.
The aim of the study was to compare the efficacy, safety and tolerability of oral topotecan compared with standard i.v. topotecan in patients with relapsed ovarian cancer. Randomisation (1 : 1 ratio) was carried out by telephone (no further details reported) and stratified by prior taxane exposure, interval from previous platinum therapy and tumour diameter (< 5 cm vs. ≥ 5 cm). Three categorisations of response to first-line chemotherapy were defined: platinum refractory (PD or SD during initial chemotherapy); platinum resistant (initial response followed by relapse within 6 months); and platinum sensitive (initial response with subsequent relapse at > 6 months).
Outcomes assessed included response rate (as per WHO criteria), time to response, TTP, survival and toxicity. Median follow-up was not stated. Although open label in design, all responses that were claimed to be confirmed or partial were subject to independent, blinded radiological review. The only outcome evaluated for the subgroups categorised by extent of sensitivity to platinum was response rate.
Duration of treatment with topotecan was determined by response to therapy and was at the discretion of the clinician. It was recommended that patients with SD receive a minimum of four cycles of treatment and that patients responding to treatment receive at least two cycles of treatment beyond response. Patients assigned to oral topotecan received a median of four (range 1–23) cycles and those assigned i.v. topotecan received a median of six (range 1–26) cycles. Dose reductions were permitted for grade 3 or 4 AEs, with about 10% of patients in each group requiring a reduction in dose.
Topotecan administered on five consecutive days (conventional regimen) compared with topotecan administered weekly
Sehouli et al.23 report the results of a randomised, multicentre, Phase II trial in Germany involving 194 patients with platinum-resistant recurrent epithelial ovarian or primary peritoneal cancer after radical surgery and at least one platinum-containing chemotherapy. Patients with disease measurable by CT or MRI, or disease evaluable by CA125 according to GCIG criteria, were eligible. Platinum resistance was defined as clinical disease progression after a TFI of < 6 months after a platinum-based regimen. Inclusion criteria with regards to number of previous lines of chemotherapy were not specified.
The primary goal of the trial was to compare weekly administration of topotecan at a dose of 4.0 mg/m2 each week, applied on days 1, 8 and 15 of a 28-day cycle (n = 97) compared with the conventional regimen of 1.25 mg/m2 for five consecutive days (n = 97). The rationale for the trial was that weekly administration of topotecan is considered to be less toxic and is widely used in clinical practice, despite the lack of an evidence base of effectiveness. It should be noted that the dose used in the ‘conventional’ 5-day regimen is lower than the dose recommended in the Summary of Product Characteristics (SmPC) for topotecan (1.5 mg/m2).
Randomisation was central with permutated blocks in a 1 : 1 ratio and was carried out by phone and facsimile. However, the level of masking in the trial is unclear. The primary outcome evaluated was the clinical benefit rate, which was defined as the composite of CR, PR and SD. Response was determined according to RECIST for measurable disease or GCIG criteria for serum CA125 levels. Use of the CA125 marker or scans to evaluate response was at investigators’ discretion, with all responses confirmed by a second examination. Secondary end points were toxicity, PFS and OS; QoL was also explored. All analyses were based on the ITT principle. No sample size calculation was reported but it is stated that the study was not powered for a direct comparison between the dosing schedules or to reveal differences in response rates.
Treatment in each group was continued until intolerable toxicity or disease progression or until the patient refused further therapy, with maximum treatment duration of 12 months. Dose of topotecan could be reduced by 25% for any grade 3 or 4 adverse effects according to the NCI-CTC.
Median follow-up was 23.4 months (range 12.7–41.4 months). Patients in the weekly topotecan group received statistically significantly fewer cycles of chemotherapy than the group receiving topotecan at the conventional dosing regimen (3.5 with weekly topotecan vs. 4.8 with conventional topotecan; p = 0.002).
Paclitaxel high dose (250 mg/m2) compared with paclitaxel standard dose (175 mg/m2)
Omura et al.68 conducted a Phase III, randomised, multicentre trial comparing two doses of paclitaxel (250 mg/m2 vs. 175 mg/m2) involving patients with recurrent or persistent histologically confirmed epithelial ovarian cancer despite prior platinum therapy. A third group, paclitaxel 135 mg/m2, was closed early because of inadequate patient accrual. Eligible patients had received not more than one prior platinum-based regimen, had adequate bone marrow, kidney and liver function; and a Gynecologic Oncology Group (GOG) performance status of 0, 1 or 2.
The aim of the trial was to evaluate whether increasing dose of paclitaxel was associated with an increase in response. The primary outcome measures were PFS and OS. Objective response (CR or PR) rates were recorded in patients with measurable disease (pleural effusion or elevated CA125 level were not regarded as measurable disease). The study also assessed whether prophylactic filgrastim 10 µg/kg was more effective than filgrastim 5 µg/kg at reducing the incidence of febrile neutropenia in patients receiving paclitaxel 250 mg/m2. The TAG considers that the administration of filgrastim is unlikely to influence comparative clinical effectiveness.
Sequential, permuted block randomisation was used to assign patients to paclitaxel 175 or 250 mg/m2 by 24-hour i.v. infusion every 3 weeks. Both treatments were administered for a minimum of six cycles. Patients could continue treatment indefinitely if there was no clinical progression or excessive toxicity after six cycles. Paclitaxel dose intensity could be reduced for some grade 3 or greater toxicities (not otherwise specified). Patients experiencing neutropenic fever while receiving paclitaxel 175 mg/m2 were allowed filgrastim during subsequent therapy cycles.
Based on the sample size calculation, it was estimated that 540 patients, followed until approximately 80% had died, would provide an 80% chance of detecting a true HR of 1.4 between paclitaxel 135 mg/m2 and either of the more intense regimens (type I error p = 0.025 for one-tailed test). However, the study failed to enrol sufficient patients in the paclitaxel 135 mg/m2 arm and a decision was made to ‘allocate all of the type I error to the comparison of the two higher-dose regimens’. Initially designed to evaluate effects of the two paclitaxel regimens in platinum-resistant clinically measurable disease, owing to slow accrual, after commencement of the trial, the eligibility criteria were expanded to include patients with platinum-sensitive disease and without clinically measurable disease.
Of the 184 women randomly assigned to paclitaxel 175 mg/m2 and the 188 to paclitaxel 250 mg/m2, 164 (89%) and 166 (88%), respectively, were eligible. Ten eligible women (three in the paclitaxel 175 mg/m2 group and seven in the paclitaxel 250 mg/m2 group) were not assessed for tumour response because of death, toxicity or withdrawal but were classified as not responding for an ITT analysis among eligible patients. The primary survival outcomes were restricted to eligible patients.
Median duration of follow-up is not reported. The proportion of women receiving six or more cycles of therapy was similar between the two groups, with 58% and 55% of patients in the paclitaxel 175 mg/m2 and paclitaxel 250 mg/m2 group, respectively, receiving six or more cycles. One patient refused to take any dose of the allocated treatment.
Paclitaxel weekly compared with paclitaxel every 3 weeks
Rosenberg et al.60 report the results of a randomised bifactorial multicentre study carried out at sites in Sweden and Finland. The aim of the study was to assess the efficacy and toxicity of paclitaxel given at the same dose intensity administered either weekly or every 21 days. Patients were randomised to paclitaxel 67 mg/m2 every 7 days or paclitaxel 200 mg/m2 every 21 days. Enrolled patients (n = 208) had advanced ovarian cancer (histologically proven) that had progressed during or relapsed after administration of a platinum-based regimen. To be eligible, patients had to have measurable disease that had been documented clinically and/or radiologically. Only one prior platinum-containing regimen was permitted. In addition, all patients were taxane naive.
The RCT was of a bifactorial design. In addition to randomisation to either paclitaxel weekly or every 21 days, patients were also randomised to oral dexamethasone (20 mg) taken 12 hours and 6 hours before paclitaxel infusion or administration of i.v. dexamethasone (20 mg) 30 minutes before paclitaxel infusion. Results in the full publication cited here focus on treatment with paclitaxel. Premedication with clemastine 2 mg (Tavegel®, Novartis) and cimetidine 300 mg (or ranitidine 50 mg) was given intravenously to all patients 30 minutes prior to paclitaxel infusion.
The primary endpoint of the study was clinical response rate as per WHO criteria, with TTP and OS evaluated as secondary outcomes. Randomisation was reported to have been carried out at the BMS office in Stockholm and patients were randomised in a 1 : 1 ratio. Patients were stratified by platinum resistance, with a differentiation at 6 months (randomisation strata: relapse at ≤ 6 months vs. > 6 months after primary platinum-based treatment). No further details on the method of randomisation are reported. The level of masking in the trial is unclear.
Median duration of follow-up was 27 months (range 7–47+ months). Patients to whom paclitaxel was administered weekly at a dose of 67 mg/m2 received a median of 5.7 courses of treatment (range 1–16 courses) compared with a median of seven courses in the group receiving paclitaxel 200 mg/m2 every 21 days (range 1–17 courses). More patients in the paclitaxel weekly arm (32 vs. 20) were taken off the study early (within 9 weeks) owing to either early progression or for administrative reasons. The difference in early progressions could be because of a low initial weekly dose or some patients may have had a more aggressive tumour biology.
The sample size calculation estimated that 318 patients would be required to detect the prespecified relative difference between groups of 54% with 80% power. To ensure a sufficient number of evaluable patients, it had been planned to recruit a total of 350 patients. Owing to slow recruitment of taxane-naive patients with recurrent disease, the study closed early after inclusion of 208 patients. The study may therefore have been underpowered to detect a difference between the two regimens.
Quality assessment of studies included in clinical effectiveness review
Pegylated liposomal doxorubicin hydrochloride plus carboplatin compared with paclitaxel plus carboplatin
The trial carried out by Bafaloukos et al.29 is generally well designed with the primary analysis based on the ITT population. However, limited details on trial methodology are provided in the full publication. Randomisation is reported to have been carried out at the central HeCOG Data Office in Athens but a description of the method of randomisation is not reported. The level of masking within the trial is unclear. The primary outcome is response rate, which is determined by radiological scan or CA125 level. Assessment of response is associated with disparity in interpretation of scan results, both across different assessors and within categorisation (CR or PR) by an individual assessor. It is unclear whether radiological scans were evaluated by an independent review panel. In addition, TTP was measured from date of treatment initiation rather than date of randomisation, which is a more commonly used definition for TTP in clinical trials. The evaluation of the quality of the trial is presented in .
Summary of quality assessments of studies included in review of clinical effectiveness
The CALYPSO trial31 is a well-designed and well-conducted trial. Progression and response were reviewed independently. Although the methods indicate that analyses are based on the ITT principle, three randomised patients (one in the PLDH plus carboplatin group and two in the paclitaxel plus carboplatin group) were judged to be ineligible because of absence of evidence of ovarian cancer post randomisation and were excluded from analyses of clinical effectiveness. Thus, the analyses are not strict ITT analyses. The evaluation of the quality of the trial is presented in .
Pegylated liposomal doxorubicin hydrochloride plus carboplatin compared with carboplatin alone
The Alberts et al.28 trial seems to be generally well designed, although limited details on the methodology of the trial are provided in the full publication. The method of randomisation and level of masking are unclear. As the primary outcome is OS, masking, or lack of masking, is unlikely to introduce bias into the evaluation of treatment effect. The key issue associated with trial design is that the study is likely to have been underpowered as a result of early closure owing to slow patient accrual (61 patients recruited out of a planned 900 patients). The authors identify several factors that could have contributed to slow accrual, including dissolution of the SWOG Gynecologic Cancer Committee after initiation of the trial and publication of results from the larger ICON4/AGO-OVAR 2.2 trial.61 The evaluation of the quality of the trial is presented in .
Trabectedin plus pegylated liposomal doxorubicin hydrochloride compared with pegylated liposomal doxorubicin hydrochloride alone
The OVA-301 trial30 was a well-conducted trial. Methodologically, the design of the trial was robust, with clinical effectiveness analyses based on the ITT population, and progression and response reviewed by an independent radiologist who was masked to treatment allocation. A secondary analysis of the primary outcome of PFS was carried out based on review by an independent oncologist (radiological assessment in conjunction with prespecified clinical data) who was also masked to treatment allocation. The methods of the trial are well reported. As noted in the Final Appraisal Determination (FAD) for the assessment of trabectedin plus PLDH as part of the Technology Appraisal process (TA222),73 one potential area that affects the external validity of the trial is the omission of a platinum-based chemotherapy as a comparator, particularly as a large proportion of patients enrolled had platinum-sensitive disease. The authors commented that the inclusion of platinum-resistant patients contributed to the decision against use of a platinum-based control, as platinum-based therapy would have been inappropriate in this setting. The evaluation of the quality of the trial is presented in .
Pegylated liposomal doxorubicin hydrochloride compared with topotecan
The trial carried out by Gordon et al.49 was generally a well-designed trial. Although open label in design, scans for assessment of disease response and progression underwent independent radiological review. Although the methods state that analyses are based on the ITT principle, in the first publication, results are based on patients who received at least a partial dose of study drug (474 patients out of 481 randomised), which is a modified ITT analysis. However, in the publication describing longer-term follow-up of OS, analysis of OS is based on the ‘all randomised’ population and, as such, is a true ITT analysis. The evaluation of the quality of the trial is presented in .
Pegylated liposomal doxorubicin hydrochloride compared with paclitaxel
Technology appraisal no. 9113 reports that the study carried out by Schering-Plough (Trial 30–57) was a reasonably good-quality, randomised, open-label comparative trial. The key issue noted was that approximately 50% of the planned number of patients were recruited (216 recruited out of planned 438 patients). It is therefore likely that the trial is underpowered to detect a difference between PLDH and paclitaxel in treatment effect. TA91 also notes that the results of the trial ‘are likely to be preliminary and the longer term implications of any differences observed in the treatment effect at the time of data analysis are unclear’. The evaluation of the quality of the trial is presented in .
Topotecan compared with paclitaxel
A key strength of the trial evaluating topotecan compared with paclitaxel (ten Bokkel Huinink et al.21) is that, for the primary outcome of response rate, all claimed responses were evaluated by an independent radiologist who was masked to treatment allocation. As a sample size calculation was not reported, there is uncertainty whether the trial was adequately powered to detect a difference between treatments. Furthermore, results are not based on the ITT principle, with only patients who received at least one dose of study drug being included in the final analysis. The trial design allowed patients to cross over to the alternative treatment should they fail to respond to their allocated treatment. The switch in treatment during the trial generates confounding in the final analysis of OS. The evaluation of the quality of the trial is presented in .
Gemcitabine plus carboplatin compared with carboplatin alone
The trial carried out by Pfisterer et al.50 is generally a well-designed and well-conducted trial, with efficacy analyses based on the ITT principle. With PFS as a primary outcome and an open-label design, there is potential for bias. It is unclear from the full publication whether radiological assessment of progression was reviewed by an independent panel. The evaluation of the quality of the trial is presented in .
Paclitaxel plus carboplatin compared with platinum-based therapy alone
ICON4 and AGO-OVAR 2.2 are well-conducted parallel trials.61 Comprehensive details on most aspects of trial methodology are provided in the full publication.61 The level of masking is unclear but OS is the primary outcome and therefore awareness of treatment allocation is unlikely to influence results of this outcome. Analyses of clinical effectiveness are based on the ITT population. The evaluation of the quality of the trial is presented in .
The trial carried out by Gonzalez-Martin et al.48 was a Phase II trial of a ‘pick-the-winner’ design, which the authors state has a ‘90% chance of selecting the better treatment if the difference is at least 15% and the smaller response rate is assumed to be 30%’. Therefore, no sample size calculation was carried out. A ‘pick-the-winner’ trial is designed as a screening trial to facilitate a selection between promising experimental regimens in a Phase II setting, and, as such, do not typically include the standard of care. Trials with a ‘pick the winner’ design are underpowered for hypothesis testing or comparisons of treatment effect on the outcomes of interest, such as survival.74 Therefore, as the authors comment, all reported statistical analyses are exploratory and reported p-values should be interpreted with caution. Limited details on trial methodology are reported and the level of masking in the trial is unclear. Although it is reported that randomisation was carried out in a central data centre, the method of randomisation is not described. The evaluation of the quality of the trial is presented in .
Paclitaxel plus carboplatin compared with paclitaxel alone
Limited details of the methodology of the CARTAXHY trial62 are available in the publication presenting the results of the trial. A key strength of the trial is that clinical efficacy analyses were based on the ITT principle. Although it is stated that the study is randomised, details on the method of randomisation are not reported. As an open-label trial, there is potential for bias in the assessment of progression and response. It is unclear whether radiological scans underwent independent radiological review. The evaluation of the quality of the trial is presented in .
Paclitaxel compared oxaliplatin
The trial reported by Piccart et al.63 is generally a well-designed trial. The primary outcome was objective confirmed response. As an open-label design, the outcome of confirmed response could potentially be open to bias. The descriptions of the methods states that response was verified by two independent radiologists. However, it is unclear whether the independent radiologists were truly independent and masked to treatment allocation. Although limited details are reported on the method of randomisation, it is reported that the treatment allocation was assigned by the EORTC. The key issue associated with the trial is the uncertainty around the power of the trial. The evaluation of the quality of the trial is presented in .
Topotecan oral compared with topotecan intravenous
The trial reported by Gore et al.24 is generally well designed, with analysis based on the ITT population. In addition, although open label in design, assessments of CR and PR were validated by masked independent radiological review.
It is stated that randomisation was carried out by telephone but no details on the method of randomisation are reported. No power calculation is reported and thus it is unclear whether the study is adequately powered. The population is clinically homogeneous in that all patients randomised had measurable disease at baseline and also had received only one prior chemotherapeutic treatment. The evaluation of the quality of the trial is presented in .
Topotecan administered on five consecutive days (conventional regimen) compared with topotecan administered weekly
There are several factors that impact on the quality of the design and conduct of the trial carried out by Sehouli et al.23 Although it is stated that all analyses are carried out based on the ITT principle, the analysis of clinical benefit does not include all patients randomised. There is no discussion of the omission of patients from this analysis. The dose used for the ‘conventional’ regimen for topotecan is lower than that recommended in the SmPC. The authors comment that the reduced dose is widely accepted by many international cancer societies but go on to highlight that there are no RCTs evaluating the comparative effectiveness of 1.25 mg/m2 compared with 1.5 mg/m2 of topotecan. In addition, use of radiological scans or CA125 level to determine response was at the discretion of the investigator. It is widely accepted that CA125 level is not sufficient to confirm response to treatment. Examination of the results for response indicates that a large proportion of patients were evaluated by CA125 level alone (80.1%). Moreover, it is unclear whether the investigator was masked to treatment allocation. Although responses had to be confirmed by a second examination, it is unclear whether response was confirmed by the same investigator or by independent review. The trial was not adequately powered to detect a difference between groups. These factors potentially limit the comparison of the results from this trial with similar trials in ovarian cancer. The evaluation of the quality of the trial is presented in .
Paclitaxel high dose (250 mg/m2) compared with paclitaxel standard dose (175 mg/m2)
Limited methodological details were reported in Omura et al.68 The method of randomisation was robust, with treatment regimens sequentially assigned from stratified, permuted blocks. The level of masking in the trial is unclear. Although the methods state that the analyses are based on the ITT principle, patients identified post randomisation to be ineligible for participation in the trial were excluded from all analyses, and, therefore, analyses are not based on the ITT population. A key issue with the trial is the sample size, with only 265 patients recruited from a planned 540, even after expansion of the protocol to include platinum-sensitive patients and those with measurable disease. Thus, the study is likely to be underpowered to detect a true difference between the treatment regimens for which results are reported.
Paclitaxel weekly compared with paclitaxel every 3 weeks
The trial carried out by Rosenberg et al.60 is of reasonable quality. Efficacy analyses are based on the ITT principle. Limited details are reported on trial methodology in terms of method of randomisation and level of masking. The key issue with the trial is that it is potentially underpowered to detect a difference in the primary outcome of response rate between the paclitaxel regimens evaluated. The evaluation of the quality of the trial is presented in .
Comparability of baseline characteristics
Within most of the trials identified, the treatment groups were well matched in terms of population baseline characteristics, including age, TFI, the number of previous chemotherapy agents received, disease measurability (for those trials including patients with measurable and non-measurable disease), and performance status. Differences between groups that were reported to be significant are described below; imbalances that were reported to be non-significant or for which the significance of the difference was not assessed in the trial are not discussed. Detailed baseline characteristics of the individual trials are available in the data abstraction forms presented in Appendix 2.
An unanticipated imbalance in PFI was noted in a retrospective analysis of OVA-301.30 Patients in the PLDH monotherapy group had a significantly longer mean PFI than patients in the trabectedin plus PLDH group (mean PFI: 13.3 months with PLDH alone vs. 10.6 months with trabectedin plus PLDH; p = 0.009). Longer PFI is correlated with increased likelihood of response to treatment. Therefore, the potential direction of bias in analysis of treatment effect is against trabectedin plus PLDH. To account for this imbalance, the authors carried out additional exploratory analyses based on PFI as a continuous covariate (discussed in Chapter 4, Trabectedin for the treatment of patients with relapsed platinum-sensitive ovarian cancer). The analyses were not prespecified and as such were hypothesis generating.
Baseline characteristics of key prognostic factors (based on expert advice) are summarised in . Also, based on expert advice, the TAG has focused on the subgroups of platinum-sensitive ovarian cancer and PRR ovarian cancer rather than the full trial population. Baseline characteristics are considered in terms of comparability within platinum-sensitive patients and PRR patients.
Considering patients with platinum-sensitive disease, a potential source of heterogeneity within the trials is the proportion of patients with FPS (relapse at > 12 months after last platinum-based treatment) ovarian cancer compared with PPS (relapse at ≥ 6–12 months after last platinum-based treatment) at baseline. The greater the duration of PFI, the more favourable the prognosis. In trials involving patients with only platinum-sensitive disease,28,29,31,48,50,61 the proportion of patients with PPS ovarian cancer ranges from 28.6% to 43.0%. Considering the large trial ICON4/AGO-OVAR 2.2,61 the proportion of patients with PPS compared with FPS is 74.7% and 25.3%, respectively. ICON4/AGO-OVAR 2.261 has been reported to have longer median PFS and OS for both groups compared with other trials involving platinum-sensitive patients, which is thought to be attributable to the comparatively larger proportion of patients with FPS ovarian cancer who have an improved prognosis compared with those who are PPS. Given that the NMA is based on relative treatment effects (HR), and that most trials are well balanced between groups in FPS ovarian cancer compared with PPS, the TAG considered the trials sufficiently clinically homogeneous to compare treatments in a NMA.
Number of prior lines of chemotherapy is another source of potential heterogeneity. Increasing number of previous chemotherapy regimens is associated with a decrease in response to treatment. Of the 16 trials identified, seven included patients who had received two or more prior lines of chemotherapeutic treatment. In trials involving only patients with platinum-sensitive disease, the proportion of patients with more than one line of prior chemotherapy in each trial is generally small, ranging from 4% to 15.5%. By contrast, as could be expected, in trials involving patients with PRR ovarian cancer, the proportion of patients with two or more chemotherapy regimens is larger, at about 30% in all trials. In all trials, the number of patients with multiple lines of prior chemotherapy is well balanced within the trial. It is possible that inclusion of trials in which patients received two or more chemotherapy regimens is likely to underestimate the effects of the evaluated treatments in patients with first recurrence of disease, and thus potentially bias the results of an indirect comparison towards treatments that are given as second line. Again, as the HR used in the NMA is a relative treatment effect, the impact of these trials on the overall result could be minimal.
Scales evaluating performance status are used to assess disease progression and how a patient’s daily living abilities are affected by their disease. On the ECOG scoring system (also referred to as the Zubrod or WHO score), the lower a patient’s performance score, the greater their capacity for physical activity; a score of ‘0’ or ‘1’ indicates that the patient is ambulatory. In the Karnofsky scale, which scores from ‘100 to 0’, higher performance score is favourable: a score of > 80 indicates that a patient is able to carry on normal activity and to work with no special care required. Good performance status has been shown to be an important prognostic factor in several types of cancer.75
In the identified trials, the proportion of patients with unfavourable baseline performance score (ECOG/Zubrod/WHO score of ≥ 2; Karnofsky score of < 80) is small, ranging from 0% to 16% across the trials. Including patients with less favourable performance scores is likely to underestimate the effect of the treatments. For example, in Gonzalez-Martin et al.,48 the 12.3% increase in proportion of people with ECOG score 2 in the platinum treatment group may limit the benefit received by people receiving platinum monotherapy when compared with paclitaxel plus platinum (i.e. paclitaxel plus platinum may have less benefit over platinum monotherapy). In addition, in Rosenberg et al.60 the 9% increase in WHO score 2 in the three-weekly paclitaxel group may limit the benefit received by three-weekly paclitaxel monotherapy compared with weekly paclitaxel monotherapy (i.e. the benefit of paclitaxel weekly may have less benefit over three-weekly paclitaxel). However, in those trials that include patients with a less favourable performance score, the proportion of patients in each treatment group is well balanced and thus the impact on the overall result could be minimal.
Diagnosis of recurrent disease based on raised CA125 levels alone has been found to predate evidence of disease progression from clinical examinations or radiological scans by a median of 4 months in 70% of patients with ovarian cancer.76 Thus, there is uncertainty whether patients diagnosed as having recurrent disease by only CA125 level would have the same diagnosis on radiological scan. In addition, it is also possible that the degree of sensitivity to platinum could differ. For example, based on CA125 level alone, a patient could be categorised as PPS at baseline but as FPS 4 months later with radiological confirmation. Of the trials identified, seven RCTs23,28,29,31,48,62,68 reported that patients with only CA125 level as an indicator of recurrent disease were enrolled. In trials in patients with platinum-sensitive disease, there was considerable variation across the trials in the proportion of patients with non-measurable disease at baseline, ranging from 8.5% to 38.2%. In some trials, patients with non-measurable disease were not included in analyses of response rate. Despite the identified disparity in methods used to diagnose recurrent disease at baseline, as the proportion of patients in each group within the individual trials was well balanced, the TAG considered that the heterogeneity could have a minimal impact on the NMA.
Considering heterogeneity among treatments evaluated, it is important to note that ICON461 evaluated the efficacy of adding paclitaxel to ‘conventional’ platinum-based chemotherapy compared with platinum-based therapy alone. A large proportion of patients in each treatment group received carboplatin as the platinum component of their regimen (80% in the paclitaxel plus platinum-based therapy group vs. 29% in the platinum-based chemotherapy alone group). Of the remaining 20% of patients in the paclitaxel plus platinum group, 10% were administered cisplatin, and 5% received paclitaxel plus carboplatin or cisplatin, switching between the two platinum monotherapies. In the conventional platinum-based monotherapy group, 4% of patients received cisplatin alone, and a further 2% received either carboplatin or cisplatin monotherapy, switching between the two platinum monotherapies. Moreover, 17% of patients in the conventional chemotherapy group received the triple therapy of cyclophosphamide, doxorubicin and cisplatin, which the ICON investigators had compared against carboplatin in an earlier trial and found no statistically significant difference between the treatments in effect on OS.61 Although a small proportion of patients received platinum treatment other than carboplatin, there is evidence that the regimens received have similar efficacy.
Although differences in key prognostic factors across the trials have been identified, when considering the trials that would inform the NMA for platinum-sensitive disease and for PRR disease, the TAG considers the trials sufficiently clinically homogeneous to compare clinical effectiveness of treatments.
