Chemotherapy for metastatic and recurrent cervical cancer

Kate Scatchard; Jennifer L Forrest; Maxine Flubacher; Paul Cornes; Chris Williams

doi:10.1002/14651858.CD006469.pub2

. 2012 Oct 17;2012(10):CD006469. doi: 10.1002/14651858.CD006469.pub2

Chemotherapy for metastatic and recurrent cervical cancer

Kate Scatchard ^1,^✉, Jennifer L Forrest ², Maxine Flubacher ³, Paul Cornes ⁴, Chris Williams ⁵

Editor: Cochrane Gynaecological, Neuro‐oncology and Orphan Cancer Group

PMCID: PMC6457617 PMID: 23076924

Abstract

Background

Cervical cancer is the second most common cancer among women up to 65 years of age and is the most frequent cause of death from gynaecological cancers worldwide. A woman's risk of developing cervical cancer by 65 years of age ranges from 0.69% in developed countries to 1.38% in developing countries. Although screening by Pap smear should mean early detection at a curable stage for most women, many still present with advanced or metastatic disease with a worse prognosis. The addition of platinum‐based chemotherapy to radiotherapy has improved outcome compared to radiotherapy alone; however, 30% to 50% fail to respond to treatment or develop recurrent disease. There are no standard treatment options for these patients, although platinum‐based chemotherapy is frequently used and trials are on‐going.

Objectives

To compare different types and combinations of cytotoxic chemotherapy for the treatment of metastatic/recurrent cervical cancer.

Search methods

We searched the Cochrane Gynaecological Cancer Group Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL, Issue 1, 2012), MEDLINE (1950 to January 2012) and EMBASE (1980 to January 2012). The reference lists from these and those of review articles were also checked.

Selection criteria

All randomised controlled trials (RCTs) involving chemotherapy for metastatic/recurrent cervical cancer. Trials involving radiotherapy, chemoradiotherapy, intra‐arterial chemotherapy, biological agents or immunomodulators were excluded.

Data collection and analysis

Three review authors independently reviewed trials for inclusion and data extraction and assessed risk of bias.

Main results

There were no data comparing best supportive care with chemotherapy. Cisplatin‐based regimens are the most widely used and therefore we have concentrated on these trials. In terms of response rates some non‐platinum regimens are equivalent but toxicity is higher. The most common cisplatin regimen was 50 mg/m² day 1 q21days. Higher doses had similar survivals. There was no direct comparison between single‐agent cisplatin and carboplatin. Overall survival (OS) and progression‐free survival (PFS) were not adequately reported and quality of life (QoL) outcomes were incompletely documented. Combination regimens were more toxic than single agents, but in the limited reported data this did not appear to adversely affect QoL.

No significant difference in response rate by site of recurrence was found, although there was a trend towards improved response when the main site of disease was beyond the previously irradiated pelvis.

Authors' conclusions

Combination cisplatin‐based chemotherapy could be a viable option for patients of good performance status with recurrent/metastatic cervical cancer, but further trials that report adequate survival and QoL data are sought. Response rates and improvements in survival are low. Cisplatin‐based combinations have significant toxicity. Outcomes are poor and novel cytotoxic/biological agents and optimal scheduling need further investigation. Future trials need to stratify for and perform planned subgroup analysis with respect to previous treatment and site of recurrence.

Plain language summary

Chemotherapy primarily aimed at improving length of life while maintaining quality of life for incurable cervical cancer

Cervical cancer often affects young women. Cancer that has come back after initial treatment (recurrent) or has already spread around the body at diagnosis (metastatic) is incurable so chemotherapy is aimed at improving length of life, while maintaining good quality of life. A literature search was conducted identifying 30 potential trials; four trials were excluded. The 26 clinical trials included in this review encompass a large range of different drugs, doses and combinations in a mixed group of patients over a long time period (1976 to 2011), making it difficult to compare treatment options. Although there are no trials directly comparing chemotherapy with symptomatic management alone, chemotherapy is widely used in this setting and assumed to be of benefit. Cisplatin and carboplatin chemotherapy were shown to shrink the cancer in 10% to 30% of patients and are widely used in current practice. Cisplatin chemotherapy when combined with other drugs has been shown to prolong survival by a few months compared with cisplatin alone, but with the cost of increased side effects. Other chemotherapy has been used, but has been found to be less effective or more toxic. Quality of life for patients on chemotherapy appears to be similar for cisplatin and cisplatin‐based combinations. Nearly all patients in these studies were relatively fit and well prior to starting treatment, despite their cancer; these results may not be the same in patients who are not fit and well.

Background

Description of the condition

Cervical carcinoma arises in the uterine cervix and 60% to 70% of cases are squamous cell carcinomas, approximately 25% to 30% are adenocarcinomas or adenosquamous carcinomas, with a very small number of rarer cancers, such as small cell tumours and neuroendocrine tumours (Jemal 2008; Meta‐analysis Collaboration 2008).

Cervical cancer is the second most common cancer among women aged up to 65 years and is the most frequent cause of death from gynaecological cancers worldwide. A woman's risk of developing cervical cancer by 65 years of age ranges from 0.69% in developed countries to 1.38% in developing countries (GLOBOCAN 2008). In Europe, about 60% of women with cervical cancer are alive five years after diagnosis (EUROCARE‐3). The stage of disease at diagnosis determines survival rates; lesions less than 4 cm in diameter and confined to the cervix (FIGO (International Federation of Gynecology and Obstetrics) Stage Ib1) at presentation have a five‐year survival rate of 92%, while women with cancer spread beyond the true pelvis to adjacent organs (FIGO Stage IVa) have a five‐year survival rate of only 17% (Jemal 2008; Meta‐analysis Collaboration 2008).

The overwhelming risk factor for development of cervical cancer is the presence of human papilloma virus (HPV), particularly subtypes 16 and 18. Vaccines against HPV have become available and introduced in some countries. In the future it is hoped this will significantly reduce the incidence of cervical cancer. HPV causes the cervical epithelium to become increasingly abnormal (graded as cervical intra‐epithelial neoplasia (CIN) grades I to III), this then becomes invasive in 30% to 70% of women over 10 to 12 years, although this process can be much faster in a small minority (< one year). Invasive cancer then spreads directly by invading adjacent structures and metastasising via regional lymph nodes and, less commonly, via the bloodstream to distant sites such as the lungs.

Description of the intervention

Early‐stage cancer confined to the cervix, or with extension into upper vagina (Stage I to IIa), can be successfully treated by radical surgery (with or without neoadjuvant chemotherapy) or concomitant chemoradiation, giving five‐year survival rates of 80% to 90% (Eifel 2001). A European Organisation for Research and Treatment of Cancer (EORTC) trial (EORTC 55994) of chemoradiotherapy versus neoadjuvant chemotherapy plus surgery is on‐going. For more advanced cancer, Stages IIb to IVa the treatment of choice is chemoradiotherapy; as a Cochrane meta‐analysis has shown that the addition of platinum‐based chemotherapy significantly improves survival compared to radiotherapy alone (Green 2005). Combined chemoradiotherapy will cure 50% to 70% of patients with locally advanced carcinoma of the cervix (Meta‐analysis Collaboration 2008; Willmott 2009) and can be used for local control of metastatic disease.

Palliative chemotherapy is used for the management of Stage IVb patients whose disease has spread to distant sites, or for patients with inoperable recurrent or persistent disease (de la Motte Rouge 2006; Pectasides 2008). A number of chemotherapeutic options have been used either as single agents or in combination including cisplatin, adriamycin, ifosfamide, paclitaxel, irinotecan, topotecan, vinorelbine and gemcitabine with responses in the range of 15% to 46% (Moore 2006; Tewari 2005).

How the intervention might work

The aim of chemotherapy is to slow cancer growth, improve survival with minimal toxicity so there is an improvement in quality of life (QoL).

Why it is important to do this review

There are no standard treatments for patients with persistent, recurrent or metastatic cervical cancer. In this group of patients, who are often young and otherwise fit and well, it is important to establish chemotherapy regimens that have the greatest chance of response with tolerable side effects and an improvement in QoL. In the era of chemoradiation as primary management for locally advanced cervical cancer the role of chemotherapy in this setting becomes even more important to establish.

Objectives

To investigate the effectiveness of single‐agent and combination chemotherapy in the treatment of patients with metastatic or recurrent cervical cancer with regards to progression‐free survival (PFS), overall survival (OS), adverse effects of treatment and QoL.

Methods

Criteria for considering studies for this review

Types of studies

Randomised clinical trials (RCTs).

Types of participants

Adult women (aged 18 years or over) with metastatic (FIGO Stage IVb) or recurrent cervical carcinoma who are suitable for chemotherapeutic treatment.

Types of interventions

Single‐agent cytotoxic chemotherapy.
Multi‐agent cytotoxic chemotherapy.
Best supportive care.

For the purposes of this review chemotherapy refers to any cytotoxic drug given intravenously or orally with the intent of producing tumour regression as defined by World Health Organization (WHO) criteria for assessing response (Miller 1981); thus biological and immunomodulators were excluded. Trials involving concomitant radiotherapy were excluded because this could complicate assessment of response to chemotherapy, which is the primary objective of the review, particularly as most trials involving radiotherapy compare radiotherapy with radiotherapy plus chemotherapy. Intra‐arterial chemotherapy regimens were excluded as this was felt to be a local rather than systemic therapy and so not comparable. We included trials that used both platinum and non‐platinum chemotherapy and planned to subgroup based on whether the chemotherapy was platinum or non‐platinum based (see Subgroup analysis and investigation of heterogeneity). Best supportive care is not a clearly defined term, but is generally considered to include any interventions that palliate symptoms and optimise QoL, without treating the underlying cancer, such as pain relief or psychological support (www.WHO.int). Where there is no standard treatment option for a particular disease new treatments are often compared with best supportive care (i.e. symptom control) alone.

Types of outcome measures

Primary outcomes

Response rate (percentage of patients with evidence of reduction in tumour size following treatment, usually assessed by computerised tomography (CT) scan within one to three months).
OS (length of time from completing treatment to death from whatever cause).

Secondary outcomes

Time to progression (length of time from start of chemotherapy to evidence of cancer progression).
QoL, measured using a scale that has been validated through reporting of norms in a peer‐reviewed publication.
Toxicity, classified according to CTCAE 2006:
- haematological (leukopenia, anaemia, thrombocytopenia, neutropenia, haemorrhage);
- gastrointestinal (nausea, vomiting, anorexia, diarrhoea, liver, proctitis);
- genitourinary;
- skin (stomatitis, mucositis, alopecia, allergy);
- neurological (peripheral and central);
- pulmonary;
- other.

Search methods for identification of studies

Papers in all languages were sought and translations carried out when necessary.

Electronic searches

See Cochrane Gynaecological Cancer Group methods used in reviews.  The following electronic databases were searched:

the Cochrane Gynaecological Cancer Review Group's Trial Register
the Cochrane Central Register of Controlled Trials (CENTRAL), Issue 1, 2012
MEDLINE to January 2012
EMBASE to January 2012

The MEDLINE, EMBASE and CENTRAL search strategies based on terms related to the review topic are presented in Appendix 1, Appendix 2 and Appendix 3, respectively.

All relevant articles found were identified on PubMed and using the 'related articles' feature, a further search was carried out for newly published articles.

Searching other resources

Unpublished and grey literature

Meta‐Register, Physicians Data Query, www.controlled‐trials.com/rct, www.clinicaltrials.gov and www.cancer.gov/clinicaltrials were searched for ongoing trials.

Reference lists and Correspondence

The citation lists of included studies were checked and experts in the field contacted to identify further reports of trials.

Data collection and analysis

Selection of studies

All titles and abstracts retrieved by electronic searching were downloaded to the reference management database Endnote, duplicates were removed and the remaining references were examined by three review authors (KS, JF and MF) independently. Those studies which clearly did not meet the inclusion criteria were excluded and copies of the full text of potentially relevant references were obtained. The eligibility of retrieved papers were assessed independently by three review authors (KS, JF and MF). Disagreements were resolved by discussion between the three review authors. Reasons for exclusion are documented. No blinding of review authors to article author or journal title occurred.

Data extraction and management

For included studies, data were abstracted as recommended in Chapter 7 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This included data on the following:

author, year of publication and journal citation (including language);
country;
setting;
inclusion and exclusion criteria;
study design, methodology;
patient characteristics (age, histology, grade, extent of disease, previous therapy, performance status, whether disease lies within an area previously treated with radiotherapy);
cervical cancer details at diagnosis:
- FIGO Stage;
- histological cell type;
- tumour grade;
number of participants in each arm of the trial;
number excluded from analysis;
type of intervention (drug, dose, regimen, frequency, number of cycles);
proportion of participants who received all/ part/none of the intended treatment;
delays in treatment;
risk of bias in study (see below);
length of follow‐up;
outcomes: response rate, OS, time to progression, QoL and toxicity:
- for each outcome: outcome definition (with diagnostic criteria if relevant);
- unit of measurement (if relevant);
- for scales: upper and lower limits, and whether high or low score is good;
- results: number of participants allocated to each intervention group;
- for each outcome of interest: sample size; missing participants;
- the time points at which outcomes were collected and reported will be noted.

Data on outcomes were extracted as below:

for time to event (OS and time to progression) data, we planned to extract the log of the hazard ratio [log(HR)] and its standard error from trial reports; if these were not reported, we attempted to estimate them from other reported statistics using the methods of Parmar 1998;
for dichotomous outcomes (e.g. response rate, toxicity), we extracted the number of patients in each treatment arm who experienced the outcome of interest and the number of patients assessed at end point, in order to estimate a risk ratio (RR);
for continuous outcomes (e.g. QoL), we planned to extract the final value and standard deviation of the outcome of interest and the number of patients assessed at end point in each treatment arm at the end of follow‐up, in order to estimate the mean difference (if trials measured outcomes on the same scale) or standardised mean difference (if trials measured outcomes on different scales) between treatment arms and its standard error.

Where possible, all data extracted were those relevant to an intention‐to‐treat analysis, in which participants were analysed in groups to which they were assigned.

Data were abstracted independently by three review authors (KS, JF and MF) onto a data abstraction form specially designed for the review. Differences between review authors were resolved by discussion or arbitration of a third party (CW or PC).

Assessment of risk of bias in included studies

The risk of bias in included RCTs was assessed using The Cochrane Collaboration's tool and the criteria specified in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This included assessment of:

selection bias:
- random sequence generation;
- allocation concealment;
performance bias:
- blinding of participants and personnel (patients and treatment providers);
detection bias:
- blinding of outcome assessment;
attrition bias:
- incomplete outcome data: we recorded the proportion of participants whose outcomes were not reported at the end of the study; we coded a satisfactory level of loss to follow‐up for each outcome as:
  - low risk of bias, if fewer than 20% of patients were lost to follow‐up and reasons for loss to follow‐up were similar in both treatment arms;
  - high risk of bias, if more than 20% of patients were lost to follow‐up or reasons for loss to follow‐up differed between treatment arms;
  - unclear risk of bias if loss to follow‐up was not reported;
reporting bias:
- selective reporting of outcomes;
other possible sources of bias.

The 'Risk of bias' tool was applied independently by three review authors (KS, JF and MF) and differences resolved by discussion or by appeal to a third review author (CW or PC). Results are presented in a 'Risk of bias' summary graph. Results of meta‐analyses were interpreted with consideration of the findings with respect to risk of bias.

Measures of treatment effect

We used the following measures of the effect of treatment:

for time to event data, we planned to use the HR;
for dichotomous outcomes, we used the RR;
for continuous outcomes, we planned to use the mean difference between treatment arms.

Dealing with missing data

We did not impute missing outcome data for the primary outcome.

Assessment of heterogeneity

Heterogeneity between trials was assessed by visual inspection of forest plots, by estimation of the percentage heterogeneity between trials that cannot be ascribed to sampling variation (Higgins 2003), and by a formal statistical test of the significance of the heterogeneity (Deeks 2001). If there was evidence of substantial heterogeneity, the possible reasons for this were investigated and reported.

Assessment of reporting biases

There was an insufficient number of included trials to assess the potential for small study effects such as publication bias adequately.

Data synthesis

If sufficient, clinically similar studies were available, their results were pooled in meta‐analyses.

For time‐to‐event data, HRs would have been pooled using the generic inverse variance facility of RevMan 5 (RevMan 2011).
For any dichotomous outcomes, the RR was calculated for each trial and these were then pooled.
For continuous outcomes, the mean differences between the treatment arms at the end of follow‐up would have been pooled if all trials measured the outcome on the same scale, otherwise standardised mean differences would have been pooled.

Random‐effects models with inverse variance weighting was used for all meta‐analyses (DerSimonian 1986).

Subgroup analysis and investigation of heterogeneity

We included trials that used both platinum and non‐platinum chemotherapy and planned subgroup analyses, grouping the trials based on whether the chemotherapy was platinum or non‐platinum based.

Sensitivity analysis

We planned to perform sensitivity analyses excluding trials at high risk of bias

Results

Description of studies

Results of the search

Five hundred and sixty eight references were identified by searching the electronic databases (excluding duplicates). Reviewing the abstracts for these references identified 30 potentially eligible studies. Three trials were excluded (Baker 1978; Greggi 2000; Kumar 1998) owing to non‐randomisation of participants. One trial was excluded as it was a trial of differing Bryostatin‐1 schedules showing no activity (Armstrong 2003). Thus 26 eligible trials were included in this review (see Characteristics of included studies), comprising of 3766 randomised participants of which 631 were excluded from analysis in the published papers or abstracts. Thirteen trials were sufficiently similar and were included in meta‐analyses. No unpublished data from completed trials were identified.

Included studies

The RCTs included in this review dated from the 1970s to the present day with the number of patients ranging from five to 581. The trials included patients with locally recurrent, persistent or metastatic disease. Most of the patients had previously received radiotherapy to the primary site; however, since 1999 chemoradiotherapy has been the gold standard treatment for locally advanced disease and so patients included in more recent trials were likely to have previously received concurrent chemotherapy (generally platinum‐based) with radiotherapy.

The trials used a wide variety of designs and chemotherapy schedules. Five trials compared single‐agent platinum with a platinum‐containing combination (Alberts 1987; Cadron 2005; Long 2005; Moore 2004; Omura 1997) generally cisplatin, four trials compared single‐agent platinum with different formulations or doses (Bonomi 1985; Lira‐Puerto 1991; McGuire 1989; Thigpen 1989), two trials compared single‐agent platinum with a non‐platinum (Long 2006; Pfeiffer 1998), 10 trials compared a variety of non‐platinum‐containing regimens (Bond 1976; Freedman 1980; Greenberg 1977; Malkasian 1981; Moseley 1976; Omura 1978; Omura 1981; Palo 1976; Wallace 1978; Weiss 1992), four trials compared differing platinum combinations (Bloss 2002; Edmonson 1988; Monk 2008; Mountzios 2009) and the remaining trial compared non‐platinum single‐agent with platinum‐containing combination (Bezwoda 1986).

Outcomes were incompletely reported, with only response rate and some adverse event components being reported to an adequate level to include in meta‐analyses. Time to event data were poorly reported with most trials presenting median and quartile statistics. Median survival in all trials ranged from four months to 17 months. The intention was to compare survival data and time to progression data by comparing HRs, but these were only reported in two studies. Where data were not reported, we had hoped to be able to calculate them using the method described by Parmar 1998, by extracting the log rank P value, number of events and the number of subjects in the treatment and control groups; however, insufficient trial data (mean survival, probabilities or individual patient data) had been reported for this. Michiels 2005 showed that median survival times are not reasonable surrogate measures for meta‐analysis of survival. Eighteen of the 26 trials did not perform intention‐to‐treat analyses.

It had been anticipated that the majority of trials would be reporting toxicity with reference to Common Toxicity Criteria version 3.0 (CTCv3.0); however, this only became available in 2003 (replacing version 2.0, which ran from 1998 to 2003) and the majority of included trials (18/26) were published prior to 2003. Of these pre‐1998 studies eight made no attempt to grade toxicity (Bezwoda 1986; Bond 1976; Edmonson 1988; Freedman 1980; Greenberg 1977; Malkasian 1981; Moseley 1976; Omura 1978), two used an unspecified grading system (Palo 1976; Wallace 1978), and remainder used Eastern Cooperative Oncology Group (ECOG) (Lira‐Puerto 1991), Southwest Oncology Group (SWOG) (Alberts 1987; Weiss 1992) and Gynecologic Oncology Group (GOG) (Bonomi 1985; McGuire 1989; Omura 1981; Omura 1997; Thigpen 1989). None of the included trials published after 2003 (8/23) used CTCv3.0 and only two (Long 2005; Monk 2008) used CTCv2.0. The remainder using WHO (Mountzios 2009; Pfeiffer 1998), SWOG (Weiss 1992) or GOG (Bloss 2002; Moore 2004) scales. This wide variety made it very difficult to compare toxicities between regimens despite an attempt to convert all the differing grading scales into the CTCv2.0 equivalent.

Excluded studies

Four references were excluded, after obtaining the full text, for the following reasons:

Baker 1978, Greggi 2000 and Kumar 1998 were excluded as they were not randomised trials;
Armstrong 2003 was excluded as trial drugs were not cytotoxic agents.

For further details of all the excluded trials see the Characteristics of excluded studies.

Risk of bias in included studies

For risk of bias summary see Figure 1 and Figure 2.

'Risk of bias' graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.

'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study.

Allocation

All included trials were described as randomised or involved the randomised assignment of treatment; however, there was only sufficient detail given in 10 (Bloss 2002; Greenberg 1977; Long 2005; Long 2006; McGuire 1989; Moore 2004; Omura 1978; Omura 1997; Thigpen 1989; Wallace 1978) out of the 26 trials to confirm that the randomisation procedure was adequate. Of these trials only five also gave details of allocation concealment, which was by central randomisation (Greenberg 1977; Long 2005; Omura 1997) or closed envelope technique (Omura 1978; Wallace 1978).

Blinding

Adequacy of blinding could not be assessed in any of the trials. Many trials reported using physical examination as part of the assessment of response introducing potential for significant bias. Whether or not the outcome assessor was blinded was not reported in any of the trials.

Incomplete outcome data

Twelve trials excluded 5% or less of their participants (Alberts 1987; Bloss 2002; Edmonson 1988; Freedman 1980; Greenberg 1977; Lira‐Puerto 1991; Long 2005; Long 2006; Malkasian 1981; Mountzios 2009; Omura 1997; Weiss 1992), eight trials excluded 6% to 19% (Bezwoda 1986; Cadron 2005; McGuire 1989; Monk 2008; Moore 2004; Omura 1981; Pfeiffer 1998; Thigpen 1989) and five trials excluded 20% to 49% (Bonomi 1985; Moseley 1976; Omura 1978; Palo 1976; Wallace 1978). In one trial (Bond 1976) the total number of participants was not given neither was it stated how many patients were excluded from the final analysis. Frequent reasons for exclusion were given as ineligibility, loss to follow‐up, inadequate data, lack of easily assessable disease and missing data.

In the older trials particularly there was a large drop‐out rate, such that in three trials nearly half the patients randomised were not evaluated (Bond 1976; Omura 1978; Wallace 1978). Patients were sometimes excluded because of a lack of measurable disease.

Selective reporting

All trials reported response rates but survival data reporting was inconsistent. Twenty‐one of 26 trials reported OS with two further trials reporting median survival (Alberts 1987; Bond 1976), but were not reported in sufficient detail to allow inclusion in meta‐analyses. Survival data were not reported in four trials (Moseley 1976; Omura 1981; Omura 1997; Weiss 1992). PFS was only reported in 11 trials (Bloss 2002; Bonomi 1985; Monk 2008; Moore 2004; Mountzios 2009; Omura 1981; Omura 1997; Palo 1976; Pfeiffer 1998; Thigpen 1989; Wallace 1978).

Where full survival data were not reported there was no justification for the choice of end points. Only two trials (Long 2005; Monk 2008) reported HRs and there was insufficient detail in the other trials to calculate using Parmars methods (Michiels 2005; Parmar 1998). This is discussed in more detail in the Results section.

Other potential sources of bias

A variety of first‐line treatment was given prior to these trials with a historical bias. In the older trials (patients enrolled pre‐1999) the majority of first‐line treatment was radiotherapy whereas post‐1999 most patients received primary chemoradiotherapy. In some trials (Bond 1976; Edmonson 1988; Palo 1976) cross‐over was allowed between arms.

Effects of interventions

In the meta‐analysis comparing various forms of chemotherapy, two trials (Bezwoda 1986; Greenberg 1977) assessing response rate and three trials (Alberts 1987; Omura 1997; Moore 2004) assessing toxicity had no responses in one of the arms, so we added 0.5 to these cells to allow calculation of an RR. This is the default zero‐cell correction within RevMan (RevMan 2011), and biases the result of the meta‐analysis towards no difference between the two types of chemotherapy being compared.

The methods of assessing response rates were variable, many of the older trials used clinical evaluation to a large extent with some radiological imaging (e.g. pelvic x‐ray) while more modern trials focused on CT or magnetic resonance imaging (MRI). Criteria for determining response also vary; however, three (Long 2005; Moore 2004; Omura 1997) of the five trials included in the meta‐analysis used the GOG definition based on the best response achieved at any point during treatment that had to be maintained for at least four weeks (complete response: disappearance of all lesions for at least four weeks; partial response: a greater than 50% reduction in the product of bi‐dimensional measurement of each lesion for at least four weeks; progressive disease: a greater than 50% increase in the product of the bi‐dimensional measurement of any lesion). One study (Alberts 1987) used SWOG criteria where the difference in size of lesions had to be 25% or greater to be progressive or partially responsive disease. Cadron 2005 used WHO criteria for determining response but gave no further details as to what this entailed. Response rates are early, easily measurable indicators often used as a surrogate for benefit but do not necessarily correlate with a survival advantage.

In Analysis 1.3, Analysis 1.4, Analysis 1.5, Analysis 1.6 and Analysis 1.7 complete and partial response were grouped together and were deemed 'response'.

1.3 — Comparison 1 Response rate (complete response + partial response), Outcome 3 Cisplatin single agent vs. cisplatin doublet.

1.4 — Comparison 1 Response rate (complete response + partial response), Outcome 4 Platinum‐containing regimen vs. non‐platinum‐containing regimen.

1.5 — Comparison 1 Response rate (complete response + partial response), Outcome 5 Platinum‐containing regimen vs. non‐platinum‐containing regimen: sensitivity analysis.

1.6 — Comparison 1 Response rate (complete response + partial response), Outcome 6 Platinum + paclitaxel vs. platinum combination (non‐paclitaxel).

1.7 — Comparison 1 Response rate (complete response + partial response), Outcome 7 In‐field vs. out‐field recurrence patients with platinum‐containing regimen.

Single‐agent versus combination chemotherapy

Meta‐analysis of 10 RCTs (Alberts 1987; Bezwoda 1986; Cadron 2005; Greenberg 1977; Long 2006; Malkasian 1981; Moore 2004; Omura 1981; Omura 1997; Wallace 1978), assessing 1438 participants, did not find any statistically significant difference in response rate between women who received single‐agent chemotherapy and those who received combination therapy (RR 0.94; 95% CI 0.57 to 1.55). The percentage of the variability in effect estimates that was because of heterogeneity between studies rather than chance may represent substantial heterogeneity (I² = 67%). A wide variety of chemotherapy agents and doses were used across trials (see Table 1). Many of the trials were from an era when activity of agents was not known and so included inactive agents. The trials of Bezwoda 1986 and Greenberg 1977 were removed because of this in a sensitivity analysis, but the results were largely unchanged (RR 0.94; 95% CI 0.59 to 1.50; I² = 66%). In addition many of these older trials were of poor quality, where more stringent reporting guidelines were not introduced until after these trials were completed.

1. Comparison of median survival: any single agent vs. any combination.

Study	Single agent			Combination
Study	n	Median OS (months)	Median PFS (months)	n	Median OS (months)	Median PFS (months)
Alberts 1987 C vs. MC vs. MVBC	9	17 (C)	nr	54 (MVBC) 51 (MC)	6.9 (MVBC) 7 (MC)	5.4 (MVBC) 7.2 (MC)
Bezwoda 1986 H vs. CMeth	13	4 (H)	nr	37	9 (CMeth)	nr
Cadron 2005 C vs. CIF	11	13 (C)	nr	10	12.3 (CIF)	nr
Greenberg 1977 A vs. AB	9	4 (A)	nr	11	4.3 (AB)	nr
Long 2005 C vs. CTop vs. MVAC	146	6.5 (C)	2.9 (C)	63 (MVAC) 147 (CTop)	9.4 (MVAC) 9.4 (CTop)	4.4 (MVAC) 4.6 (CTop)
Malkasian 1981 A vs. VF	20	5.9 (A)	2.8 (A)	21	9 (VF)	2.9 (VF)
Moore 2004 C vs. CP	134	8.9 (C)	3	130	9.9 (CP)	4.9
Omura 1981 cyclo vs. AFV	30	6.5 (cyclo)	nr	31	7.6 (AFV)	nr
Omura 1997 C vs. CMito vs. CI	140	8 (C)	3.2	147 (CMito) 151 (CI)	7.3 8.3	nr (CMito) 4.6 (CI)
Wallace 1978 A vs. AV vs. Acyclo	63	5.9 (A)	3.3	63 (AV) 52 (Acyclo)	5.5 (AV) 7.3 (Acyclo)	3.4 (AV) 3.9 (Acyclo)

Study	Single‐agent cisplatin			Cisplatin combination
Study	n	Median OS (months)	Median PFS (months)	n	Median OS (months)	Median PFS (months)
Alberts 1987 C vs. MC vs. MVBC	9	17	nr	54 (MVBC) 51 (MC)	6.9 (MVBC) 7 (MC)	5.4 (MVBC) 7.2 (MC)
Cadron 2005 C vs. CIF	11	13	nr	10	12.3	nr
Long 2005 C vs. CT	146	6.5	2.9	147	9.4	4.6
Moore 2004 C vs. CP	134	8.9	3	130	9.9	4.9
Omura 1997 C vs. CMito vs. CI	140	8	3.2	147 (CMito) 151 (CI)	7.3 8.3	nr (CMito) 4.6 (CI)

Study	Platinum containing			Non‐platinum containing
Study	n	Median OS (months)	Median PFS (months)	n	Median OS (months)	Median PFS (months)
Bezwoda 1986 H vs. CMeth	37	11 (CMeth)	nr	13	4 (H)	nr
Long 2005/Long 2006 C vs. CT vs. MVAC	146 147	6.5 (C) 9.4 (CT)	2.9 (C) 4.6 (CT)	63	9.4 (MVAC)	4.4
Pfeiffer 1998 Carbo vs. Ten	12	9.3 (Carbo)	4.6	14	9.6 (Ten)	4.0

Study	Cisplatin + other			Cisplatin + taxane			P value
Study	n	Median OS (months)	Median PFS (months)	n	Median OS (months)	Median PFS (months)	P value
Monk 2008 CV vs. CT vs. CG vs. CP	117 (CV) 118 (CT) 119 (CG)	10‐10.3	4.9 (CV) 4.6 (CT) 4.7 (CG)	118 (CP)	12.9	5.8	ns
Mountzios 2009 CI vs. CIP	74 (CI)	13.2	6.3	79 (CIP)	15.4	7.9	0.048 (OS) 0.023 (PFS)

Study	Platinum / dose (mg/m²)	n	RR (%)	OS (months)	PFS (months)	TP (months)
Alberts 1987	C 50 d1 q21	9	33	17	nr	nr
Bonomi 1985	C 50 d1 q21 C 100 d1 q21 C 20 d1‐5 q21	150 166 128	20 31.4 25	7.1 7.0 6.1	nr	3.7 4.6 3.9
Cadron 2005	C 37.5 d1, 2 q28	11	9	13	nr	nr
Lira‐Puerto 1991	Carbo 400 d1 q28 I 300 d1 q 28	48 41	26 30	7.5 7.6	nr	5.5 6
Long 2005	C 50 d1 q21	146	13	6.5	2.9	nr
McGuire 1989	Carbo 400/340 d1 q28 I 270/230 d1 q28	175 177	15.4 10.8	6.5 5.6	nr	2.7 3.0
Moore 2004	C 50 d1 q21	134	19	8.9	3.0	nr
Omura 1997	C 50 d1 q21	140	17.8	8	nr	nr
Thigpen 1989	C 50 d1 (24 h) q21 C 50 d1 (1 mg/minute) q21	163 168	18 17	6.4 6.2	nr	3.5 2.9
Pfeiffer 1998	Carbo 400 d1 q28	12	33	9.3	4.6	4

Regimen	Hazard ratio PFS	Hazard ratio OS
CV vs. CP	1.357	1.147
CG vs. CP	1.394	1.322
CT vs. CP	1.268	1.255

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Single agent vs. combination	10	1438	Risk Ratio (IV, Random, 95% CI)	0.94 [0.57, 1.55]
2 Single agent vs. combination: sensitivity analysis	8	1368	Risk Ratio (IV, Random, 95% CI)	0.94 [0.59, 1.50]
3 Cisplatin single agent vs. cisplatin doublet	5	1114	Risk Ratio (IV, Random, 95% CI)	0.60 [0.44, 0.81]
4 Platinum‐containing regimen vs. non‐platinum‐containing regimen	3	413	Risk Ratio (IV, Random, 95% CI)	1.33 [0.50, 3.54]
5 Platinum‐containing regimen vs. non‐platinum‐containing regimen: sensitivity analysis	2	363	Risk Ratio (IV, Random, 95% CI)	0.93 [0.58, 1.49]
6 Platinum + paclitaxel vs. platinum combination (non‐paclitaxel)	2	587	Risk Ratio (IV, Random, 95% CI)	1.47 [1.01, 2.15]
7 In‐field vs. out‐field recurrence patients with platinum‐containing regimen	6	1062	Risk Ratio (IV, Random, 95% CI)	0.62 [0.46, 0.83]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Neutropenia G3/G4 in single‐agent cisplatin vs. combination	4	1073	Risk Ratio (IV, Random, 95% CI)	0.04 [0.02, 0.12]
2 Thrombocytopenia G3/4 in single‐agent cisplatin vs. combination	4	1104	Risk Ratio (IV, Random, 95% CI)	0.16 [0.05, 0.48]
3 Infection G3/4 in single‐agent cisplatin vs. combination	2	552	Risk Ratio (IV, Random, 95% CI)	0.42 [0.22, 0.81]
4 Renal dysfunction G3/G4 in single‐agent cisplatin vs. combination	3	980	Risk Ratio (IV, Random, 95% CI)	0.81 [0.46, 1.41]
5 Neuropathy G3/G4 in single‐agent cisplatin vs. combination	2	552	Risk Ratio (IV, Random, 95% CI)	1.39 [0.45, 4.33]

Methods	Randomised Phase II trial
Participants	119 women with incurable squamous cell carcinoma unsuitable for surgery or radiotherapy. No previous chemotherapy exposure
Interventions	Arm 1: mitomycin C 10 mg/m² iv d2, 44, vincristine 0.5 mg/m² iv d2, 4, 44, 46, bleomycin 30 mg continuous iv infusion over 24 h for 4 days d1‐4, 43‐46, cisplatin 50 mg/m² iv d1, 22, 43, 64 Arm 2: mitomycin C 12 mg/m² iv d1, 43, cisplatin 50 mg/m² iv d1, 22, 43, 64 Arm 3: cisplatin 50 mg/m² iv d1, 22, 43, 64
Outcomes	Response rate Response duration Median survival Toxicity (SWOG)
Notes	5 ineligible patients (reasons not stated). Arm 3 stopped early owing to slow accrual. Not clear whether ITT analysis performed
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	High risk	Not documented
Incomplete outcome data (attrition bias)   All outcomes	Low risk	5 ineligible patients (reasons not stated). Arm 3 stopped early owing to slow accrual. Not clear whether ITT analysis performed
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	High risk	Cisplatin arm stopped early despite appearing to have better results so only 9 patients evaluable

Methods	Randomised Phase II trial 1982 to 1984
Participants	50 women with recurrent or metastatic cervical cancer
Interventions	Arm 1: hydroxyurea 1.5 g/m² po d1‐10, 14‐d rest period then maintenance with 1 g/m² po d1‐14 q28 Arm 2: cisplatin 20 mg/m² iv daily d1‐3 and methotrexate 100 mg/m² iv d3 with leucovorin rescue
Outcomes	OS Response rates Toxicity
Notes	3 patients were excluded (refused chemotherapy after randomisation). No ITT analysis
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Low risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	3 patients were excluded
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	High risk	Hydroxyurea (arm 1) arm stopped early owing to lack of response, only 12 patients randomised to cisplatin + methotrexate arm (arm 2), the remainder entered that arm as only treatment option once the hydroxyurea arm closed. Thus many randomised and trial results not used in meta‐analyses

Methods	Randomised Phase III trial
Participants	24 women with histologically confirmed cervical cancer with distant metastases after surgery or recurrence after radiotherapy
Interventions	Arm 1: cisplatin 37.5 mg/m² iv d1‐2 q28 Arm 2: cisplatin 37.5 mg/m² iv d1‐2 and Ifosfamide 2 g/m² iv d1‐2, mesna 0.5 g/m², 5‐fluorouracil 500 mg/m² d1‐2, folinic acid 30 mg/m² d1‐2 q28
Outcomes	Response rates Toxicity (WHO) OS
Notes	3 patients ineligible (did not receive chemotherapy owing to rapid progression). ITT analysis performed. Trial stopped early owing to poor accrual
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase II trial 1980 to 1985
Participants	91 women with advanced cervical carcinoma
Interventions	Arm 1: bleomycin 20 U/m² iv 3‐d infusion, cisplatin 60 mg/m² d1 iv, doxorubicin 40 mg/m² iv D1, cyclophosphamide 400 mg/m² iv d1 q28 Arm 2: bleomycin 20 U/m² iv 3‐d infusion, cisplatin 60 mg/m² d1 iv q21 Bleomycin omitted after 4 cycles in both arms
Outcomes	Response rates OS
Notes	1 patient excluded as ineligible owing to second primary cancer. No ITT analysis
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Dynamic allocation system
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase II trial 1975
Participants	59 women with recurrent squamous cell cervical cancer
Interventions	Arm 1: sequential cyclophosphamide 8 mg/kg iv d1‐5 q28 then adriamycin 80 mg/m² iv d1 q28 then hexamethylmelamine 8 mg/kg po daily continuously Arm 2: sequential adriamycin 80 mg/m² iv d1 q28 then cyclophosphamide 8 mg/kg iv d1‐5 q28 then hexamethylmelamine 8 mg/kg po daily continuously Arm 3: sequential hexamethylmelamine 8 mg/kg po daily continuously then cyclophosphamide 8 mg/kg iv d1‐5 q28 then adriamycin 80 mg/m² iv d1 q28 Not stated prospectively what criteria would be for switching treatment (i.e. number of cycles versus progressive disease)
Outcomes	Response rates Survival time after chemotherapy start Survival from time of treatment change
Notes	1 patient excluded (died prior to starting chemotherapy). No ITT analysis performed. Cross‐over between arms allowed. Reasons for switching chemotherapy agents varied and included disease progression, toxicity and cumulative doxorubicin dose
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Stratified by ease of disease evaluation then randomised
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Low risk	Not documented, but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	High risk	Patients allowed to switch between arms for variety of reasons including toxicity, cumulative doxorubicin dose and disease progression

Methods	Randomised Phase II trial 1984 to 1987
Participants	89 women with recurrent measurable squamous cell cervical cancer
Interventions	Arm 1: carboplatin 400 mg/m² iv d1, 1 dose escalation if haematological toxicity ≤ grade 1 Arm 2: iproplatin 300 mg/m² iv d1, 1 dose escalation if haematological toxicity ≤ grade 1 Cycle length not stated
Outcomes	Response rates Toxicity OS
Notes	3 patients excluded (1 patient refused treatment after randomisation, 2 patients excluded as recurrence subsequently deemed to be benign fibrosis). ITT analysis not performed but excluded patients eligible for toxicity evaluation
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase III trial 1999 to 2002
Participants	294 women with histologically confirmed advanced (Stage IVb), recurrent or persistent squamous cell, adenocarcinoma or adenosquamous cervical cancer unsuitable for curative treatment with surgery or radiotherapy, or both
Interventions	Arm 1: cisplatin 50 mg/m² iv d1 q21 Arm 2: cisplatin 50 mg/m² iv d1 and topotecan 0.75 mg/m² iv d1‐3 q21
Outcomes	Response rates Toxicity OS
Notes	8 patients excluded (1 institutional review board approval error, 2 second active malignancy present, 2 wrong histological type, 1 primary tumour other than cervical cancer, 2 inadequate tissue to confirm metastases). ITT analysis performed Trial originally had 3 arms but third arm (MVAC) was closed early owing to toxicity and is reported in detail in separate papers
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients assigned with equal probability using a fixed‐block design
Allocation concealment (selection bias)	Low risk	GOG Statistical and Data Centre performed randomisation
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	8 patients excluded but balanced across groups. ITT analysis performed
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Unclear risk	MVAC arm closed early owing to toxic deaths

Methods	Randomised Phase III trial 1999 to 2001
Participants	186 women with histologically confirmed advanced (Stage IVb), recurrent or persistent squamous cell, adenocarcinoma or adenosquamous cervical cancer unsuitable for curative treatment with surgery or radiotherapy, or both
Interventions	Arm 1: cisplatin 50 mg/m² iv d1 q21 Arm 2: cisplatin 50 mg/m² iv d1 and topotecan 0.75 mg/m² iv d1‐3 q21 Arm 3: methotrexate 30 mg/m² iv d1, 15, 22, vinblastine 3 mg/m² iv d2, 15, 22, doxorubicin 30 mg/m² iv d2, cisplatin 70 mg/m² iv d2 q28 Maximum 6 cycles given unless maximum cumulative dose of doxorubicin achieved or progressive disease or unacceptable toxicity
Outcomes	Response rates Toxicity OS
Notes	6 patients ineligible (1 clerical error, 2 wrong cell type, 1 wrong primary, 1 inadequate pathology tissue, 1 second primary tumour. ITT analysis performed This paper reports the third arm of trial reported by Long 2006
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients assigned with equal probability using a fixed‐block design
Allocation concealment (selection bias)	Low risk	GOG Statistical and Data Centre randomly assigned patients
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	6 patients excluded balanced across arms. ITT analysis
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Unclear risk	MVAC arm closed early owing to toxic deaths

Methods	Randomised Phase III trial 1984 to 1987
Participants	394 patients with histologically confirmed squamous carcinoma of the cervix, recurrent disease following primary therapy or metastases. No previous chemotherapy allowed
Interventions	Arm 1: carboplatin 400 mg/m² d1 q28 Arm 2: iproplatin 270 mg/m² d1 q28
Outcomes	Response rates Response duration Toxicity (GOG) OS
Notes	23 ineligible (7 non‐squamous histology, 4 other primaries, 8 previous other malignancy, 3 no pathological documentation of primary tumour, 1 previous chemotherapy). 10 additional patients not evaluable for toxicity or response. 7 did not receive treatment. 3 lost to follow‐up. ITT analysis not performed
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block design with balanced assignments
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	23 ineligible and 10 additional patients not evaluable for toxicity or response balanced across groups. ITT analysis not performed
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase III trial 2003 to 2007
Participants	513 women with advanced, recurrent or persistent squamous cell cervical cancer not suitable for surgery or radiotherapy
Interventions	Arm 1: cisplatin 50 mg/m² d2 plus paclitaxel 135 mg/m² over 24 h d1 q21 Arm 2: cisplatin 50 mg/m² d1 plus vinorelbine 30 mg/m² d1, 8 q21 Arm 3: cisplatin 50 mg/m² plus gemcitabine 1000 mg/m² d1, 8 q21 Arm 4: cisplatin 50 mg/m² plus topotecan 0.75 mg/m² d1‐3 q21
Outcomes	OS PFS Response rates Toxicity (WHO) QoL
Notes	First 41 patients enrolled were excluded from primary analysis. another 38 patients later found to be ineligible. 434 evaluable for efficacy. 9 patients never treated so excluded from toxicity assessment
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Low risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	High risk	First 41 patients enrolled were excluded from primary analysis. Another 38 patients later found to be ineligible
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase II clinical trial
Participants	153 women with metastatic, persistent or recurrent cervical cancer. performance status 0 to 2
Interventions	Arm 1: ifosfamide 1.5 g/m² d1‐3 plus cisplatin 70 mg/m² d2 q28 Arm 2: ifosfamide 1.5 g/m² d1‐3 plus cisplatin 70 mg/m² d2 plus paclitaxel 175 mg/m² d1 q28
Outcomes	Response rates PFS OS Toxicity
Notes	4 patients excluded (3 had had previous chemotherapy, 1 previous primary cancer)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase II trial March 1990 to May 1994
Participants	28 women with recurrent or advanced cervical cancer
Interventions	Arm 1: carboplatin 400 mg/m² iv d1 q28 Arm 2: teniposide 125 mg/m² iv d1‐3 q28
Outcomes	Response rates OS PFS Toxicity (WHO)
Notes	2 patients ineligible (1 severe renal impairment, 1 poor performance status). No ITT analysis performed
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Methods	Randomised Phase II trial April 1982 to July 1985
Participants	331 women with advanced or recurrent squamous cell cervical cancer no longer amenable to control by surgery or radiotherapy
Interventions	Arm 1: cisplatin 50 mg/m² iv rate 1 mg/minute q21 days Arm 2: cisplatin 50 mg/m² iv rate over 24 h q21 days
Outcomes	Response rates OS PFS Toxicity
Notes	63 patients deemed ineligible (7 other primary site, 4 second primary, 26 wrong histological subtype, 26 no documentation of cervical primary, 1 elevated blood nitrogen urea at entry, 1 GOG performance status 4 at entry). 6 further patients with unevaluable (1 clerical error at entry, 1 never treated, 1 inadequate pathology, 2 inadequate data submitted, 1 not documented reason for unevaluability)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Latin square arrangement balancing the sequence of assigned regimens within and across institutions
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	High risk	63 patients excluded as ineligible and further 6 patients were unevaluable balanced across groups
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

PERMALINK

Chemotherapy for metastatic and recurrent cervical cancer

Kate Scatchard

Jennifer L Forrest

Maxine Flubacher

Paul Cornes

Chris Williams

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Unpublished and grey literature

Reference lists and Correspondence

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

Included studies

Excluded studies

Risk of bias in included studies

1.

2.

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

1.3. Analysis.

1.4. Analysis.

1.5. Analysis.

1.6. Analysis.

1.7. Analysis.

Single‐agent versus combination chemotherapy

1. Comparison of median survival: any single agent vs. any combination.

1.1. Analysis.

1.2. Analysis.

Cisplatin single‐agent versus cisplatin combination chemotherapy

2. Comparison of median survival: cisplatin single agent vs. cisplatin combination.

Platinum versus non‐platinum chemotherapy

3. Comparison of median survival: platinum vs. non‐platinum‐containing regimens.

Cisplatin combination with or without paclitaxel

4. Comparison of median survival: cisplatin combination including or excluding a taxane.

In‐field versus out‐of‐field responses with platinum‐containing chemotherapy

5. Comparison of different single‐agent cisplatin regimens.

Overall survival, progression‐free survival and time to progression

6. Comparison of four cisplatin‐containing doublets.

Toxicity

Methods	Randomised Phase II trial September 1974 to June 1976
Participants	326 women with advanced, recurrent or persistent squamous cell cervical cancer not suitable for surgery or radiotherapy
Interventions	Arm 1: adriamycin 60 mg/m² iv q21 Arm 2: adriamycin 60 mg/m² iv, vincristine 1.5 mg/m² iv q21 Arm 3: adriamycin 60 mg/m² iv, cyclophosphamide 500 mg/m² iv q21
Outcomes	OS PFS Response rates Toxicity (WHO)
Notes	65 patients ineligible (more than half for wrong histological subtype, remainder because of wrong primary, 2nd primary, elevated urea or creatinine, poor risk at entry or no evidence of progression after previous treatment; breakdown of numbers not given). Further 69 patients deemed to be not evaluable (inadequate data collection, therapy not initiated, insufficient follow‐up after first course, improper randomisation, clerical errors and others; no specific breakdown in numbers)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Low risk	Closed envelope technique
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	Not documented but OS unlikely to be affected by blinding
Incomplete outcome data (attrition bias)   All outcomes	High risk	65 patients ineligible and further 69 patients deemed to be not evaluable balanced across groups
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes

Methods	Randomised Phase II
Participants	44 women with persistent, recurrent or metastatic Squamous cell cancer cervix 1987 to 1990
Interventions	Arm 1: didemnin B 2.6‐3.5 mg/m² iv over 30 minutes d1 q28 (subsequent dose escalations if no toxicity) Arm 2: trimetrexate 8 mg/m² if previous radiotherapy or 12 mg/m² if no radiotherapy iv over 10 minutes d1‐5 q21 (dose escalations to 15 mg/m² and 18 mg/m² respectively permitted on subsequent cycles if no toxicity)
Outcomes	Overall response rate
Notes	0% response rate in either arm at any dose 1 patient ineligible as baseline imaging not done within 14 days of starting treatment
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not documented
Allocation concealment (selection bias)	Unclear risk	Not documented
Blinding (performance bias and detection bias)   All outcomes	High risk	Not documented
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data
Selective reporting (reporting bias)	Low risk	Published report included all pre‐specified outcomes
Other bias	Low risk	No hint at any other possible biases

Study	Reason for exclusion
Armstrong 2003	Comparison of 2 Bryostatin‐1 schedules, neither active
Baker 1978	Not an RCT
Greggi 2000	Not an RCT
Kumar 1998	No evidence that patients randomised between treatment arms

Trial name or title	JCOG0505
Methods	Phase III RCT
Participants	Women with IVB persistent or recurrent cervical cancer not amenable to curative treatment with local therapy
Interventions	Arm 1: paclitaxel 135 mg/m² over 24 h d1 plus carboplatin AUC 5 d1 q21 Arm 2: paclitaxel 135 mg/m² over 24 h d1 plus cisplatin 50 mg/m² 2 h d2 q21
Outcomes	OS PFS Toxicity Unplanned hospital admissions (surrogate for QoL)
Starting date	21 February 2006
Contact information	kitagawa.ryo@east.ntt.co.jp
Notes	Planned sample size 250