This review concluded that use of available molecular-targeted therapies combined with gemcitabine was not supported for treating advanced pancreatic cancer as they did not significantly benefit overall survival and were associated with higher grade toxicities compared with gemcitabine alone. The authors' conclusions should be treated cautiously given the small number of included trials and the diverse molecular-targeted therapies studied.

To compare the effectiveness and safety of gemcitabine chemotherapy plus molecular-targeted therapies versus standard palliative chemotherapy plus gemcitabine for advanced pancreatic cancer.

Seven databases (including PubMed, EMBASE, Cochrane Central Register of Controlled Trials, and the Hepatobiliary Group in The Cochrane Library) were searched up to November 2011 with no restriction on publication stage or language; search terms were reported. Bibliographies of potentially relevant articles were hand searched. Reviews were excluded.

Phase III randomised controlled trials (RCTs) that compared standard palliative chemotherapy plus gemcitabine alone or in combination with placebo versus gemcitabine plus molecular-targeted therapies for advanced pancreatic cancer were eligible for inclusion. Molecular targeted therapies were defined in the paper. Trials that included radiation therapy were excluded. The primary outcome was overall survival in patients receiving at least one cycle of chemotherapy, defined as the time interval between enrolment date and death. Secondary outcomes included progression-free survival, overall response rate, and grades 3, 4 and 5 toxicities according to the National Cancer Institute Common Toxicity Criteria (details were reported).

All the included trials were multi-centred. Over half of the trials (57%) were international; the remaining individual trials were conducted in the USA, the USA and UK, and the USA and Canada. The intervention arms of trials included gemcitabine combined with molecular-targeted therapies of axitinib, bevacizumab, cetuximab, erlotinib, marimastat, pemetrexed and tipifarnib. The control groups in most trials received gemcitabine plus placebo (71% trials) and gemcitabine alone in the other trials. No dosage details were reported.

Two independent reviewers performed the study selection. Disagreements were resolved by consensus after receipt of the full text.

Trial quality was assessed by two independent reviewers using the Cochrane Collaboration risk of bias assessment tool. Criteria assessed included: randomisation method including allocation sequence generation and allocation concealment; power calculation; methods used to deal with missing data; description of attrition and drop-outs; intention-to-treat analysis performed; clear definitions of primary outcomes; and whether a cross-over study design was used. Blinding of caregivers, patients and assessors was considered.

The numbers of events were used to calculate hazard ratios (HRs), with 95% confidence intervals (CIs), for overall survival and progression-free survival (from median survival times); odds ratio (ORs), with 85% confidence intervals, were calcuated for response rates and drug toxicities. The principal authors of articles were contacted for missing data.

Two independent reviewers performed the data extraction.

Trial results were pooled using random-effects models, since a high level of trial heterogeneity was expected, to give hazard ratios (with 95% confidence intervals) and odds ratios (with 95% confidence intervals). Z-values were reported. Forest plots provided data in a visual form.

A sensitivity analysis was performed for the primary outcome of overall survival to find the impact of trials with a higher risk of bias.

Publication bias was assessed visually using a funnel plot.

Seven RCTs were included in the review (3,973 participants, range 239 to 745). All trials described the allocation system and performed a power calculation. However all trials had unclear handling of missing data and did not provide details of assessor blinding. Five trials reported the blinding of caregivers; six trials had patient blinding. Only one trial reported the presence of allocation concealment.

Primary outcome: There was a non-significant increase in overall survival for gemcitabine chemotherapy plus molecular-targeted therapies (HR 0.941, 95% CI 0.877 to 1.010; seven RCTs) compared with gemcitabine alone. A sensitivity analysis was performed excluding the only trial with a significant effect (median survival 6.24 months with molecular targeted therapies and 5.91 months without, so it was considered an outlier); the result was not significant.

Secondary outcomes: There was a significant improvement in progression-free survival (HR 0.862, 95% CI 0.797 to 0.933; five RCTs) and overall response rate (OR 1.356, 95% CI 1.054 to 1.744; seven RCTs) for gemcitabine plus molecular-targeted therapies compared with versus gemcitabine alone. However, gemcitabine plus molecular-targeted therapies versus gemcitabine alone showed significantly more common grade 3 and 4 toxicities (OR 1.793, 95% CI 1.508 to 2.132; seven RCTs) and a significantly higher risk of treatment-related mortality (grade 5 toxicities; (OR 2.192, 95% CI 1.198 to 4.009; six RCTs).

Sensitivity analyses for study quality did not change the results.

The funnel plot showed no evidence of publication bias.

Palliation of pancreatic cancer with gemcitabine plus molecular-targeted therapies did not provide a significant survival benefit compared with gemcitabine alone and was associated with increased grade 3 and 4 toxicities and, more importantly, grade 5 toxicities. It did improve progression-free survival and overall response rate.

The review addressed a well-defined question for study design, interventions and outcomes, although eligible participants were less clearly defined. Relevant databases were searched; the search included unpublished trials. Publication bias was assessed and no evidence was found. Generally, efforts were made to reduce error and bias in the review process.

Trial quality was assessed using suitable criteria and was generally moderate to high. Some relevant study details were reported, but definitions of advanced pancreatic cancer and relevant heterogeneity data were not reported; this, along with with few participant details, made it difficult to assess the analysis decisions. The synthesis performed may not have been appropriate in view of the diverse nature of the molecular-targeted therapies used. Limitations recognised by the reviewers included the diverse nature of the molecular-targeted therapies in their structures and specific targets, and the small number of included trials (although they were adequate in size).

The authors' conclusions should be treated with caution given the variation in the molecular-targeted therapies used and the small numbers of included trials.

Practice: The authors proposed that the findings of the review did not support the use of available molecular-targeted therapies in combination with gemcitabine for the palliation of patients affected by advanced pancreatic cancer.

Research: The authors recommended that cost effectiveness should be an outcome considered in further studies of the addition of molecular-targeted therapies to other cytotoxic agents in the treatment of pancreatic cancer. They also recommended further RCTs on combining molecular-targeted therapies with oxaliplatin, irinotecan, leucovorin and fluorouracil (FOLFIRINOX), a promising combination of drugs.

Not stated.