Meta-analyses of chemotherapy for locally advanced and metastatic pancreatic cancer: results of secondary end points analyses

In advanced pancreatic cancer, level one evidence has established a significant survival advantage with chemotherapy, compared to best supportive care. The treatment-associated toxicity needs to be evaluated. This study examines the secondary outcome measures for chemotherapy in advanced pancreatic cancer using meta-analyses. A systematic review was undertaken employing Cochrane methodology, with search of databases, conference proceedings and trial registers. The secondary end points were progression-free survival (PFS)/time to progression (TTP) (summarised using the hazard ratio (HR)), response rate and toxicity (summarised using relative risk). There was no significant advantage of 5FU combinations vs 5FU alone for TTP (HR=1.02; 95% CI=0.85–1.23) and toxicity. Progression-free survival (HR 0.78; CI 0.70–0.88), TTP (HR=0.85; 95% CI=0.72–0.99) and overall response rate (RR=0.56; 95% CI=0.46–0.68) were significantly better for gemcitabine combination chemotherapy, but offset by the greater grade 3/4 toxicity thrombocytopenia (RR=1.94; 95% CI=1.32–2.84), leucopenia (RR=1.46; 95% CI=1.15–1.86), neutropenia (RR=1.48; 95% CI=1.07–2.05), nausea (RR=1.77; 95% CI=1.37–2.29), vomiting (RR=1.64; 95% CI=1.24–2.16) and diarrhoea (RR=2.73; 95% CI=1.87–3.98). There is no significant advantage on secondary end point analyses for administering 5FU in combination over 5FU alone. There is improved PFS/TTP and response rate, with gemcitabine-based combinations, although this comes with greater toxicity.

Advanced pancreatic cancer has a poor prognosis, with a median survival of 2 -6 months for metastatic disease and 6 -11 months for locally advanced disease (Cancer Research, 2006). Chemotherapy with fluoropyrimidines, gemcitabine, either alone or in combination with other agents (Rocha Lima and Flores, 2006), and chemoradiation are all used in the palliative setting (Mancuso et al, 2006). Overall survival meta-analyses, using relative risk (Yip et al, 2006) or the hazard ratio (HR) (Fung et al, 2003;Sultana et al, 2007), have established a role for chemotherapy over best supportive care. Questions have arisen as to the cost at which this survival advantage is gained, in particular, the toxicity profile. Following from our previous survival meta-analysis (Sultana et al, 2007), we present the results of the secondary outcome measures meta-analysis.
There has only been one fully published meta-analysis evaluating secondary outcome measures, with no pooling of the results of these end points (Yip et al, 2006). Other published reports have assessed this only for the comparison of gemcitabine combinations vs gemcitabine. (Liang, 2005;Milella et al, 2006;Heinemann et al, 2006a;Xie et al, 2006a, b;Bria et al, 2007;Heinemann et al, 2007). To fully evaluate the risks vs the benefits of treatment, a comprehensive evaluation including assessment of several composite end points is required.

METHODS
Detailed description of the methodology of the systematic review has already been described (Sultana et al, 2007).
The secondary outcome measures evaluated were progressionfree survival (PFS -time from randomisation to progression or death) or time to progression (TTP -time from randomisation to disease progression), overall response rate (ORR -number of partial and complete responses) and toxicity (as published by the trialists, was recorded, with the most frequently reported events analysed).
Individual trial level time to event data (PFS/TTP) were summarised by the log HR and its variance was approximated using previously reported methods (Parmar et al, 1998;Williamson et al, 2002). Trial level log HRs and their variances were pooled using an inverse variance, weighted average and results presented as a HR and 95% confidence interval.
Dichotomous data (ORR and toxicity) were summarised using relative risks and 95% confidence intervals and pooled using the Mantel -Haenszel method for combining trials (Deeks et al, 2001). Heterogeneity was assessed by visual inspection of the Forrest plot, the Cochran's w 2 test (using a 10% significance level, in view of the low power of tests for heterogeneity (Paul and Donner, 1992)) and interpretation of the I 2 statistic (percentage of variation due to heterogeneity with higher values indicating a greater degree of heterogeneity) (Deeks et al, 2004). A fixed effect approach was adopted unless there was evidence of significant unexplained heterogeneity in which case a random effects approach was used.

RESULTS
Results are presented for the comparisons with adequate data to assess the secondary outcome measures.
Overall response rate appeared better in the gemcitabine arm; however, the wide confidence interval suggests a benefit for either gemcitabine or 5FU (one trial; 126 patients; RR ¼ 0.14; 95% CI ¼ 0.01 -2.66). In the Burris trial (Burris et al, 1997), haematological toxicity was seen more frequently following gemcitabine therapy (grades 3 and 4 neutropenia in 25% of gemcitabine and 4.9% of 5FU patients; Po0.001).

DISCUSSION
5FU combinations did not prolong TTP over 5FU alone, despite significantly better response rate with the former. The study of Yip et al (2006) assessed the parameters described in our analyses, but did not pool the results unlike our approach. In the two trials that had assessed PFS, the overall summary estimate favoured 5FU combination chemotherapy, but there was significant inter-trial heterogeneity. This may be due to the differences in dosing. The dose of 5FU administered was lower in the Maisey et al (2002) study (300 mg m À2 day À1 in both arms) compared to the Ducreux et al (2002) study (500 mg m À2 day À1 used in the single-agent arm and 1000 mg m À2 used in the combination arm).
As overall survival is a better indicator of efficacy than response rate (Maisey et al, 2002), the evidence from these end points, interpreted alongside the overall survival result (Sultana et al, 2007), do not support the use of 5FU combinations over 5FU single agent.
Meta-analyses of the secondary end points were not possible in the gemcitabine vs 5FU comparison, as these results were only available for one randomised trial.    Figure 5 Results for gemcitabine vs gemcitabine-based combination chemotherapy -response rate.

Chemotherapy in advanced pancreatic cancer A Sultana et al
Previous meta-analyses of secondary end points evaluating gemcitabine-based combinations vs gemcitabine employed differing survival analyses methodology (Liang, 2005;Heinemann et al, 2006a;Milella et al, 2006;Xie et al, 2006a). In contrast to these reports, our survival analyses were conducted using the HR, which is the ideal measure for time-to-event analyses, as it accounts for both censoring of data and the time it takes for the event (such as death or progression) to occur (Parmar et al, 1998).
For gemcitabine-based chemotherapy vs gemcitabine alone, our findings of improved PFS/TTP are in agreement with the metaanalyses of Xie et al (2006b). Better ORR with the combination regimens was in keeping with the studies of Xie et al and Milella Legend: n = number of toxicity events N = total number of patients Figure 6 Results for gemcitabine vs gemcitabine-based combination chemotherapy -haematological toxicity.
Chemotherapy in advanced pancreatic cancer A Sultana et al et al (Xie et al, 2006b), while increased toxicity profile was noted by Xie et al (2006b). The meta-analyses that examined gemcitabine plus a platinum agent vs gemcitabine alone found better PFS/TTP in the combination arm (Xie et al, 2006a;Heinemann et al, 2007), significant improvement in ORR (Heinemann et al, 2007) and greater toxicity (Xie et al, 2006a). We have done our utmost to cover most reported end points in the randomised controlled trials. We could not address quality of life due to the different methods used for reporting quality of life. Although we have pooled the response rate and adverse events data across studies to permit a clinically relevant analysis, reporting of these parameters varied. Response rates were reported using clinical parameters, the WHO and RECIST criteria, whereas the CTC, WHO and ECOG scales were used for toxicity data.
To conclude, there is insufficient evidence to suggest a TTP, response rate and toxicity advantage in administering 5FU in combination with other chemotherapy agents over 5FU alone. There is a small but significant TTP/PFS advantage, as well as improved response rate, with gemcitabine-based combinations, and this provides a justification for the use of these agents, despite their greater toxicity. An area for further randomised controlled trials to assess is which gemcitabine-based combination chemotherapy regimens are least toxic, while retaining all the other advantages of the combination approach. Legend: n = number of toxicity events N = total number of patients Figure 7 Results for gemcitabine vs gemcitabine-based combination chemotherapy -gastrointestinal toxicity.
Chemotherapy in advanced pancreatic cancer A Sultana et al