To evaluate three technologies for the management of advanced colorectal cancer: (1) first-line irinotecan combination [with 5-fluorouracil (5-FU)] or second-line monotherapy; (2) first- or second-line oxaliplatin combination (again, with 5-FU); and (3) raltitrexed, where 5-FU is inappropriate. To examine the role of irinotecan and oxaliplatin in reducing the extent of incurable disease before curative surgery (downstaging).
Ten electronic bibliographic databases covering the period up to August 2004.
Searches identified existing studies of the effectiveness and economics of the technologies and any studies that evaluated any of the indications outlined above were included. Data were extracted and assessed generic components of methodological quality. Survival outcomes were meta-analysed.
Seventeen trials were found, of varying methodological quality. Compared with 5-FU, first-line irinotecan improved overall survival (OS) by 2-4 months (p=0.0007), progression-free survival (PFS) by 2-3 months (p<0.00001) and response rates (p<0.001). It offered a different toxicity profile and no quality of life (QoL) advantage. However, second-line irinotecan compared with 5-FU improved OS by 2 months (p=0.035) and PFS by 1 month (p=0.03), and provided a better partial response rate, but with more toxicities and no QoL advantage. Compared with second-line best supportive care, irinotecan improved OS by 2 months (p=0.0001), had a different toxicity profile and maintained baseline QoL longer, but with no overall difference. The addition of oxaliplatin to second-line 5-FU is associated with a borderline significant improvement in overall survival (p<0.07); a significantly higher response rate (<0.0001); and more serious toxicities. There is no evidence for a significant difference in QoL. Schedules with treatment breaks may not reduce clinical effectiveness but reduce toxicity. The addition of oxaliplatin to second-line 5-FU also saw no improvement in OS (p<0.07), better PFS (by 2.1 months, p=0.0001), an 8.9% higher response rate (p<0.0001), more toxicities and no QoL advantage. There was no significant difference in OS or PFS between first-line irinotecan and oxaliplatin combinations except when 5-FU was delivered by bolus injection, when oxaliplatin provided better OS (p=0.032) and response rates (p=0.032), but not PFS (p=0.169). The regimens had different toxicity profiles and neither conferred a QoL advantage. When compared to 5-FU, raltitrexed is associated with no significant difference in overall or progression-free survival; no significant difference in response rates; more vomiting and nausea, but less diarrhoea and mucositis; no significant difference in, or worse QoL. Raltitrexed treatment was cut short in two out of four included trials due to excess toxic deaths. 5-FU followed by irinotecan was inferior to any other sequence. First-line irinotecan/5-FU combination improved OS and PFS, although further unplanned therapy exaggerated the OS effect size. Staged combination therapy (combination oxaliplatin followed by combination irinotecan or vice versa) provided the best OS and PFS, although there was no head-to-head comparison against other treatment plans. In the only trial to use three active chemotherapies in any staged combination, median OS was over 20 months. In another study, the longest median OS from a treatment plan using two active agents was 16.2 months. Where irinotecan or oxaliplatin were used with 5-FU to downstage people with unresectable liver metastases, studies consistently showed response rates of around 50%. Resection rates ranged from 9 to 35% with irinotecan and from 7 to 51% with oxaliplatin. In the one study that compared the regimens, oxaliplatin enabled more resections (p=0.02). Five-year OS rates of 5-26% and disease-free survival rates of 3-11% were reported in studies using oxaliplatin. Alone or in combination, 5-FU was more effective and less toxic when delivered by continuous infusion. Existing economic models were weak because of the use of unplanned second-line therapies in their trial data: the survival benefits in patients on such trials cannot be uniquely attributed to the allocated therapy. Consequently, the economic analyses are either limited to the use of PES (at best, a surrogate outcome) or are subject to confounding. Weaknesses in cost components, the absence of direct in-trial utility estimates and the limited use of sensitivity analysis were identified. Improvements to the methodologies used in existing economic studies are presented. Using data from two trials that planned treatment sequences, an independent economic evaluation of six plans compared with first-line 5-FU followed on progression by second-line irinotecan monotherapy (NHS standard treatment) is presented. 5-FU followed on progression by irinotecan combination cost 13,174 pounds per life-year gained (LYG) and 10,338 pounds per quality-adjusted life-year (QALY) gained. Irinotecan combination followed on progression by additional second-line therapies was estimated to cost 12,418 pounds per LYG and 13,630 pounds per QALY gained. 5-FU followed on progression by oxaliplatin combination was estimated to cost 23,786 pounds per LYG and 31,556 pounds per QALY gained. Oxaliplatin combination followed on progression by additional second-line therapies was estimated to cost 43,531 pounds per LYG and 67,662 pounds per QALY gained. Evaluations presented in this paragraph should be interpreted with caution owing to missing information on the costs of salvage therapies in the trial from which data were drawn. Irinotecan combination followed on progression by oxaliplatin combination cost 12,761 pounds per LYG and 16,663 pounds per QALY gained. Oxaliplatin combination followed on progression by irinotecan combination cost 16,776 pounds per LYG and 21,845 pounds per QALY gained. The evaluation suggests that these two sequences have a cost-effectiveness profile that is favourable in comparison to other therapies currently funded by the NHS. However, the differences in OS observed between the two trials from which data were taken may be a result of heterogeneous patient populations, unbalanced protocol-driven intensity biases or other differences between underlying health service delivery systems.
Treatment with three active therapies appears most clinically effective and cost-effective. NHS routine data could be used to validate downstaging findings and a meta-analysis using individual patient-level data is suggested to validate the optimal treatment sequence.