This review assessed the effect on survival of adjuvant hepatic arterial infusion chemotherapy after curative hepatic resection for colorectal cancer liver metastases. The authors found a small but non-statistically significant improvement in 1- and 2-year survival. The authors' conclusions appear reliable, although they did not address the impact of variation in treatment regimens and duration.

To assess the effect on survival of using adjuvant hepatic arterial infusion (HAI) of chemotherapy after curative hepatic resection for colorectal cancer in people with liver metastases.

The authors searched MEDLINE, EMBASE, the Cochrane Library and the Cochrane Controlled Trials Register for papers published in peer-reviewed journals, with no limits on publication year. They also contacted experts in the field and reviewed reference lists to identify additional studies. Studies published in a language other than English were excluded.

Study designs of evaluations included in the review

Prospective clinical trials with a control arm were eligible for inclusion.

Specific interventions included in the review

Studies were eligible for inclusion if they compared HAI chemotherapy after curative resection (R0 resection) with a control group. Any form of control was eligible as long as it did not include HAI treatment. All of the included studies used some form of 5-fluorouracil or floxuridine. There were variations in other adjunctive treatments, dosages, routes and durations of HAI (range: 11 weeks to 11 years). The control groups used no treatment, oral tegafur or uracil, and intravenous systemic chemotherapy.

Participants included in the review

Eligible studies included people undergoing curative hepatic resection for colorectal cancer liver metastases. The mean or median age of the participants was similar in all studies, about 60 years.

Outcomes assessed in the review

Studies that presented survival information numerically or as survival curves, at 1 and 2 years after surgery, were included.

How were decisions on the relevance of primary studies made?

The final inclusion of studies was determined by consensus, with adjudication by a third author when required.

Jadad scores were used to assess the validity of randomised controlled trials (RCTs) on the basis of randomisation, blinding, and descriptions of withdrawals and drop-outs. Two authors assessed the studies for validity.

Two authors extracted the data into a standardised evidence table. The data were double-entered. The percentages of patients surviving at 1 and 2 years were extracted from the text, where available, or estimated from survival curves. Standard deviations of survival differences between control and treatment arms were derived from calculated variances. In addition, data on sample size, treatment regimens, median survival and disease-free survival, and response rates were also extracted.

How were the studies combined?

The authors pooled the studies in a meta-analysis using a random- effects model.

How were differences between studies investigated?

The authors performed a sensitivity analysis to assess heterogeneity between the studies by assessing the effect of removing one study at a time.

Seven studies were included in the review: six RCTs (n=628) and one non-randomised prospective clinical trial (n=58). The numbers of participants ranged from 19 to 226 in the individual studies.

The pooled analysis found no significant increase in survival at 1 year or 2 years post surgery for those receiving HAI: the differences in the percentage surviving at 1 and 2 years were 1.8 (95% confidence interval, CI: -4.8, 8.5, P=0.59) and 9.6 (95% CI: -2.2, 21.4, P=0.11), respectively.

The sensitivity analysis found that no single study appeared to influence the observed survival difference.

Most studies were of similar quality with a Jadad score of 3 (5 studies); one study had a score of 2 and one had a score of 1.

HAI of chemotherapy was associated with a small but non significant improvement in survival at 2 years after resection of colorectal liver metastases.

This well conducted review clearly stated research questions, had defined inclusion and exclusion criteria, and followed QUORUM guidelines. All selections and assessments were made in duplicate, thereby reducing the chance of reviewer error or bias. The authors chose to focus only on studies published in peer-reviewed journals and in the English language. This might have resulted in publication bias, which could affect the validity of the conclusions. The rationale for excluding non-English language studies was unclear. The methods used to select and assess the validity of the included studies seemed appropriate.

The authors did not use formal statistical methods to assess heterogeneity, although they assessed the effect of each study on the pooled results in a sensitivity analysis and found no evidence that removing any individual study changed the overall conclusion. The random-effects model used to pool the studies appeared appropriate, although given that the outcome was survival, the hazard ratio would have been the most relevant outcome to pool. The authors did not report why they did not use the hazard ratio or if hazard ratios were reported by any of the studies.

The authors only reported results up to 2 years when most of the studies had a follow-up of 5 years. They stated that future research should include longer term follow-up, but did not explain why they had excluded existing evidence about longer term follow-up from the meta-analysis.

The authors stated that there was variation in the treatment regimens used and in the duration of treatment. It was unclear whether this variation was taken into full consideration when drawing conclusions from the analysis. The authors' conclusions seem appropriate based on the results presented in this review. However, it should be noted that the results presented in the paper have been incorrectly reported as the survival difference in months between treatment groups, when it is the difference in the percentage of patients surviving.

Practice: The authors stated that the routine use of HAI after resection of colorectal metastases to the liver cannot be recommended based on this meta-analysis.

Research: The authors stated that longer term follow-up and a properly powered clinical trial are required to examine the existence of a survival benefit. Further research may also examine the impact of HAI on disease-free survival.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.