The XELOX regimen comprising capecitabine and oxaliplatin requires less intravenous administration than FOLFOX4, comprising 5-FU, folinic acid and oxaliplatin. Therefore it might be expected that it would cost less overall when costs of drug delivery are taken into account. This study showed that this was indeed the case and that, when the fewer cycles of therapy required with XELOX (8 versus 12) were also taken into account, the savings in other costs outweighed the higher purchase price of the XELOX regimen. However, the publicly funded provider in HK subsidised the use of FOLFOX4 and not XELOX, thus encouraging the use of a regimen that resulted in higher use of hospital resources. Those patients willing to bear the cost of the alternative therapy because of the benefits they perceived to themselves were saving resource costs for the provider. This is, of course, because historically FOLFOX4 has been the approved treatment and XELOX, the newer treatment, had a higher purchase price per dose and had not yet been considered for inclusion into the list of approved drugs for subsidy.
This cost analysis was a relatively simple study designed to obtain information to allow a more considered assessment of the two regimens. Therefore, it must be asked whether the information collected is adequate for decision-making. Cost minimisation analysis (CMA) is defined by Drummond et al  as a design used where the consequences are the same with a disparity in costs only and it is relatively low cost, simple and quick to conduct. This is an advantage for countries which have few people with economic modelling skills. However, it depends on the major assumption that the clinical effects of the comparison treatments are the same. Such a finding of similar effectiveness is often extrapolated from 'negative' trials, i.e. where neither drug is found to be superior but these studies may have insufficient power to test equivalence of effects . In the case of these two drugs, there was some evidence from trials designed to determine that the new drug was not inferior to the alternatives [26–28] but we did not subject that evidence to a full review. One concern, when the benefits have not been included in the economic modelling, is that any negative impacts of the drugs may not have been fully accounted for . Therefore in a costing study such as this one, it would be important to consider what negative outcomes might occur in the particular clinical situation and to consider doing a full cost-effectiveness study if there should be any doubts. For example, with a treatment such as XELOX which is taken on an outpatient basis, we might be concerned that reduced frequency of contact with the doctor could delay identification of side effects. Careful monitoring of the patients or specific patient advice plus a contact number to call might help to resolve this question.
A cost analysis is highly dependent on good local data on costs and sufficiently completed medical records from which resource use can be extracted. Hong Kong has a computerised clinical information system which covers all public medical services; cross-referencing was conducted in this study between different sections of the record to ensure completeness of the data captured.
We did not apply any discounting in this study because the time course for each patient was less than a year. All costs relate to the base year of 2009 even though the actual resource use was in the years between 2004 and 2008.
This study found that, despite XELOX having a 64% higher purchase price than FOLFOX4, from the provider's perspective it could be a more cost-effective option in Hong Kong since the total cost of its use was 70% of that of FOLFOX4 when all healthcare resource use was taken into account. This was mainly because of the reduced need for hospital-based intravenous drug delivery, a finding that is consistent with Garrison et al  and Scheithauer et al  in comparisons of the same two drug regimens, as well as Perrocheau et al  when comparing XELOX against FOLFOX6.
The finding of lower cost for XELOX differs from the US results where the XELOX regimen was more expensive than FOLFOX4. This is mainly the result of capecitabine being approximately 8 times the cost of 5-FU in US while in Hong Kong it was just over twice the price. Our sensitivity analysis also showed that capecitabine would have to be four times as expensive for the total resource cost in the XELOX group to equal that in the FOLFOX group.
The relative differences in other costs also combine to make XELOX less costly even when the US data on side-effects and subsequent resource use were modelled together with Hong Kong costs. Taking a societal perspective increased the cost difference between XELOX and FOLFOX4 as also found in the US . This study demonstrates the fact that cost-effectiveness evidence cannot be easily generalised across jurisdictions [22–25, 41–43] and that local cost data are required.
We estimated the total costs of the treatment regimens and this should now provide very useful information for the healthcare providers to determine the impact on their budgets of using one drug rather than the other. Of course, the total costs of a treatment may be spread across different budgets, for example a drug procurement budget and a hospital bed provision budget, including staffing, testing and other items. While there may be higher costs in one budget e.g. drug procurement this might be more than compensated for by decreases in another. We hope that the information provided would allow such specific budget impacts to be estimated. We also hope that our study will encourage decision makers to take a more comprehensive view of costs which might encompass more than one budget, rather than focussing solely on the price of the drug.
Limitations of This Paper
Our study was a retrospective and non-randomised study. It reflects the actual resource use of the subjects' care but, of course, is subject to bias. Allocation to the specific treatment might have been biased by some patient characteristics and, since one drug was substantially self-paid and the other one subsidised, the socio-economic status of the patients could have differed. However we had no information on subjects' socio-economic status for comparison. On the other hand, the comparison of subjects' characteristics found a statistically significant difference only in the proportion with liver metastases and the number was higher in the XELOX group, perhaps increasing the costs in that group. However, with only 30 subjects in each group, the statistical power of such comparisons is low and so we cannot rule out the possibility of some bias in treatment allocation.
Due to the greater number of patients treated on XELOX than FOLFOX4 in Hong Kong in recent years it was not possible to match the two chemotherapy groups in terms of treatment years so some FOLFOX4 patients' treatments date back to 2004-2005 while all XELOX patients were 2006-2008. However, we do not believe that the clinical management practices have changed much over this time period.
We may have underestimated the benefits of XELOX. Shiroiwa et al  compared the two drug regimens in Japan using utility values collected during the clinical trials and estimated quality-adjusted progression-free survival days. They demonstrated better quality of life in the XELOX group which they said enhanced the regimen's cost-effectiveness relative to FOLFOX4.
We had no information on costs of carers and have therefore underestimated the societal costs but including these would predictably make XELOX even less costly than FOLFOX4 because of the time spent attending hospital by the carer as well as the patient.