Reader Comments

Post a new comment on this article

Dubious assumptions main driver behind results

Posted by pietervanbaal on 26 Nov 2010 at 15:14 GMT

Having read the paper two questions immediately arise that seriously cast doubt on the main conclusions of the paper.

First of all, although screening may result in earlier diagnosis and therefore in a more favorable CRC stage distribution this does not necessarily result in an increase in life expectancy. At it worst, this does not result in any gain in life expectancy but only in an increase in the period lived with the diagnosis of CRC. The authors do not mention how they correct for this so-called ‘lead-time’ bias.

Secondly, under the strong assumption that screening does indeed lead to prolonged life, incremental health care costs are underestimated and incremental QALYs are overestimated by the authors. If people live longer due to the screening interventions they are exposed to other risks and diseases besides CRC in their added life years. These competing risks decrease quality of life and increase health care utilization. The effects of competing risks in life years gained on health care costs and quality of life are ignored by the authors. Since the life years gained occur at older ages these mechanisms can not be ignored if the goal is to estimate the costs and benefits of the screening strategies.

No competing interests declared.

RE: Dubious assumptions main driver behind results

sjheitman replied to pietervanbaal on 10 Dec 2010 at 19:09 GMT

Dr van Baal raises two concerns with our conclusions; the first being that screening does not lead to prolonged life (i.e. screening only resulting in lead-time bias), the second being that incremental health care costs are underestimated and incremental QALYs are overestimated for survivors who would not have survived without screening. With respect, we disagree with both points. However, this is a complex model presented within the context of a relatively brief manuscript and thus we acknowledge that part of these concerns may relate to lack of clarity of study methods on our part.

We do not feel that our results are due to lead-time bias. Lead time bias in the case of a screening intervention occurs when a test appears to prolong survival from the disease when in fact it only results in an earlier diagnosis. In fact, a number of controlled studies (cited below) have shown that colorectal cancer (CRC) screening does result in a reduction in CRC-related mortality. CRC is curable when identified at an early stage and it is widely held that it can be prevented through the removal of pre-cancerous adenomatous polyps. There has been recent debate whether this benefit in fact applies to neoplasia in the entire colon or just the left colon, but CRC screening does appear to lower CRC-related mortality – this is reflected in additional QALYs gained within the screening strategies of our model.

While we agree with Dr. van Baal that living longer through a reduction in CRC-related mortality exposes one to other competing risks and diseases in addition to added health care costs, it is not true that living longer simply leads to a reduction in one’s quality of life. It is important to note that our model does in fact account for competing mortality risks. Those who “escape or survive” CRC in our model remain at risk of the background age-dependent mortality rates, which increase as individuals age. Moreover, while quality of life may be lower for survivors of CRC (and this is accounted for in our analysis), it is unlikely that quality of life is lower for patients with adenomas (i.e. precancerous lesions) which are removed during therapeutic colonoscopy. Lastly, we acknowledge that for survivors of CRC we did not account for future health care costs that are unrelated to CRC care (we only accounted for “CRC-related” health care costs in survivors). However, as Dr van Baal is aware, this is a controversial area within economic evaluation. Nevertheless, both the costs and health benefits in our model were discounted according to standard practice, and as such, it is unlikely that inclusion of future unrelated health care costs for additional survivors would impact our results. Therefore, as best as possible within the context of a model, we feel our results are an accurate reflection of the cost-effectiveness of CRC screening, within the limits of available evidence.

1. Atkin WS, Edwards R, Kralj-Hans I, et al. Once-only flexible sigmoidoscopy screening in prevention of colorectal cancer: a multicentre randomised controlled trial. Lancet. May 8 2010;375(9726):1624-1633.

2. Hardcastle JD, Chamberlain JO, Robinson MH, et al. Randomised controlled trial of faecal-occult-blood screening for colorectal cancer. Lancet. Nov 30 1996;348(9040):1472-1477.

3. Kronborg O, Fenger C, Olsen J, Jorgensen OD, Sondergaard O. Randomised study of screening for colorectal cancer with faecal-occult-blood test. Lancet. Nov 30 1996;348(9040):1467-1471.

4. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. N Engl J Med. May 13 1993;328(19):1365-1371.

No competing interests declared.

RE: RE: Dubious assumptions main driver behind results

pietervanbaal replied to sjheitman on 13 Dec 2010 at 10:34 GMT

While I am glad that the authors took my comments seriously, I am under the impression that they did not grasp the nature of my comments.

With respect to survival benefits. First, the authors should have provided a more in depth account of the literature surrounding the survival benefits of colorectal screening. A Cochrane review concluded that although screening reduces CRC mortality there is no evidence that it reduces all cause mortality. Secondly, if there is evidence that indeed screening increases life expectancy then simply taking survival stratified by stage derived from data from a non-screened population is the wrong way to model this.

With respect to competing risks and quality of life. Just as the mortality of competing risks is highly age dependent, this is also the case for quality of life: on average a life year gained at older ages results in less QALYs than a life year gained at lower ages. This mechanism is not captured in the model.

With respect to competing risks and health care costs. Although there is a debate on the inclusion of medical costs in life years gained, economic arguments to exclude them are ill founded (see recent discussion in the Journal of Health Economics). That the main results would probably not have been different if these costs would have been included is probably also a bit premature. Although in an absolute sense these costs may not be be very high they may outweigh the potential savings. As a result, it may well be the case that the screening strategies would turn out not to be cost saving anymore.

1. Hewitson P, Glasziou P, Irwig L, Towler B, Watson E.
Cochrane Database Syst Rev. 2007 Jan 24;(1):CD001216.
Screening for colorectal cancer using the faecal occult blood test, Hemoccult.

2. Lee RH. Future costs in cost effectiveness analysis.
J Health Econ. 2008 Jul;27(4):809-18.

3. Meltzer D. Response to "Future costs and the future of cost-effectiveness analysis". J Health Econ. 2008 Jul;27(4):822-5.

4. Feenstra TL, van Baal PH, Gandjour A, Brouwer WB. Future costs in economic evaluation. A comment on Lee.
J Health Econ. 2008 Dec;27(6):1645-9

5. de Kok IM, Polder JJ, Habbema JD, Berkers LM, Meerding WJ, Rebolj M, van Ballegooijen M. The impact of healthcare costs in the last year of life and in all life years gained on the cost-effectiveness of cancer screening. Br J Cancer. 2009 Apr 21;100(8):1240-4.

No competing interests declared.

RE: RE: RE: Dubious assumptions main driver behind results

sjheitman replied to pietervanbaal on 17 Dec 2010 at 03:59 GMT

Indeed, we did comprehend the nature of the comments made, but question their overall impact.

1) Once again, we acknowledge that some methodological detail could not be developed in full given the nature of our paper in a general medical journal. The CRC screening literature is extremely vast and a more in depth account of the literature was not the aim of our manuscript. We are aware of the Cochrane review published in 2007 and agree that this review and others for that matter have not shown CRC screening via guaiac-based FOBT to reduce all-cause mortality. Please note that we have not made this claim anywhere in our article. Nevertheless, two points are worth mentioning. First, FITs are more sensitive than g-FOBTs, which might translate into reductions in all-cause mortality, even if relatively small. Second, a reduction in CRC-related mortality is still relevant given that many jurisdictions remain committed to treating CRC, the costs of which are on an upward trajectory.

2) We agree that a life year gained at older ages results in less QALYs than a life year gained at lower ages. However, we would like to remind the readership that the population that was modeled is average-risk and screen eligible, and thus by definition is healthy without excessive co-morbidity. Furthermore, screening stops in our model at the age of 75.

3) Once again, we did account for competing health care risks in our model. Furthermore, we agree with the comment that in an absolute sense the competing costs would be unlikely to be very high, especially when discounted over time and considering that a large proportion would be effectively canceled out across strategies.

Our results are based on a complex model that was carefully validated. Extensive sensitivity analyses indicate that our results are robust. Our base case analysis suggests that CRC screening using FIT for average risk individuals is dominant over no screening, but at a minimum is highly attractive. While we appreciate the interest and criticism of our article, we feel that the claim that our results are mainly driven by dubious assumptions to be potentially inaccurate.

No competing interests declared.

RE: RE: RE: RE: Dubious assumptions main driver behind results

pietervanbaal replied to sjheitman on 17 Dec 2010 at 16:12 GMT

1) The only way to gain life years is by assuming a reduction in all cause mortality. Assuming a decline in CRC mortality and that other cause mortality is unaffected by the intervention is equivalent to assuming an effect on all cause mortality.
Furthermore, I understand that is difficult to present a model in full in a paper. However, given these difficulties, I would have expected supplementary online appendices on the model and, preferably, peer reviewed publications on the structure of this particular model. More importantly, I would have expected more in depth answers to my concerns instead of merely stating that the model is complex.

2) The screening age is irrelevant for my criticism. The relevant age is the age at which the life years are added to someone's life due to screening.

3) Competing health risks are addressed in the model only with respect to mortality. Not with respect to health care costs and quality of life.

Finally, I was a bit baffled that the authors questioned the overall impact of my comments on the outcomes of their analysis. The implications of my comments could well be that colorectal cancer screening does not lead to prolonged life and even if it does, it will probably not result in cost savings.

No competing interests declared.