HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Correspondence to:

ARTICLES: P. J. Kelly, J. Rosand, J. P. Kistler, V. E. Shih, S. Silveira, A. Plomaritoglou, and K. L. Furie Homocysteine, MTHFR 677CT polymorphism, and risk of ischemic stroke: Results of a meta-analysis Neurology 2002; 59: 529-536 [Abstract] [Full text] [PDF]

Correspondence: Submit a response to this article

Correspondence published:

Reply to Letter to the Editor

Peter J Kelly, Jonathan Rosand, Vivian Shih, J. Philip Kistler and Karen L. Furie   (13 November 2002)

Homocysteine, MTHFR 677CT polymorphism, and risk of ischemic stroke: Results of a

Una Fallon, Yoav Ben-Shlomo   (13 November 2002)

Reply to Letter to the Editor 13 November 2002
Peter J Kelly Massachusetts General Hospital Boston MA, Jonathan Rosand, Vivian Shih, J. Philip Kistler and Karen L. Furie Send Correspondence to journal: Re: Reply to Letter to the Editor PJKELLY{at}PARTNERS.ORG Peter J Kelly, et al.	Drs. Fallon and Ben-Shlomo mainly address the second analysis in our paper. As others and we have emphasized, there are several decision- points, which must be addressed when performing meta-analysis of observational data, and well-described caveats, which apply when interpreting the result [1]. The didactic tone of their discussion is interesting, as no consensus exists in the field on a single standard methodology to resolve these decisions. A primary aim of our analysis was to provide reliable summary estimates of tHcy-associated stroke risk which could be clearly communicated to and applied by practicing physicians, and which were derived from studies in which cases were well-characterized. The practical utility of previous analyses has been limited, as the results have been difficult to apply in a clinical context. In some cases, investigators have pooled studies of stroke with those of heart disease, or reported results in terms which are not readily communicated to clinicians and patients, such as the log-transformed or raw increase in odds ratio (OR) per 1- or 5-µmol/L increase in tHcy [2, 3]. When counseling patients, there are intuitive benefits to expressing disease risk associated with a continuous variable in terms of "high" or "normal" levels, as is commonly done with blood pressure and cholesterol. We summarized the relationship between tHcy and stroke risk in terms of the case-control difference in arithmetic mean and the OR. These measures have the advantages of being straightforward and readily interpreted in clinical practice, and have been used in earlier similar analyses [4]. As we emphasized in the text and Table 2, a disadvantage of this approach is that no consistent threshold for high tHcy has been applied across published studies. However, the differences in threshold values (most 12- 20µmol/L) are of relatively little clinical significance when viewed in the context of the range of tHcy values (>100µmol/L) associated with stroke in patients with acquired or inherited disorders of tHcy metabolism [5]. The result is interpreted as an estimate of stroke risk in individuals with elevated tHcy in the mild or low-moderate range, as is common in the general population. We agree that the OR from a nested case-control study approximates the hazard ratio from cohort studies. In the original manuscript submission, we combined cohort and nested case-control studies in a single analysis, but separated them by reported effect measure at the request of the reviewers. The combined pooled risk estimate of cohort and nested case -control studies was 1.5 (95% CI 1.2, 1.89, p<0.001). To facilitate the interpretation of our results, we minimized statistical manipulation of published data. As described in the Methods section, we excluded studies, which reported geometric means, as the standard deviations may not be accurately calculated. The study by Alfthan et al. [6] fulfilled inclusion criteria for the analysis of arithmetic means, but was excluded from our second analysis as a dichotomous OR was not provided and could not be accurately derived. Detailed exploratory analyses of between-study heterogeneity, while of theoretical interest, have limited clinical utility and were not a primary aim of our analysis. References: 1. Petitti DB. Meta-Analysis, Decision Analysis, and Cost- Effectiveness Analysis. Methods for Quantitative Synthesis in Medicine. 2nd ed. New York, Oxford University Press, 2000. 2. Ford ES, Smith SJ, Stroup DF, Steinberg KK, Mueller PW, Thacker SB. Homocyst(e)ine and cardiovascular disease: a systematic review of the evidence with special emphasis on case-control studies and nested case- control studies. Int.J Epidemiol. 2002;31:59-70. 3. Brattstrom L, Wilcken DEL, Ohrvik J, Brudin L. Common methylenetetrahydrofolate reductase gene mutation leads to hyperhomocysteinemia but not to vascular disease. Circulation 1998;98:2520 -2526. 4. Boushey CJ, Beresford S, Omenn GS, Motulsky AG. A quantitative assessment of plasma homocysteine as a risk factor for vascular disease: Probable benefits of increasing folic acid intakes. JAMA 1995;274:1049- 1057. 5. Kelly PJ, Furie KL, Kistler JP, et al. Stroke in young patients with hyperhomocysteinemia due to cystathionine beta-synthase deficiency. Neurology (in press). 6. Alfthan G, Pekkanen J, Jauhiainen M, Pitkaniemi J, Karvonen M, Tuomilehto J et al. Relation of serum homocysteine and lipoprotein(a) concentrations to atherosclerotic disease in a prospective Finnish population based study. Atherosclerosis 1994;106:9-19.
Homocysteine, MTHFR 677CT polymorphism, and risk of ischemic stroke: Results of a 13 November 2002
Una Fallon University of Bristol United Kingdom, Yoav Ben-Shlomo Send Correspondence to journal: Re: Homocysteine, MTHFR 677CT polymorphism, and risk of ischemic stroke: Results of a una.fallon{at}bristol.ac.uk Una Fallon, et al.	Unlike randomised controlled trials, meta-analysis of observational studies remains controversial because bias and confounding contribute to between study heterogeneity. The meta-analysis by Kelly et al. [1] is further complicated by other methodological shortcomings. A meta-analysis of a continuous exposure should standardise the estimate of effect for a uniform change in that exposure. For example other meta-analyses of total plasma homocysteine (tHcy) and cardiovascular disease have expressed the pooled odds ratio in terms of a 5mmol change in tHcy [2, 3]. Kelly et al. states that it is clinically useful to express the risk of stroke associated with tHcy in a dichotomous manner. However, they do not state the chosen cut-off for defining hyperhomocysteinaemia. In the supplementary tables on the Neurology website, the definition of hyperhomocysteinaemia is different for each study. By combining estimates based on these different definitions, the authors have rendered their pooled effect estimate meaningless. They also pooled studies on the basis of the reported measure e.g. odds ratios, hazard ratio or arithmetic means rather than on methodological rigour or study design. For example, case control studies and nested case control studies are pooled because they both report odds ratios. Case control studies are subject to selection bias whereas nested case control studies, essentially an efficient use of prospective cohort data, are not. This is a logical source of possible heterogeneity between studies. In this scenario, it may be more appropriate to pool nested case control studies and cohort studies. An odds ratio from a nested case control study is a close approximation to the true risk or hazard ratio as long as there has been appropriate sampling of controls. Pooling based on arithmetic means, selectively excluded valid analysis of studies of geometric means and again resulted in different study designs being combined. The authors also pooled estimates which were adjusted for different confounders (table 4A) and omitted one major nested case control study [4], the first 'null' study, from two of their analyses (Figure 4), without any justification. It is tempting to pool numbers without thinking about important sources of heterogeneity between studies, particularly when underpowered statistical tests of heterogeneity show no effect. It would be more productive to examine these differences in detail using sensitivity analysis and meta-regression5. This may provide insights into why results from observational studies of homocysteine and stroke are so different and help produce a more valid estimate rather than a potentially misleading one. References: 1. Kelly PJ, Rosand J, Kistler JP, Shih VE, Silveira S, Plomaritoglou A et al. Homocysteine, MTHFR 677C-->T polymorphism, and risk of ischemic stroke: Results of a meta-analysis 1. Neurology 2002;59:529-36. 2. Ford ES, Smith SJ, Stroup DF, Steinberg KK, Mueller PW, Thacker SB. Homocyst(e)ine and cardiovascular disease: a systematic review of the evidence with special emphasis on case-control studies and nested case- control studies. Int.J Epidemiol. 2002;31:59-70. 3. Ueland PM, Refsum H, Beresford SA, Vollset SE. The controversy over homocysteine and cardiovascular risk. Am.J.Clin.Nutr. 2000;72:324-32. 4. Alfthan G, Pekkanen J, Jauhiainen M, Pitkaniemi J, Karvonen M, Tuomilehto J et al. Relation of serum homocysteine and lipoprotein(a) concentrations to atherosclerotic disease in a prospective Finnish population based study. Atherosclerosis 1994;106:9-19. 5. Egger M, Davey Smith G, Schneider M. Systematic reviews of observational studies. In Egger M, Davey Smith G, Altman DG, eds. Systematic Reviews in Health Care. Meta-analysis in context, pp 211-27. London: BMJ Publishing Group, 2001.