analysis (2), although other authorities recommend the need forsummary measures for the best estimate of the impact of anintervention, albeit that correct methodological techniques areused to investigate the differences among single trials (3).

Furthermore, Moreno et al. (1) tried to assess, by means oflinear regression analysis, the possible association of several angio-graphic variables with the benefit of stents, finding a significantinverse relationship between reference vessel diameter (RVD) andthe risk reduction of angiographic restenosis after stent placementwith respect to balloon angioplasty (i.e., the smaller the vessel, thelarger the benefit of stenting). However, we believe the correctmethodological tool to address whether a covariable may have asignificant effect on the outcome in a meta-analysis is meta-regression (4). Using this technique (which weights each studyaccording to its statistical weight [i.e., the inverse of the variance])to analyze the very same data presented in their study, we did notfind any significant relation between RVD and risk reduction ofangiographic restenosis after stent placement.

Moreover, another recent trial, published only as an abstract (5),did not confirm the investigators’ hypothesis. Indeed, in front ofthe smallest mean RVD of all the trials on small-vessel disease(2.17 mm), there was a nonsignificantly increased risk of restenosis(relative risk 1.14 [95% confidence interval 0.89 to 1.48]) afterstenting.

Finally, meta-analytic techniques allow quantitative assessmentof treatment effects from pooled data. With the widening of theiruse and because of potential errors, improper analyses may result inmisleading conclusions; thus, optimal methodological proceduresshould be utilized to validate findings.

*Pierfrancesco Agostoni, MDGiuseppe G. L. Biondi-Zoccai, MDAntonio Abbate, MD

*Catheterization LaboratorySection of CardiologyDepartment of Biomedical and Surgical SciencesUniversity of VeronaPiazzale A. Stefani 137126 VeronaItalyE-mail: agostonipf@genie.it

doi:10.1016/j.jacc.2004.10.022

REFERENCES

1. Moreno R, Fernandez C, Alfonso F, et al. Coronary stenting versusballoon angioplasty in small vessels. A meta-analysis from 11 random-ized trials. J Am Coll Cardiol 2004;43:1964–72.

2. Petitti DB. Meta-Analysis, Decision Analysis and Cost-EffectivenessAnalysis: Methods for Quantitative Synthesis in Medicine. 2nd ed.New York, NY: Oxford University Press, 2000.

3. Lau J, Ioannidis JP, Schmid CH. Summing up the evidence: one answeris not always enough. Lancet 1998;351:123–7.

4. Thompson SG, Higgins JP. How should meta-regression analyses beundertaken and interpreted? Stat Med 2002;21:1559–73.

5. Hausleiter J, Kastrati A, Mehilli J, et al. A randomized trial comparingphosphorylcholine-coated stents with balloon angioplasty in smallcoronary arteries in patients with symptomatic coronary artery disease.The ISAR–SMART-2 trial (abstr). Circulation 2003;108 SupplIV:IV569.

REPLY

We acknowledge Dr. Agostoni and colleagues for their interest inour work, but we do not agree with their comments. Heterogeneitybetween studies is not a contraindication to quantitative poolinganalysis. When analyzing the effect of a therapeutic mean, twodifferent types of analysis can be performed: random-effect model(between-trial variance and within-trial variance) and fixed-effectmodel (1,2). The first type is used when significant heterogeneityexists, whereas the fixed-effect model assumes that between-trialvariance is zero. As we clearly specified in our Methods, we usedthe random-effect model (3). As Agostoni and colleagues empha-size, controversy abounds concerning the best approach whenheterogeneity exists, but we believe the investigators consider thismethodology correct. In fact, in a very recently published meta-analysis, they concluded that radial access yields to lower proce-dural success than does femoral access, despite having foundsignificant heterogeneity between studies when evaluating this endpoint (4).

Also, it is important to consider the causes of heterogeneityamong trials. In our study, heterogeneity was not due to clinical ordesign discrepancies between the studies, but rather to differencesin the number of patients included (treatment favored balloon onlyin the COMPASS trial (5) and in the study by Park et al. (6), andthese had the lowest number of patients).

The second criticism raised by Agostoni and colleagues is thatthe relationship between risk ratio for restenosis and referencevessel diameter (RVD) should have been adjusted by the inverse ofvariance. Apparently they were unable to find (data not provided)any significant relation between both variables. Accordingly, wehave analyzed our data after adjusting this relation for the inverseof variance (7–9). After analyzing it, the association between RVDand risk reduction remained significant, and in fact the beta-coefficient for RVD was even higher (Y � �6.514 � 2.674 [RVD]� 2.156 [varRR]); p value for RVD � 0.017).

Moreover, ever since we wrote up the study, several randomizedtrials comparing stent and balloon in small vessels have beenreported and even published (10–12). Dr. Agostoni and colleaguesonly mention the recently reported ISAR SMART-2 trial, inwhich coronary stenting was nonsuperior to balloon in smallcoronary vessels. The ISAR–SMART-2, however, has one impor-tant limitation not recognized by Agostoni et al: the very high rateof cross-over from balloon to stent (�40%, in comparison with19% in our meta-analysis) (10). Unfortunately, they do notmention the already published study by Kinsara et al. (12), inwhich restenosis rate decreased from 49% to 30% (balloon andstent, respectively; p � 0.009) in very small vessels (2.09 mm). Thesame occurs with the LASMAL-II trial, in which angiographicrestenosis significantly decreased from 45% to 28%, respectively (p� 0.043) (11).

Interventional cardiology is one of the areas of medicine inwhich most randomized trials are being performed. Because ofthat, we may be tempted to use meta-analytic techniques withoutprofound knowledge of the trials included and the methodologiesemployed. To avoid improper conclusions, it is necessary tounderstand adequately the meta-analytic techniques and theirlimitations, to obtain advice from expert statisticians, and toevaluate thoroughly all the trials included. This methodologyallowed us to conclude that coronary stenting reduces the reste-nosis rate in comparison to balloon angioplasty in small coronaryarteries (3).

324 Correspondence JACC Vol. 45, No. 2, 2005January 18, 2005:318–28

*Raúl Moreno, MD, FESCCristina Fernandez, MDCarlos Macaya, MD, FESC

*Interventional CardiologyCardiovascular InstituteHospital Clı́nico San CarlosMartı́n Lagos, s/n28040 MadridSpainE-mail: raulmorenog@terra.es

doi:10.1016/j.jacc.2004.10.023

REFERENCES

1. Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: acomparison of methods. Stat Med 1999;18:2693–708.

2. Thompson SG. Why and how sources of heterogeneity should beinvestigated. In: Egger M, Smith GD, Altman DG, editors. System-atic Reviews in Health Care: Meta-Analysis in Context. London: BMJBooks, 2001:157–75.

3. Moreno R, Fernandez C, Alfonso F, et al. Coronary stenting versusballoon angioplasty in small vessels. A meta-analysis from 11 random-ized trials. J Am Coll Cardiol 2004;43:1964–72.

4. Agostoni P, Biondi-Zoccai GGL, De Benedictis ML, et al. Radialversus femoral approach for percutaneous coronary diagnostic andinterventional procedures. Systematic overview and meta-analysis ofrandomized trials. J Am Coll Cardiol 2004;44:349–56.

5. Tsuchikane E, Takeda Y, Nasu K, Awata N, Kobayashi T. Balloonangioplasty plus cilostazol administration versus primary stenting ofsmall coronary artery disease: final results of COMPASS. CatheterCardiovasc Interv 2004;63:44–51.

6. Park SW, Lee CW, Hong MK, et al. Randomized comparison ofcoronary stenting with optimal balloon angioplasty for treatment oflesions in small coronary arteries. Eur Heart J 2000;21:1785–9.

7. Normand SL. Meta-analysis: formulating, evaluating, combining, andreporting. Stat Med 1999;18:321–59.

8. Ahlbom A. Biostatistics for Epidemiologists. Boca Raton, FL: LewisPublishers, 1993:64–5.

9. Rothman KJ, Greenland S. Types of epidemiologic studies. In:Rothman KJ, Greenland S, editors. Modern Epidemiology. Boston,MA: Little, Brown, 1986:67–78.

10. Hausleiter J, Kastrati A, Mehilli J, et al. A randomized trial comparingphosphorylcholine-coated stents with balloon angioplasty in smallcoronary arteries in patients with symptomatic coronary artery disease.The ISAR-SMART-2 trial (abstr). Circulation 2003;108 Suppl.IV:IV69.

11. Rodriguez AE, Rodriguez M, Fernandez C, et al. Latin AmericaSmall Vessel Randomized Study in Diabetic Patients (LASMAL-II):clinical and angiographic follow-up data (abstr). J Am Coll Cardiol2004;43 Suppl A:58A.

12. Kinsara AJ, Niazi K, Patel I, Amoudi O. Effectiveness of stents insmall coronary arteries. Am J Cardiol 2003;92:584–7.

Thrombolysis Should BeRegarded as First-Line Therapyfor Prosthetic Valve Thrombosisin the Absence of ContraindicationsI read with great interest the retrospective, multicentric ProstethicValve Thrombolysis-Role of Transesophageal Echocardiography(PRO-TEE) study in a recent issue of the Journal (1). The resultsof the registry appear to confirm our previous and rather conser-vative recommendations (2). According to these guidelines ob-structive thrombi in patients presenting in NYHA functional classIII to IV should be treated with thrombolysis if the surgical risk is

high. However, the basis of these recommendations was aninternational consensus conference held in 1994, when thrombol-ysis of left-sided prosthetic valves was a rather debated approach.A review of 53 studies cited in the published reports indicates theresults of thrombolysis being inferior in the period between 1974and 1995 than between 1996 and 2003: success rate 77% versus90%, embolism rate 13% versus 4%, death rate 7.5% versus 2.5%,while the number of treated episodes was virtually the same (235vs. 234). Recent reports and reviews recommend thrombolysis asthe first-line approach in the management of prosthetic valvethrombosis irrespective of functional class or thrombus size (3–5).

In contrast, contributors to the registry conclude that throm-bolysis should be limited to patients with thrombus size �0.8 cm2

as measured by transesophageal echocardiography (TEE). Thisconclusion was based on a small number of patients coming from14 centers over 16 years. The selection of patients into the registrymay have been biased (1). Moreover, in obstructive cases the fixedthrombus usually cannot be measured within the valve orifice (3,4).

In contrast, mobile nonobstructive thrombi are easily measur-able and in most cases large, at least in length, but thrombolysis ofthese thrombi can be performed with negligible embolic risk (3–5).The role of TEE is particularly essential to reveal thrombus innonobstructive cases (3) and serial TEE is required to monitor theefficacy of thrombolysis (4,6). Assessment of “hemodynamic” and“clinical results” without “thrombolytic results” does not allow theevaluation of treatment results in nonobstructive cases. In sum-mary, despite our previous recommendations I consider use ofthrombolysis in all patients with prosthetic valve thrombosisirrespective of functional class and thrombus size unless contrain-dications exist. Thus, TEE must be performed before and duringtreatment.

*Maria Lengyel, MD, FACC, FESC

*Hungarian Institute of CardiologyHaller 29Budapest, 1096HungaryE-mail: vargakat@kardio.hu

doi:10.1016/j.jacc.2004.10.028

REFERENCES

1. Tong TA, Roudaut R, Özkan M, et al., on behalf of the ProstheticValve Thrombolysis-Role of Transesophageal Echocardiography(PRO-TEE) Registry Investigators. Transesophageal echocardiographyimproves risk assessment of thrombolysis of prosthetic valve thrombosis:results of the international PRO-TEE registry. J Am Coll Cardiol2004;43:77–84.

2. Lengyel M, Fuster V, Keltai M, et al. Guidelines for the managementof left-sided prosthetic valve thrombosis: a role for thrombolytictherapy. J Am Coll Cardiol 1997;30:1521–6.

3. Lengyel M, Vándor L. The role of thrombolysis in the management ofleft-sided prosthetic valve thrombosis: a study of 85 cases diagnosed bytransesophageal echocardiography. J Heart Valve Dis 2001;10:636–49.

4. Ozkan M, Kaymaz C, Kirma C, et al. Intravenous thrombolytictreatment of mechanical prosthetic valve thrombosis: a study using serialtransesophageal echocardiography. J Am Coll Cardiol 2000;35:1881–9.

5. Rinaldi CA, Heppell RM, Chambers JB. Treatment of left-sidedprosthetic valve thrombosis: thrombolysis or surgery?. J Heart Valve Dis2002;11:839–43.

6. Alpert JS. The thrombosed prosthetic valve. J Am Coll Cardiol2003;41:659–60.

325JACC Vol. 45, No. 2, 2005 CorrespondenceJanuary 18, 2005:318–28