Biological or Mechanical Prostheses in Tricuspid Position? A Meta-Analysis of Intra-institutional Results

Transcription

1 Biological or Mechanical Prostheses in Tricuspid Position? A Meta-Analysis of Intra-institutional Results CARDIOVASCULAR Giulio Rizzoli, MD, FETCS, Igor Vendramin, MD, Georgios Nesseris, MD, Tomaso Bottio, MD, Cosimo Guglielmi, MD, and Laura Schiavon, DStat Istituto di Chirurgia Cardiovascolare and Centro Informativo di Ateneo, Università di Padova, Padova, Italy Background. Tricuspid valve replacement (TVR) is an uncommon procedure. The use of biological vs mechanical prostheses in TVR has pros and cons. Therefore, we debate the choice between the different types of valves by means of a meta-analysis of studies of the last decade. Methods. The heading tricuspid valve replacement and (bio* or mec*) was used to retrieve studies from Medline, Current Contents, and Embase. Eight out of 11 studies met the preset strict criteria: intra-institutional comparison of results of biological or mechanical TVR. Survival of hospital-discharged patients was recalculated to reduce the effect of unbalanced perioperative risk factors on overall survival. Hazard ratio was obtained from actuarial survival graphics comparison and at-risk groups, according to the method described by Parmar. If missing, the number of patients at risk was approximated assuming constant and noninformative censoring. Hazard pooling was done according to study heterogeneity. Bioprostheses were assumed as the gold standard and mechanical prostheses assumed as the challenging device. Therefore, a hazard more than 1 pointed to a higher risk of mechanical prostheses. Our 1998 study was updated for this analysis. Results. In this study, 1,160 prostheses and 6,046 follow-up years were analyzed. The pooled survival hazard ratio of mechanical prostheses versus bioprostheses was 1.07 (0.84 to 1.35, p 0.60). The pooled freedom from reoperation hazard ratio was 1.24 (0.67 to 2.31, p 0.67). Pooled survival differences were trivial, favoring mechanical prostheses at 1 ( 0.04%) and 15 years ( 1.1%) and favoring bioprostheses ( 1.8%) at 10 years. Conclusions. There is not a gold standard in tricuspid prostheses replacement. Prosthetic choice is left to the surgeon s clinical judgment, taking into consideration each patient s characteristics and needs. (Ann Thorac Surg 2004;77: ) 2004 by The Society of Thoracic Surgeons Tricuspid valve replacement (TVR) is a r?ther uncommon operation. Most institutional series involve small numbers of cases and a large variety of prosthetic models. Furthermore, tricuspid prostheses are frequently combined with left heart valve prostheses, either on the same operation or on reoperation. These circumstances, along with the institutional differences and the risk factors of the individual patient, including sex, age, and clinical status, make identification of the most suitable prosthesis difficult [1 3]. Does this multifaceted target and heterogeneity prevent conclusions? We try to gain a better insight from a meta-analysis of the literature results comparing biological and mechanical prostheses within the same institution. Material and Methods Accepted for publication Oct 2, Address reprint requests to Dr Rizzoli, Cardiochirurgia, Via Giustiniani 2, Padova, Italy; giulio.rizzoli@unipd.it. The literature published in the last 10 years, available online through the Medline, Current Contents, and Embase, using the key words tricuspid valve replacement and (Bio* or Mec*), was reviewed. The guidelines and techniques for research synthesis of observational studies [4] were applied and our published data [5] were reviewed, excluding very old prosthetic models applied before Meta-analysis of results obtained with the surgical use of biological or mechanical tricuspid valve prostheses was organized in steps as previously detailed [6]. The main end point was the late survival of patients after operation. Survival rate of patients discharged from hospital was analyzed considering that it was less influenced by operative and patient specific risk factors than overall survival. This required recalculating the survival rates after exclusion of early mortality. Other minor points were freedom from reoperation and survival without reoperation. Statistics The most appropriate effect index, to evaluate time related effects of surgical procedures, was the hazard ratio and its confidence limits. The extraction of this index was performed according to the techniques suggested by Parmar [7]. We assumed biological prostheses as the gold 2004 by The Society of Thoracic Surgeons /04/$30.00 Published by Elsevier Inc doi: /j.athoracsur

2 CARDIOVASCULAR 1608 RIZZOLI ET AL Ann Thorac Surg BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION 2004;77: Table 1. Selected Studies of the Literature Study (Publication Date) Location Operative era No. Bio-All-Mec M/F ratio Age (Years) Bio-All-Mec Operative Death(%) Van Nooten (1995) [11] Belgium Scully (1995) [20] Canada Munro (1995) [18] Canada Farinas (1996) [19] France Hayashi (1996) [16] Japan Ratnatunga (1998) [14] UK Dalrymple (1999) [17] UK Do (2000) [12] France Kaplan (2002) [15] Turkey Carrier (2003) [3] Canada Local data (2003) Italy Total (mean) (0.37) (49.3) (19.2) Bio biological; Mec mechanical; M/F male/female. standard and mechanical prostheses as the challenging device. Therefore, a hazard ratio less than 1 indicates better results of mechanical prostheses because of larger reduction of late mortality and a hazard ratio greater than 1 indicates a higher risk of mechanical prostheses. Hazard ratio pooling was done according to the heterogeneity test [6]. The STATA package (Stata Corp., College Station, TX) was used in the analysis [8]. Survival probability with mechanical or biological prostheses was averaged and compared at 1, 5, 10, and 15 years and survival differences were reported with their confidence limits. An attempt to explain the hazard ratio variability among the studies was made utilizing the Poisson regression or negative binomial maximumlikelihood regression [9] of the independent variables summarized in Tables 1, 2, and 3. The hazard ratio was the dependent variable and the total follow-up time was the weighting exposure variable. Results Results of the Cardiac Surgery in Our Institution Between January 1, 1970 and December 31, 2002, 101 tricuspid prostheses (71 porcine, 7 pericardial, 15 tilting disk, and 8 bileaflet) were implanted on 84 patients. The median follow-up time was 7.3 years (6.9 biological, 7.3 mechanical) and total follow-up time was 646 years (511 biological, 135 mechanical). Late survival, freedom from reoperation, and survival without reoperation at 1, 5, 10, and 15 years are summarized in the Appendix. Age corrected estimates of the hazard ratio of survival (0.84, 0.36 to 1.93, p 0.67), of freedom from reoperation (0.86, 0.25 to 3.04, p 0.72), and of survival without reoperation (0.84, 0.42 to 1.69, p 0.82) were used for comparison with literature studies. Ninety-seven percent of the biological, and all the mechanical, prostheses are receiving anticoagulant treatment. Table 2. Follow-Up and Survival Data Study Mean Follow-up (Years) Total Follow-up (Years) Mec total follow-up (Years) Bio total Follow-up (Years) Biological Valve 5 yr-surv. Mechanical Valve 5 yr-surv. Van Nooten [11] a 70 a Scully [20] Munro [18] Farinas [19] Hayashi [16] a 71 a Ratnatunga [14] Dalrymple [17] Do [12] Kaplan [15] Carrier [3] Local data Total (mean) (6.8) (66.7) (66.4) a survival of hospital discharged patients. Bio biological; Mec mechanical; 5 yr-surv. overall 5 year postoperative survival.

3 Ann Thorac Surg RIZZOLI ET AL 2004;77: BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION Table 3. Risk Factors and Complication Incidence NYHA 3 4 ratio Bio-all-Mec Reoperation Ratio Urgent Ratio Bio-all-Mec Single Valve Ratio Bio-all-Mec Triple Valve Ratio Bio-all-Mec Thromb. Incidence % pt/years 1609 SVD Incidence % pt/years CARDIOVASCULAR Van Nooten [11] Scully [20] Munro [18] Farinas [19] Hayashi [16] Ratnatunga [14] Dalrymple [17] Do [12] Kaplan [15] Carrier [3] Local data Total (mean) (0.81) (0.53) (0.19) (0.36) (0.26) Bio biological; Mec mechanical; SVD structural valve deterioration; Thromb. thrombosis. Meta-Analysis According to our research of the literature, 12 studies reported an intrainstitutional comparison of the biological versus mechanical prostheses. The studies of Van Nooten and colleagues [10, 11] and of Do and colleagues [12, 13] were replicated on two different journals, with the intent that the duplicated data were excluded from metaanalysis, to avoid estimation bias. One registry-based study [14], although not strictly intrainstitutional, was included for its surgical relevance. Therefore a total of 11 studies, with 646 mechanical and 514 biological tricuspid prostheses, were further analyzed. They were summarized in Tables 1, 2, and 3 according to time of publication, geographical location of the study, operative time limits, number of prostheses, demographic factors, and perioperative deaths. Demographic Factors Mean age was 49.3 years and the median age was 50 years (interquartile range 46 to 55) (Table 1). It ranged from very young patient cohorts (36 years in the Turkish study [15] and 40 years in the Japanese study [16]) to much older cohorts (59 years in the UK study [17]). Four studies [3, 5, 15, 18] included separate mean ages for mechanical and biological tricuspid prostheses. Males were 36% (range, 2.6% to 80%). Mean male to female ratio was 0.37, median ratio 0.58 (interquartile range, 0.27 to 0.46); the studies of Munro and colleagues [18] and Dalrymple and colleagues [17] have very high female prevalence. Follow-Up The total follow-up summed 6,046 years (Table 2). Mean duration was 6.8 years, median 6.5 (interquartile range, 5.6 to 7.8 years): the studies of Munro and colleagues [18] and Carrier and colleagues [3] have the shortest follow-up (3.7 to 4.0), while those of Farinas and colleagues [19] and Dalrymple and colleagues [17] have the longest (9.5 to 8.1). Total follow-up of biological and mechanical prostheses was calculated assuming a follow-up split proportional to the prosthetic prevalence. In two recent studies [3, 15] total follow-up and follow-up of the mechanical or the biological prostheses was calculated from the number of patients at risk at the stated time, as suggested in CTSNET Surgeon s Forum: Dilemmas in Adult Cardiac Surgery [1]. Risk Factors Risk factors are summarized in Table 3 for each study as prevalence ratios: New York Heart Association (NYHA) class 3 to 4; previous operation; urgent operation; isolated tricuspid valve replacement; and triple valve replacement. By Poisson regression analysis, none of these variables was significant to explain the hazard ratio variability among studies with the exception of the male-to-female ratio (p 0.055). In fact, hazard ratio was in studies with a lower male proportion ( 46%) and in patients with a higher male prevalence. Complications THROMBOSIS. There was a large variability of this complication among the series with a median of 1.28% pt/year episodes (interquartile range, 0.50% to 2.22% pt/year) (Table 3). The series of Ratnatunga and colleagues [14] and Farinas and colleagues [19] report the lowest estimate; Do and colleagues [12] and Carrier and colleagues [3] report the highest. Van Nooten and colleagues [11] and our series observed one case of biological valve thrombosis. Anticoagulant treatment of biological valves was lifelong in the series of Scully and Armstrong [20]; it was limited to three months in the series of Van Nooten and colleagues [11], Hayashi and colleagues [16], Kaplan and colleagues [15], and our series, and was missing in the other papers. STRUCTURAL VALVE DETERIORATION. The median valve period was 1.71% pt/year (interquartile range, 0.42% to 2.27% pt/year). The series of Ratnatunga and colleagues [14],

4 CARDIOVASCULAR 1610 RIZZOLI ET AL Ann Thorac Surg BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION 2004;77: Table 4. Hazard Ratio of Mechanical Versus Biological Prostheses At Risk Biological At Risk Mechanical HR UCL p value Survival Ratnatunga [14] Scully [20] Van Nooten [11] Hayashi [16] Kaplan [15] Dalrymple [17] Carrier a [3] Local data a Overall Reoperation freedom Munro [18] Carrier a [3] infinite Kaplan [15] Dalrymple [17] Local data a Overall Reoperation free survival Van Nooten [11] Local study a Overall a age corrected values. HR hazard ratio; low confidence limits; UCL upper confidence limits. Farinas and colleagues [19], and Munro and colleagues [18] had the lowest incidence; Do and colleagues [12] and Carrier and colleagues [3] reported the highest. Overall, 21 mechanical valve thromboses and 37 deterioration episodes were reported. The incidence rate of thrombosis was 0.87% pt/year and the incidence of the deterioration was 1.02% pt/year. The difference was not significant (p 0.25). Hazard Ratio Seven studies [3, 12, 14 17, 20], plus our series, made possible the computation of the survival hazard ratio (Table 4). Two of these studies [17, 20] did not report the at-risk group at the stated follow-up times and this figure was generated according to Parmar [7]. A third study [16] compared survival, excluding three late deaths due to suicide, sepsis, and pneumonia. The study was utilized because our interest was on survival rate comparison and not on absolute survival values. The pooled hazard ratio estimate was 1.07 (0.84 to 1.35, p 0.60) (Fig 1). Four papers [3, 15, 17, 18] and ours presented data on the freedom from reoperation (Table 4) Among these studies, Carrier and colleagues [3] compared prostheses for a follow-up of five years, while the study of Munro and colleagues [18] had a very small number of mechanical prostheses followed beyond two years. In both series it was impossible to obtain a reasonable estimate of the variance and confidence limits. In the three remaining studies the pooled hazard ratio estimate was 1.24 (0.67 to 2.31, p 0.50) (Fig 2). A single study [11] plus ours enabled the comparison of the survival without reoperation (Table 4), with both having similar results. The pooled estimate was 0.86 (0.70 to 1.05, p 0.14). The survival of the biological and mechanical prostheses was available for 7 series [3, 12, 14, 15 17, 20] plus ours (Fig 3). Among 391 patients discharged with mechanical prostheses, the pooled 1, 5, 10, and 15 year survival of the hospital-discharged patients was 86.9%, 73.5%, 60.2%, and 47.8%, respectively; among 477 patients discharged Fig 1. Survival hazard ratio of each study with related confidence limits. The pooled hazard ratio is indicated from the dotted line.

5 Ann Thorac Surg RIZZOLI ET AL 2004;77: BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION Fig 2. Reoperation free hazard ratio of each study with related confidence limits. The pooled hazard ratio is indicated from the dotted line. (HR hazard ratio.) with tricuspid bioprostheses, survival was 86.5%, 73.6%, 62%, and 46.7%, respectively. Five-year survival was identical. Differences were trivial and favoring mechanical prostheses at 1 year ( 0.04, ) and at 15 years ( 1.1, 24 26), favoring biological prostheses at 10 years (1.8, 16 12). Comment Tricuspid valve replacement was performed for the first time in Fortunately, this operation is rarely required and it is performed in less than 2% of all valve operations [6]. Therefore, most institutional series involve a small number of patients so this makes difficult any statistical evaluation of the performance of the prostheses. Therefore, as randomized studies are not possible in this setting, observational studies are the only available source of information. This limitation increases the confounding due to intrinsic patient factors, to extrinsic operative variability, and to data reporting. Biological prostheses were initially considered ideal in Fig 3. Average survival of the hospital-discharged patients, at stated follow-up years, from the studies of Table 1. The difference in survival between biological and mechanical prostheses is reported with the confidence limits the tricuspid position because they would not require anticoagulation and were expected to have a slower degeneration than in the mitral or aortic position. However, this belief was contradicted by the finding of pannus formation on the mural cusp [21, 22] and by limited durability of about 7 to 9 years [3 5]. On the other hand, the new bileaflet mechanical valves have an appealing low gradient, decreased turbulence, and optimal durability [15]. From neither our previous experience in 1998 [5], nor from the review of the literature, was a clear superiority of biological over mechanical prostheses observed, suggesting that mechanical prostheses should be preferred in young patients and in patients with left sided mechanical prostheses. Two papers have recently reopened the discussion, one suggesting superior results with the use of mechanical devices [15] and the other with the use of biological devices [3]. In commenting on the paper by Kaplan and colleagues [15], we anticipated that a meta-analysis approach, based on the collection of data from the literature comparing the results within the same institution [1], should help to clarify the discordance. It has been written that meta-analysis of nonrandomized studies presents methodologic limits and the results are not reliable [23]. However, we share the opinion that some numerical information is better than none. Therefore, we utilized the methodologic adjustments proposed by the meta-analysis of observational studies in epidemiology (MOOSE) group [4]. We limited our review of the literature to the most recent years and have reanalyzed our results to limit confusion from the most outdated prosthetic models. With the same purpose, we selected only papers comparing mechanical and biological devices within the same institution in order to restrict the errors depending upon intrinsic and extrinsic variables. Calculation of the hazard ratio was selected because it is the most appropriate index to compare late results of different techniques [7], independent of the rate of success. The ratio was obtained by comparing survival rates of hospitaldischarged patients, because it is less influenced by operative and patient specific risk factors than overall survival. Intrainstitutional comparison has the further advantage of ensuring the same accuracy in the collection of the survival data. Similar accuracy in the follow-up is important mainly in recollection to the most common complications: Thrombosis occurs at random but it is most probable in the early postoperative periods because of difficulties in rapidly reaching a good anticoagulation control. Conversely, degeneration of biological prostheses requires time to develop so that its real occurrence depends on follow-up time. The median incidence of mechanical valve thrombosis in this composite series, with a rather large mixture of mechanical valves, was 1.28% pt-years: this rate is intermediate between the initial reports with mono-leaflet valves [24] and the more recent reports with the use of the St. Jude valve (St. Jude Medical, Inc, St. Paul, MN) [25]. CARDIOVASCULAR

6 CARDIOVASCULAR 1612 RIZZOLI ET AL Ann Thorac Surg BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION 2004;77: Analysis of freedom from reoperation indicated a larger hazard of the mechanical prosthesis. This is surprising, because the mean incidence of reoperation due to degeneration of the bioprosthesis is similar to the incidence of reoperation due to thrombosis of mechanical prostheses. This effect was strongly influenced by the study of Kaplan and colleagues [15], which reports 5 mechanical valves and 2 biological events within a short 5 year follow-up. Nonetheless, this group has the most favorable survival with the use of mechanical prostheses and these data support their use, due to their favorable hemodynamic characteristics and durability. In their experience, only 3 thromboses required reoperation. In fact, bileaflet valve thrombosis is rarely fatal and most often is successfully treated by increased anticoagulation or thrombolysis [26]. Conversely, Carrier and colleagues [3] advocate the use of biological prostheses in young patients because of limited life expectancy, unrelated to the type of tricuspid prostheses at long term follow-up. In our review, the pooled late survival of hospital-discharged patients was 60% at 10 years and 50% at 15 years so that many patients are exposed to the risk of valve deterioration. The same authors observed that six patients with first generation bioprostheses were reoperated within 9 years, but they mentioned the event as rare (5%) with the use of the Carpentier-Edwards pericardial prosthesis (Edwards Lifesciences, Irvine, CA). We may argue that the increased durability of second-generation porcine and pericardial prostheses is commonly used to strongly support the use of the biological devices. Comparison of durability of nonconcomitant valve series is, for the moment, suspicious, when unsupported by a propensity score analysis [27]. In our meta-analysis we choose to eliminate the weakest group, with high perioperative mortality, and we recalculated the survival including only the hospitaldischarged patients. We assume that this patients cohort is not affected by a variety of risk factors and is more homogeneous. Despite these restrictions the meta-analysis shows a large heterogeneity among the selected studies, due to several problems. With the exception of the UK Ratnatunga study [14], which is registry-based, the person-time products are small. Some studies did not publish the size of the group at-risk at stated survival times and did not provide a division of the total follow-up according to the use of biological or mechanical devices. Therefore, these data had to be approximated. Confounding of survival estimates due to operative age was not taken into account, with the exception of a single study [3] together with our series. The search of sources of variability of the hazard estimate among studies, with the use of Poisson and negative binomial distribution models, was inconclusive. The single factor identified as source of variability was the proportion shown among sexes and a higher risk of mechanical prostheses in studies with a larger male prevalence (possibly due to sex related etiology and habits). Pooling of the results suggests an identical risk for both prosthetic types. With 249 biological and 186 mechanical valves, followed for 5 years, we have an 85% power to recognize a difference of survival of 10% with a p value of 0.2 (possibly significant). The absence of any difference in the survival data supports the opinion that there is no gold standard for prosthetic tricuspid valve replacement. Nonetheless, results like the meta-analysis presented here have serious limits because they look to the entire forest and not to the single pine. Therefore, we believe that the choice between mechanical or biological prostheses in the tricuspid position should be individualized according to the clinical judgment, the age, the cardiac disease, the etiology, and the habits of the patient, as pointed out by Del Campo and Sherman [2]. It has been stated that the exercise of personal preference by the implanting surgeon in the choice of biological or mechanical prosthesis for tricuspid position seems reasonable [14]. We believe that this statement should be attenuated by several considerations: (1) Tricuspid position makes no exception to the rule that patients more than 65 to 70 years obtain the largest advantage from bioprostheses and younger patients from mechanical prostheses. (2) The extent and the severity of cardiac disease might suggest, in some cases, a limited expectation of life and therefore might favor the use of biological prostheses in younger patients, as concluded by Carrier [3]. On the other hand, concomitant use of left-sided mechanical prostheses favors the same valve for the right heart. (3) The male sex, as opposed to the female sex, is more prone to nonrheumatic etiology, drug addiction, and endocarditis so that it could be favored in the use of biological prostheses. Small size patients with small right ventricles could take advantage of the better hemodynamics of the low profile bileaflet valve as opposed to the largest bioprosthesis, which is prone to develop mural cusp pannus and thrombosis. (4) Finally, by reviewing our data, we found that 97% of living patients with biological tricuspid prostheses are receiving anticoagulant treatment, making the need for anticoagulation a not reliable criterion for the choice of the tricuspid prosthesis. We also found a prosthetic thrombosis in a 16-year-old patient who was not anticoagulated. The occasional report of bioprosthesis thrombosis [11, 29] and pulmonary embolism suggests that anticoagulation treatment should be prolonged indefinitely, even in young patients, unless sinus rhythm [16] and a right atrium less than 5.5 cm [15] exist. This project was supported by MIUR 2003 ROME, Italy, Metaanalysis of the studies comparing long term results of biological and mechanical prostheses. References 1. CTSNET surgeon s forum. Dilemmas in adult cardiac surgery: cfm?catid 8&threadid 4390&highlight_key y&keyword1 tricuspid.

7 Ann Thorac Surg RIZZOLI ET AL 2004;77: BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION 2. Del Campo C, Sherman JR. Tricuspid valve replacement: results comparing mechanical and biological prostheses. Ann Thorac Surg 2000;69: Carrier M, Hebert Y, Pellerin M, et al. Tricuspid valve replacement: an analysis of 25 years of experience at a single center. Ann Thorac Surg 2003;75: Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology. JAMA 2000;283: Rizzoli G, De Perini L, Bottio T, Minutolo G, Thiene G, Casarotto D. Prosthetic replacement of the tricuspid valve: biological or mechanical. Ann Thorac Surg 1998;66:S Rizzoli G, Schiavon L, Bellini P. Does the use of bilateral internal mammary artery (IMA) grafts provide incremental benefit relative to the use of a single IMA graft? A metaanalysis approach. Eur J Cardiothorac Surg 2002;22: Parmar MKB. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Statist Med 1998;17: StataCorp Stata Statistical Software. release 7.0 College Station, TX: Stata Corporation. 9. Grunkemeier LG, Wu Y. Our complication rates are lower than theirs: statistical critique of heart valve comparisons. J Thorac Cardiovasc Surg 2003;125: Van Nooten GJ, Caes FL, Francois KJ, et al. The valve choice in tricuspid valve replacement: 25 years of experience. Eur J Cardiothorac Surg 1995;9: Van Nooten G J, Caes F, Taeymans Y, et al. Tricuspid valve replacement: postoperative and long term results. J Thorac Cardiovasc Surg 1995;110: Do QB, Pellerin M, Carrier M, et al. Clinical outcome after isolated tricuspid valve replacement: 20-year experience. Can J Cardiol 2000;16: Do QB, Pellerin M, Carrier M, et al. Remplacement valvulaire tricuspide isole. Resultats a long terme. Arch Mal Coeur Vaiss 2000;93: Ratnatunga CP, Edwards MB, Dore CJ, Taylor KM. Tricuspid valve replacement: UK Heart Valve Registry mid-term results comparing mechanical and biological prostheses. Ann Thorac Surg 1998;66: Kaplan M, Kut MS, Demirtas MM, Cimen S, Ozler A. Prosthetic replacement of tricuspid valve: bioprosthetic or mechanical. Ann Thorac Surg 2002;73: Hayashi J, Saito A, Yamamoto K, Watanabe H, Ohzeki H, 1613 Eguchi S. Is a bioprosthesis preferable in tricuspid valve replacement? Thorac Cardiovasc Surg 1996;44: Dalrymple Hay MJ, Leung Y, Ohri SK, et al. Tricuspid valve replacement: bioprostheses are preferable. J Heart Valve Dis 1999;8: Munro AI, Jamieson WR, Tyers GF, Germann E. Tricuspid valve replacement: porcine bioprostheses and mechanical prostheses. Ann Thorac Surg 1995;60:S Farinas JM, Leclerc Y, Antchouey AM, Mercier LA. Le remplacement valvulaire tricuspidien: suivi clinique et echographique a long terme. Annales de Chirurgie 1996;50: Scully HE, Armstrong CS. Tricuspid valve replacement. Fifteen years of experience with mechanical prostheses and bioprostheses. J Thorac Cardiovasc Surg 1995;109: Nakano K, Eishi K, Kosakai Y, et al. Ten year experience with the Carpentier-Edwards pericardial xenograft in the tricuspid position. J Thorac Cardiovasc Surg 1996;111: Guerra F, Bortolotti U, Thiene G, et al. Long-term performance of the Hancock porcine bioprosthesis in the tricuspid position. A review of forty-five patients with fourteen-year follow-up. J Thorac Cardiovasc Surg 1990;99: Shapiro S. Is meta-analysis a valid approach to the evaluation of small effect in observational studies? Am J Epidemiol 1997;50: Thorburn CW, Morgan JJ, Shanahan MX, Chang VP. Longterm results of tricuspid valve replacement and the problem of prosthetic valve thrombosis. Am J Cardiol 1983;51: Nakano K, Koyanagi H, Hashimoto A, Ohtsuka G, Nojiri C. Tricuspid valve replacement with the bileaflet St. Jude medical valve prosthesis. J Thorac Cardiovasc Surg 1994;108: Shapira Y, Sagie A, Jortner R, Adler Y, Hirsch R. Thrombosis of bileaflet tricuspid valve prosthesis: clinical spectrum and the role of non-surgical treatment. Am Heart J 1999;137: Blackstone EH. Comparing apples and oranges. J Thorac Cardiovasc Surg 2000;123: Okada Y, Nasu M, Nishiuchi S, Shomura T. Long-term echocardiographic follow-up of patients with a tricuspid bioprosthesis. ASAIO Trans 1990;36:M (Appendix follows on p. 1614) CARDIOVASCULAR

8 CARDIOVASCULAR 1614 RIZZOLI ET AL Ann Thorac Surg BIOLOGICAL OR MECHANICAL PROSTHESES IN TRICUSPID POSITION 2004;77: Appendix Results of Local Follow-Up Study Freedom From Reoperation (Biological) Year No Reoperated Reoperation Freedom UCL No Reoperated (Mechanical) Reoperation Freedom UCL Survival Without Reoperation (Biological) (Mechanical) Year No. Fail Survival UCL No. Fail Survival UCL Survival (Biological) (Mechanical) Year No. Dead Survival UCL No. Dead Survival UCL Fail dead or reoperated; low confidence limits; UCL upper confidence limits.