- Email: [email protected]
Likelihood of underreporting of outlet strut fracture from examination of the Dutch Björk-Shiley CC cohort
Likelihood of Underreporting of Outlet Strut Fracture from Examination of the Dutch Bjo ¨ rk-Shiley CC Cohort Marjon Kallewaard, PhD, Ale Algra, MD, PhD, Jo Defauw, MD, Diederick Grobbee, MD, PhD, and Yolanda van der Graaf, MD, PhD, for the Bjo ¨ rk-Shiley Study Group* The Dutch Bjo¨rk-Shiley convexo-concave (BScc) cohort serves as a reference population on the risk of outlet strut fracture and is being used to formulate guidelines for prophylactic replacement. Fractures, however, may be undetected at death. The aim of this study was to quantify the degree of underestimation of strut fracture in the Dutch BScc cohort. Multivariate Cox regression analysis was used to assess the relative and absolute risk of death from different causes within 14 years. The unexplained “excess” mortality among 70& BScc valve recipients was attributed to unreported fatal strut fractures and used to estimate its extent in this group, which then was extrapolated to the 60& BScc valve recipients. For 70& BScc valve recipients, the adjusted hazard ratio for death from all causes except strut fracture was 1.2 (95% confidence interval [CI] 1.0 to 1.5). The 14-year
absolute risks for 70& and 60& BScc valve recipients were 44% and 37%, respectively. Among 70& and 60& BScc valve recipients, underreporting of fracture was estimated to be 25% (95% CI 0 to 49) and 26% (95% CI 0 to 52), respectively. Estimates based on sudden death and fatal congestive heart failure yielded essentially the same results. Thus, underreporting of fatal strut fracture in the Dutch BScc cohort is estimated to be approximately 25%. Hence, the risk and lethality of fracture of BScc valves are underestimated and indications for prophylactic replacement should be adjusted accordingly. For example, the advantage of valve replacement in a 40-year-old patient with a 29-mm 60& BScc mitral valve would almost double to 0.82 years. Q1998 by Excerpta Medica, Inc. (Am J Cardiol 1998;82:768 –773)
jo¨rk-Shiley convexo-concave (BScc) heart valves carry an increased risk of mechanical failure with B potential catastrophic consequences for the implan-
ients in the Netherlands.1 The increased risk could not be explained by documented strut fractures and factors known to be associated with death after valvular surgery.1 If strut fractures indeed remain undetected, the risk estimates of outlet strut fracture are underestimations of the true incidence,1,2,4 – 6 and the recommendations for prophylactic replacement would have to be adjusted accordingly.6 This study quantifies the degree of underestimation of outlet strut fracture in the Dutch BScc cohort.
tee.1,2 Worldwide, approximately 35,000 living BScc valve recipients face the risk of outlet strut fracture. Only 2 cohort studies, a Swedish institution– based study and a Dutch nationwide retrospective cohort study, have examined the risk and lethality of outlet strut fracture.1,2 To date the Dutch BScc cohort study is the only study to follow-up actively both 60° and 70° BScc valve recipients. Consequently, the Dutch cohort serves as a reference population of the risk and lethality of strut fracture worldwide and is being used to formulate guidelines for prophylactic replacement.3 However, even if follow-up is complete, some strut fractures may not be detected.1,2,4 – 6 Strut fracture results in an almost instantaneous death, especially among BScc aortic valve recipients, and most patients die at home (only a few of whom undergo necropsy).1,2,4,6 – 8 Compared to 60° BScc valve recipients, undetected strut fractures may also be the reason for increased risk of death among 70° BScc valve recipFrom the Julius Center for Patient Oriented Research, Clinical Epidemiology Unit, Utrecht University, Medical School, Utrecht, The Netherlands; and Department of Cardiothoracic Surgery, the St. Antonius Hospital, Nieuwegein, The Netherlands. Manuscript received February 19, 1998; revised manuscript received and accepted May 5, 1998. Address for reprints: Yolanda van der Graaf, MD, PhD, Julius Center for Patient Oriented Research, Utrecht University, Medical School, P.O. Box 05500, 3508 GA TA Utrecht, The Netherlands. *Members of the Netherlands Bjo¨rk-Shiley study group are listed in the Appendix.
768
©1998 by Excerpta Medica, Inc. All rights reserved.
METHODS
Patients: The Dutch BScc Follow-up Study is a nationwide retrospective cohort study of all Dutch BScc valve recipients. All Dutch BScc valve recipients were identified and actively followed until outlet strut fracture, death, or the end of the study; 2,264 BScc valve recipients received 2,254 60° and 279 70° valves from 1979 to 1985.1 For each patient a detailed clinical profile was obtained from the clinical records (Table I). Presence of acute endocarditis was defined as surgery before the completion of a course of intravenous antibiotics; left ventricular function was classified as good, reduced, or poor from the right oblique view of the ventricular angiogram; presence of coronary artery disease was defined as a diameter lumen narrowing .50% on the coronary angiogram, and classified as 1-2-, or multiple-vessel (.3) disease. The methods of the first Dutch BScc follow-up study have been described in detail previously.1,9 Clinical characteristics of the patients are summarized in Table I. Follow-up: Information on the vital status was obtained from municipal registries and on reoperation 0002-9149/98/$19.00 PII S0002-9149(98)00461-5
classified as sudden unexplained deaths (ICD-9 code 798).1,2 However, because mitral valve fractures 60° BScc Valve 70° BScc Valve have a protracted course and as many † Recipients Recipients heart valve recipients have some de(n 5 2,000) (n 5 263) gree of congestive heart failure, unAge (yr) (mean 6 SD) 54.2 6 12.6 55.3 6 12.5 detected strut fractures may also be Sex (men) 1,140 (57.0%) 153 (58.2%) misclassified as fatal congestive Rheumatic pathogenesis 399 (20.0%) 69 (26.2%) heart failure (ICD-9 code 428). Recent infective endocarditis 210 (10.5%) 21 (8.0%) Previous valve replacement 90 (4.5%) 14 (5.3%) Moreover, many patients with conLeft ventricular function gestive heart failure die suddenly, Good 1,562 (78.1%) 225 (85.6%) making the distinction between strut Reduced/poor 94 (4.7%) 10 (3.8%) fracture and arrhythmogenic death Unknown 344 (17.2%) 28 (10.6%) extremely difficult. Postoperative Coronary artery disease None 1,345 (67.3%) 180 (68.4%) mortality was defined as death occur1- or 2-vessel 283 (14.1%) 40 (15.2%) ring within 30 days after operation.11 $3-vessel 115 (5.8%) 21 (8.0%) Late mortality was defined as death Unknown 257 (12.9%) 22 (8.4%) during follow-up but after 30 days. History of myocardial infarction 60 (3.0%) 4 (1.5%) Emergency procedure 60 (3.0%) 15 (5.7%) The primary end points of this study Size of BScc prosthesis ($29 mm) 746 (37.3%) 112 (42.6%) were death from all causes except Valve replacement strut fracture and the combined event Aortic 1,009 (50.5%) 113 (43.0%) of sudden death or fatal congestive Mitral 684 (34.2%) 116 (44.1%) heart failure. Aortic and mitral‡ 307 (15.4%) 34 (12.9%) Concomitant surgery Data analysis: Cumulative surCoronary artery bypass surgery 579 (29.0%) 83 (31.6%) vival curves according to the Tricuspid valvuloplasty 29 (1.5%) 1 (0.4%) Kaplan-Meier product limit method Aortic root replacement 34 (1.7%) 0 (0.0%) and the log-rank test were used for *In the present report, the only Dutch BScc valve recipient who underwent mitral and tricuspid valve comparison.12 To estimate the degree replacement, and who experienced a strut fracture, was excluded from analysis. of underreporting we assumed the † Seventeen patients who underwent aortic and mitral valve replacement and received a 60° and 70° following: BScc valve were included here as the 70° BScc valve recipients only. ‡ Seventy-three patients who underwent aortic and mitral valve replacement of a BScc valve and a 1. First, we compared the survival non-BScc valve. of 60° and 70° BScc valve recipients, while censoring documented strut fractures and adjusting for factors known from all Dutch centers for cardiothoracic surgery (end to be associated with death after valvular surgery. We date of the study: July 1, 1996). Information on causes assumed that any remaining “excess mortality” among of death was obtained from the patient’s general prac- Dutch 70° BScc valve recipients could be attributed to titioner. Mean duration of follow-up was 10.0 years deaths due to undetected strut fractures. We first exam(range from 0 to 17). Forty-six patients (2.0%) were ined the independent contribution of valve opening angle lost to follow-up. and risk factors, known to be associated with late morOutcome events: The causes of death were coded tality after valvular surgery (as listed in Table I), to the independently by 2 of the authors (AA and YvdG) several causes of death in a multivariate Cox proportionaccording to the 9th revision of the International Clas- al-hazards regression model (entry criterion p ,0.2).13 sification of Diseases.10 For the purpose of this study Because there are different risk factors associated with we identified the following categories: death from death during the first year after valvular surgery and outlet strut fracture, sudden death or fatal congestive death during the subsequent years,2,14,15 we restricted our heart failure (ICD-9 codes 798 and 428), other causes analyses to BScc valve recipients who at least survived 1 of cardiovascular death (ICD-9 codes 391-427 and 429-460), death due to malignancy (ICD-9 codes 150- year after valvular surgery (n 5 1,949). For each of the 240), remaining causes of death, and a category that variables entered in the model, the proportional-hazards assumption was tested using time-dependent covariates. included deaths from unknown causes. We decided in advance to combine sudden death Hazard ratios are presented with 95% confidence limits and fatal congestive heart failure as the class most (95% CI) derived from the Cox model. 2. Next, we determined the absolute 14-year risks likely to contain missed fatal strut fractures. Outlet of death from all causes except strut fracture and from strut fracture of an aortic BScc valve is followed by almost instantaneous death,1,2,7 and fracture of a mitral the combined event of sudden death or fatal congesvalve results in sudden onset of symptoms of severe tive heart failure for “average” 60° and 70° BScc left ventricular failure with a duration of up to 48 valve recipients who survived at least 1 year as S0(t 5 exp (PI 1 beta z valve opening angle [0 5 60°, 1 5 70°]) , where hours, and persisting without hemodynamic improve- 14) ment until death.1,2,7 Given the acute nature of the PI is the summed product of the coefficients derived complaints following outlet strut fracture, it has been from the Cox model with the mean cohort levels of suggested that undetected strut fractures may be mis- variables of the prognostic factors used in the model. S0 TABLE I Clinical Characteristics of 2,263 60° and 70° Bjo ¨ rk-Shiley CC (BScc) Valve Recipients in the Dutch BScc Cohort*
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
769
FIGURE 1. Age-adjusted survival curves of Dutch BScc valve recipients according to valve opening angle and mode of death. Upper lines, death from all causes except strut fracture; lower lines, death from all causes.
(t 5 14) is the baseline survival function of the models at 14 years. The excess mortality, possibly attributable to undetected strut fractures among 70° BScc valve recipients, was then calculated by subtracting the 14-year absolute risk estimates of 70° and 60° BScc valve recipients. 3. To finally estimate the number of undetected fractures among 60° valve recipients we assumed that the hazard ratio for the valve opening angle of unreported strut fractures is the same as the hazard ratio for reported strut fractures in this group of patients (hazard ratio 7.8, 95% CI 4.3 to 14.2). Standard errors for the percentage of underreporting of outlet strut fracture were estimated with bootstrap resampling (random sampling with replacement);16 calculations were repeated 500 times. All calculations were performed using SPSS and the functions of the Design Library which works in S-plus software (Data Analysis Products Division of MathSoft Inc., Seattle Washington).17
RESULTS Operative mortality was the cause of death in 137 (6.9%) 60° BScc and 17 (6.5%) 70° BScc valve recipients. Lethality of strut fracture was 88% for aortic valve (7 of 8) and 55% for mitral valve (23 of 42) recipients. Among the 1,863 60° BScc valve recipients who survived at least 30 days after surgery, 17 (0.9%) died of strut fracture, 257 (13.8%) of the combined event of sudden death or fatal congestive heart failure, 183 (9.8%) of cardiac death, 102 (5.5%) of malignancies, and 131 (7.0%) of other causes. Among the 246 770 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 82
70° BScc valve recipients, these numbers were 12 (4.9%), 51 (20.7%), 17 (6.9%), 13 (5.3%), and 7 (2.8%), respectively. The cause of death was unknown for 163 (8.7%) 60° BScc and for 22 (8.9%) 70° BScc valve recipients. Necropsy was performed on 188 (16.7%) of the deceased patients, most of whom died of operative causes (n 5 96). The 14-year cumulative probabilities for BScc valve recipients for different causes of death are presented in Table II. The age-adjusted Kaplan-Meier curves show a better survival among 60° BScc valve recipients than among 70° BScc valve recipients (p log-rank 0.003; Figure 1). The difference was still apparent, although somewhat smaller, after correction for death due to strut fracture (p log-rank 0.04). Further analyses were restricted to Dutch BScc valve recipients who survived 1 year after implantation. Multivariate analysis showed that the risk of death from all causes except strut fracture was 20% higher among patients with 70° BScc valves (hazard ratio 1.2, 95% CI 1.0 to 1.5). The multivariate model, which studied sudden death or fatal congestive heart failure, showed that 70° BScc valve recipients had an almost twofold higher risk (hazard ratio 1.8, 95% CI 1.3 to 2.5) than 60° BScc valve recipients. Valve opening angle was no longer an independent predictor when death from other cardiovascular causes (hazard ratio 0.9, 95% CI 0.5 to 1.5), death from noncardiovascular causes (hazard ratio 0.8, 95% CI 0.5 to 1.4), and death from unknown causes (hazard ratio 1.3, 95% CI 0.8 to 2.0) were studied multivariately. Table SEPTEMBER 15, 1998
0.79 26.2% (19.1–57.8)
DISCUSSION
*Comparison of 60° and 70° BScc valves; p value log-rank test. Values are expressed as cumulative probability (%) (95% CI). CI 5 confidence interval; CHF 5 congestive heart failure.
13.4% (11.2–15.9) Other cardiovascular causes
12.3% (6.1–23.9)
0.45
15.0% (12.0–18.6)
14.0% (7.3–26.0)
0.89
20.7% (15.8–26.8)
0.50 39.9% (14.5–80.8) 24.7% (21.0–28.9) 12.9% (10.7–16.0) Sudden death/CHF
25.7% (16.8–38.2)
,0.001
40.5% (30.0–52.9)
,0.001
24.3% (18.9–31.0)
0.20 75.1% (54.4–91.5) 53.3% (49.5–57.3) 40.9% (37.9–44.1) Any cause except fracture
53.1% (43.2–63.8)
0.02
57.7% (47.6–68.1)
0.22
57.5% (51.8–63.4)
0.03 80.1% (61.9–93.3) 41.1% (38.1–44.3) Any cause
55.2% (45.3–65.6)
0.007
54.3% (50.4–58.2)
60.2% (50.4–70.2)
0.10
58.3% (52.6–64.1)
p Value* 70° BScc (n 5 34) 60° BScc (n 5 307) Causes of Death
60° BScc (n 5 1,009)
70° BScc (n 5 113)
p Value*
60° BScc (n 5 684)
70° BScc (n 5 116)
p Value*
Aortic and Mitral Valve Mitral valve (n 5 800) Aortic Valve (n 5 1,122)
TABLE II Fourteen-Year Cumulative Probabilites for Death from Any Cause, Death Due to Any Cause Except Strut Fracture, Sudden Death, and Cardiac Death in the Dutch BScc Cohort
III summarizes the average 14-year absolute risk of death from all causes except strut fracture and of the combined event of sudden death or fatal congestive heart failure for the 60° and 70° BScc valve recipients. Among 70° BScc valves underreporting was estimated to be 25% (95% CI 0 to 49) if deaths from all causes except strut fracture were considered, and 18% (95% CI 3 to 33) if the combined event of sudden death or fatal congestive heart failure was used. Underreporting of strut fracture among 60° BScc valves was estimated to be 26% (95% CI 0 to 52) and 19% (95% 4 to 35), respectively (Table III). Analysis of the data according to valve position yielded essentially the same results. Although the degree of underreporting was somewhat higher in aortic valves this difference was not statistically significant. The findings of our study suggest that the degree of underreporting of outlet strut fractures in the Dutch BScc cohort may be as high as 25% (95% CI 0 to 49). The current risk estimates of outlet strut fracture are therefore substantial underestimations of the true incidence. Several groups have published guidelines for prophylactic replacement of BScc valves,4,6,18,19 each of which debated the validity of the estimates of the risk and lethality for outlet strut fracture; however, adjustment of these risk estimates had considerable effect on the indications for prophylactic replacement.6,19 For example, Steyerberg et al19 showed that for a hypothetical 40-year-old male recipient of a 29-mm 60° BScc mitral valve, the advantage of valve replacement would be 0.45/year. Applying their method it can be calculated that the advantage of surgery would almost double to 0.82 year if the estimates for the risk and lethality of outlet strut fracture were 25% higher. In contrast, because of the enormous underreporting of fractures in other countries, fracture rates are multiplied by 2.4 (USA) or 3.7 (non-USA, nonNetherlands) to approximate Dutch fracture rates.3,18 Data from this study suggest that even higher multiplication factors should be used. Given the low incidence of necropsy in the Netherlands, complete assessment of the true incidence of outlet strut fracture was impossible. We therefore approximated the degree of underreporting of outlet strut fracture by quantification of the proportion of the “unexplained excess” deaths among 70° BScc valve recipients. We assumed that because the higher mortality among 70° BScc valve recipients could not be explained by clinical characteristics known to be associated with death after valvular surgery or by documented strut fractures, the excess had to be attributed to unreported outlet strut fractures. It may be debated whether all relevant clinical variables explaining different mortality in 60° and 70° BScc valve recipients are included in the multivariate models. However, because it is not plausible that patients were assigned to have a 60° or 70° BScc valve because of their clinical profile, we consider it unlikely that the 30% higher risk of death among 70° BScc valve recipients may entirely be explained by variables not included in
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
771
TABLE III Estimation of the Degree of Underreporting of Outlet Strut Fractures Among Dutch BScc Valve Recipients Who Survived at Least One Year After Implantation During a Two- to 14-Year Period After Implantation At Risk
Absolute 14-Year Risk* (95% CI)
Deaths
Cause of death
Valve Type
All causes except strut fracture Combined event of sudden death Or fatal congestive heart failure
70° BScc 60° BScc 70° BScc
44.0% (37.5–51.1) 37.2% (34.7–39.8) 23.5% (17.8–30.8)
96 692 45
60° BScc
13.7% (11.8–15.8)
215
Patients
Person– Years
Documented Strut Fractures
Unreported Strut Fractures†
Percentage Underreporting‡ (95% CI)
224 1,755 224
1,855 17,795 1,855
20 23 20
6.5 8.0 4.4
25% (0–49) 26% (0–52) 18% (3–33)
17,795
23
5.4
19% (4–35)
1,755
*Adjusted (see Table II) absolute-14 year risks. † The number of undetected strut fractures of 70° BScc valves was estimated by multiplying the difference of the 14th-year absolute risks between 70° and 60° BScc valve recipients with the number of deaths. For example (44.0%–37.2%) z 96 deaths 5 6.5. Assuming that the hazard ratio for valve opening angle of unreported fractures is 7.8, the number of unreported strut fractures of 60° BScc valves could be calculated. For example (6.5/1855)/(?/17,795) 5 7.8 . ? 5 8.0. ‡ [Underreported fractures/(documented 1 underreported fractures)] z 100%.
our multivariate models. Because of the small number of postmortem examinations, we were not able to differentiate valve-related complications such as valve thrombosis and arrhythmias from outlet strut fracture. A higher incidence of these fatal complications may be another explanation for the higher mortality among 70° BScc valve recipients. If so, this would be rather ironic since the 70° BScc valve was designed to be less thrombogenic.20 The superior quality of the 70° BScc valve compared with the 60° BScc valve, however, has never been proven.21,22 To calculate the number of undetected outlet strut fractures among 60° BScc valve recipients, we assumed that the hazard ratio for the valve opening angle for unreported strut fractures is the same as that for reported strut fractures. If so, the degree of underreporting of outlet strut fracture would approximately be the same among 60° and 70° BScc valve recipients, an assumption that was also made by others.2 However, because the necropsy rate among the Dutch BScc patients who died during follow-up was higher among the 70° BScc valve recipients (14.8% vs 8.7% among the 60° BScc valve recipients; p 5 0.03), it could be argued that the degree of underreporting of strut fracture in 60° valves in reality is even higher than the estimated 20% to 30%. The findings of this study agree with those by Lindblom and colleagues2 who estimated the percentage of undetected strut fractures to be approximately 28% (with a possible maximum of 47%). They extrapolated the percentage of fractures reported among BScc valve recipients who died suddenly or from unexplained causes, and who underwent autopsy, to patients who died suddenly or from unexplained causes without necropsy2; postmortem examination is routine practice in Sweden (75% of all deceased BScc valve recipients underwent autopsy). In view of these findings, the argument to facilitate autopsy in heart valve recipients is even stronger than it was. It is of the utmost importance to evaluate valvular prostheses carefully after longtime use. The BScc strut fracture problem would have come out in the open significantly earlier if such a policy for artificial heart valve recipients had been in effect. 772 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 82
Finally, if 25% of fatal strut fractures are not reported, the current estimates of lethality of strut fracture would be too low. In our study, the lethality of strut fracture was 54% (23 of 42) for mitral valves and 88% (7 of 8) for aortic valves. If the number of fatal strut fractures of BScc mitral valves were to be 25% percent higher, lethality would increase to 64% (34 of 53) for mitral valves and to 90% (9 of 10) for aortic valves. Higher estimates of the risk and lethality of strut fracture should lead to adjustment of the indications for prophylactic replacement, i.e., more BScc valve recipients should be considered for reoperation.4,6,19 Acknowledgment: We thank Ewout Steyerberg, MD, for his valuable contribution to the analysis.
APPENDIX Participating Centers: Academisch Medisch Centrum (Bas de Mol), Onze Lieve Vrouwe Gasthuis Amsterdam (Leon Eijsman), Medisch Centrum de Klokkenberg (ThR van Geldorp), Academisch Ziekenhuis Leiden (Hans Huysmans), Sint Antonius Ziekenhuis Nieuwegein (Jo Defauw), Academisch Ziekenhuis Utrecht (Jaap Bredee), Academisch Ziekenhuis Groningen (Tjark Ebels), Dijkzigt Ziekenhuis Rotterdam (Lex van Herwerden), Sint Radboud Ziekenhuis (LK Lacquet).
1. van der Graaf Y, de Waard F, van Herwerden LA, Defauw J. Risk of strut
fracture of Bjo¨rk-Shiley valves. Lancet 1992;339:257–261. 2. Lindblom D, Bjo¨rk VO, Semb BKH. Mechanical failure of the Bjo¨rk-Shiley
valve. Incidence, clinical presentation, and management. J Thorac Cardiovasc Surg 1986;92:894 –907. 3. Brookmeyer R, Yasui Y. Statistical analysis of passive surveillance disease registry data. Biometrics 1995;51:831– 842. 4. Birkmeyer JD, Marrin CAS, O’Connor GT. Should patients with Bjo¨rk-Shiley valves undergo prophylactic replacement? Lancet 1992;340:520 –523. 5. Ostermeyer J, Horstkotte D, Bennett J, Huysmans H, Lindblom D, Olin C, Orinius E, Semb G. The Bjo¨rk-Shiley 70° convexo-concave prosthesis strut fracture problem (present state of information). J Thorac Cardiovasc Surg 1987; 35:71–77. 6. van der Meulen JHP, Steyerberg EW, van der Graaf Y, van Herwerden LA, Verbaan CJ, Defauw JJAMT, Habbema JDF. Age thresholds for prophylactic replacement of Bjo¨rk-Shiley convexo-concave heart valves. A clinical and economic evaluation. Circulation 1993;88:156 –164. 7. Hiratzka LF, Kouchoukos NT, Grunkemeier GL, Miller DC, Scully HE, Wechsler AS. Outlet strut fracture of the Bjo¨rk-Shiley 60° convexo-concave valve: current information and recommendations for patient care. J Am Coll Cardiol 1988;11:1130 –1137. 8. Ericsson A, Lindblom D, Semb G, Huysmans HA, Thulin LI, Scully HE, Bennett JG, Ostermeyer J, Grunkemeier GL. Strut fracture with Bjo¨rk-Shiley 70° convexo-concave valve. An international multi-institutional follow-up study. Eur J Cardiothorac Surg 1992;6:339 –346.
SEPTEMBER 15, 1998
9. Kallewaard M, Algra A, van der Graaf Y. Welder identity, weld date, and the
risk of outlet strut fracture in Bjo¨rk-Shiley convexo-concave valves: the Dutch cohort study. Heart 1996;76:510 –512. 10. World Health Organization. World Health Organization (ed). International classification of diseases. Manual of the international statistical classification of diseases, injuries, and causes of death. Ninth Revision. Geneva, Switzerland, 1977. 11. Edmunds LH Jr, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD. Guidelines for reporting morbidity and mortality after cardiac valvular operations. Ad Hoc Liaison Committee for Standardizing Definitions of Prosthetic Heart Valve Morbidity of The American Association for Thoracic Surgery and The Society of Thoracic Surgeons. J Thorac Cardiovasc Surg 1996;112: 708 –711. 12. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457– 481. 13. Cox DR. Regression models and life tables. J R Stat Soc B 1972;34:187–202. 14. Lindblom D, Lindblom U, Qvist J, Lundstrom H. Long-term relative survival rates after heart valve replacement. J Am Coll Cardiol 1990;15:566 –573.
15. Blackstone EH, Kirklin JW. Death and other time-related events after valve
replacement. Circulation 1985;72:753–767. 16. Efron B, Tibshani RJ. An Introduction to the Bootstrap. London: Chapmann
& Hall Inc., 1993. 17. Harrell FE. Design: S-plus functions for biostatistical/epidemiological mod-
elling, testing, estimation, validation, graphics, prediction, and typesetting by storing enhanced model design attributes in the fit. Programs available at internet: lib.stat.cmu.edu. 1996. 18. Blackstone EH, Kirklin JW. Recommendations for prophylactic removal of heart valve prostheses. J Heart Valve Dis 1992;1:3–14. 19. Steyerberg EW, van der Meulen JHP, van Herwerden LA, Habbema JDF. Prophylactic replacement of Bjo¨rk-Shiley convexo-concave heart valve: an easyto-use tool for decision support. Heart 1996;76:264 –268. 20. Bjo¨rk VO. The optimal opening angle of the Bjo¨rk-Shiley tilting disc valve prosthesis. Scand J Thorac Cardiovasc Surg 1981;15:223–227. 21. Lindblom D. Long-term clinical results after aortic valve replacement with the Bjo¨rk-Shiley prosthesis. J Thorac Cardiovasc Surg 1988;95:658 – 667. 22. Lindblom D. Long-term clinical results after mitral valve replacement with the Bjo¨rk-Shiley prosthesis. J Thorac Cardiovasc Surg 1988;95:321–333.
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
773
absolute risks for 70& and 60& BScc valve recipients were 44% and 37%, respectively. Among 70& and 60& BScc valve recipients, underreporting of fracture was estimated to be 25% (95% CI 0 to 49) and 26% (95% CI 0 to 52), respectively. Estimates based on sudden death and fatal congestive heart failure yielded essentially the same results. Thus, underreporting of fatal strut fracture in the Dutch BScc cohort is estimated to be approximately 25%. Hence, the risk and lethality of fracture of BScc valves are underestimated and indications for prophylactic replacement should be adjusted accordingly. For example, the advantage of valve replacement in a 40-year-old patient with a 29-mm 60& BScc mitral valve would almost double to 0.82 years. Q1998 by Excerpta Medica, Inc. (Am J Cardiol 1998;82:768 –773)
jo¨rk-Shiley convexo-concave (BScc) heart valves carry an increased risk of mechanical failure with B potential catastrophic consequences for the implan-
ients in the Netherlands.1 The increased risk could not be explained by documented strut fractures and factors known to be associated with death after valvular surgery.1 If strut fractures indeed remain undetected, the risk estimates of outlet strut fracture are underestimations of the true incidence,1,2,4 – 6 and the recommendations for prophylactic replacement would have to be adjusted accordingly.6 This study quantifies the degree of underestimation of outlet strut fracture in the Dutch BScc cohort.
tee.1,2 Worldwide, approximately 35,000 living BScc valve recipients face the risk of outlet strut fracture. Only 2 cohort studies, a Swedish institution– based study and a Dutch nationwide retrospective cohort study, have examined the risk and lethality of outlet strut fracture.1,2 To date the Dutch BScc cohort study is the only study to follow-up actively both 60° and 70° BScc valve recipients. Consequently, the Dutch cohort serves as a reference population of the risk and lethality of strut fracture worldwide and is being used to formulate guidelines for prophylactic replacement.3 However, even if follow-up is complete, some strut fractures may not be detected.1,2,4 – 6 Strut fracture results in an almost instantaneous death, especially among BScc aortic valve recipients, and most patients die at home (only a few of whom undergo necropsy).1,2,4,6 – 8 Compared to 60° BScc valve recipients, undetected strut fractures may also be the reason for increased risk of death among 70° BScc valve recipFrom the Julius Center for Patient Oriented Research, Clinical Epidemiology Unit, Utrecht University, Medical School, Utrecht, The Netherlands; and Department of Cardiothoracic Surgery, the St. Antonius Hospital, Nieuwegein, The Netherlands. Manuscript received February 19, 1998; revised manuscript received and accepted May 5, 1998. Address for reprints: Yolanda van der Graaf, MD, PhD, Julius Center for Patient Oriented Research, Utrecht University, Medical School, P.O. Box 05500, 3508 GA TA Utrecht, The Netherlands. *Members of the Netherlands Bjo¨rk-Shiley study group are listed in the Appendix.
768
©1998 by Excerpta Medica, Inc. All rights reserved.
METHODS
Patients: The Dutch BScc Follow-up Study is a nationwide retrospective cohort study of all Dutch BScc valve recipients. All Dutch BScc valve recipients were identified and actively followed until outlet strut fracture, death, or the end of the study; 2,264 BScc valve recipients received 2,254 60° and 279 70° valves from 1979 to 1985.1 For each patient a detailed clinical profile was obtained from the clinical records (Table I). Presence of acute endocarditis was defined as surgery before the completion of a course of intravenous antibiotics; left ventricular function was classified as good, reduced, or poor from the right oblique view of the ventricular angiogram; presence of coronary artery disease was defined as a diameter lumen narrowing .50% on the coronary angiogram, and classified as 1-2-, or multiple-vessel (.3) disease. The methods of the first Dutch BScc follow-up study have been described in detail previously.1,9 Clinical characteristics of the patients are summarized in Table I. Follow-up: Information on the vital status was obtained from municipal registries and on reoperation 0002-9149/98/$19.00 PII S0002-9149(98)00461-5
classified as sudden unexplained deaths (ICD-9 code 798).1,2 However, because mitral valve fractures 60° BScc Valve 70° BScc Valve have a protracted course and as many † Recipients Recipients heart valve recipients have some de(n 5 2,000) (n 5 263) gree of congestive heart failure, unAge (yr) (mean 6 SD) 54.2 6 12.6 55.3 6 12.5 detected strut fractures may also be Sex (men) 1,140 (57.0%) 153 (58.2%) misclassified as fatal congestive Rheumatic pathogenesis 399 (20.0%) 69 (26.2%) heart failure (ICD-9 code 428). Recent infective endocarditis 210 (10.5%) 21 (8.0%) Previous valve replacement 90 (4.5%) 14 (5.3%) Moreover, many patients with conLeft ventricular function gestive heart failure die suddenly, Good 1,562 (78.1%) 225 (85.6%) making the distinction between strut Reduced/poor 94 (4.7%) 10 (3.8%) fracture and arrhythmogenic death Unknown 344 (17.2%) 28 (10.6%) extremely difficult. Postoperative Coronary artery disease None 1,345 (67.3%) 180 (68.4%) mortality was defined as death occur1- or 2-vessel 283 (14.1%) 40 (15.2%) ring within 30 days after operation.11 $3-vessel 115 (5.8%) 21 (8.0%) Late mortality was defined as death Unknown 257 (12.9%) 22 (8.4%) during follow-up but after 30 days. History of myocardial infarction 60 (3.0%) 4 (1.5%) Emergency procedure 60 (3.0%) 15 (5.7%) The primary end points of this study Size of BScc prosthesis ($29 mm) 746 (37.3%) 112 (42.6%) were death from all causes except Valve replacement strut fracture and the combined event Aortic 1,009 (50.5%) 113 (43.0%) of sudden death or fatal congestive Mitral 684 (34.2%) 116 (44.1%) heart failure. Aortic and mitral‡ 307 (15.4%) 34 (12.9%) Concomitant surgery Data analysis: Cumulative surCoronary artery bypass surgery 579 (29.0%) 83 (31.6%) vival curves according to the Tricuspid valvuloplasty 29 (1.5%) 1 (0.4%) Kaplan-Meier product limit method Aortic root replacement 34 (1.7%) 0 (0.0%) and the log-rank test were used for *In the present report, the only Dutch BScc valve recipient who underwent mitral and tricuspid valve comparison.12 To estimate the degree replacement, and who experienced a strut fracture, was excluded from analysis. of underreporting we assumed the † Seventeen patients who underwent aortic and mitral valve replacement and received a 60° and 70° following: BScc valve were included here as the 70° BScc valve recipients only. ‡ Seventy-three patients who underwent aortic and mitral valve replacement of a BScc valve and a 1. First, we compared the survival non-BScc valve. of 60° and 70° BScc valve recipients, while censoring documented strut fractures and adjusting for factors known from all Dutch centers for cardiothoracic surgery (end to be associated with death after valvular surgery. We date of the study: July 1, 1996). Information on causes assumed that any remaining “excess mortality” among of death was obtained from the patient’s general prac- Dutch 70° BScc valve recipients could be attributed to titioner. Mean duration of follow-up was 10.0 years deaths due to undetected strut fractures. We first exam(range from 0 to 17). Forty-six patients (2.0%) were ined the independent contribution of valve opening angle lost to follow-up. and risk factors, known to be associated with late morOutcome events: The causes of death were coded tality after valvular surgery (as listed in Table I), to the independently by 2 of the authors (AA and YvdG) several causes of death in a multivariate Cox proportionaccording to the 9th revision of the International Clas- al-hazards regression model (entry criterion p ,0.2).13 sification of Diseases.10 For the purpose of this study Because there are different risk factors associated with we identified the following categories: death from death during the first year after valvular surgery and outlet strut fracture, sudden death or fatal congestive death during the subsequent years,2,14,15 we restricted our heart failure (ICD-9 codes 798 and 428), other causes analyses to BScc valve recipients who at least survived 1 of cardiovascular death (ICD-9 codes 391-427 and 429-460), death due to malignancy (ICD-9 codes 150- year after valvular surgery (n 5 1,949). For each of the 240), remaining causes of death, and a category that variables entered in the model, the proportional-hazards assumption was tested using time-dependent covariates. included deaths from unknown causes. We decided in advance to combine sudden death Hazard ratios are presented with 95% confidence limits and fatal congestive heart failure as the class most (95% CI) derived from the Cox model. 2. Next, we determined the absolute 14-year risks likely to contain missed fatal strut fractures. Outlet of death from all causes except strut fracture and from strut fracture of an aortic BScc valve is followed by almost instantaneous death,1,2,7 and fracture of a mitral the combined event of sudden death or fatal congesvalve results in sudden onset of symptoms of severe tive heart failure for “average” 60° and 70° BScc left ventricular failure with a duration of up to 48 valve recipients who survived at least 1 year as S0(t 5 exp (PI 1 beta z valve opening angle [0 5 60°, 1 5 70°]) , where hours, and persisting without hemodynamic improve- 14) ment until death.1,2,7 Given the acute nature of the PI is the summed product of the coefficients derived complaints following outlet strut fracture, it has been from the Cox model with the mean cohort levels of suggested that undetected strut fractures may be mis- variables of the prognostic factors used in the model. S0 TABLE I Clinical Characteristics of 2,263 60° and 70° Bjo ¨ rk-Shiley CC (BScc) Valve Recipients in the Dutch BScc Cohort*
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
769
FIGURE 1. Age-adjusted survival curves of Dutch BScc valve recipients according to valve opening angle and mode of death. Upper lines, death from all causes except strut fracture; lower lines, death from all causes.
(t 5 14) is the baseline survival function of the models at 14 years. The excess mortality, possibly attributable to undetected strut fractures among 70° BScc valve recipients, was then calculated by subtracting the 14-year absolute risk estimates of 70° and 60° BScc valve recipients. 3. To finally estimate the number of undetected fractures among 60° valve recipients we assumed that the hazard ratio for the valve opening angle of unreported strut fractures is the same as the hazard ratio for reported strut fractures in this group of patients (hazard ratio 7.8, 95% CI 4.3 to 14.2). Standard errors for the percentage of underreporting of outlet strut fracture were estimated with bootstrap resampling (random sampling with replacement);16 calculations were repeated 500 times. All calculations were performed using SPSS and the functions of the Design Library which works in S-plus software (Data Analysis Products Division of MathSoft Inc., Seattle Washington).17
RESULTS Operative mortality was the cause of death in 137 (6.9%) 60° BScc and 17 (6.5%) 70° BScc valve recipients. Lethality of strut fracture was 88% for aortic valve (7 of 8) and 55% for mitral valve (23 of 42) recipients. Among the 1,863 60° BScc valve recipients who survived at least 30 days after surgery, 17 (0.9%) died of strut fracture, 257 (13.8%) of the combined event of sudden death or fatal congestive heart failure, 183 (9.8%) of cardiac death, 102 (5.5%) of malignancies, and 131 (7.0%) of other causes. Among the 246 770 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 82
70° BScc valve recipients, these numbers were 12 (4.9%), 51 (20.7%), 17 (6.9%), 13 (5.3%), and 7 (2.8%), respectively. The cause of death was unknown for 163 (8.7%) 60° BScc and for 22 (8.9%) 70° BScc valve recipients. Necropsy was performed on 188 (16.7%) of the deceased patients, most of whom died of operative causes (n 5 96). The 14-year cumulative probabilities for BScc valve recipients for different causes of death are presented in Table II. The age-adjusted Kaplan-Meier curves show a better survival among 60° BScc valve recipients than among 70° BScc valve recipients (p log-rank 0.003; Figure 1). The difference was still apparent, although somewhat smaller, after correction for death due to strut fracture (p log-rank 0.04). Further analyses were restricted to Dutch BScc valve recipients who survived 1 year after implantation. Multivariate analysis showed that the risk of death from all causes except strut fracture was 20% higher among patients with 70° BScc valves (hazard ratio 1.2, 95% CI 1.0 to 1.5). The multivariate model, which studied sudden death or fatal congestive heart failure, showed that 70° BScc valve recipients had an almost twofold higher risk (hazard ratio 1.8, 95% CI 1.3 to 2.5) than 60° BScc valve recipients. Valve opening angle was no longer an independent predictor when death from other cardiovascular causes (hazard ratio 0.9, 95% CI 0.5 to 1.5), death from noncardiovascular causes (hazard ratio 0.8, 95% CI 0.5 to 1.4), and death from unknown causes (hazard ratio 1.3, 95% CI 0.8 to 2.0) were studied multivariately. Table SEPTEMBER 15, 1998
0.79 26.2% (19.1–57.8)
DISCUSSION
*Comparison of 60° and 70° BScc valves; p value log-rank test. Values are expressed as cumulative probability (%) (95% CI). CI 5 confidence interval; CHF 5 congestive heart failure.
13.4% (11.2–15.9) Other cardiovascular causes
12.3% (6.1–23.9)
0.45
15.0% (12.0–18.6)
14.0% (7.3–26.0)
0.89
20.7% (15.8–26.8)
0.50 39.9% (14.5–80.8) 24.7% (21.0–28.9) 12.9% (10.7–16.0) Sudden death/CHF
25.7% (16.8–38.2)
,0.001
40.5% (30.0–52.9)
,0.001
24.3% (18.9–31.0)
0.20 75.1% (54.4–91.5) 53.3% (49.5–57.3) 40.9% (37.9–44.1) Any cause except fracture
53.1% (43.2–63.8)
0.02
57.7% (47.6–68.1)
0.22
57.5% (51.8–63.4)
0.03 80.1% (61.9–93.3) 41.1% (38.1–44.3) Any cause
55.2% (45.3–65.6)
0.007
54.3% (50.4–58.2)
60.2% (50.4–70.2)
0.10
58.3% (52.6–64.1)
p Value* 70° BScc (n 5 34) 60° BScc (n 5 307) Causes of Death
60° BScc (n 5 1,009)
70° BScc (n 5 113)
p Value*
60° BScc (n 5 684)
70° BScc (n 5 116)
p Value*
Aortic and Mitral Valve Mitral valve (n 5 800) Aortic Valve (n 5 1,122)
TABLE II Fourteen-Year Cumulative Probabilites for Death from Any Cause, Death Due to Any Cause Except Strut Fracture, Sudden Death, and Cardiac Death in the Dutch BScc Cohort
III summarizes the average 14-year absolute risk of death from all causes except strut fracture and of the combined event of sudden death or fatal congestive heart failure for the 60° and 70° BScc valve recipients. Among 70° BScc valves underreporting was estimated to be 25% (95% CI 0 to 49) if deaths from all causes except strut fracture were considered, and 18% (95% CI 3 to 33) if the combined event of sudden death or fatal congestive heart failure was used. Underreporting of strut fracture among 60° BScc valves was estimated to be 26% (95% CI 0 to 52) and 19% (95% 4 to 35), respectively (Table III). Analysis of the data according to valve position yielded essentially the same results. Although the degree of underreporting was somewhat higher in aortic valves this difference was not statistically significant. The findings of our study suggest that the degree of underreporting of outlet strut fractures in the Dutch BScc cohort may be as high as 25% (95% CI 0 to 49). The current risk estimates of outlet strut fracture are therefore substantial underestimations of the true incidence. Several groups have published guidelines for prophylactic replacement of BScc valves,4,6,18,19 each of which debated the validity of the estimates of the risk and lethality for outlet strut fracture; however, adjustment of these risk estimates had considerable effect on the indications for prophylactic replacement.6,19 For example, Steyerberg et al19 showed that for a hypothetical 40-year-old male recipient of a 29-mm 60° BScc mitral valve, the advantage of valve replacement would be 0.45/year. Applying their method it can be calculated that the advantage of surgery would almost double to 0.82 year if the estimates for the risk and lethality of outlet strut fracture were 25% higher. In contrast, because of the enormous underreporting of fractures in other countries, fracture rates are multiplied by 2.4 (USA) or 3.7 (non-USA, nonNetherlands) to approximate Dutch fracture rates.3,18 Data from this study suggest that even higher multiplication factors should be used. Given the low incidence of necropsy in the Netherlands, complete assessment of the true incidence of outlet strut fracture was impossible. We therefore approximated the degree of underreporting of outlet strut fracture by quantification of the proportion of the “unexplained excess” deaths among 70° BScc valve recipients. We assumed that because the higher mortality among 70° BScc valve recipients could not be explained by clinical characteristics known to be associated with death after valvular surgery or by documented strut fractures, the excess had to be attributed to unreported outlet strut fractures. It may be debated whether all relevant clinical variables explaining different mortality in 60° and 70° BScc valve recipients are included in the multivariate models. However, because it is not plausible that patients were assigned to have a 60° or 70° BScc valve because of their clinical profile, we consider it unlikely that the 30% higher risk of death among 70° BScc valve recipients may entirely be explained by variables not included in
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
771
TABLE III Estimation of the Degree of Underreporting of Outlet Strut Fractures Among Dutch BScc Valve Recipients Who Survived at Least One Year After Implantation During a Two- to 14-Year Period After Implantation At Risk
Absolute 14-Year Risk* (95% CI)
Deaths
Cause of death
Valve Type
All causes except strut fracture Combined event of sudden death Or fatal congestive heart failure
70° BScc 60° BScc 70° BScc
44.0% (37.5–51.1) 37.2% (34.7–39.8) 23.5% (17.8–30.8)
96 692 45
60° BScc
13.7% (11.8–15.8)
215
Patients
Person– Years
Documented Strut Fractures
Unreported Strut Fractures†
Percentage Underreporting‡ (95% CI)
224 1,755 224
1,855 17,795 1,855
20 23 20
6.5 8.0 4.4
25% (0–49) 26% (0–52) 18% (3–33)
17,795
23
5.4
19% (4–35)
1,755
*Adjusted (see Table II) absolute-14 year risks. † The number of undetected strut fractures of 70° BScc valves was estimated by multiplying the difference of the 14th-year absolute risks between 70° and 60° BScc valve recipients with the number of deaths. For example (44.0%–37.2%) z 96 deaths 5 6.5. Assuming that the hazard ratio for valve opening angle of unreported fractures is 7.8, the number of unreported strut fractures of 60° BScc valves could be calculated. For example (6.5/1855)/(?/17,795) 5 7.8 . ? 5 8.0. ‡ [Underreported fractures/(documented 1 underreported fractures)] z 100%.
our multivariate models. Because of the small number of postmortem examinations, we were not able to differentiate valve-related complications such as valve thrombosis and arrhythmias from outlet strut fracture. A higher incidence of these fatal complications may be another explanation for the higher mortality among 70° BScc valve recipients. If so, this would be rather ironic since the 70° BScc valve was designed to be less thrombogenic.20 The superior quality of the 70° BScc valve compared with the 60° BScc valve, however, has never been proven.21,22 To calculate the number of undetected outlet strut fractures among 60° BScc valve recipients, we assumed that the hazard ratio for the valve opening angle for unreported strut fractures is the same as that for reported strut fractures. If so, the degree of underreporting of outlet strut fracture would approximately be the same among 60° and 70° BScc valve recipients, an assumption that was also made by others.2 However, because the necropsy rate among the Dutch BScc patients who died during follow-up was higher among the 70° BScc valve recipients (14.8% vs 8.7% among the 60° BScc valve recipients; p 5 0.03), it could be argued that the degree of underreporting of strut fracture in 60° valves in reality is even higher than the estimated 20% to 30%. The findings of this study agree with those by Lindblom and colleagues2 who estimated the percentage of undetected strut fractures to be approximately 28% (with a possible maximum of 47%). They extrapolated the percentage of fractures reported among BScc valve recipients who died suddenly or from unexplained causes, and who underwent autopsy, to patients who died suddenly or from unexplained causes without necropsy2; postmortem examination is routine practice in Sweden (75% of all deceased BScc valve recipients underwent autopsy). In view of these findings, the argument to facilitate autopsy in heart valve recipients is even stronger than it was. It is of the utmost importance to evaluate valvular prostheses carefully after longtime use. The BScc strut fracture problem would have come out in the open significantly earlier if such a policy for artificial heart valve recipients had been in effect. 772 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 82
Finally, if 25% of fatal strut fractures are not reported, the current estimates of lethality of strut fracture would be too low. In our study, the lethality of strut fracture was 54% (23 of 42) for mitral valves and 88% (7 of 8) for aortic valves. If the number of fatal strut fractures of BScc mitral valves were to be 25% percent higher, lethality would increase to 64% (34 of 53) for mitral valves and to 90% (9 of 10) for aortic valves. Higher estimates of the risk and lethality of strut fracture should lead to adjustment of the indications for prophylactic replacement, i.e., more BScc valve recipients should be considered for reoperation.4,6,19 Acknowledgment: We thank Ewout Steyerberg, MD, for his valuable contribution to the analysis.
APPENDIX Participating Centers: Academisch Medisch Centrum (Bas de Mol), Onze Lieve Vrouwe Gasthuis Amsterdam (Leon Eijsman), Medisch Centrum de Klokkenberg (ThR van Geldorp), Academisch Ziekenhuis Leiden (Hans Huysmans), Sint Antonius Ziekenhuis Nieuwegein (Jo Defauw), Academisch Ziekenhuis Utrecht (Jaap Bredee), Academisch Ziekenhuis Groningen (Tjark Ebels), Dijkzigt Ziekenhuis Rotterdam (Lex van Herwerden), Sint Radboud Ziekenhuis (LK Lacquet).
1. van der Graaf Y, de Waard F, van Herwerden LA, Defauw J. Risk of strut
fracture of Bjo¨rk-Shiley valves. Lancet 1992;339:257–261. 2. Lindblom D, Bjo¨rk VO, Semb BKH. Mechanical failure of the Bjo¨rk-Shiley
valve. Incidence, clinical presentation, and management. J Thorac Cardiovasc Surg 1986;92:894 –907. 3. Brookmeyer R, Yasui Y. Statistical analysis of passive surveillance disease registry data. Biometrics 1995;51:831– 842. 4. Birkmeyer JD, Marrin CAS, O’Connor GT. Should patients with Bjo¨rk-Shiley valves undergo prophylactic replacement? Lancet 1992;340:520 –523. 5. Ostermeyer J, Horstkotte D, Bennett J, Huysmans H, Lindblom D, Olin C, Orinius E, Semb G. The Bjo¨rk-Shiley 70° convexo-concave prosthesis strut fracture problem (present state of information). J Thorac Cardiovasc Surg 1987; 35:71–77. 6. van der Meulen JHP, Steyerberg EW, van der Graaf Y, van Herwerden LA, Verbaan CJ, Defauw JJAMT, Habbema JDF. Age thresholds for prophylactic replacement of Bjo¨rk-Shiley convexo-concave heart valves. A clinical and economic evaluation. Circulation 1993;88:156 –164. 7. Hiratzka LF, Kouchoukos NT, Grunkemeier GL, Miller DC, Scully HE, Wechsler AS. Outlet strut fracture of the Bjo¨rk-Shiley 60° convexo-concave valve: current information and recommendations for patient care. J Am Coll Cardiol 1988;11:1130 –1137. 8. Ericsson A, Lindblom D, Semb G, Huysmans HA, Thulin LI, Scully HE, Bennett JG, Ostermeyer J, Grunkemeier GL. Strut fracture with Bjo¨rk-Shiley 70° convexo-concave valve. An international multi-institutional follow-up study. Eur J Cardiothorac Surg 1992;6:339 –346.
SEPTEMBER 15, 1998
9. Kallewaard M, Algra A, van der Graaf Y. Welder identity, weld date, and the
risk of outlet strut fracture in Bjo¨rk-Shiley convexo-concave valves: the Dutch cohort study. Heart 1996;76:510 –512. 10. World Health Organization. World Health Organization (ed). International classification of diseases. Manual of the international statistical classification of diseases, injuries, and causes of death. Ninth Revision. Geneva, Switzerland, 1977. 11. Edmunds LH Jr, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD. Guidelines for reporting morbidity and mortality after cardiac valvular operations. Ad Hoc Liaison Committee for Standardizing Definitions of Prosthetic Heart Valve Morbidity of The American Association for Thoracic Surgery and The Society of Thoracic Surgeons. J Thorac Cardiovasc Surg 1996;112: 708 –711. 12. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457– 481. 13. Cox DR. Regression models and life tables. J R Stat Soc B 1972;34:187–202. 14. Lindblom D, Lindblom U, Qvist J, Lundstrom H. Long-term relative survival rates after heart valve replacement. J Am Coll Cardiol 1990;15:566 –573.
15. Blackstone EH, Kirklin JW. Death and other time-related events after valve
replacement. Circulation 1985;72:753–767. 16. Efron B, Tibshani RJ. An Introduction to the Bootstrap. London: Chapmann
& Hall Inc., 1993. 17. Harrell FE. Design: S-plus functions for biostatistical/epidemiological mod-
elling, testing, estimation, validation, graphics, prediction, and typesetting by storing enhanced model design attributes in the fit. Programs available at internet: lib.stat.cmu.edu. 1996. 18. Blackstone EH, Kirklin JW. Recommendations for prophylactic removal of heart valve prostheses. J Heart Valve Dis 1992;1:3–14. 19. Steyerberg EW, van der Meulen JHP, van Herwerden LA, Habbema JDF. Prophylactic replacement of Bjo¨rk-Shiley convexo-concave heart valve: an easyto-use tool for decision support. Heart 1996;76:264 –268. 20. Bjo¨rk VO. The optimal opening angle of the Bjo¨rk-Shiley tilting disc valve prosthesis. Scand J Thorac Cardiovasc Surg 1981;15:223–227. 21. Lindblom D. Long-term clinical results after aortic valve replacement with the Bjo¨rk-Shiley prosthesis. J Thorac Cardiovasc Surg 1988;95:658 – 667. 22. Lindblom D. Long-term clinical results after mitral valve replacement with the Bjo¨rk-Shiley prosthesis. J Thorac Cardiovasc Surg 1988;95:321–333.
VALVULAR HEART DISEASE/UNDERREPORTING OF FRACTURE OF BSCC VALVES
773