Ann Thorac Surg 2003;76:482-485
© 2003 The Society of Thoracic Surgeons
Original article: cardiovascular
Is 30-day mortality an adequate outcome statistic for patients considering heart valve replacement?
Maria-Benedicta Edwards, MPhila*,
Kenneth M. Taylor, FRCSa
a United Kingdom Heart Valve Registry, Department of Cardiothoracic Surgery, Hammersmith Hospital, London, United Kingdom
Accepted for publication February 15, 2003.
* Address reprint requests to Dr Edwards, UK Heart Valve Registry, Department of Cardiothoracic Surgery, Hammersmith Hospital, Du Cane Road, London W12 0NN, United Kingdom
e-mail: m.b.edwards{at}ic.ac.uk
 |
Abstract
|
|---|
BACKGROUND: In-hospital mortality is widely used by clinicians as a benchmark measure of outcome for determining risks/benefits of cardiac surgery. Patients, however, may wish to have information on estimated longer-term outcomes. Mortality risk by 1 year after the operation may be a more meaningful outcome statistic. We therefore undertook to determine 30-day and 365-day postoperative mortality rates in a large series of consecutive patients who have undergone heart valve replacement (HVR) surgery in the United Kingdom since 1986.
METHODS: Data on 80,757 patients registered on the UK Heart Valve Registry were analyzed. Kaplan–Meier actuarial survival analysis was calculated to determine 30-day (group 1) and 365-day (group 2) mortality. Cox proportional hazards were calculated for each group to identify significant risk factors for mortality less than 1 year.
RESULTS: Thirty-day mortality represents around half (56%) of the 365-day mortality. This ratio was robust for most subdivisions of the total population. Cox proportional hazards demonstrated female sex, age older than 70 years, single tricuspid valve replacement, multiple valve replacement regardless of type of valve, and subsequent valve operation as significant risk factors of early mortality (group 1). However, men were at significantly greater risk of late mortality (group 2). All other factors remained significant with the exception of subsequent valve operation, which was dropped from the model (group 2).
CONCLUSIONS: The robustness of these data would allow cardiologists and cardiac surgeons to provide preoperative patients with a reasonably accurate estimate of survival rates at 1 year after valve replacement surgery in addition to the customary short-term 30-day outcome.
|
Introduction
|
|---|
The use of in-hospital mortality as a benchmark measure of outcome is still widely used by cardiologists and cardiac surgeons as they discuss with patients the potential risks and benefits of a proposed cardiac operation. Although previous authors have rightly encouraged use of the standardized 30-day mortality data (where available), and continued procedural related risk beyond 30-days has previously been highlighted in coronary artery studies [1–3], this very short-term outcome measure may not satisfy the discerning patient who may wish a reasonable estimate of a longer-term outcome. For many patients the mortality risk by 1 year after the operation may be a more meaningful outcome statistic. We therefore undertook to determine 30-day and 365-day postoperative mortality rates in a large series of consecutive patients who had undergone heart valve replacement surgery in the United Kingdom since 1986.
|
Material and methods
|
|---|
The UK Heart Valve Registry, set up in 1986 and based at the Hammersmith Hospital London, is a national database funded by the Department of Health, which prospectively collects data from all UK cardiac surgical centers. Certain preoperative, implant, and postoperative data are collected for all patients who undergo heart valve replacement surgery in the United Kingdom. All patients are followed up for the occurrence of death through the national agencies responsible for registering all deaths of UK residents and nationals (Office for National Statistics [England and Wales], General Register Office [Scotland], and Central Services Agency [Northern Ireland]). These agencies return to the Registry a copy of the patient’s death certificate, which includes date, place, and certified cause of death, and identifies patients who underwent autopsy. Thus, the Registry is provided with precise data from implant to death for all patients registered on the database. Analyzed data included patient age at operation, sex, number of valves replaced at operation, valve position, type of valve implant, sequence of valve operation, follow-up time, and date and cause of death.
From the outset, the emphasis of the Registry has been on achieving maximal cooperation and completeness of patient registration. The data requested therefore are intentionally brief and simple to obtain. The Registry does not seek to obtain intercurrent morbidity data or follow-up data on patients apart from the occurrence of reoperation and death. This limitation was accepted from the planning stage as inevitable to maintain simplicity and completeness of data collection, rather than risk reduced cooperation and incomplete data from implanting centers. It is therefore acknowledged that the UK Heart Valve Registry data cannot be used to analyze outcomes in highly specific details as can be undertaken in smaller series from individual institutions. However, the Registry is able to report accurate mortality data and time-related survival as well as trends in heart valve replacement surgery within the United Kingdom as a whole.
Between January 1, 1986, and December 31, 2000, 80,757 consecutive patients underwent valve replacement surgery in the United Kingdom and were registered on the UK Heart Valve Registry database. Kaplan–Meier actuarial survival curves were constructed to determine 30-day (group 1) and 365-day mortality (group 2) for the total population and for each variable separately. Cox proportional hazard function was conducted separately for each group to identify significant risk factors for mortality within 1 year and risk factors with a p valve considered to be significant (p < 0.005). Data from the UK Heart Valve Registry database were analyzed using Stata 6 Release Software (Stata Corporation, College Station, TX).
|
Results
|
|---|
The study was 97.5% complete with 2,059 patients lost to follow-up, giving a total of 440,009 patient follow-up years.
Patient population and characteristics
The patient population comprised 43,294 men and 37,463 women. The mean age for first-time operation was 62.9 ± 12.0 years (range 18 to 95 years). The total number of valves implanted was 90,184, of which 87,410 were implanted in first-time operations and a further 2,774 valves were implanted in subsequent valve reoperations.
Aortic valve replacements (AVR) accounted for 57,948 operations, compared with 31,597 mitral (MVR), 459 tricuspid (TVR), and 180 pulmonary valve replacements. There were a total of 76,370 single valve replacement operations and 6,907 double operations. Eighty-eight patients had undergone triple valve replacement; however, due to the small numbers these triple valve patients had been excluded from the analyses.
Mechanical heart valves remain the valve of choice and account for two-thirds (n = 59,885, 66.4%) of all valve implants. In comparison, there were 29,578 (32.8%) bioprosthetic valves and 721 (0.8%) human tissue valves implanted.
Overall mortality within 1 year
Overall, 8,278 (10.3%) patients died within the first postoperative year, and 52% (n = 4,287) of these deaths were cardiac-related. The death rate was slightly higher in early mortality (deaths occurring up to 30 days, 53%) compared with later mortality (occurring between 31 and 365 days, 50%), although this difference was not significant. Similarly, valve-related deaths accounted for 10% (n = 818) of all deaths in the first year, with only 2% of deaths up to 30 days after the operation being classified as valve-related, compared with 21% for deaths occurring between 31 and 365 days.
Figure 1
shows deaths according to the number of days after the operation for 1 year after the operation (in 30-day increments) expressed as a percentage of the total deaths at 365 days. The five major causes of death were cardiac failure (n = 2,862, 35%), myocardial infarction (n = 897, 11%), cerebrovascular accident (n = 847, 10%), pneumonia (n = 363, 4%), and infective endocarditis (n = 283, 3%). The overall 1-year mortality data showed 4,696 (56.7%) deaths occurred within 30 days of the operation. A further 3,582 (43.3%) deaths occurred between 31 to 365 days afterward.
Table 1
shows factors associated with a higher mortality. For death less than 30 days after the operation, these factors included female sex, age at operation older than 70 years, double valve replacement, single MVR or TVR, replacement with a human tissue valve, and reoperation (ie, re-replacement of a prosthetic heart valve). These factors were the same for patients who died between 31 and 365 days after the operation with the exception of sex and subsequent valve surgery.
Group 1: 30-day mortality
Deaths less than 30 days were mainly cardiac-related (n = 2,487, 53%) and non–valve-related (n = 4,613, 98%) [1]. Hazard function ratios of deaths at less than 30 days after the operation confirmed the mortality findings in Table 1 and identified the following risk factors as significant: age at operation older than 70 years, single MVR and TVR, valve replacement with either a single bioprosthesis or human tissue, double valve replacement (MVR+TVR; AVR+TVR), and reoperation (re-replacement of a prosthetic heart valve) (Table 2).
Although women were more likely to die than men, this difference was not significant (Table 2).
Group 2: mortality between 31 and 365 days
Thirty-one to 365 days after the operation a further 3,582 patients died (43%). Half of these patients died from cardiac-related illnesses (n = 1,800, 50%) and half from non–cardiac-related diseases (n = 1,782, 50%). Valve-related deaths, however, accounted for 21% (n = 736) of deaths. In contrast to the earlier period, men demonstrated a significantly greater risk of dying during this period. Also at significant risk of mortality 31 to 365 after the operation were patients who were older than 70 years; had undergone single AVR, TVR, or double valve replacement with either combined aortic and mitral valves or mitral and tricuspid valves; or had received a single bioprosthetic valve or double mechanical or double bioprosthesis. Reoperation was no longer considered to be a predictive risk factor (Table 2).
|
Comment
|
|---|
Patients referred for heart surgical procedures are, in general, increasingly well informed. In preoperative discussions with their cardiologist or cardiac surgeon, the risk-to-benefit ratio of the proposed procedure is central. Traditionally, the outcome measure used to address this issue has been that of early mortality, usually in-hospital mortality or perhaps 30-day mortality. However, as Figure 1 demonstrates, the mortality rate is relatively high over the first 30 days and continues to increase before leveling off. Thus, the use of a 30-day cut-off underestimates the persistent risk of mortality within the first year and raises the question, is the expectation that a patient will be discharged from a hospital alive, or be alive 30 days after the proposed operation, an appropriately beneficial outcome? Does that statistic represent a "meaningful extension of life?" Is such an outcome measure sufficient for the patient to assess the ratio of risks to benefits, or would a reasonably accurate indicator of longer-term outcome be more helpful for the patient to decide whether to agree to the operation?
The present study has determined mortality in heart valve replacement surgery across the first year after the operation, using the UK Heart Valve Registry database of more than 80,000 consecutive patients. The Registry receives copies of the death certificates from the UK government agencies responsible for registering all deaths of UK residents and nationals. The data are therefore extremely robust in determining when patients die after a heart valve operation. These data have allowed analysis of survival at 30 days and at 365 days to determine the disparity and study the ratio of 30-day to 365-day postoperative survival. Although the finding that 365-day mortality is substantially higher than 30-day mortality is no surprise, the consistency of the relative mortality difference is striking. The 30-day mortality rate represents around 57% of the 365-day rate for the total study population and for most of the patient subgroups (Table 1).
Within patient factors, only sex shows a small but statistically significant difference, with 30-day mortality accounting for 60% of 1-year mortality in women compared with 54% in men. This difference may reflect the fact that 30-day mortality after valve replacement surgery has been consistently higher in women than in men. Longitudinal analysis of data from the UK Heart Valve Registry’s most recent report indicates that cumulative survival rates for men and women equalize at around 2 to 3 years after operation, and remain remarkably constant thereafter [4]. Age at operation appears to have little or no effect on the 30-day to 365-day mortality ratio, which was 56.7% for patients younger than 60 years and 56.8% for patients older than 70 years.
Within procedure-related factors, the site of the valve replacement has no effect for single implants in the aortic, mitral, and tricuspid positions, although the 30-day to 365-day mortality ratio for double implants was slightly higher than for single implants. No difference was noted in the ratio for first-time operations versus reoperations at the same site. Single implants with bioprosthetic valves showed a small but significant increase in the ratio compared with mechanical valves (57.5% versus 54.9%) with p < 0.000. Although the ratio is higher for homograft valves (either alone or in combination with a mechanical valve) the numbers are too small to draw any meaningful conclusions.
These data indicate that, as a broad benchmark statistic applicable across sex, age, valve position, primary or redo, and mechanical or bioprosthetic, 30-day mortality rates for heart valve replacement procedures represent around 56% to 57% of the 1-year postoperative mortality rate. The robustness of this ratio would allow cardiologists and cardiac surgeons to provide patients with a reasonably accurate estimate of survival rates at 1-year after the proposed operation, in addition to the customary short-term 30-day outcome. We believe that for most patients, 1-year postoperative survival represents a more meaningful extension of life, rather than simply the chance of leaving hospital alive after their valve replacement surgery.
|
Acknowledgments
|
|---|
The authors thank Linda Morton of the Department of Cardiothoracic Surgery, Hammersmith Hospital London, for her help in formulating ideas and for giving simple and comprehensive explanations of the statistical methodology.
|
References
|
|---|
- Edmunds L.H., Clark R.E., Cohn L.H., et al. Guidelines for reporting morbidity and mortality after cardiac valvular operations. Ann Thorac Surg 1996;62:932-935.[Abstract/Free Full Text]
- Osswald B.R., Blackstone E.H., Tochtermann U., et al. The meaning of early mortality after CABG. Eur J Cardiothoracic Surg 1999;15:401-407.
- Sergeant P.T., Blackstone E.H. Closing the loop: optimizing physicians’ operational and strategic behavior. Ann Thorac Surg 1999;68:362-366.[Abstract/Free Full Text]
- United Kingdom Heart Valve Registry Report 2000. London, UK: A UK Heart Valve Registry publication, 2001
This article has been cited by other articles:
|
|
|
|
 
D. L. Ngaage, M. E. Cowen, and A. R. Cale
Cardiopulmonary bypass and left ventricular systolic dysfunction impacts operative mortality differently in elderly and young patients
Eur. J. Cardiothorac. Surg.,
February 1, 2009;
35(2):
235 - 240.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. S. Gammie, J. W. Brown, J. M. Brown, R. S. Poston, R. N. Pierson III, P. N. Odonkor, C. S. White, J. S. Gottdiener, and B. P. Griffith
Aortic valve bypass for the high-risk patient with aortic stenosis.
Ann. Thorac. Surg.,
May 1, 2006;
81(5):
1605 - 1610.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
