In low and middle-income countries, heart valve disease (HVD) is caused mainly by rheumatic heart disease (RHD) [1‐6]. Although incidence of acute rheumatic fever has been declining over recent decades in some of these countries, RHD is still considered a major public health problem worldwide [5]. In high-income countries, HVD remains common, despite decreasing prevalence of RHD, because the decline in inflammatory HVD aetiology has been accompanied by an increase in degenerative valve diseases [7]. Most studies describing HVD patterns have been performed in developed countries, which contain less than 20% of the world’s population [4]. Large-scale nationwide studies in developing countries are scarce [4, 5].

In Brazil, although the incidence of acute rheumatic fever has been reduced in recent decades [8], a large number of patients need to perform heart valve surgery (HVS) to repair or replace valves that have suffered rheumatic injury. In these cases, the mitral or aortic valves, or both simultaneously, are the most affected [9]. Moreover, as in high-income countries, population ageing has led to an increased number of injuries caused by cardiac valve degeneration. As a result, Brazil has a double burden of diseases connected with poverty, such as rheumatic fever, and degenerative diseases associated with an aging population [10].

Little is known about HVS patient survival in low- and middle-income countries. Most surgical risk evaluations have been based on high-income country data. Similarly, health managers who need to allocate scarce resources rationally to achieve quality in health services are severely challenged. Moreover, disclosure of variations in cardiac surgery outcomes is known to offer opportunities to improve the quality of health care and to be an essential tool for quality assessment.

Maximum and median follow-up were 7.7 and 2.8 years, respectively (0.002–7.707). The 78,808 patients included in the study were similarly distributed by sex, with 40,136 (51.0%) females and 38,702 (49.0%) males. Median age was 50.0 (35.9–62.5) years. Most were residents of Brazil’s southeast region. Median hospital stay was 10 (7–13, 16–18) days. The most common surgery procedure was VRp, in 54,475 (69.1%) patients. The baseline characteristics of the final cohort are described in Table 1.

Valve injury aetiology, when identified, is shown in Table 2. The most prevalent cause was rheumatic heart disease. The second group was classified as “insufficient aetiology description” (IAD). This group included situations when there is no specific ICD10.0 code to describe the aetiology; for example, degenerative aortic stenosis, probably coded just as “aortic stenosis”. IAD group median age was 60.7 years. Median age was lower (44.2 years) in RHD patients than in the remaining cohort (57.5 years) (p < 0.001). In mitral stenosis patients, aetiology was RHD in 94.7% (p < 0.001). As aetiology was improperly encoded in 35.1% of patients undergoing HVS, this variable was not used in risk assessment.

In-hospital mortality was 7.6% (3585 patients) and overall mortality at close of study period was 22.1% (17,383 patients) from all causes of death. Most causes were classified by ICD codes as diseases of the circulatory system (13,449 deaths; 77.6%), followed by diseases of the respiratory system (813 deaths; 4.7%).

Table 3 shows a subset of patients (n = 47,515) where the type of mitral or aortic valve disease was identified and where there was no concomitant CABG. The most common valve injury encountered was mitral stenosis (38.9%). Patients with combined aortic and mitral lesions were the youngest (median 43.3 years) and those with aortic stenosis were the oldest (median 58.0 years) (p < 0.001). Repair was performed predominantly in females (65.9%; p < 0.0001).

When stratified by age group, highest in-hospital mortality was observed in patients aged ≥80 years, and the lowest was for those between 20 and 39.9 years (p < 0.001). In-hospital mortality rates differed significantly by sex; they were higher among women (7.8% versus 7.3%; p < 0.001), although this difference was small (Table 4). In-hospital mortality by valve procedure was 3.5% for VRr, 6.9% for VRp, 8.2% for multiple valve repair and/or replacement and 14.6% for concomitant CABG (p < 0.001).

Univariate and multivariate logistic regressions (Tables 5 and 6) were performed to identify independent per-operative predictors for in-hospital mortality. All variables that were significant (<0.10) in univariate analyses were added to the multivariate logistic regression model. Corresponding odds ratios (ORs) and 95%CIs were calculated. Because of their low frequency, the following aetiologies have been grouped as “other”: Kawasaki disease, Takayasu arteritis, endomyocardial fibrosis and hypertrophic cardiomyopathy.

The following tables show multivariate analysis by stepwise logistic regression for in-hospital mortality including the variables age and female sex. Table 7 compares concomitant CABG with the other surgeries studied in our cohort. In this comparison, concomitant CABG increases risk (OR = 2.0366). Table 8 compares valve repair against replacement or multiple valve surgeries. In this comparison, repair reduces risk (OR = 0.4586).

Shortest hospital stays were for repair procedures (median = 9 days) (p < 0.001) and showed no difference when analysed by sex (p = 0.21). They were shorter for patients who died during index procedure admission (7 days) than for survivors (10 days) (p < 0.001). Significantly longer hospital stays were observed among patients who underwent concomitant CABG (11 days) and those over 80 years old (12 days).

Overall survival of HVS patients is shown in Table 9 and Fig. 1a. There were 78,808 HVS patients, and survival was 69.9% (69.38–70.47%) at the end of follow-up time. Table 10 shows overall survival of patients with rheumatic heart disease. Figure 1b shows survival by procedure: at 1 and 5 years, this was 92.5% (95%CI 91.8–93) and 85.3% (95%CI 84.2–86.4) for valve repair; 85.7% (95%CI 85.4–86.0) and 75.5% (95%CI 75.1–76) for valve replacement; 83.8% (95%CI 83.1–84.4) and 74.4% (95%CI 73.4–75.5) for multiple valves surgery; and 72.2 (95%CI 71.1–73.2) and 58.1% (95%CI 56.2–60) for concomitant CABG. Survival, by type of valvular lesion showed no differences.

Survival was observed to differ significantly by sex (Fig. 2a). Long-term survival by sex at year 1 and 5 was, respectively, 85.0% and 76.5% for females and 84.2% and 72.5% for males.

Significant differences by age group are illustrated in Fig. 2b. Survival was worst among patients aged ≥80 years and best in the 20 to 39 year age group. Survival rates after year 1 and 5 of follow-up were, respectively, 64.7%, and 46.4% in the ≥80 year age group and 91.5% and 85.1% in the 20 to 39 year group.

Long-term survival rates differed significantly between patients who underwent HVS and concomitant CABG and those who underwent HVS only (Fig. 3a): at years 1 and 5 of follow-up, these were, respectively, 72.1% and 58.1%, against 85.9% and 76.0%.

Figure 3b shows long-term survival of patients submitted to HVS with non-concomitant CABG. Long-term survival rates for VRp and multiple valve surgery were similar at year 1 (85.7% versus 83.8%) and year 5 (75.5% versus 74.4%) of follow-up, while VRr returned highest long-term survival rates (92.5% and 85.3%), respectively.

The proportional hazard assumption was violated for sex and geographical region of residence; accordingly, these variables were stratified in multivariate survival analysis. For the other variables, the proportional hazard assumption was satisfied 30 days after hospital discharge; thus this is the timeframe considered in multivariate analysis. Patients who underwent valve surgery with concomitant CABG showed greater risk than those who underwent valve surgery alone (HR 1.3123; 95%CI 1.2271–1.4033; p < 0.0001), while patient age contributed, as secondary factor, to increases in this set (HR 1.0216; 95%CI 1.0203–1.0229; p < 0.0001). Non-concomitant CABG patients submitted to valve repair showed lower risk than those with VRp or multiple surgeries (HR 0.6937; 95%CI 0.6296–0.7645; p < 0.0001), but age contributed, as a secondary factor, to increases in this set (HR 1.0209; 95%CI 1.0195–1.0222; p < 0.0001).

To the best of our knowledge, this retrospective HVS cohort, comprising 78,808 patients, is one of the largest ever published. Sex distribution was balanced. Valve replacement accounted for 69.1% of procedures performed. The most common valve injury was mitral stenosis (38.9% of the total). In 94.7% of mitral stenosis patients, aetiology was rheumatic heart disease. In-hospital mortality was 7.6% and was higher for females who had undergone the concomitant CABG procedure and for the elderly. Overall survival was 69.9% at the end of follow-up time. Survival was worst among elderly, male and concomitant coronary artery bypass grafting patients. Median age was 50.0 years, a lot younger than in studies in developed countries, such as in the Euro Heart Survey on Valvular Heart Disease, where mean age was 64 years [21]. Proportions of valve lesion types and aetiologies were also different from most published studies based on North American and European populations [22‐24]. In the “valve lesion identified” subset of the study population, mitral stenosis was the most frequent lesion. RHD was the most prevalent aetiology, accounting for more than half of the subjects studied. This finding is similar to those of other studies in lower- and middle-income countries [10, 25‐27].

Rheumatic fever is the leading cause of mitral stenosis worldwide and, in 40% of cases, it presents as a single lesion [28]. In our cohort, the aetiology underlying 94.7% of mitral stenosis lesions was RHD. VRp accounted for more than two thirds of procedures, while multiple valve surgery accounted for 15.48%. Multiple valve surgery was observed more frequently than reported in the literature (3–14%) [29], which probably relates to the prevalence of RHD in our cohort (53.7%).

In developing countries, although repair procedures offer the advantage of averting long-term anticoagulation therapy and achieve better outcomes, VRp is usually performed more than VRr in cases of rheumatic disease [25, 30, 31]. Patients are less likely to be suitable for VRr in developing countries, where waiting times are long and RHD is common; as a result, access to surgery tends to occur at an advanced stage of disease. The likelihood of successful mitral repair is associated with the timing of surgical referral, annual volume and expertise of cardiac surgery centres, underlying aetiologies and type of lesion [26, 27]. In RHD, valve repair may be rendered impossible, even for experienced surgeons, by various degrees of fibrotic process, involvement of chordae tendineae and calcification of the mitral leaflets or annulus [25]. In developed countries, at institutions with good performance, valve repair is standard therapy for most patients with mitral regurgitation [32]. Note, however, that the most common aetiologies for valve dysfunction in low- and middle-income countries are different from those studied and often published in developed countries. Accordingly, valve repair entails different levels of challenge and outcomes in these different real settings. Although overall survival has been better among RHD patients than among those with valve lesions with other aetiologies (probably as a result of younger age and fewer comorbidities), valve lesions from rheumatic fever are still a undesirable outcome since RHD is largely avoidable by means of primary and secondary prevention.

Valve injury aetiology was not clearly stated in some groups of our cohort and hence we classified it as “insufficient aetiology description”. The lack of a specific ICD 10.0 code may be one of the underlying causes of insufficient coding. In Brazilian public health databases, the “diagnosis” field draws on the ICD 10.0 codex to classify diseases and other health problems. In the literature, degenerative aetiology plays a very important role in other HVS cohorts [33, 34]. Although this study population is younger, it must be assumed that many patients grouped as “insufficient aetiology description” presented injuries caused by a degenerative process of the heart valves. The median age of these miscoded patients was 60.7 years, in contrast to the remaining group where the median was 44.2 years. Corroborating this assumption, almost 80% of patients over 80 years old were classified as “aetiology insufficiently described”. In addition, the proportion of Brazil’s population over 65 years old was 4.0% in 1980, increasing to 7.4% in 2010. Life expectancy in Brazil increased 10 years in that period and, therefore, aging-related diseases can be expected to emerge [35].

Note also that, in one of the few studies evaluating the burden of HVD in developing countries, the most frequent aetiology was RHD (60.3%), in line with this study, while degenerative valve disease, detected in 15% of patients, was the second most common aetiology [10].

Overall in-hospital mortality was 7.6%, but higher for female patients who had undergone concomitant CABG and for the elderly. Concomitant CABG increases mortality substantially [22, 29, 33]. Here, the study population was younger (age has been identified as an important independent predictor of HVS outcomes in adults) [29, 33, 34]. This study involved more patients than those of Leavitt (1057) and Nicolini (1167), although the number of concomitant CABG surgeries was smaller and involved a younger population. Leavitt and Nicolini reported overall in-hospital mortality of 15.5% and 6.9%, respectively. Hannan (14,190) reported 3.3% for CCV without concomitant CABG and 18.7% for multiple valve surgery with concomitant CABG. Both Hannan and Nicolini observed that in-hospital mortality rates were better for VRr than for replacements, similar what is reported in other studies [22, 36‐39]. Ribeiro [40] reported cardiac surgery in-hospital mortality in Brazil as 8.9%, from 2000 to 2003; this was not solely for HVS, however, but included other procedures.

Unlike other published studies [22, 29, 33], this study found better overall long-term survival for women. Although VRr was performed more often in women, it cannot explain this favourable result, because VRr contributed only 6.3% of total procedures. In this study, overall long-term survival rates at 1 and 5 years were 84.6% and 74.5%, which are lower than those reported for a double valve surgery cohort, showing that our outcomes seem to be less satisfactory [29]. Seventy per cent of patients were alive at close of follow-up, which agrees with a previous study with similar procedure mix [41]. Long-term survival for VRr was better than for replacements, as reported in another study [39], but different from Leavitt and Nicolini. The latter author does not supported the theory that, for patients undergoing HVS, VRr is better than replacement. He reported comparable long-term survival at 5 years for VRr and replacement. Different outcomes in different studies can probably be ascribed to cohorts with different demographic characteristics, comorbidities, valve lesion aetiologies, surgeons’ skills and repair techniques, especially in observational studies such as this.

The group of patients who underwent multiple valve surgery comprised a great diversity of clinical features and possible combinations of surgeries, making it difficult to analyse this group in comparison with the VRp group, although they displayed similar results and no statistically significant difference in long-term survival.

Concomitant CABG reduced long-term survival considerably. For patients who underwent concomitant CABG, long-term survival at years 1 and 5 of follow-up was, respectively, 72.1% and 58.1% and, for those who underwent HVS alone, 85.9% and 76.0%.

These rates were similar to those reported in the literature [29, 32, 42, 43]. Hannan and Nicolini have observed that patients who underwent HVS and concomitant CABG were older, had more comorbidities and had more often undergone other previous cardiovascular surgeries. These features could account for lower overall survival.

Taken together with the in-hospital mortality rates and long-term survival found here, length of hospital stay for repair procedures reinforces the assumption that, in Brazil, patients selected for VRr procedures are likely to present better valve and general health status than those submitted to other procedures, which is in line with other studies [34, 39]. Comorbidities and old age predicted lesser likelihood of repair, although good results have been reported for elderly patients undergoing mitral valve surgery, especially mitral VRr, in a large series [34]. Gillinov observed that mitral valve repair was more likely to be performed if valve leaflets were morphologically normal, with functional mitral valve regurgitation, while mitral valve replacement was more often performed in the event of preoperative atrial fibrillation, severe mitral valve stenosis and calcified mitral leaflets or if a mechanical aortic prosthesis was in place [39].

It is clear that old age influences in-hospital mortality, length of hospital stay and long-term survival of patients undergoing HVS. This was also found in this study (OR = 1.0338 for each additional year during hospitalisation). In-hospital mortality was the highest, length of hospital stay, the longest and long-term survival at year 5 of follow-up, the shortest for patients older than 80 years, in agreement with other studies [33, 34, 41, 44].

Length of hospital stay was similar by sex [33] and shorter in patients who died during index procedure admission than in survivors. Most in-hospital deaths (53.9%) occurred up to day 7 of hospitalisation and 25.03% up to day 2. These rates suggest that half of patients died in the operating theatre or in the very early postoperative period. This finding serves to highlight the need to improve the quality of surgical care during the operation or to improve the appropriateness of surgical indications for HVS in Brazil.

The scarcity of long-term survival series drawn from large databases in middle-income countries makes it difficult to compare between studies of populations exposed to similar risk factors and similar patterns of diseases, which reinforces the importance of our findings.

This study was based on a nationwide cohort of patients in the real world setting, which is its main strength. Nevertheless, its conclusions should be read in the light of certain limitations. The two large databases employed offered few clinical variables. This means that we were unable to identify all possible confounding variables. We adjusted for all the available variables, but we should not underestimate the possibility that the omission of important covariates may affect our results.

The limitations of the present study include also the inherent constraints of retrospective, nonrandomised data collection. Although case matching and multivariate analysis can help to account for some differences in the groups, the effects of unmeasured variables affecting selection of therapy may confound the analysis.

In Brazil at the beginning of the twenty-first century, the most common valve lesion is still mitral stenosis, while the most common underlying aetiology is rheumatic heart disease. Single valve replacement accounts for the overwhelming majority of the procedures performed, whereas only a very small number of patients with heart valve diseases benefited from valve repairs. In-hospital mortality and overall survival rates of patients who have undergone heart valve surgery were less satisfactory than those reported in high-income countries. The findings of this study can help guide decision-making processes in middle-income countries similar to Brazil and those concerned with improving the quality of care.