INTRODUCTION

Arterial hypertension (AHT) is a major health problem since it is a known risk factor for developing cardiovascular disease.1 One of the organ disorders derived from AHT is chronic kidney disease (CKD), defined initially by the presence of kidney damage and in later stages, as a decrease in glomerular filtration rate (GFR).

Various publications have shown that CKD is responsible for increased cardiovascular morbidity and mortality, which is proportional to kidney function deterioration.2-4 For this reason, various professional associations and organisations5-8 include reduced GFR as one of the consequences consider in hypertensive patients. They also state that it is useful in therapeutic decision tables. Current recommendations9,10 are aimed more towards minimising the progression of kidney deterioration and treating the complications inherent in kidney failure, thus reducing the cardiovascular risk associated with CKD.

CKD is a condition that can be diagnosed in primary care (PC) in its early stages by estimating GFR through various formulas (hidden CKD) since serum creatinine is not usually altered until more advanced stages. The most widely used and validated are the Cockcroft-Gault11 and MDRD12 formulas, which have various advantages and limitations. They are recommended depending on the stage of renal function alteration in the literature.13-15 In addition, their association with the appearance of cardiovascular events (CVE) has been discovered.

There have been few studies carried out in Spain which have analysed the evolution of long-term cardiovascular morbidity and mortality in cohorts of hypertensive patients monitored in PC, according to renal function. Since AHT is detected and controlled mainly in PC, it stands to reason that the general hypertensive population is better represented by the population cared for in PC.

In 1993, a prospective study was conducted in which the prevalence of left ventricular hypertrophy was determined in a general hypertensive population that was free of cardiovascular disease and that was being treated in a primary care centre.16 The cohort of patients was monitored from that year onwards.17

The aim of this study was to analyse the association between cardiovascular morbidity/mortality and initial renal function, according to the Cockcroft-Gault and MDRD formulas, in a cohort of hypertensive individuals followed during 12 years.

MATERIAL AND METHOD

The "Gòtic" prospective study of a cohort of hypertensive patients treated at a health centre in Barcelona (Spain) was the source of data for this study. Monitoring began in 1993 and lasted 12 years. The selection criteria of the population and the variables analysed have been previously published.17 For this study, we also excluded patients with extremes of body weight (body mass index [BMI] less than 19 kg/m2 or greater than 35 kg/m2), significant alterations in muscle mass, acute renal failure, pregnancy, severe liver disease, generalised oedema or ascites, and those who continued treatment with drugs that blocked secretion of creatinine. Finally, the study cohort included a total of 223 hypertensive individuals.

GFR was calculated according to the Cockcroft-Gault (GFR=[140-age]*weight(kg)/[serum creatinine*72]*0.85, if female)11 and MDRD (GFR=186*serum creatinine—1.154*age—0.203*1.212, if black*0.742, if female)12 formulas and was classified according to the CKD stage.9

A CVE was considered to be the appearance of the following episodes during the follow-up period: heart failure; ischaemic cardiopathy (angina, acute myocardial infarction); stroke (permanent or temporary); peripheral vasculopathy (symptoms of intermittent claudication or confirmation through eco-Doppler); death due to cardiovascular event and sudden death. Sudden death was defined as death occurring within one hour after onset of symptoms or individuals dying without any witnesses who had no prior diagnosis of coronary heart disease or other presumably deadly diseases. The recording of the onset of CVE was done during AHT follow-up visits. All CVEs recorded were confirmed with the patients’ doctor and with the information and registration systems in hospitals and PC centres. 

Strategies were developed to minimise losses during the follow-up period and to recover information related to mortality.17

The CVEs and the causes of death were evaluated by an external committee of doctors who were unaware of the patients' situations regarding their kidney function.

The study was approved by the Ethics Committee for Clinical Research of IDIAP Jordi Gol of the Catalan Institute of Health.

Survival time was measured from the date when the subject was included in the study until the occurrence of a CVE, if such an event occurred. The censoring variable was the presence or absence of a CVE from the inclusion of subjects in the study (1993) to the end date of the follow-up (2005). Only the initial event was taken into account if there were various episodes of the same event in the same patient. Survival and incidence rates were estimated using the actuarial method. For the bivariate analysis, we used the Student's t-test or its corresponding nonparametric test in the case of continuous variables, and the chi-squared test in the case of categorical variables. We used the kappa index to evaluate the correlation between the estimates of GFR carried out with both formulas.18 The prognostic value of GFR calculated with each of the two formulas was evaluated using Cox proportional hazards regression models, as well as the prognostic value of the adjustment factors. One model was evaluated for each of the formulas used.

The covariate adjustments for the initial models were: age (years), sex (male/female), time of AHT diagnosis (months), mean systolic and diastolic blood pressure of the last two visits, obesity at the start of the study (BMI>30 kg/m2), diabetes mellitus (DM), dyslipidaemia and tobacco consumption. The variables included in the final model were selected by combining statistical and substantive criteria.19,20 We evaluated the confounding and interaction effects between the model's variables. Furthermore, we checked the proportional hazards assumption for the various covariates and for the group of those included in the final model, as well as the presence of multicollinearity. Quantitative data were described by estimating the mean and standard deviation when they followed a normal distribution. Otherwise, the median was used along with the interquartile range (IQR). The estimates related directly with the objective of the study are accompanied by their respective 95% confidence intervals (CI). The accepted level of statistical significance was P<.05. The data analysis was performed using the SPSS 15.0 statistical software for Windows.

RESULTS

Of the 265 individuals in the cohort at baseline, GFR could only be measured in 223 (84.2%). The relationship and reasons for exclusion were: 25 (9.4%) had a BMI>35 kg/m2 and 17 (6.4%) for other reasons, such as consumption of drugs that blocked creatinine secretion, presence of severe liver disease, and creatinine readings were not available.

Table 1 shows the characteristics of the participants at the start of monitoring, as well as the mean analytical values, the frequency of different cardiovascular risk factors (CRF), the distribution of GFR estimates according to the Cockcroft-Gault and MDRD formulas, and the proportion of patients treated with antihypertensive drugs.

The median follow-up time was 10.7 (IQR 6.5 to 12.1) years. Follow-up was complete in 191 participants (85.7%) and incomplete in a total of 32 (14.3%). The latter group differed from the first group in mean age (68.5 years versus 64.2 years) and mean time to diagnosis of hypertension (149 months versus 82 months). The distribution of the remaining study variables was similar in both groups.

The cumulative survival, or the proportion of participants who remained free of cardiovascular events by the end of the study, was 64.7% (95% CI, 57.9-71.6). The mean incidence of CVE during the entire follow-up period was 3.6 (95% CI, 2.7-4.4) per 100 hypertensive individuals/year.

The distribution of estimated GFR with both formulas at the start of the study is shown in Table 2. The agreement between the two estimates was moderate (kappa=0.501; 95% CI, 0.399-0.604). Most patients were classified in stages 2 and 3 by both formulas. Within stage 3, stage 3a was more frequent: 86.8% for calculations performed using the Cockcroft-Gault formula and 76.1% when using the MDRD formula.

Table 3 shows the results obtained in the bivariate analysis, which compared participants according to whether or not they suffered a CVE during follow-up. Only the presence of DM was related with a higher probability of presenting at least one CVE during follow-up. Conversely, female sex and a GFR<60 ml/min/1.73 m2 according to the MDRD formula were associated with a lower risk of CVE. However, there was no association between the occurrence of CVE and a GFR as estimated by the Cockcroft-Gault formula.

In the final multivariable model (Table 4), only a GFR≥60 ml/min/1.73 m2 estimated using the MDRD formula and the presence of DM at the start of the study were associated with an increased cardiovascular risk.

DISCUSSION

In this study, a moderately reduced GFR was not associated with increased cardiovascular risk. Even patients with a GFR≥60 ml/min/1.73 m2 according to the MDRD formula had a higher incidence of events. This was the opposite to what has been described in the literature.2,21 These results are not confirmed by the measurement of GFR according to the Cockcroft-Gault formula.

The MDRD formulas is useful to estimate GFR below 60 ml/min/1.73 m2 and has a tendency to underestimate higher values.14,15,22 Its author has recently explained12 the new equation CKD-EPI that has a greater precision than MDRD for GFR≥60 ml/min/1.73 m2.12 The Cockcroft-Gault formula, which is also appropriate for these values of GFR, underestimates renal function less for values above 60 ml/min/1.73 m2.

Other studies have produced conflicting results regarding the performance of the MDRD formula depending on the stages of kidney failure.

In a study performed in Spain with a retrospective cohort of the general population between 35 and 75 years of age and with stage 3 renal function (GFR between 30 and 60 ml/min/1.73 m2), Buitrago et al14 reported that the Cockcroft-Gault formula detected more hidden CKD in men, with higher cardiovascular risk and greater age. MDRD, on the other hand, did this more in women with higher obesity, higher diastolic blood pressure and higher triglyceride readings. The MDRD equation could therefore avert significantly important cases from the point of view of cardiovascular event prevention. These authors found only moderate consistency between the estimates performed with both formulas, which was similar to the results in our study. As with our study, they also did not find any evidence of a relationship between hidden CKD detected by one or the other formula and cardiovascular morbidity and mortality.23 Similar observations have been reported in other studies with larger sample sizes, such as the NHANES-I24 study in the general U.S. population.

This questions the importance of GFR as a predictive tool for cardiovascular risk by itself, or at least the usefulness of considering Stage 3 as a single group.25 Indeed, various studies found that the risk increased significantly with GFR below 45 ml/min/1.73 m2 (stage 3b), with the risk of patients with higher GFR (stage 3a) comparable to those with normal renal function.2,26 In our study, most patients in stage 3 corresponded to stage 3a, both by the Cockcroft-Gault and the MDRD formulas.

The controversy between formulas persists even in more advanced stages of kidney failure (4 and 5) for which some authors advocate the superiority of the Cockcroft-Gault formula over the MDRD.27

Furthermore, ageing reduces the relationship between estimated GFR and morbidity and mortality as shown in some studies.28 Other authors29 did not report any increase in mortality in elderly patients with GFR between 45 and 59 ml/min/1.73 m2. Other factors, such as sex, may also affect the prognostic value of GFR. A lower correlation has been reported in women25,30 and it is females that make up the majority of our sample.

As with other studies2-4 that did find a relationship between reduced GFR and cardiovascular morbidity and mortality, a first group corresponded to the general population. The mean age in our study was higher than those of patients in the Go et al. and Hallan et al. studies. This may contribute to a different cardiovascular risk and may change the importance of the renal function prognosis. The Keith et al. study, with a mean sample age similar to ours, found that the prognostic differences between stage 2 and 3 were less than 5%, with statistical significance in large samples but not in smaller ones such as ours. Furthermore, the Go et al. study reported that the greatest prognostic differences occurred when GFR<45 ml/min/1.73 m2 (stages 3b to 5), especially between 60 and 80 years of age, and with an GFR<30 ml/min/1.73 m2 in those over 80 years of age, which was a lot less among stages with better renal function.

In studies of hypertensive patients, Ruilope et al.31 found that GFR had a predictive value, but in patients with high cardiovascular risk.

Our study has various limitations. One of them lies in the greater likelihood of losses due to the long duration of the follow-up (longer than in other studies). This could increase the possibility of a selection bias that would affect internal validity. However, a number of strategies used17 meant that in the end the percentage of losses was 14.3%.

Another drawback was the lack of monitoring of certain parameters such as the evolution of GFR over time. This could be of great importance since it has been recently reported that the rate of decline in renal function for individuals over 65 years old is higher in relation to cardiovascular risk than baseline creatinine.32

Also, albuminuria and proteinuria were not recorded systematically at the start of the study. In 1993, their measurement was not included in the assessment protocols for hypertensive patients. Several studies have emphasised the importance of this parameter, which has a greater prognostic value than GFR estimated by MDRD. Hemmelgarn et al.26 found an increased cardiovascular risk and progression of renal deterioration in patients in stages 1-2 with proteinuria compared to patients in stage 3a without proteinuria. Furthermore, the PREVEND study33 showed that individuals with GFR<60 ml/min/1.73 m2 with no proteinuria did not have a higher cardiovascular risk than those with higher GFR. Combining reduced GFR with proteinuria may improve risk prediction.25,26,34

Other markers, such as cystatin C, have recently been shown to be predictors of cardiovascular risk and renal disorders in elderly patients with GFR≥60 ml/min/1.73 m2 to a greater extent than GFR according to MDRD25,35 as occurs with the aforementioned CKD-EPI.15 However, there is little access to both of these methods in PC.

Lastly, the expected influences of antihypertensive treatment were not evaluated during the follow-up, nor were other treatments such as hypolipidaemic or antiplatelet therapy.

In conclusion, this study does not demonstrate an increased cardiovascular risk in hypertensive patients with moderately reduced GFR.

The formula estimates of GFR on its own, currently recommended by clinical practice guidelines, may be less useful for predicting CVE in hypertension as they do not reflect the real risk of some important populations. It may be advisable to consider redefining the values of risk factors of renal function (stages 3 or 3b) and assess the possible benefit of incorporating gender and the value of albuminuria to improve the predictive value of estimated GFR. Hypertensive patients in stages 2-3a are the most prevalent in PC and are therefore the ones who need the best strategies to prevent cardiovascular events.

ACKNOWLEDGEMENTS

This study was made possible thanks to the collaboration of the entire "Gothic" PCT of Barcelona who contributed to the optimal monitoring of the patients. Furthermore, the study was supported by the IDIAP Jordi Gol, as well as assistance (FIS Exp. PI040356) from the Carlos III Health Institute of the Ministry for Science and Innovation. The authors of this study declare that there are no conflicts of interest.

10297_108_11173_en_10297_18107_1784_es_10297_16032_1784_es_tabla_1_en.doc

Table 1. Characteristics of the study cohort at the start of follow-up (n=223)

10297_108_11174_en_10297_18107_1785_es_10297_16032_1785_es_tabla_2_en.doc

Table 2. Distribution of the study population according to glomerular filtration rates (GFR) at the start of the study, estimated using the MDRD and Cockcroft-Gault formulas

10297_108_11175_en_10297_18107_1786_es_10297_16032_1786_es_tabla_3_en.doc

Table 3. Characteristics at the start of the study of patients who completed the follow-up (n=191) according to the presence or absence of cardiovascular events (CVE) during that period

10297_108_11176_en_10297_18107_1787_es_10297_16032_1787_es_tabla_4.doc

Table 4. Final Cox regression model to determine the predictive power of cardiovascular events on the study variables at 12 years follow-up of the cohort of hypertensive individuals (n=223)