INTRODUCTION

Early detection of patients with occult chronic kidney disease (CKD) is one of the measures proposed by the Spanish Society of Nephrology (SEN) and other institutions to fight against the announced epidemic of renal failure.1,2,3 Primary care physicians and teams have a significant responsibility in early identification of patients with CKD and in optimization of their treatment in the early stages, in order to decrease  the  risk of progression of  this disease and  its associated cardiovascular  morbidity.  While  measurement  of plasma creatinine  levels  is  the most universal routine method  to assess kidney  function,  it  is a well known  fact  that its  relationship  to glomerular  filtration  rate  (GFR)  is poor, with  great  creatinine  clearance  losses  (higher  than  50%) being  required  to detect minimum  increases  in  plasma  levels.4 In fact, creatinine increases above the values considered normal are usually seen when marked decreases in kidney  function have already occurred, particularly  in elderly women and subjects with a  reduced muscle mass.5 This limitation may  be  overcome  by measuring  creatinine  clearance  in  24-hour  urine.  This  method,  however,  not  only overestimates  the  true GFR, but also  requires an adequate urine  collection,  which  represents  a  serious  disadvantage for  routine  use  in  primary  care  practices.  The  K-DOQI (Kidney Disease Outcomes Quality Initiative) clinical practice guidelines6 and  the SEN7 therefore  recommend use of predictive equations  to estimate GFR. Among  these, particular mention should be made of  the Cockcroft-Gault8 and MDRD (derived from the Modification of Diet in Renal Disease  study)  equations.9 Most  comparisons  favour  the MDRD equation,10 but controversy  still exists, and  studies have reported conflicting results.12-15

Based on  the  foregoing,  a  study was designed  to  achieve two  objectives.  1)  to  analyse  the  agreement  between Cock-croft-Gault and MDRD equations in diagnosis of occult CKD in  a  cohort of patients  aged 35-74 years  attending  an urban health  centre with  no  history  of  cardiovascular  disease  and normal plasma  creatinine  values  followed  up  for  10  years, and 2) to ascertain the coronary risk profile of patients diagnosed of occult CKD by each of these two equations to estimate GFR.

PATIENTS AND METHODS

An observational,  follow-up study was conducted on a  retrospective cohort of 845 patients  (4.4% of  the population assigned  to  the  health  centre)  aged 35-74  years  with  no known history of ischemic heart disease or other cardiovascular diseases followed up for 10 years and with recording in  the  clinical history  of  all  of  them,  between  1-01-1990 and 31-12-1994, of the variables required to estimate GFR using  the  Cockcroft-Gault8 and MDRD9 equations  and to calculate coronary risk using the original Framingham16 and the  REGICOR17 equations.  In  the  Cockcroft-Gault  equation, GFR was corrected for a body surface area of 1.73 m2, calculated by the Dubois and Dubois equation.18 The REGICOR  equation was  selected because  this  formula,  adapted from  the original Framingham  equation, has been  calibrated and recently validated in a wide sample of the Spanish population.19 Some equations to calculate risk of cardiovascular death, such as the one derived from the INDANA project,20 include serum creatinine among the eleven variables used to estimate risk, but this equation has not been validated  and  its  applicability  to  the Spanish population  is unknown. It was therefore not included in this study. Moreover, equations derived from Framingham, despite not including CKD among  their  risk  factors, appear  to discriminate adequately  the probability of coronary events  in patients with this disease.21

Cardiovascular  events  investigated  in  cohort  follow-up included coronary events  (acute myocardial  infarction and documented  angina),  stroke,  and cardiovascular  death (from  coronary,  cerebrovascular,  or  other  cardiovascular causes). Acceptance of an event as being of cardiovascular origin required confirmation of diagnosis in the specialized setting  or  at  a  reference  hospital  using  the  relevant  tests (stress  test,  thallium,  coronary  angiography,  etc.). Patients with  a  high  coronary  risk were  defined  as  those  having  a risk > 20% in the original Framingham table16 and > 10% in the  Framingham-REGICOR  tables.22 Qualified  staff  completed  a  case  report  form  for  each  patient  after  the  following variables were defined: systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol, high density  lipoprotein  (HDL)  cholesterol,  smoking, blood glucose, body mass index (BMI), triglycerides, low density lipoprotein (LDL) cholesterol, use of lipid lowering drugs, and use of  antihypertensive drugs. CKD was  classified  in  five stages,6,23 and  the  term  occult CKD was  applied when  the estimated  glomerular  filtration  rate  was  less  than  60 mL/min  (stage 3-5 CKD)1 with plasma creatinine values < 1.4 mg/dL in males and < 1.3 mg/dL in females.24 Patients with severe liver disease, morbid obesity (BMI > 40 kg/m2), malnutrition (BMI £ 18.5 kg/m2), or limb amputations were excluded.7 A total of 845 patients  (8.3% of  the population aged 35-74 years) finally met the inclusion criteria.

Statistical analysis

Data analysis and management was performed using the statistical software SPSS 12.0 for Windows, the R environment (version 2.5.0), and Epi Dat 3.1 software. Parameters used as representative  of  the  sample  for  the  univariate  descriptive analysis  included  mean,  standard  deviation,  observed  frequencies,  and proportions  for normal distributions,  and median and quartiles 1 and 3 for non-normal distributions. Normality  of  variables  was  verified  using  a Kolmogorov-Smirnov  test  and  normality  diagrams,  and  homoscedasticity was studied using a Levene  test. In  the bivariate analysis of normal distributions, a t test for independent samples was used for quantitative variables, and a Chi-square test or a Fisher¿s exact  test  for categorical variables. A nonparametric Mann-Whitney U  test was  used  in  the  bivariate study  of  variables with  a  non-normal  distribution. A kappa index was used to analyse agreement between the two equations  to  estimate GFR  in diagnosis of occult CKD.25 Agreement  was  considered  excellent  when  values  ranging  from 0.81 and 1 were obtained, good for values ranging from 0.61 and  0.80,  and  moderate  for  values  ranging  from  0.41  and 0.60. For the analysis of the correlation and agreement between the Cockcroft-Gault and MDRD equations as procedures for quantitative measurement of GFR, the Bland-Altman method and the Shapiro-Wilk test were used to assess whether the difference in means was significantly different from zero. Finally, scatter plots, Pearson¿s correlation coefficient, and  the regression line were also obtained.

RESULTS

Table I shows the clinical characteristics of patients included in  the cohort. Women  represented 56.7% of  the sample, and showed  a  higher  rate  of  occult CKD  as  compared  to  men when the MDRD formula was used (17.7% versus 3.6%, p < 0.001). Women were also older and showed higher SBP, HDL cholesterol, and BMI values, and a  lower coronary  risk and proportion of smokers as compared to men (table I). An 8.3% of the population would have CKD using the Cockcroft-Gault formula, and 11.6% using the MDRD equation. Patients diagnosed  occult  CKD  by  the  Cockcroft-Gault  equation  were older (67.4 versus 64.4 years, p < 0.001) and had a greater coronary  risk  according  to  both  the  original  Framingham  and the REGICOR  equations, while  patients diagnosed  of CKD by  the MDRD  formula had a higher BMI  (29.6 versus 26.3 kg/m2, p < 0.001) and were more commonly women (86.7% versus 67.1%, p < 0.01) (table II). 

The kappa  index of  agreement between  the  two  formulas for  estimating  GFR  in  diagnosis  of  occult  CKD  was  0.55, suggesting  a  moderate  agreement. This  conclusion  is  also drawn by looking at the Bland-Altman plot (fig. 1), where a great number of measurements are seen not to lie around the zero difference (p < 0.001). The Pearson¿s correlation coefficient  between  the  GFRs  estimated  by  the  Cockcroft-Gault and the MDRD equations in the whole cohort was 0.828 (p < 0.001), and 0.142 (p = 0.327) when the relationship was limited to patients with CKD according to both equations. The regression  line between both formulas  (GFR estimated by  the Cockcroft-Gault equation = 4.90 + 1.02 MDRD) suggests that estimation of GFR using the Cockcroft-Gault equation gives higher values than estimated with the MDRD equation; specifically, the Cockcroft-Gault formula tends to give values 4.90 units higher than the MDRD formula.

The number and percentage of patients included in the different intervals of GFR values estimated by both formulas is shown in table III, where agreement between the two formulas is seen to be higher in the lower GFR categories. For instance, 71.4% of the 70 patients with GFR < 60 mL/min/1.73 m2 using  the  Cockcroft-Gault  equation  are  included  in  the same GFR  interval  in MDRD, while only 43.4% of patients with an estimated GFR > 90 mL/min/1.73 m2 were included. It is also seen that the Cockcroft-Gault formula usually allocates more patients  to higher GFR  intervals as compared  to the MDRD  formula  (212 versus 108  patients), whereas  the MDRD equation assigned a greater proportion of patients  to lower  GFR  categories  as  compared  to  the  Cockcroft-Gault formula (table III).

To summarize, only 42.4% (50 patients) of the 118 patients considered to have CKD were diagnosed the condition using both formulas. Of the 68 patients diagnosed of occult CKD by only  one  of  the  equations,  20  were  rated  as  having  occult CKD by the Cockcroft-Gault equation, and 48 by the MDRD formula (fig. 2).

Eighty-four percent of concordant patients (those with occult CKD according to both formulas to estimate GFR, table IV) were women. A low proportion of smokers (10.0%) and a moderate coronary risk according to the original Framingham (18.8%)  and REGICOR  (7.9%)  equations were  seen  in  this female group. A comparison of these patients and discordant patients (those found occult CKD in only one of the equations for estimating GFR)  revealed a  lower age  (64.0 versus 66.7 years, p < 0.05) and a higher BMI (29.7 kg/m2 versus 27.2, p < 0.01) in the latter group (table IV).

When  discordant  patient  groups  were  compared  to  each other,  i.e.  patients  considered  to  have  occult  CKD  by  the Cockcroft-Gault equation versus patients diagnosed CKD by the MDRD formula alone, patients in this second group were seen  to  be mainly women  (89.6%) with  obesity  (BMI  32.0 kg/m2) and higher DBP and triglyceride values (table V). By contrast,  patients  diagnosed  occult  CKD  by  the  Cockcroft- Gault equation only (table V) were mainly males (75.0%) of an older age  (69.1 versus 61.9 years, p < 0.001) and with a high coronary risk according to both the original Framingham (32.7%) and the REGICOR (13.1%) formulas.

Males  showed  higher  rates  of  events,  both  coronary (13.9% versus 6.9%  in  females, p < 0.001) and cardiovascular (16.4% versus 9.8%, p < 0.01) during follow-up. However, there were no significant differences in the coronary or  cardiovascular  event  rates  between  patients  diagnosed occult CKD by one or the other formula for estimating GFR (tables I and II).

DISCUSSION

CKD is one of the main public health problems due to both its high prevalence26-28 and  its  significant  cardiovascular morbidity and social and financial costs.2 The prevalence of occult CKD detected  in our population  (8.3% and 11.6% using  the Cockcroft-Gault  and MDRD  equations  respectively)  agrees with  that reported by other studies  in our area27,29-31 and other countries.32 The greater prevalence of occult CKD in females has  also  been  confirmed  by  different studies.29-31 The  EROCAP study,31 conducted in patients seen at primary care centres, found a 7.9% prevalence of occult CKD. This study considered  normal  plasma  creatinine  values  lower  than  1.1 mg/dL in females and 1.2 mg/dL in males and did not exclude patients with extreme BMIs. In our study, if we had disregarded BMI restrictions and accepted the same normal range for plasma creatinine, the occult CKD rate would be 5.7% (with all cases occurring in females). These small differences may be explained by the different mean age and prevalence of cardiovascular  risk  factors  in both populations, with older ages and higher diabetes rates in patients from the EROCAP study. In any case, the prevalence of occult CKD found in the population seen at health centres in both studies could support the need  for  its  early  detection  in  order  to  decrease  the  risk  of progression  and  the  cardiovascularmorbidity  associated  to CKD.2 In fact,  early  identification  of CKD  in  primary  care would allow for early intervention on this disease and its risk factors, would  facilitate  the  start  of  treatment limiting  progression of renal damage, and would help avoid use of drugs that  impair kidney  function,  thus  reducing  the serious social and health implications of CKD.1 However, the high proportion  of  hypertensive  (66.7%)  and  diabetic  (28.2%)  patients enrolled  into  both  the EROCAP study31 and  our  own  study (including 79.4% and 23.2% of hypertensive and diabetic patients  respectively), as well as  the mean age  in both  studies (60.6 and 55.5 years respectively), lead us to question whether screening for occult CKD should maybe not be limited to patients with diabetes or high blood pressure, or older than 55 years, as advocated by other authors until studies showing a better cost-effectiveness ratio in the general population are reported.33

Isolated measurement of plasma creatinine is the most universal  routine method  for  assessing  kidney  function,  but  its relationship to GFR is poor, particularly in the elderly and especially in females.26 These data justify the interest of scientiwith  that reported by other studies  in our area27,29-31 and other countries.32 The greater prevalence of occult CKD in females has  also  been  confirmed  by  different  studies.29-31 The  EROCAP study,31 conducted in patients seen at primary care centres, found a 7.9% prevalence of occult CKD. This study considered  normal  plasma  creatinine  values  lower  than  1.1 mg/dL in females and 1.2 mg/dL in males and did not exclude patients with extreme BMIs. In our study, if we had disregarded BMI restrictions and accepted the same normal range for plasma creatinine, the occult CKD rate would be 5.7% (with all cases occurring in females). These small differences may be explained by the different mean age and prevalence of cardiovascular  risk  factors  in both populations, with older ages and higher diabetes rates in patients from the EROCAP study. In any case, the prevalence of occult CKD found in the population seen at health centres in both studies could support the need  for  its  early  detection  in  order  to  decrease the  risk  of progression  and  the  cardiovascular  morbidity  associated  to CKD.2 In  fact,  early  identification  of CKD  in  primary  care would allow for early intervention on this disease and its risk factors, would  facilitate  the  start  of  treatment  limiting  progression of renal damage, and would help avoid use of drugs that  impair kidney  function,  thus  reducing  the serious social and health implications of CKD.1 However, the high proportion  of  hypertensive  (66.7%)  and  diabetic  (28.2%)  patients enrolled  into  both  the EROCAP study31 and  our  own  study (including 79.4% and 23.2% of hypertensive and diabetic patients  respectively), as well as  the mean age  in both  studies (60.6 and 55.5 years respectively), lead us to question whether screening for occult CKD should maybe not be limited to patients with diabetes or high blood pressure, or older than 55 years, as advocated by other authors until studies showing a better cost-effectiveness ratio in the general population are reported.33

Isolated measurement of plasma creatinine is the most universal  routine method  for  assessing  kidney  function,  but  its relationship to GFR is poor, particularly in the elderly and especially in females.26 These data justify the interest of scientific organizations and societies  in systematic  inclusion of  the GFR  value, estimated  through  equations,  in  routine  laboratory reports. The MDRD formula  is  the best validated equation34 and the one recommended by the SEN,7 but other organizations  accept  use  of  the  Cockcroft-Gault  equation  to estimate kidney function as an alternative.35 The lack of standardization of methods to measure creatinine and the characteristics of  the populations analysed may explain  these disagreements.36

Correlation between both equations was  low  in our study, which represents a significant problem for detection of occult CKD because of the trend to select, at least according to these results, different populations, as has also been previously reported.14 This  dispersion  of  the  values  obtained  with  both equations estimating GFR  is not surprising because  they are different  approaches  to  the  study  of  kidney  function  from serum creatinine measurements: the MDRD equation estimates  clearance  of  I125 iothalamate,  and  the  Cockcroft-Gault equation  estimates  creatinine  clearance. On  the  other  hand, the presence or absence of weight in one and the other formula may condition results and partly account for the populational differences seen, though in our study values were adjusted for  body  surface  area.18 However,  even  if  the  difference  in measurements could be accounted for by the different origin of  equations,  this explanation would  not  avoid  its  implications when the physician has to decide whether or not a given patient should be considered to have renal failure.

The dispersion of values seen in the Bland-Altman agrees with  the modest  index of agreement between both equations in  identification of patients with occult CKD criteria  (kappa index  of  0.55). All  patients  from  our  study with  CKD  has GFR  values  ranging  from  59  and  30 mL/min/1.73 m2,  i.e. they had stage 3 CKD, the traditionally called renal failure.5,7 If the MDRD equation is accepted to have a greater diagnostic  accuracy7 for  GFR  values  ranging  from  15  and 60 mL/min/1.73 m2, it could be considered that patients with occult CKD  in our  study were mainly women  (86.7%) with  a mean age of 64.4 years, a BMI near  to obesity 29.6 kg/m2), and  a moderate  coronary  risk  according  to  the original Framingham and the REGICOR equations (table III). However, analysis of the discordant groups, i.e. patients defined as having occult CKD by only one of the two equations (table V), provides  revealing  results.  Patients  rated  as  having  occult CKD by the MDRD equation only continued to be mainly females (89.6%) with obesity (BMI 32.0 kg/m2) and a moderate coronary  risk. By contrast, patients with occult CKD according  to  the Cockcroft-Gault equation  (table V) were mainly males (75.0%) of an older age (69.1 years), including a high proportion  of  smokers  (30.0%)  and  diabetics  (35.0%),  and with a high coronary risk according to both the original Framingham  equation  (32.7%)  and  the calibrated  REGICOR equation  (coronary  risk, 13.1%). This patient group, consisting mainly of males, has  sufficient  cardiovascular  risk  factors to make biologically likely a high probability of developing CKD. Twenty percent of  them  experienced  a  coronary event and 25% some cardiovascular event during 10-year follow-up  (table V), which  also  indirectly  suggests  that CKD behaves  as  a  significant  cardiovascular  risk  factor  at  individual level.21 It is therefore very possible that these are not patients  erroneously  rated  as  carriers  of  occult  CKD  by  the Cockcroft-Gault equation. If the MDRD equation was used in laboratory  reports,  this would exclude  this group of patients who are rated as carriers of occult CKD and with a high true probability of  suffering  it by the Cockcroft-Gault only, who could  be  deprived  of  the measures  aimed  at  slowing  CKD progression  that  the  introduction  of  routine  reporting  of  the estimated GFR intends to promote. These data partly support the recently reported findings of better results with the Cockcroft-Gault equation in patients with advanced CKD (stages 4 and 5).15 The fact that, in discordant groups, patients rated as having  occult  CKD  by  the  Cockcroft-Gault  equation  were mainly males (75%) could also be related to the low proportion of women  (4%)  enrolled  in  the original  study by  these authors.8 On the other hand, the MDRD equation contains no anthropometric parameters, and  it  is  therefore not surprising that its precision may be different in a population with other anthropometric characteristics  despite  adjustment  for  body surface area, with up to 60% of discordant estimates in the elderly population.37 Our data  should  encourage, in  agreement with recommendation by the SEN7 and various authors,36,38 to develop new equations to estimate GFR with a greater diagnostic accuracy from standardized methods to measure creatinine  and/or  other  biological  variables,  and  to  validate  such equations  in  independent  populations  versus reference methods to measure GFR.

Our study has significant limitations. The population enrolled was not selected at random, but based on the availability of a clinical history and the required variables to estimate coronary  risk and GFR using  the Cockcroft-Gault and MDRD equations. A computerized  clinical  history  system  was  not available  at our  centre.  However,  all  clinical  records  were systematically reviewed and followed up for 10 years, using a tracking and monitoring  strategy  that prevented missing patients or cardiovascular events. Our  results  therefore correspond to a population seen in an urban health centre, and may not be extrapolable to the general population, though this has no influence on the comparison of the two equations for GFR estimation analysed. The fact that this study, as the EROCAP study,31 was based on a single time point, so that patients with a transient kidney function impairment could not be differentiated from those with established CKD, should not influence the  comparison  either. A significant  limitation  of  the  study could be an inadequate identification and quantification of the cardiovascular events occurring during cohort follow-up. However, the methods established for searching and confirming cardiovascular  events,  consulting clinical  records,  hospital files,  and  the  registry  office,  and  contacting  with  relatives were  very  rigorous.  Moreover,  it  would  be  difficult  that events  of this  nature were  overlooked  during  a  10-year  follow-up. Finally, our  study was observational  in nature, and the existence of other unidentified confounding factors cannot  be  completely  excluded. Also,  and  as  occurs  with  this type of studies, its results may only be used to generate new hypotheses.

To  summarize,  this  study confirmed a high prevalence of occult stage 3 CKD (11.6% and 8.3% when  the MDRD and Cockcroft-Gault  equations  are used respectively)  in patients aged 35-74 years with no cardiovascular disease. The  studyalso  revealed a moderate agreement between  the Cockcroft-Gault and MDRD formulas, with a different profile of excluded patients depending on whether one or the other formula is chosen.  The  choice  of  the MDRD  formula would  exclude from diagnosis of CKD a population group consisting mainly of elderly (69 years) males (75%) with a high cardiovascular risk, both estimated by  the original Framingham and REGICOR equations and confirmed by 10-year follow-up.

ACKNOWLEDGEMENTS

This  study was  supported by  redIAPP (Innovation and  Integration of Prevention and Health Promotion in Primary Care), thematic  cooperative  research  network  G03/170,  and  by  a grant from the Program for Promotion of Research in Primary Care  of  the  Instituto  de  Salud  Carlos  III.  The  second  and fourth authors of the article also received a predoctoral scholarship  from  the Spanish Society of Family and Community Medicine.