#NephJC Chat
Tuesday December 10, 2019 at 9 pm EST
Wednesday December 11, 2019 at 9 pm IST
Wednesday December 11, 2019 at 9 pm GMT, 1 pm PST
Nat Med. 2019 Nov;25(11):1753-1760. doi: 10.1038/s41591-019-0627-8. Epub 2019 Nov 7.
Glomerular filtration rate by differing measures, albuminuria and prediction of cardiovascular disease, mortality and end-stage kidney disease.
Lees JS, Welsh CE, Celis-Morales CA, Mackay D, Lewsey J, Gray SR, Lyall DM, Cleland JG, Gill JMR, Jhund PS, Pell J, Sattar N, Welsh P, Mark PB.
PMID: 31700174
Introduction
While cardiovascular disease accounts for over 50% mortality in ESKD, the risk of cardiovascular events increases sharply with declining eGFR well before people need renal replacement therapy
Figure from Go et al, NEJM 2004
In fact, in this cohort study of community dwelling adults over the age of 65 with CKD, they were 6 times more likely to die from cardiovascular-related causes than to progress to ESKD.
From Dalrymple et al , JGIM, 2011
Despite its importance as a risk factor for adverse cardiovascular outcomes, CKD has not been integrated into modern cardiovascular risk equations.
Cystatin C is a protein produced at a constant rate by all nucleated cells. Cystatin C is freely filtered and, unlike creatinine, is not secreted. In contrast to creatinine, there is little cystatin C in the urine because the proximal tubular cells reabsorb and catabolise cystatin C. Cystatin C levels stabilise after one year of life and, as cystatin C is not related to muscle mass, are similar to those of adults. Although cystatin C was hailed as not being subject to non-GFR determinants that creatinine is, in fact cystatin C is affected by non-GFR factors, they just differ from the ones that influence creatinine levels. Evidence now suggests glucocorticoids, thyroid dysfunction, obesity, diabetes, smoking, and inflammatory states affect cystatin C levels. Before the derivation of CKD-EPI for cystatin C, a study using the Cardiovascular Health Study cohort showed cystatin C was linearly associated with risk of death, whereas creatinine and eGFR had a J-shaped association with risk of death.
Cystatin C was a strong and independent predictor of all-cause and cardiovascular mortality. Higher cystatin C levels were independently associated with risk of incident myocardial infarction and stroke. How well did creatinine predict all-cause mortality? You had to look at the participants in the highest 7% of the cohort for creatinine readings to find a significantly increased risk of all-cause mortality, and there was no independent association with this category of creatinine with cardiovascular mortality or incident cardiovascular events
Creatinine is the measure used in routine clinical practice, cystatin c being the preserve or research, living donor assessment or rare clinical cases where extra precision is needed, or concerns exist about the accuracy of creatinine in a specific patient - due to concerns regarding body size extremes and so on.
Lees, et al. was a large prospective cohort study to determine whether eGFR (determined with serum creatinine (eGFRcr), cystatin C (eGFRcys) or combined cystatin C-creatinine (eGFRcr-cys)) and albuminuria improve risk prediction for all-cause mortality and cardiovascular disease.
The Study
Figure from Shlipak et al, NEJM 2005
Methods
The UK Biobank recruited 502,219 volunteer participants from 37 to 73 years of age between 2007 to 2010 from all around the United Kingdom. Volunteers gave urine, blood and saliva samples as well as information about themselves via questionnaire. The ultimate aim of the project is to improve prevention, diagnosis and treatment of a wide range of cardiovascular and other diseases. Biological samples were analysed at a central laboratory. It is noted in the methods that samples were collected at various times of day. Ethnicity was initially coded as white, black, south Asian or other, but was coded as black or other for purposes of eGFR calculators.
Participants in the cohort were excluded in this analysis if they had the following:
Self-reported ESKD
Calculated eGFR <15ml/min/1.73m2
History of CVD: self-reported angina, myocardial infarction, stroke or transient ischaemic attack.
Data were adjusted by UK Biobank centrally before release.
The outcomes of interest are listed as follows:
All-cause mortality: death from any cause; data obtained from death certificates from NHS registries
Composite of fatal and non-fatal CVD events: fatal CVD events, non-fatal MI, stroke or heart failure; data from linkage with routine hospital data as well as death certificates
Fatal CVD events: identified from fatal CVD ICD10 codes and from death certificates.
ESKD was defined as reaching CKD stage G5 or requirement for renal replacement therapy – determined using hospital admission ICD10 codes.
The date of first assessment was also the start of the follow-up period.
Statistical Analysis
eGFR was categorized according to KDIGO CKD stage. Urine ACR measures were grouped in according to KDIGO clinical practice guidelines (<3, 3-30, >30 mg/mmol). The distribution of atherosclerotic risk factors and urine ACR were calculated across each CKD stage and across each outcome.
Kaplan-Meier curves were used for event-free survival and outcome event rates per 100,000 person years in each CKD stage. Cox proportional hazard ratios, adjusted for age, sex, ethnicity, smoking, systolic and diastolic blood pressure, antihypertensive medications, statins, total and HDL cholesterol, were created for the outcomes of interest. Harrell’s C statistics were used to assess the discrimination of the model with eGFR alone and with eGFR and albuminuria.
Results
There were 502,219 participants initially included and participants were excluded in cases of:
Missing data (31,283)
Cases of prevalent ESKD (260) or eGFR <15 ml/min/1.73 m2 (38)
Prior CVD events (30,112)
That left 440,526 participants who were followed for a median of 8.9 years (Q1–Q3 8.2–9.5 years). The events were as follows:
15,469 deaths from any cause
2,552 deaths from CVD
8,662 incident fatal or nonfatal CVD events (with 2552 CV deaths, this equals 6110 non fatal CV events?)
336 cases of incident renal replacement therapy.
Since the paper does not have a figure 1, we created one for our readers, below:
CVD events were defined by AHA/ACC definition
Participants with lower eGFR measures had typical CKD comorbid features – more likely to be on antihypertensive and statin medications, to report diabetes and unsurprisingly had the highest category of ACR. In terms of demographics, participants with lower eGFR measure tended to be older, male, and smokers. They had higher systolic and lower diastolic blood pressures as well.
Table 1 from Lees et al Nature Med 2019
A greater proportion of participants were classified as having CKD G3 – 5 when using cystatin eGFR estimates, as eGFR-cystatin based estimates provided lower estimated GFR than both creatinine- and combined creatinine-cystatin eGFRs.
For correlation between eGFR estimates, the strongest correlation was between eGFRcys and eGFRcr-cys (r=0.925, P<0.001) and the weakest between eGFRcr and eGFRcys (r=0.599, P<0.001).
The adjusted hazard ratios for outcomes plotted against each eGFR measure were largely linear and negative, and this was most notable in eGFRcys. Of note, this relationship between eGFR decline and adjusted hazard ratio for events progressed in a linear fashion starting at eGFR 90 ml/min/1.73 m2 for eGFRcys, where as the trajectory wavered between 70 and 90 ml/min/1.73 m2 for eGFRcr and then progressed linearly after 70 ml/min/1.73 m2 .
Figure 1 from Lees et al Nature Med 2019
Unsurprisingly, adjusted hazard ratios for each eGFR outcome combination were consistent with lower eGFR being associated with higher risk of CVD, all-cause mortality, fatal CVD and ESKD.
The adjusted HR of every 10ml/min/1,73m2 increase in eGFR was strongest for measures of eGFRcys.
Table 2 from Lees et al Nature Med 2019
Decreasing eGFR was associated with an increased hazard ratio for all outcomes, and this was sustained across urine ACR groups, although the magnitude of the association was similar in higher and lower urine ACR groups. For ESKD risk, higher urine ACRs categories were associated with higher risk.
Heat maps for prediction of all-cause mortality and CVD events using AHA and SCORE criteria incorporating eGFRcys and urine ACR groups were generated.
Figure 2 from Lees et al Nature Med 2019
The net reclassification index attempts to quantify how well a new model reclassifies subjects compared to an old model. Incorporating eGFRcys and albuminuria did not improve the net reclassification index across 7.5% threshold used in AHA/ACC guidelines.
Table 3 from Lees et al Nature Med 2019
The c-statistic (equivalent to the area under the Receiver Operating Characteristic curve) is a standard measure of the predictive accuracy for binary outcomes in a logistic regression model. For all-cause mortality, atherosclerotic risk factors yielded a C-statistic of 0.7157 (95% confidence interval (CI) 0.7115–0.7200). Addition of both eGFRcys and eGFRcr-cys, but not eGFRcr, significantly improved discrimination, with the biggest improvement seen with eGFRcys (C-statistic+0.0103, 95% CI 0.0087–0.0121).
For both the composite fatal/nonfatal CVD outcome and fatal CVD outcome, the C statistic for atherosclerotic risk factors was improved by addition of eGFRcys (C-statistic + 0.0039, 95% CI 0.0025–0.0052 and + 0.0085 (0.0049–0.0122) respectively). Similar results were not seen with eGFRcr.
Table 4 from Lees et al Nature Med 2019
Figure 3 from Lees et al Nature Med 2019
There was at least a 20% difference between eGFRcr and eGFRcys measure-ments in over 40% of the cohort
Interestingly, there was absolute discordance ≥20% between eGFRcr and eGFRcys measurements in 183,867 participants – this was just over two fifths of the cohort (41.47%). Baseline characteristics did not discriminate between those who had concordance between eGFRcys and eGFRcr and those who did not. Amongst those with discordant results, eGFRcys remained the marker with the greatest improvement in C-statistic, adjusted for atherosclerotic risk factors and urine ACR, across mortality and both CVD outcome measures.
Discussion
This study demonstrated that eGFRcys is more strongly associated with future CVD events than eGFRcr for both CVD outcomes and future ESKD risk. They also showed that eGFRcys identified a greater proportion of participants with CKD G3–5 than both eGFRcr and eGFRcr-cys, and that it is more strongly associated with clinical outcome than either of the other measures.
The authors note that the UK Biobank population may not be representative of the UK population as this is in lower prevalence rates of CKD and albuminuria in group compared to published rates in England and other published groups. This group self-selected and may represent a motivated group with better health outcomes overall versus an average population. As the authors note, applying this to a population with a greater burden of kidney disease, the benefit of measuring eGFRcys could be increased further. The population was predominantly white. Interestingly, going back to the race question (discussed here on NephJC), in this study eGFRcys was not predictive of CVD outcome in nonwhite ethnic groups, although there were limited nonwhite participants.
Outside of the nephrology community, do healthcare professionals really recognise CKD as a cardiovascular risk factor? How many medical students when asked to list cardiovascular risk factors say hypertension, diabetes, hyperlipidaemia and even things like homocystinuria before even mentioning CKD? Do we explain to patients that they are at elevated cardiovascular risk by virtue of having CKD or do discussions focus more on risk of requiring dialysis being high or low in the future?
The Kidney Risk Factor Equation has changed how we can think about risk of ESKD for our patients and how we can frame conversations about risk. It performed better than nephrologists at predicting risk in this study, although it seemed nephrologists were poor at communicating their understanding of risk to patients given the difference between their and the patients’ risk estimations.
But given the excess cardiovascular morbidity and mortality in the CKD population, shouldn’t we should be calculating and discussing this additional and perhaps more important risk for patients? If we want to do that better, we should move towards cystatin C. And perhaps we do not need to measure cystatin C monthly or three-monthly like we do with creatinine, but for screening high risk individuals or at first detection of CKD there is a clear role for cystatin C to be incorporated into clinical care.
Summary by Sinead Stoneman,
Irish Nephrology Specialist Registrar, Beaumont Hospital Dublin