routine BNP assessment: helpful for patients with CKD

 a recent article found that routinely measuring B-type Natriuretic Peptide (BNP) levels longitudinally in patients with kidney failure was associated with lower risk of both kidney replacement therapy (dialysis or kidney transplant) as well as heart failure hospitalizations (see heart failure CKD follow BNP AJKD2023 in dropbox or doi: 10.1053/j.ajkd.2023.05.003).

Details:

-- 2998 outpatients with nondialyzed chronic kidney disease (CKD) stages 3-5 from a nephrology clinic at an academic hospital in Osaka, Japan, in this retrospective cohort study conducted between 2005 and 2019 

-- inclusion criteria: at least 20yo, follow-up at least 90 days, eGFR between 10 and <60, no history of KRT (kidney replacement therapy: dialysis or transplant), no history of hospitalization for heart failure

-- mean age 66, 65% male, BMI 23, BP 131/75

-- mean eGFR based on creatinine: 38.1 (ie, average CKD stage 3b, with CKD stage 3 in 67%, stage 4 in 25%, stage 5 in 8%)

-- comorbidities: diabetes 40%, coronary heart disease 8%, valvular heart disease 4%, PAD 0.5%, cerebral infarction 2%

-- meds: ACE/ARB 9%, loop diuretics 3%, thiazides 2%, mineralocorticoid receptor antagonists (MRAs) 2%

-- high-sensitivity CRP 0.7, urine protein-creatinine ratio 0.54 g/g

-- BNP 60.2 pg/mL (upper limit of normal 18.4 pg/mL)

-- main outcome: kidney replacement therapy (KRT)

-- secondary outcomes: acute kidney injury (AKI, defined as a creatinine value increase in the last 7 days, per the specific KDIGO criteria:  increase of creatinine to at least 1.5 times baseline within the last seven days), and heart failure hospitalization. 

-- additional secondary outcomes: composite of all-cause mortality or KRT initiation, urgent dialysis, and planned dialysis initiation

 

-- mean follow-up 5.9 years, with BNP being measured 4,807 times in 632 patients (21% of the 2998 in the study), with each patient having a median of 3 BNP analyses

-- statistical evaluation: quite complex since patients were not chosen randomly for BNP monitoring, with mathematical modeling to attempt to compensate for the high likelihood of selection bias. they used "marginal structural models at each time point, applied to account for potential time-dependent confounders”; inverse probability weighted pooled logistic regression models were used to create a balanced “pseudo population” to estimate hazard ratios

-- in addition, they used ordering intact PTH levels as a negative control exposure, noting that:

    -- the role of PTH monitoring in patients with CKD has not been established

    -- the cost of the PTH test is similar to the BNP cost

   -- there is no documented benefit to treating hyperparathyroidism with vitamin D or its analogs in terms of kidney function

 

Results: 

-- those patients who had BNP measured at baseline were: older (68yo vs 66yo), more likely to have diabetes (49.1% vs 37.2%) or heart disease (14.4% vs 6.8%), had a lower eGFR by creatinine (35.8 vs 38.5) and had a higher urinary protein level  (0.65 vs 0.52 g/g)

-- at study onset, 5.0% were on a diuretic, but during the follow-up period, 50.4% were on a diuretic. (63.7% who had BNP measured)

 

-- during the followup period: 

    -- 449 patients started KRT, 765 had AKI, 236 were hospitalized for heart failure, 271 died

    -- ACE/ARB were given to 1,739 patients, loop diuretics to 1,268, and MRAs to 622

    -- diuretics were changed in 15.4% of those who had BNP results and 1.7% of those without

    

-- with the adjustments as per the statistical evaluation noted above, primary outcomes with monitoring BNP levels vs no BNP monitoring:

    -- KRT: 56% reduction, HR 0.44 (0.21-0.92)

    -- composite of death or KRT: 77% reduction, HR 0.23 (0.10-0.54)

    -- AKI: 64% reduction, HR 0.36 (0.18-0.72)

    -- heart failure hospitalization: 63% reduction, HR 0.37 (0.14-0.95)

 

-- if adjusted for AKI or heart failure, there was not a significant reduction in KRT (ie, it seemed that the diagnoses of AKI or heart failure were associated with the KRT)

 

-- urgent dialysis initiation: 75% decreased, HR 0.25 (0.08-0.80), but no statistically significant relationship for planned dialysis initiation (ie, the only KRT difference was in those receiving urgent dialysis)

 

-- no significant relationship between PTH monitoring and KRT initiation

 

--  sensitivity analyses: excluding in-hospital data negated the significance of the associations above

    -- no significant difference by considering only patients whose BNP measurements were <1x/yr vs never, or reclassifying the analysis dataset with a visit-to-visit time interval of 90 days

 

Commentary:

-- chronic kidney disease and heart failure are remarkably common:

    -- the US HANES survey found that 20.1% of people 60-69yo have CKD, increasing to 42.6% in those at least 70yo 

    -- the lifetime risk of heart failure is 24%

    -- and diabetes, a common disease associated with both of these, is increasing quite quickly in the US

 

--measuring BNP has the advantage that this is a simple and inexpensive lab test

    --the value of BNP does change some in the setting of CKD, though much less so than the value of NT-proBNP levels. This study found benefit of BNP levels without adjusting for the stage of CKD, just using the baseline BNP level and changes over time

--the finding that adjusting for AKI and heart failure rendered KRT nonsignificant suggests that KRT may result from volume overload often related to heart failure (and finding increased BNP by regular BNP analyses might lead to more aggressive earlier diuresis, which might then decrease congestive nephropathy) or AKI (and regular BNP analyses would suggest that the AKI was not from volume overload but from volume contraction, thereby perhaps leading to earlier treatment with fluids)

    -- measuring BNP levels could then decrease heart failure and AKI, leading to lower levels of the need for dialysis

    -- and one finding congruent with this mechanism is that urgent dialysis decreased significantly, though planned dialysis did not (it seems that the urgent dialysis was typically associated with sicker patients being admitted to the hospital with heart failure or AKI: perhaps many of these dialyses could have been prevented by earlier detection/modified treatment through BNP measurements)

-- of note, clinical exam for volume overload is not so great. A study of critically ill patients who got pulmonary artery catheterization found a poor correlation between clinical exam and the cath findings: 58% of patients had their planned therapy altered by the cath results. And this was in an ICU setting where clinicians are more likely to have lots of experience assessing volume status clinically): https://journals.lww.com/ccmjournal/abstract/1984/07000/clinical_evaluation_compared_to_pulmonary_artery.1.aspx

-- i should add here that there is a pretty compelling argument that creatinine is significantly inferior to cystatin C in predicting both renal and more global bad outcomes such as atherosclerotic cardiovascular disease, heart failure, cardiovascular death, and all-cause mortality  (see https://gmodestmedblogs.blogspot.com/2023/12/cystatin-c-better-predictor-of-bad.html )

Limitations:

-- this was a retrospective cohort study, not an RCT, so one cannot determine that BNP monitoring was causally related to these outcomes. for example, did those clinicians who did aggressive BNP monitoring also do more compulsive evaluation/treatment of other risk factors for heart failure (eg, decreasing sodium intake, monitoring fluid intake, other dietary changed) or developing worsening CKD (eg, better control of blood pressure or urinary protein excretion)

-- there was no consistency in how often BNP levels were assessed. or what prompted more levels to be done

-- there was no consistency in how clinicians responded to changes in BNP determinations. was there a specific level of increase or decrease that led to management changes (eg with diuretics)?

    -- however, the bias here would likely be that sicker patients had more BNP levels done. so, despite their being sicker (and presumably more likely to have more BNP testing), this cohort that got BNP testing did better, and had less urgent dialysis

    -- but there was also no breakdown of results by initial stage of CKD, which might have been helpful in which patients should be monitored and how frequently

-- i wonder whether checking PTH levels is a valid negative control exposure because:

    -- if in Japan PTH testing is not recommended, why would clinicians there be checking it, allowing for the comparison of BNP vs PTH analyses in this study? and, wouldn't these clinicians ordering BNP testing make them different than other clinicians who "know better"? and, would they therefore be comparable in their overall clinical choices and care of patients with CKD?

    -- though the clear indication for tracking PTH levels in patients with CKD is lacking, many in the US do treat these levels (typically with calcitriol) if the PTH levels are increasing a lot or reach a threshold of 150 or higher.

    -- another article in this same edition of Am Journal of Kidney Disease did find that bone mineral density did increase with calcitriol in patients with CKD, though the followup time was insufficient to document changes in fractures (see ckd vit D therapy may not help AJKD2023 in dropbox, or doi: 10.1053/ j.ajkd.2023.04.003)

-- I am also concerned that a study such as this one, where there is a further progression in the intensity/complexity of the statistical analysis, would challenge its validity (in this case, with multiple imputations in order to “create a balanced pseudo population in which time-dependent confounding and collider stratification bias are minimized at each time point and the effect of the exposure on outcomes is the same as that in a real-world population”.

    -- I am not a statistical nihilist, it’s just that as studies delve deeper into mathematical manipulations (creating a “pseudo population” and gets further from the actual patient conditions). to me this raises a concern: do these models really reflect the live human beings they try to emulate? (ie, the ones we are actually taking care of…..)

    -- for example, a study on tramadol for osteoarthritis, using statistical methods of propensity-matched scoring to make the tramadol users and nonusers mathematically “equivalent”, found that tramadol led to higher mortality. BUT subsequent critical analysis of their methodology revealed the primary analysis to be faulty, and on reworking led to the opposite conclusion: http://gmodestmedblogs.blogspot.com/2020/03/tramadol-fo-oa-inc-mortalityprobs-with.html

 

so, a few issues for following patients with CKD:

-- it does seem that measuring BNP levels does help in managing patients and very likely will decrease an array of bad renal outcomes as well as heart failure events, including hospitalization for heart failure and likely all-cause mortality

    -- and, this seems to be a reasonable conclusion despite the limitations of this study, since:

        -- elevated BNP is a known marker of presymptomatic heart failure

        -- studies in those without CKD largely do show benefit of monitoring BNP levels and adjusting meds based on high BNP measurements

            -- ie, the incidence of clinical heart failure is likely associated with increasing volume overload, and more aggressive diuretic treatment is very likely to decrease hospitalization for heart failure, as well as a major indication for urgent dialysis

        -- the outcome of AKI is likely associated with volume contraction, also more likely to lead to earlier or urgent dialysis. Knowing that the BNP was not elevated in those with increasing creatinine levels could prompt early fluid repletion on an outpatient basis before it got worse (which, of course, would be contraindicated if there were volume overload per BNP assessment)

 

-- why is this article important for us in primary care (it was, after all, a study based in patients from a nephrology clinic)?

    -- we are more likely to see the patients more often than in a nephrology clinic (especially in those with earlier stages of renal failure)

    -- some patients may well be referred by renal clinic to cardiology, which may take a long time and may be unnecessary in preventing clinical heart failure (we in primary care can alter the diuretics/other meds as needed)

    -- multiple referrals may add to unnecessarily higher levels of fragmentation of care for the patients

    -- the numbers of patients with CKD is increasing a lot, in part from the increasing prevalence of obesity/diabetes/aging in the US

        -- my own experience is that basically all of my patients >70-75yo have some level of CKD

    -- and, one of our baseline missions in primary care is to prevent bad clinical outcomes when possible, and anticipate correctable problems as early as possible

-- bottom line: it does seem reasonable to follow BNP levels in patients with more advanced CKD

    -- unclear how often….  I personally will check BNP levels every 3-6 months (when I track the progression of renal failure), to see if I can alter the course of CKD progression in order to delay the progression of renal disease (especially in those with more advanced renal failure) and prevent hospitalizations for either heart failure or AKI.

    -- though, it seems that even measuring BNP <1x/yr may itself be helpful (my guess is that since 2/3 of the patients in this study had the earlier stage 3 CKD and therefore a lower associated likelihood of volume changes from the CKD, this less severe CKD might justify the value of a lower frequency of BNP assessments; those with more severe CKD might well benefit from more frequent assessments…). Would be good to have better data on this, with breakdown by severity of CKD. and probably using cystatin C as the clinical measure of estimated GFR...

 

geoff

-----------------------------------

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@bidmc.harvard.edu

  

to get access to all of the blogs:  go to http://gmodestmedblogs.blogspot.com/ to see the blogs in reverse chronological order

  -- click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

  -- or you can just click on the magnifying glass on top right, then type in a name in the search box and get all the blogs with that name in them

 

please feel free to circulate this to others. also, if you send me their emails (gmodest@bidmc.harvard.edu), i can add them to the list