Preload Reserve Is Restored in Patients With Decompensated Chronic Heart Failure Who Respond to Treatment


Address for correspondence: Pierre Squara CERIC, Clinique Ambroise Paré, 27, bd Victor Hugo, 92200 Neuilly-sur-Seine, France



The authors designed this prospective study to show the relationship between preload reserve and treatment effectiveness of chronic heart failure (CHF). Fifty patients, aged 77±24 years, with decompensated CHF (B-type brain natriuretic peptide [BNP] >1000 pg/mL) were included. Preload reserve was assessed by the changes in contraction indices during a passive leg raise (PLR). Contraction indices were assessed noninvasively using Bioreactance technology. After 4 days of optimized therapy, the same variables were reassessed and treatment-induced differences were calculated. Treatment effectiveness was assessed by the 4-day changes in BNP, body weight, and thoracic fluid content. The authors then compared treatment-induced changes in preload reserve with treatment effectiveness. Therapy was associated with an overall decrease in heart rate, blood pressure, and cardiac power index (CPi) and with an increase in all preload reserve indices. Treatment effectiveness correlated well with changes in preload reserve. The best correlation was found between treatment-induced changes in BNP and in PLR-induced changes of CPi (R=0.63, P<.001). The PLR-induced changes in CPi increased from 21±48 to 51±48 in BNP responders and decreased from 34±34 to 5±19 mW/m2 in BNP nonresponders (P<.0001). Hence, effective treatment, as indexed by a decrease in BNP, restores the preload reserve in patients with decompensated CHF.

Preload reserve is a major mechanism for adapting stroke volume and cardiac output to acute changes in afterload, especially in chronic heart failure (CHF) with limited contractility. As afterload increases, so does preload until the limits of diastolic compliance are reached. An afterload mismatch is induced if the preload is not allowed to compensate for an increased afterload, or if the limit of preload reserve has been reached.[1] In critically ill patients, it has been shown that the preload reserve may be investigated noninvasively by assessing the blood flow during a passive leg raise (PLR) using either Echo-Doppler or Bioreactance technologies.[2-7]

B-type natriuretic peptide (BNP) is a cardiac neurohormone that is specifically secreted from the left ventricle in response to volume expansion and pressure overload.[8, 9] It correlates with New York Heart Association functional class.[10, 11] Changes in the BNP serum concentration are therefore commonly used in clinical practice for the diagnosis, management, and prognosis of acute and chronic congestive heart failure.[12-18] Although it is well known that BNP secretion is proportional to the left ventricle myocardial stretch,[19] the relationship between BNP and preload reserve has not been firmly established.

We designed this study to test the hypothesis that in decompensated CHF, the response to therapy, as indexed by a change in BNP, is related to an improved preload reserve as clinically detectable during a PLR.

Patients and Methods

We included 50 consecutive patients admitted in one tertiary cardiologic center for decompensated CHF defined by a history of documented and treated CHF), New York Heart Association class IV, and BNP >1000 pg/mL. The local ethics committee approved the study (CERIC/AP N°0901) and patients gave informed consent.

Exclusion criteria were the absence of informed consent, the presence of (1) shock at the time of admission defined by a lactate blood concentration >2 mEq/L and/or a metabolic acidosis, (2) decompensated CHF requiring inotropes, (3) acute renal failure (oliguria <500 mL/24 hours) or advanced chronic renal failure (serum creatinine >30 mg/L), and (4) associated significant acute disease such as significant valve disease, unstable myocardial ischemia, decompensated diabetes, pulmonary embolism, decompensated chronic pulmonary obstructive disease, and/or pulmonary infection. We also excluded patients with severe aortic regurgitation and repaired thoracic aortic aneurysm. In these situations, the Bioreactance-derived flow measurements may be subject to error because changes in Bioreactance are believed to be mostly caused by systolic aortic volume expansion.[20, 21]

For each patient, we assessed the preload reserve by measuring the variation of different indices of the heart contractile force during a PLR. The flow-related variables were indexed to body surface area for inter-patients comparisons. We collected stroke index (SI) and cardiac index (CI) as continuously monitored noninvasively using a Bioreactance, system (NICOM; Cheetah Medical, Portland). This system has been validated for clinical[2, 22-24] and physiologic studies.[25] Proper management of the electrode wires during testing is important to ensure good signal quality and has been reported previously.[26, 27] Three averaged values of each index were considered: (1) at baseline in a semi-recumbent position, (2) during a PLR, performed by changing the bed from 45° head up to flat position with feet elevated to 45°, and (3) after return to baseline. At baselines (before and after PLR), 5 minutes of SI and CI data were averaged. During PLR, 3 minutes of measurements were averaged. Preload reserve was the difference (δ) between the value during PLR and the average of the two baseline values.

We also averaged 3 noninvasive measurements of blood pressure (BP) at each step of the protocol, using the automatic sphygmomanometer integrated in the NICOM device. We therefore were able to derive the double product in mm Hg per minute (heart rate × systolic BP), the left ventricle stroke work index in g.m/m2 (LVSWi = SI × mean BP/80) and the cardiac power index in mW/m2 (CPi = CI × mean BP/0.451).

We used the same system for monitoring and averaging similarly the mean transthoracic electrical impedance (Zo) that varies with the amount of fluid in the thorax. Thoracic impedance decreases as the thoracic fluid increases and conversely. The inverse of Zo thus varies linearly with changes in thoracic fluid. The variable 1000/Zo has therefore been referred to as a measure of thoracic fluid content (TFC) with a unit in kΩ−1.[28]


At the time of entrance to the study (day 0), patients were weighed and venous blood samples were collected for standard laboratory analysis including the serum creatinine and BNP concentration. Creatinine was measured by the Jaffre reaction (normal range 6–12 mg/L), and BNP was measured by chemoluminescent picroparticle immuno assay (normal value <100 pg/mL). We also assessed the TFC and preload reserve on day 0 or day 1.

All patients underwent echocardiographic Doppler examination. Standard therapy was then optimized, including when possible use or increase in use of: (1) vasodilatation by angiotensin-converting enzyme inhibitor and/or angiotensin II receptor blocker, (2) diuretics targeting a negative fluid balance, (3) β-blockers, and (4) aldosterone inhibitor. Other specific therapies were freely allowed when necessary to reduce or stabilize heart rate (magnesium, ivabradine, amiodarone). When appropriate, biventricular pacing was checked and HR, AV, and VV delays were optimized according to the standard protocol of our institution.

On study day 3 or 4, weight, serum creatinine, BNP, TFC, and all indices of preload reserve were reassessed. Specifically, preload reserve was indexed by the following PLR-induced variations (δ value): δSI, δCI, δLVSWi, and δCPi. The treatment-induced change of each studied variable (Δ value) was the difference between values after treatment and before treatment optimization.

We therefore estimated the treatment effectiveness by Δ weight, Δ TFC, and Δ BNP, and the preload reserve changes following treatment optimization as Δ δSI, Δ δCI, Δ δLVSWi, and Δ δCPi.

Statistical Analysis

Values are reported as mean±standard deviation. We studied two categories of variables before and after therapeutic optimization: (1) indices of treatment effectiveness and (2) indices of preload reserve. Intra-category correlations were assessed using correlation coefficients (r) and Spearman statistics. Inter-category relationships were assessed by the linear regression coefficient (R), taking treatment effectiveness indices as independent variables (x-axis). A stepwise multiple regression was performed to identify primary variables including all variables with a significant univariate relationhip. The nonparametric Wilcoxon was used to reject the null hypothesis. P<.05 was considered as significant.

For diagnostic categorization, we defined “BNP responders” as patients for whom the BNP decreased following treatment optimization and “nonresponders” as patients with unchanged or increased BNP.


We included 50 patients (34 men), aged 77±24 years. Nine patients had diastolic heart failure with an echocardiographic left ventricular (LV) ejection fraction (Simpson) 60%±10%; 41 had systolic heart failure with LV ejection fraction 31%±8%. The mean value of the variables collected at study entrance and the change after treatment optimization are shown in Table 1.

Table 1. Values Before and After Treatment Optimization (TT)
 Baseline Before TTBaseline After TT% Change
  1. Abbreviations given in text. aP<.05.

Weight, kg73.0±14.771.7±14.5a−2±1
Creatinine, mg/L19.5±13.819.6±14.40.5±0.4
BNP, pg/mL2397±15441812±1429a−26±43
TFC, 1/KΩ64±1462±15a−3±9
Mean BP, mm Hg87±1283±10a−5±9
HR, beats per min74±1570±13a−4±11
HR × systolic BP, mm Hg per min8724±23247901±1562a−7±17
SI, mL/m2 38.2±10.338.4±10.52±18
CI, L/min m22.72±0.532.63±0.56−2±16
LVSWi, g m/m241.7±12.040.1±12.9−3±18
CPi, mWatts/m2526±0.121483±115a−7±17

The correlation between initial levels of BNP and TFC was moderate (r=0.45, P<.0008). The inter-correlations between the changes in the different indices of treatment effectiveness were as follows: Δ TFC vs Δ weight (r=0.57, P<.003), Δ BNP vs Δ weight (r=0.35, P<.04), and Δ BNP vs Δ TFC (r=0.41, P<.03).

The relationship between BNP and contractile force indices at inclusion, at day 4, or when inclusion and day 4 values were pooled together was weak (R always <0.10).

The treatment optimization was followed by an increase in the preload reserve (Tables 2 and 3). These treatment-induced changes in the preload reserve (Δ δ values) were not related with the treatment-induced changes (Δ values) in any of the contractile force indices (r always <0.25).

Table 2. Preload Reserve Before and After Treatment Optimization (TT)
 Before TTAfter TT% Change
  1. Abbreviations given in text. Values are expressed as mean±%. aP<.05.

PLR-induced δmBP, mm Hg0.83±2.710.95±2.70
PPLR-induced δHR, beat per min−0.02±0.67−0.10±0.87
PLR-induced δDP, mm Hg/min67±38897±1294
PLR-induced δSI, mL/m21.21±3.612.40±3.29a99±433
PPLR-induced δCI, L/min m20.11±0.210.17±0.22a53±453
PLR-induced δLVSWi, g m/m21.47±4.373.02±4.12a106±1465
PLR-induced δCPi, mW/m224.4±49.037.8±48.8a55±302
Table 3. Preload Reserve (Increase/Decrease) Before and After Treatment Optimization (TT)
 ±δ Value Before TT±δ Value After TTTT-Induced of δ Increase/Decrease
  1. Δ, variation in values during a PLR; DP, double product; ±δ, increase/decrease in PLR-induced change. See text for all other abbreviations. P<.05 between columns. (−) value not given because the changes are not significant.

PLR-induced δSI, mL/m233/1740/1029/21
PLR-induced δCI, L/min m234/1640/1029/21
PLR-induced δLVSWi, g m/m233/1738/1229/21
PLR-induced δCPi, mW/m234/1639/1131/19

In contrast, the treatment-induced changes in the preload reserve (Δ δ values) were well linked with the treatment effectiveness (Δ values). The best regression line was found when the preload reserve improvement was expressed in terms of Δ δCPi and the treatment effectiveness in terms of Δ BNP (Figure 1; R=0.63, P<.0001). These relationships were reinforced when the analysis was restricted to patients with systolic heart failure (Figure 2; R=0.69, P<.0001). The stepwise regression including changes in BNP as x and all Δ δ values as y, selected CPi as first step (P<.0001) and HR as the second and final step (P=.04). When δCPi values after treatment optimization were cut in four quartiles (Table 4), the highest quartile showed that PLR-induced changed in CPi >25.3 mW/m2 was associated with a mean BNP concentration of 1338 pg/mL and a mean decrease in BNP of 1167 pg/mL after treatment.

Table 4. Preload Reserve After Treatment Optimization
 Limits of δCPi (mW/m2)mean δCPi (mW/m2)Post BNP (pg/mL)Change in BNP (pg/mL)
  1. δCPi, variation of CPi during a PLR after treatment optimization; post–B-type brain natriuretic peptide (BNP), BNP level after treatment; Change in BNP, post treatment - pre treatment values of BNP.

First quartile of δCPi <0−6.062874+152
2nd quartile of δCPi <14.7, >07.301701−254
3rd quartile of δCPi <25.3, >14.719.71415−971
4th quartile of δCPi >25.343.01338−1167
Figure 1.

Regression line between treatment-induced changes in B-type brain natriuretic peptide (Δ BNP) and treatment-induced changes in preload reserve as shown by the changes in passive leg raise–induced variation of cardiac power index (Δ δCPi).

Figure 2.

Regression line between treatment-induced changes in B-type brain natriuretic peptide (Δ BNP) and treatment-induced changes in preload reserve as shown by the changes in passive leg raise–induced variation of cardiac power index (Δ δCPi) in systolic heart failure.

When patients were classified into two categories, based on the BNP response to treatment optimization (BNP responders or BNP nonresponders), the inter-category initial contractile force indices were nonsignificantly different. The averaged preload reserve improved in most BNP responders (28 of 36 patients [78%]) and was impaired or unchanged in most BNP nonresponders (11 of 14 patients [79%]). Again, the overall largest difference was observed with Δ δCPi: from 20.6±48.0 to 50.6±48.2 mW m2 vs from 34.1±33.8 to 4.93±18.9, respectively (P<.001; Figure 3). Considering the BNP responder/nonresponder as the predictor and an improvement/impairment in the preload reserve as the event, we found 27 true-positives, 9 true-negatives, 10 false-negatives, and 4 false-positives. In contrast, the Δ CPi (−44.3±104 to −40.1±84.7 mW m2) as well as other treatment-induced change in the contractile force indices were not significantly different between BNP responders and nonresponders.

Figure 3.

Preload reserve as shown by a change in cardiac power index (δ CPi in mW/m2) before and after treatment optimization (TT). Red lines indicate mean values with standard deviation. BNP indicates B-type brain natriuretic peptide.

In 3 patients, the treatment optimization failed and inotropic support was necessary (one pulmonary edema, 2 oliguria). All these patients were BNP nonresponders. Two of them had a decrease in preload reserve after treatment. In the last patient, the preload reserve increased mildly but was still negative after treatment optimization. No patient had metabolic acidosis.


This study showed that in a population of decompensated CHF, preload reserve, as assessed by the PLR-induced fluid responsiveness, was related to changes of the BNP serum concentration after 4 days of optimized treatment. This relationship was established noninvasively whether considering the variations of CPi or other heart pump function indices: LVSWi, CI, and SI.

We choose changes in weight, TFC, and BNP as indicators of treatment effectiveness. The correlation between the treatment-induced changes in weight and TFC that we observed is in accordance with another study that showed a linear correlation between TFC changes and fluid removal during hemodialysis.[28] The correlation between changes in weight and BNP was weak, as previously reported.[29] Last, we found no relationship between changes in TFC and BNP. This is not really surprising as BNP is linked to myocardial wall stretch resulting from a transmural distending pressure from venous return blood volume, wall compliance, and extra wall pressure. TFC is proportional to cardiac and large vessel blood volume, which can be seen as a component of myocardial wall stretch, but also to lung, pleural, and tissue water that are poorly related to myocardial stretch. Therefore, the physiologic information provided by our 3 indices of treatment effectiveness is somewhat different. The best correlations with the preload reserve (presumably the best physiological link) were observed using the change in BNP.

The relatively large inter-patient variability found in the regression lines between the changes in preload reserve and in BNP does not really allow a good predicting model. This is not surprising since, in this clinical study, the determination of any predictability was limited by the reliability of measurements.

First, we measured BP noninvasively. Although the overall information provided by adding BP to the blood flow was real, since the stroke work and the cardiac power were more closely linked to the BNP changes as compared with blood flow alone, this technology has limited accuracy and precision.[30]

Second, we measured blood flow using chest Bioreactance. This technology is validated and was used in other physiologic studies.[2, 25] However, discordances with widely accepted reference technologies may be observed in 15% of the cases.[22] Nevertheless, Bioreactance is quite precise,[31] and changes observed in a given patient are likely to be real and not due to random error of measurement.[2, 23]

Third, the fluid responsiveness observed during a PLR is a semi-quantitative estimation of the preload reserve. Although PLR creates an internal autotransfusion and increased right ventricular venous return, the impact on LV preload (myocardial diastolic fiber stretch) is variable. The PLR-induced change in blood flow depends on (1) PLR mobilizable blood volume, (2) right ventricular function, (3) pulmonary circulation characteristics, and (4) LV function. The amount of mobilizable volume is a function of venous volume, compliance, resistance, waterfall effect, and driving pressure, all of which are under complex pathophysiologic regulations.[32-35] When the PLR is really passive, as done here by changing the bed position after clear explanation to the patient, the changes in the adrenegic tone is limited. In this study we observed minimal changes in HR and BP during PLR. Therefore, PLR represents a clear but unpredictable method for transiently expanding the volume of central circulation and the left ventricle venous return.[3]

Last, an improvement in CHF is statistically linked with a decrease in BNP but the parallelism may be impaired or delayed by several factors such as renal failure or transient change in LV function due to pulmonary disease, mitral regurgitation, LV obstruction, or coronary instability.

Despite these limitations, the observed relationships were significant enough to support the hypothesis of a clinically relevant link between the change in preload reserve and the change in BNP, from which we inferred treatment effectiveness. From these relationships, we cannot infer a direct causal pathway between change in BNP and change in preload reserve, but this is an argument that does not contradict the actual knowledge on BNP secretion and ventricle load. Therefore, we can infer a direct or indirect physiological relationship.

In previous studies, several authors failed to evidence such a relationship[36-38]; however, they investigated patients in critical care situations, requiring a fluid challenge, and looked at the acute changes of BNP during the fluid challenge. In these situations, BNP clearance may have been delayed and/or altered by different mechanisms such as kidney perfusion.[39]

Subsequently, we found that the best formulations of the preload reserve time evolution were the changes in cardiac stroke work and power indices. This is also in accordance with numerous studies showing that the cardiac power is the best indicator of heart pump function.[40, 41] Increasing the contractile force of the heart may generate pressure as well as flow. This explains that variables taking into consideration both components, such as work and power, are better indices of heart pump function.

Our findings are in accordance with the Starling law of the heart. In cases of decompensated CHF, there is a high probability that the left ventricle works on the plateau of the preload-flow relationship. Therefore, any increase in the preload is not paralleled by an increase in flow. After treatment optimization, the heart function may be displaced toward the preload-flow–dependent part of the Frank Starling relationship. In our study, this overall leftward displacement on the Franck Starling relationship was obtained by fluid removal (as shown by the overall weight loss and TFC decrease) and a decrease in the LV afterload, suspected from the significant decrease in BP after treatment optimization. Therefore, the SI was maintained and the CI was decreased proportionally to the decreased HR. The double product, LVSWi and CPi, decreased accordingly.

If we consider that the maximum CPi is stable and indicative of the severity of the CHF, the treatment optimization acts by minimizing the CPi at rest to allow further increase when necessary. In other words, the treatment optimization did not increase the afterload level at which an afterload mismatch occurred but decreased the baseline afterload to restore the preload-afterload coupling. However, a restored preload reserve was poorly linked to the magnitude of the treatment-induced decrease in CPi and other work indices. The better correlation with a decrease in BNP (a decrease in the myocardial wall stretch) probably indicates that preload reserve improvement was also linked to other hidden factors.

It has been shown that many CHF patients are discharged from the hospital with persistent signs and symptoms of congestion, leading to high risk of death or readmission. A score has been proposed to direct both current and investigational therapies designed to optimize volume status during and after hospitalization.[42] This score uses dynamic changes after orthostatic testing and Valsalva maneuver. Our study suggests that PLR could be added as another dynamic test. A restored preload reserve, as shown by a PLR-induced change in CPi >25 mW/m2, could be an additional interesting target before discharge. CPi changes can be measured by different technologies. Bioreactance is safe, precise, and inexpensive and requires no specific skills.


Despite limitations caused by noninvasive measurements, this clinical study shows that an effective treatment, as indexed by a decrease in BNP, restores the preload reserve in decompensated CHF patients by decreasing baseline CPi.

Acknowledgements and Disclosure

Pierre Squara received consulting fees from Cheetah Medical between 2006 and 2009. The authors report no specific funding in relation to this research.