A systematic review and physiology of pulmonary artery pulsatility index in left ventricular assist device therapy

Abstract OBJECTIVES Right heart failure (RHF) is a major complication following left ventricular assist device (LVAD) implantation. Pulmonary artery pulsatility index (PAPi) has been evaluated as a haemodynamic marker for RHF, but PAPi is dependent on pulmonary vascular resistance (PVR). We conducted a systematic review to assess the relationship between PAPi and RHF and death in patients undergoing LVAD implantation and examined the relationship between PAPi cut-off and PVR. METHODS We searched PubMed, EMBASE, CENTRAL and manually screened retrieved references to identify all clinical studies reporting PAPi in adult patients with a durable LVAD. Eligibility criteria were prespecified and 2 reviewers independently screened and extracted data; the Newcastle–Ottawa Scale was used to assess quality of non-randomized studies. This study was prospectively registered on PROSPERO (CRD42021259009). RESULTS From 283 unique records, we identified 16 studies reporting haemodynamic assessment in 20 634 adult patients with an implanted durable LVAD. Only 2 studies reported on mortality and in both, a lower PAPi was significantly associated with death. Fifteen studies reported RHF data and, in 10 studies, a lower PAPi was significantly associated with RHF. Six studies reported on PAPi cut-offs ranging from 0.88 to 3.3; and the cut-offs were directly related to PVR (r = 0.6613, P = 0.019). CONCLUSIONS Lower PAPi was associated with RHF and death following LVAD implantation, but a single PAPi cut-off cannot be defined, as it is dependent on PVR.


INTRODUCTION
Durable left ventricular assist devices (LVAD) have become an established therapy in patients with end-stage heart failure. Outcomes of LVAD therapy have improved with the introduction of the magnetically levitated centrifugal continuous-flow device [1], but early right heart failure (RHF) remains a major cause of morbidity and mortality following LVAD implant [2]. Therefore, preoperative assessment of the risk of RHF is central to the selection of patients for LVAD therapy.
A number of haemodynamic parameters derived from pulmonary artery catheterization have been used to assess the risk of RHF. Of these, pulmonary artery pulsatility index (PAPi) has been shown to be an independent predictor of mortality due to RHF in acute myocardial infarction, pulmonary arterial hypertension, heart failure, heart transplantation and cardiogenic shock [3][4][5][6][7]. In general, lower PAPi is associated with higher the risk of RHF. PAPi is defined as the ratio of pulmonary artery pulse pressure (PP) to right atrial (or central venous) pressure. Pulmonary artery PP is a function of stroke volume (SV) and pulmonary artery compliance (PAC), and the latter has a hyperbolic relationship with pulmonary vascular resistance (PVR) [8]. On this basis, we hypothesized that the PAPI cut-off associated with RHF would be dependent on the PVR (increase with PVR).
This systematic review first assessed the relationship between PAPi, RHF and death following LVAD implantation. Second, we evaluated the relationship between the reported PAPi cut-off and PVR from the studies identified in this systematic review.

Ethics statement
As this is a systematic review, ethics committee review was not applicable. However, this study was formally reviewed and was prospectively registered on PROSPERO (CRD42021259009). The search results are reported in accordance with the PRISMA statement [9]. All eligibility criteria and search strategies were prespecified.

Eligibility
All studies reporting measurement of PAPi in adult patients with a durable LVAD, defined as an intra-or extracorporeal device implanted in the left ventricle for the treatment of advanced heart failure, irrespective of treatment intention (destination therapy or bridge to or transplantation), and published in the English literature were included. Studies were excluded if PAPi was not reported, or data on mortality or RHF were not reported. Studies reported only as a conference abstract were excluded due to insufficient data for analysis.

Search strategy
We searched international primary research databases (PubMed, EMBASE and CENTRAL) from inception to 14 September 2021 and reference lists of relevant articles to identify all eligible studies. The following search strategy was used for all 3 databases:

Study selection and data extraction
Abstracts and then full-text articles of all identified were screened independently by 2 reviewers (Ivan H.W. Yim and Ayisha M. Khan-Kheil). All studies in patients with a durable LVAD, which contained PAPi data analysed against RHF or death, were included. RHF was author defined, including the Interagency Registry of Mechanically Assisted Circulatory Support (INTERMACS) definition of RHF [10]. Data were extracted independently by 2 reviewers (Ivan H.W. Yim and Ayisha M. Khan-Kheil) from the full-text publication and any disagreements were resolved by consensus. A full list of the data items and descriptors is available in the Supplementary Material. For all studies included, the Newcastle-Ottawa Scale [11] for assessing the quality of non-randomized studies was employed and based on the number of stars each study gained in each domain; this was then converted to the Agency for Healthcare Research and Quality standards of good, fair and poor.

Statistical analysis
Statistical analysis was performed using IBM SPSS Statistics, Version 27.0 (Armonk, NY). All continuous data are expressed as medians with interquartile ranges and all categorical data are expressed as counts and percentages where applicable.

RESULTS
The original search produced 259 unique records and we identified 14 studies reporting haemodynamic assessment in 20 634 adult patients with an implanted durable LVAD (Fig. 1). The updated search performed in August 2022 produced another 49 unique records and all full-text articles were sourced online or via national libraries; a total of 16 studies were included for analysis. Characteristics of the included studies are shown in Table 1. All studies were retrospective cohort studies originating from 4 countries, with 9 (56%) from the USA, 3 from Italy and 1 each from Japan, Germany, the Netherlands and Turkey. Articles were published in specialist heart failure, cardiothoracic surgery, transplantation or anaesthetic journals, and all were published in English.
The study periods ranged from 2004 to 2021. The median number of patients in each study was 95 (interquartile range 78.8-226.5). The types of LVAD implanted were primarily intracorporeal continuous-flow devices, HeartMate (Abbott Laboratories, USA) II or III and HeartWare (Medtronic, USA), with 1 study evaluating the NIPRO-VAD (Nipro, Osaka, Japan) extracorporeal pulsatile pump [28]. Only 6 out of the 16 studies stated the goal of LVAD therapy (bridge to transplantation/candidacy or destination therapy).
Of the 16 studies, the primary outcome was (i) RHF or the event of right ventricular assist device implantation (n = 15) and (ii) mortality associated with right atrial pressure (RAP) (n = 1). All 16 studies reported on the timing of PAPI measurement in relation to LVAD implant. Fourteen studies performed right heart catheterization prior to LVAD implantation to record routine haemodynamic parameters including PAPI (although only 6 studies gave the exact timing in hours, days or months of RHC from LVAD implantation) and 2 studies measured PAPI intraoperatively at the time of LVAD implantation. Twelve studies did not specify medical therapy at the time of haemodynamic assessment (e.g. use of inotropes or intra-aortic balloon pump).
Guglin and Omar [24] analysed the INTERMACS database including 18 733 patients and found that RAP was a haemodynamic predictor of death with an receiver operating characteristic (ROC) curve area under the curve (AUC) of 0.55 (CI 0.539-0.562, P < 0.0001) and an RAP of 13 mmHg or higher had the highest combined sensitivity and specificity in predicting mortality. PAPI and other haemodynamic parameters were also analysed and compared against RAP. Survivors were found to have a significantly higher PAPI (3 ± 3.1 vs 2.6 ± 2.7; P < 0.001) but when compared to RAP, it was found to have a lower ROC curve AUC with a difference in areas of 0.0105, P = 0.005 and therefore RAP was found to be superior at predicting death.
Gonzalez et al. [25] studied PAPI at serial time points before LVAD implantation and during the period of medical optimization, hypothesizing that the magnitude of change of PAPI and other invasive haemodynamic measurements would provide an incremental risk stratification for RHF following LVAD therapy with a secondary end-point of death at 180 days. After optimizing their patients with a combination of diuretic, intravenous sodium nitroprusside, inotropes and non-durable mechanical circulatory support where appropriate, they found that an optimized preoperative PAPi of >3.33 was associated with a significant reduction in early RHF (P < 0.001). In patients with a change in PAPi (delta PAPi) of >2.08 during the optimization period (time period not specified), there was a significant reduction in 6-month mortality following LVAD implantation presumably from reduced early RHF. In a recently published study, Cacioli et al. [26] showed that PAPi following vasodilator challenge with sodium nitroprusside provided incremental risk stratification when combined with established risk scores (EUROMACS-RHF and CRITT). In this study, PAPi alone was significantly lower in patients who developed RHF post-LVAD implantation compared with those who did not have RVF (2.2 ± 1.3 vs 3.3 ± 1.5, P = 0.008). This was even more striking following vasodilator therapy (5.3 ± 3.9 vs 2.7 ± 3, P = 0.003) suggesting a higher PAPi following vasodilator therapy indicates more right ventricular reserve. Similarly to the study performed by Gonzalez et al., this study reported a postvasodilator challenge PAPi cut-off of 3.2 (with a sensitivity of 66% and a specificity of 68%) for right ventricular failure. Furthermore, the authors combined RV fractional area change and systolic pulmonary artery pressure with post-vasodilator challenge PAPi and found that it had an AUC of 0.949.
The majority of studies used the INTERMACS definition to describe the end-point of RHF (Table 2). Fifteen studies reported RHF data relating to PAPi: in 10 [13][14][15][16][17][18][19][25][26][27] of these studies, a lower PAPi was associated with RHF, but there was no significant association between PAPi and RHF in the other 5 studies [12,[20][21][22][23]. Six studies performed ROC analyses to determine the optimal PAPi cut-off for RHF following durable LVAD implantation, ranging from 0.88 to 3.3, with a mean of 2.2 and the ROC AUC values ranged from 0.70 to 0.94 with a mean of 0.80. All 6 studies provided data on PVR in patients with and without severe RHF. There was a direct relationship between the PAPi cut-offs and PVR for patients with/without severe RHF (r = 0.6613, P = 0.019) (Fig. 2).
In assessing the quality of the included studies using the Newcastle-Ottawa Scale, all achieved an Agency for Healthcare Research and Quality grading of good, with at least 1 star in each domain.

DISCUSSION
PAPi has been evaluated as a parameter to predict the risk of RHF and death following LVAD implantation. Notable findings from this systematic review were: (i) lower PAPi measurements before LVAD implantation were associated with higher risk of RHF and/or death; (ii) the reported PAPi cut-offs for discriminating patients at risk of RHF and/or death varied by more than three-fold from 0.80 to 3.3; and (iii) the reported PAPi cut-offs were directly related to PVR.
PAPi is the ratio of PP over RAP (equation 1: PAPI = PP/RAP). Analogous to the charging of capacitors, a significant proportion of the right ventricular SV 'charges' the reservoir volume and increases pressure in the compliant pulmonary arteries in systole, which discharges during diastole. Pulmonary arterial compliance defines this relationship between an increase in blood volume (DV) and an increase in pressure in the pulmonary arterial system. In practice, PAC is difficult to measure because direct measurement of DV is not possible due to the continuous outflow from the arterial system. Therefore, in clinical practice, the ratio of SV/PP is used to determine PAC, accepting that this equation would overestimate the true PAC. Rearranging this equation, it can be appreciated that PAC and SV are the main determinants of PP (equation 2: PP = SV/PAC).
Pulmonary arterial compliance is determined by the prevailing distending pressure [i.e. mean pulmonary artery pressure (MPAP)] and by the elastic properties of the pulmonary arterial wall. The latter is mainly determined by the composition of elastin and collagen in the wall. Pulmonary arterial compliance decreases when MPAP increases in non-linear relationship [8] Records identified through other sources n = 3   This study aimed to devise the ALMA risk score for predicting RHF following LVAD implantation. Within the haemodynamic data of this study, a PAPi of <2 was found to be associated with unplanned RVAD support (P = 0.001) and on multivariable logistic regression analysis PAPi < 2 had an OR of 3.3 (CI 1.7-6.1, P = 0.001). The ALMA score employs the following 5 variables: destination therapy intention, PAPI < 2, RVSWi <300 mmHg/ml/m 2 , RV:LV ratio > 0.75 and MELD-XI score >17. The authors proposed a score of 0-1 implies low risk for RVAD requirement and a score above 4 implies very high risk for requiring RVAD following LVAD implantation. Raymer et al. [17] 216 HM2, HVAD Yes Not reported This study reported the combination of TAPSE and HeartMate risk score (HMRS) as a scoring system to predict RHF following LVAD implantation. The RHF group had a lower PAPi (P = 0.001). ROC analysis showed PAPi had an AUC of 0.63 (P < 0.001). When the haemodynamic parameters were analysed combination of TAPSE with the HMRS was the best for predicting RHF compared to HMRS + PAPi and HMRS + sRVCPI. Gudejko et al. [18] 85 HM2, HVAD Yes Not reported The data used in this study were intraoperative haemodynamic parameters and also echocardiographic data.  [29] and inverse relationship with PAWP are well characterized [30]. By extension, the pulmonary artery PP would be similarly dependent on SV, PAWP and PVR (Fig. 3). This would explain the observed relationship between the PAPi cut-off and PVR levels. We were unable to evaluate the relationship between HR, SV and PAPi, as the data were not reported in the studies.
The implications of these physiological considerations are three-fold. First, in the face of increasing PVR, the right ventricle mal-adapts to maintain relatively low RAP and normal SV; the This study compared TAPSE, CVP: PCWP ratio, RVSWI, PAPi, Pennsylvania Score, Michigan score, CRITT score, ALMA score and the EUROMACS score. PAPi was not significantly lower in the group with postop RHF (P = 0.304). They concluded that only the EUROMACS and CRITT score had an ROC AUC above 0.7 and that the combination of TAPSE and the Pennsylvania score was found to be the most sensitive (85%) whereas TAPSE + Michigan score + CVP:PCWP ratio was the most specific (97%). Benjamin et al. [22] 104 HM2, HVAD No Not reported Primary end point was duration of inotropic support and the association with RVF. They found patients who were on long-term milrinone had a significantly increased risk of developing RHF post-LVAD insertion. PAPi was not significantly different between the RHF and the group without RVF post-LVAD implant (4 ± 3.9 vs 3.2 ± 2.3, P = 0.255) Ruiz-Cano et al. [23] 80 HM3, HVAD No Not reported PAPi was not significantly different between the early RHF group versus no early RHF (2.5 vs 3, P = 0.283). This study found that blood urea nitrogen >44.5 mg/dl and CVP/PCWP > 0.55 were the parameters with the strongest association with early RHF. Guglin and Omar [24] 18608 Not reported No data Not reported This was a retrospective cohort study looking at data from the INTERMACS database and primarily assessing RAP and its ability to predict death. RAP was the main predictor of mortality in LVAD recipients. PAPi was lower in non-survivors (P < 0.001), but RAP had superior discriminatory value with a difference in AUC of 0.0105 (P = 0.0052). Gonzalez et al. [25] 315 HM2, HVAD Yes Optimal PAPi <3.3 (RHF); Delta PAPi <2.08 (Death at 6 months) This study assessed the change in PAPi during preoperative haemodynamic optimization prior to LVAD implantation. The mean optimal PAPi was lower (P < 0.001) in the group that developed early RHF. A delta PAPI of <2.08 during optimization was associated with higher mortality at 180 days (P = 0.003). Cacioli et al. [26] 75 HM2, HM3 Yes Not reported A lower PAPi was strongly associated with RHF following LVAD implant. This study also demonstrated in those who did not develop RHF post-LVAD had a significantly higher PAPi following vasodilator challenge at preop RHC (5.3 ± 3.9 vs 2.7 ± 1.3, P = 0.003). Furthermore, PAPi when combined with established risk scores provided incremental risk stratification for post-LVAD RHF. Stricagnoli et al. [27] 38 latter results in disproportionately elevated PP, and by extension an increase in PAPi. The increase in PAPi is a result of progressive adverse remodelling and should not be misconstrued as an improvement in right heart function. Second, a significant drop in PAPi would only occur when the right ventricle uncouples from the pulmonary arterial system, with resultant drop in SV and rise in RAP. Third, because of the higher PVR and lower PAC, the PAPi level would remain higher in the setting of RHF related to pulmonary vascular disease compared to RHF due to primary right ventricular cardiomyopathy with low PVR and high PAC. In this context, Essandoh et al.'s recent description of a mean weighted PAPi of 2.17 as a cut-off to predict RHF post-LVAD implantation should be interpreted and applied with caution, and extrapolation to patients with other pulmonary haemodynamic profiles should be resisted. Pulmonary artery PP is inversely related to PAC, which is inextricably related to PVR. An increase in PVR increases pulmonary artery PP at any given SV. In this regard, PVR is a determinant of pulmonary artery PP and PAPi and cannot be avoided in interpreting PAPi. The interpretation of PAPi must take into consideration the underlying disease pathophysiology.
In primary right ventricular dysfunction such as right ventricular infarction with low PVR, the compliant pulmonary arterial system would result in low pulmonary artery PP for a given SV and RAP. In contrast, PAPi must be higher at the same SV and RAP in high conditions, such as primary arterial hypertension.
As an illustration, we describe 2 clinical scenarios: Patient A, assuming a RAP of 10mmHg, has a normal PVR and normal PAC of 5ml/mmHg. Even with supranormal stroke volume of 100ml (pulse pressure=20mmHg), his PAPi would only be 2. Conversely, patient B has a high RAP of 20mmHg and a low PAC of 1ml/mmHg due to pulmonary vascular disease. Even with a RAP that is two times higher and half the stroke volume (50ml, pulse pressure=50mmHg), patient B has a higher PAPi of 2.5. Despite the lower PAPi, it is physiologically implausible to suggest that patient A has more severe right heart failure compared to patient B.
These 2 clinical scenarios highlight the key message in our systematic review-the same PAPi cut-off cannot be applied in different clinical conditions, and a single cut-off averaged from a number of heterogenous studies is likely to be misleading. Interpretation of PAPi requires an appreciation of the underlying disease and pathophysiology. We have developed a calculator available in the Supplementary Material, which shows the mechanisms for the observed PAPi depending on the SV and PVR.
It should be noted that other factors may also contribute to the variable cut-offs and discriminatory value in patients undergoing LVAD implantation. First, there are multiple causes of RHF following LVAD implantation, including technical/surgical factors. Second, the definition of severe RHF is related to the postoperative management strategy. The clinical threshold for the use of right ventricular assist device may explain the variable incidence of severe RHF reported in the literature. Thirdly, the timing of assessment relative to LVAD implantation was not uniformly described in the studies. PAPi would change depending on the effects of medical therapy, including diuretics, vasodilators and inotropes, as reported by Gonzalez et al. [25]. The assessment of PAPi closer to the time of LVAD implantation may have greater discriminatory value.

Limitations
With only 2 publications reporting on death in relation to PAPi this precludes meta-analysis and hence the descriptive nature of this systematic review. However, we believe that our description of the relationship between PAPi and PVR (and therefore a single PAPi cut-off value cannot be defined for heterogenous conditions) is not widely appreciated, yet highly relevant to clinicians, especially with greater adoption of PAPi into clinical practice.

CONCLUSION
In conclusion, lower PAPi is associated with a higher risk of RHF and mortality in patients with durable LVAD therapy. However, PAPi is inherently related to PVR and the different PVR levels may have resulted in the variable PAPi cut-offs described in the studies. The PVR level should be taken into consideration in the interpretation of PAPi.