Applicable patients
The PVI has higher accuracy for mechanically ventilated patients with a regular rhythm and nonthoracotomy [41]. The PVI reflects the degree of change in PI caused by breathing over a period of time, so it is greatly affected by cardiopulmonary exercise. The PVI has ability to reliably predict preload responsiveness, provided that the pressure changes in the chest cavity are sufficiently obvious enough and the cardiopulmonary interaction between different respiratory cycles is stable. Therefore, the PVI and other dynamic parameters of cardiopulmonary interaction are more suitable for patients with mechanical ventilation rather than spontaneous breathing. The results of the meta-analysis also showed that PVI was less reliable in the subgroup of cardiac surgery (Youden index =0.45) than in the non-cardiac surgery subgroup (Youden index =0.49).
Heterogeneity
Significant heterogeneity exists in both the overall group and most of the subgroups, which may be because the patient’s conditions are complex, the surgical methods are different and the fluid management methods are different. Meta-regression showed that intravenous colloid injection selection may be the source of heterogeneity (p=0.02). However, with the exception of 17 studies [12-18, 21-23, 25, 26, 31-35] that employed intravenous colloid injection as a method of preload responsiveness, the heterogeneity remained significant (I2=84%). The sensitivity analysis showed that 2 [18, 29] of the studies may have contributed to the heterogeneity; however, following exclusion of the two studies, the heterogeneity remained significant(I2=95%). The heterogeneity was relatively low in the subgroups undergoing noncardiac surgery(I2=63%), which may be because of the small changes of the patient’s peripheral perfusion. There was no significant heterogeneity in the subgroups of patients without undergoing surgery (I2=33%), which may be because there was no effect of surgery on patient’s hemodynamic changes. There was no significant heterogeneity in the crystalloid subgroup (I2=23%), potentially because of the small sample size (n=4).
In general, the heterogeneity within each subgroup is stable. After quality assessment of the included studies, we believe that heterogeneity will not have a fundamental impact on the reliability of the meta-analysis result.
Perfusion situation
Reliability of the PVI is largely affected by perfusion situation[42]. Peripheral perfusion deficiency can result in impaired blood flow to a stable constant partly caused by skin and other factors that signal the volume in the tissue. To date, a pulsed oximeter, which is used to calculate the PVI, will not be able to determine whether the reduction of chest pressure is caused by the variety of cardiovascular system capacity or low perfusion of the monitored site, so any influence on peripheral perfusion factors, that is, the factors that affect PI, can affect the reliability of the PVI prediction of preload responsiveness [36]. The sensitivity of the subgroup of cardiac surgery is lower than that of the other subgroups and overall (0.67 95% CI 0.40-0.87). Broch O et al. [11] reported that the PVI reliably predicted preload responsiveness only in patients with high perfusion level (PI>4%).
When using the PVI to guide goal-directed volume expansion, anesthetists should pay attention to factors that can affect perfusion situation of the monitored site (such as peripheral vascular disease, severe heart failure, application of vasoactive drugs, and damage of the monitored site).
The best threshold
The included results show that the PVI has a wide range of best thresholds for defining responders to preload responsiveness, which range from 7% to 20%. The different conditions for each study (the patient’s underlying disease, volume stroke, age, type of surgery, in operating room or in ICU), and patients’ different fluid management (the application of vasoactive drugs, rate of intravenous infusion and type of volume expansion) may contribute to high variability.
Monitored site
The monitored site could affect the morphology and respiratory variation of the PVI[43-46]. Desgranges et al. [14] compared finger, forehead and ear as monitored site, reporting that the choice of three monitored sites has no significant impact on accuracy. While Hood et al. [21] reported that the PVIfinger can reliably predict increases in SV, while the PVIearlobe can not reliably predict increases in SV in dynamic intraoperative conditions. Fischer et al. [17] demonstrated PVIforehead was more accurate than PVIfinger in predicting preload responsiveness after cardiac surgery. For safety and convenience, the PVIfinger remains the preferred choice for most patients, with the PVIforehead and PVIearlobe as stable alternatives [14].
Limitations
Our systematic review has several limitations. First, significant heterogeneity exists in both the overall group and most subgroups; thus, differences between patients and surgeries should be considered in the application of the PVI. Second, we only included mechanically ventilated patients, which limited the results extrapolated to all patients. Studies on the monitoring of the PVI on patients with spontaneous breathing must be conducted. Third, subgroup analyses of the child subgroup and the passive leg raise subgroup were not performed because of insufficient studies. Fourth, the best threshold for the PVI varied within great ranges, and the best threshold for different types of patients and surgeries remains to be studied.