Welcome
This is a web version of my PhD dissertation titled Fluid Responsiveness Prediction During Surgery – Physiological and Methodological Limitations and Considerations.
Abstract
Background
Administration of intravenous fluids is a frequent medical intervention during surgery. However, fluids have side effects, and increasing concern is being raised about the widespread use of intravenous fluids. To reduce side effects, fluids should only be administered to patients who can mobilise the added fluid via an increase in the heart’s stroke volume (SV)—this is called a fluid response. If a patient’s SV does not respond to fluid, they do not benefit from the fluid and can only experience side effects.
Several methods for predicting a patient’s response to fluid have been proposed. One of the best indicators that a patient’s SV will increase with fluid administration, is high ventilator-induced pulse pressure variation (PPV): the cyclic change in pulse pressure caused by mechanical ventilation. However, after this method was developed patients are generally being ventilated with lower tidal volumes (\(V_T\)) and higher respiratory rates (RR), and it has been shown that this lung-protective ventilation lowers the predictive accuracy of PPV.
Another approach to fluid responsiveness prediction is the mini-fluid challenge (MFC), where a patient’s fluid-responsiveness status is tested with a small amount of fluid. Only if the patient responds to the MFC, should a larger bolus of fluid be given. The hypothesis that the MFC-response predicts the response to a larger fluid bolus has been investigated in several studies.
Aims
The aim with this PhD project was to tackle challenges in fluid responsiveness prediction during surgery. Specifically, I wanted to develop methods for overcoming current clinical limitations to the use of PPV, and to understand and describe a methodological problem present in most MFC studies.
Papers
This dissertation is composed of three papers:
Paper 1 describes how the design used in most MFC studies creates a mathematical coupling between the predictor (the MFC-response) and the outcome to predict (the full fluid response). This causes an overestimation of the MFC’s predictive ability. An improved design is suggested.
Paper 2 introduces generalized additive models (GAMs) as a tool for analysing medical time series and waveforms. We demonstrate that GAMs can be used to calculate PPV in situations where the classical method fails, and that GAMs can decompose a central venous pressure waveform into physiologically meaningful components.
Paper 3 presents a clinical study of the effects of \(V_T\) and RR on PPV. The results indicate that using a GAM-derived PPV may help overcome some current limitations to the use of PPV, and that PPV’s predictive ability might be improved by adjusting PPV for \(V_T\).
Acknowledgments
First, thanks to my supervisor, Simon Tilma Vistisen. I have interrupted Simon’s work endlessly for close to four years, and he has kept his door open regardless. Many half-baked ideas, questions and frustrations have been resolved over a quick (rarely), unsolicited redecoration of Simon’s whiteboard. I suspect that few PhD students are lucky to have a supervisor who tolerates, and even engages in, biweekly musings on the wonders of mixed models.
Also, a big thanks to my co-supervisors, Thomas Scheeren, Stephen Edward Rees and Peter Juhl-Olsen. They helped design a clinical study, that was both safe, clinically relevant and practical to conduct, and were always available to answer questions.
Thanks to Thomas Scheeren for inviting me to Groningen—a stay that was canceled due to COVID19. And thanks to Stephen Rees and Dan Karbing for supervising my overly ambitious project on cardiovascular models at Aalborg University. Despite not even getting close to finishing the project, I have never learned so much in one month.
Thanks to the members of the assessment committee, Professor Hanne Berg Ravn and Professor Anders Åneman, for agreeing to assess this dissertation and the public defence, and to Professor Ebbe Bødtkjer, chairman at the public defence.
I am grateful to the patients who agreed to participate in the clinical study comprising a large part of this PhD.
Thanks to Aarhus University and Holger & Ruth Hesse’s mindefond for supporting this work.
I wish to thank all my colleagues and collaborators. Thank you for interesting scientific discussions and entertaining lunch conversation. Thanks to the clinical staff who welcomed me and my data cables in the operating room, and helped me conduct the clinical study.
Also, thanks to all the kind strangers on the internet who selflessly helped me with various statistical and technical questions issues. Special thanks to John George Karippacheril, developer of VSCapture, who helped me customize his software to work with department ventilators; essential for acquiring the data used in Papers 2 and 3. Also, special thanks to Gavin L. Simpson, who after helping out with several of the questions I posted on Stackoverflow.com and Twitter, agreed to co-write Paper 2. The template used for this dissertation is oxforddown by Ulrik Lyngs, John McManigle, Sam Evans and Keith Gillow. Thank you for sharing.
Thanks to friends and family for laying ear to both frustrations and excitement over nerdy details, and for your support in general.
Anna-Elisabeth, thank you for being a loving wife and mother, and for your admirable patience during periods with frequent absent-minded gazes from your husband. And thanks Senja and Gry for making every day meaningful, interesting and joyful.