Precision fluid therapy in sepsis: fluid responsiveness vs. fluid tolerance

Intravenous fluid is the most frequently administered drug in septic shock and one of the least precisely dosed. Two decades of trials have progressively dismantled the assumption that more fluid is better: protocolized early goal-directed therapy failed to improve on contemporary usual care, high-volume bolus protocols caused frank harm where organ support was limited, and restrictive strategies applied after an initial resuscitation have proven neutral rather than superior. What has replaced a universal volume prescription is a pair of questions that must be asked separately for each patient and repeated over time: will this patient's cardiac output actually rise with more fluid (fluid responsiveness), and can the lungs, kidneys, and venous system absorb that volume without harm (fluid tolerance)? This narrative review examines how to answer both questions in the modern intensive care unit. We ground the discussion in Guytonian venous return physiology and in the revised Starling model, in which sepsis-induced glycocalyx degradation makes infused crystalloid a short-acting drug with a cumulative side-effect profile. Using an index case of a mechanically ventilated patient with pneumococcal septic shock, atrial fibrillation, lung-protective tidal volumes, obesity, and acute kidney injury, we show why the two most widely used bedside indices, respiratory variation in inferior vena cava diameter and pulse pressure variation, are physiologically invalid in precisely the patients in whom the fluid decision is hardest, and why maintenance infusions and fluid creep deserve the same scrutiny as boluses. We make the case for direct stroke volume measurement paired with a reversible dynamic challenge, most commonly a passive leg raise, as the one approach that remains interpretable across arrhythmia, low tidal volume, spontaneous respiratory effort, and elevated intra-abdominal pressure, and we discuss dynamic arterial elastance as a complementary tool for predicting whether arterial pressure will follow flow. We review the outcome evidence, including the FRESH trial and supporting meta-analyses, the emerging literature on sepsis endotypes and subphenotypes, and the 2026 Surviving Sepsis Campaign recommendations. Finally, we address fluid tolerance and de-resuscitation, where venous congestion assessment and lung ultrasound are increasingly used, and outline where the evidence remains thin.