Fluid resuscitation in sepsis: A systemic review and meta-analysis

Rochweg B, Alhazzani W, Sindi A, Heels-Andsell D, Thabane L, Fox-Robichad A, et al.

annals of internal medicine, sep 2;161(5):347-55.

This systematic review and meta-analysis looked at the different mortality rates of various intravenous fluids and fluid protocols when used for resuscitation in sepsis. While the results were short of conclusive, there seemed to be a decreased mortality linked to the use of balanced crystalloids and albumin.

There is a long standing question relating to intravenous fluids; that of crystalloid versus colloid. The authors of this paper suggest that the question is subtler than that, seeking to explore the difference between “normal” saline, and other more balanced crystalloid fluids – those that contain an organic anion as a buffer, and have a more physiological concentration of chloride. Equally, they noted the heterogeneity of colloids, and aimed to look for a difference in mortality rates between several types of colloid; albumin, low and high molecular weight hydroxyethyl starches (HES), and gelatin. Also worth further consideration is the crystalloid component of the colloid, which varies significantly, both in terms of chloride concentration and other elements between products.

Sepsis was used as a starting point, since optimal fluid management is particularly unclear in this setting, with wide variation in fluid protocols. Studies examining sepsis have often been excluded in meta-analyses, and alongside the distinction between types of colloid and crystalloid, this formed the basis for the article.

Study selection is discussed in greater detail in the paper itself; in general, the authors sought papers which provided a direct comparison in mortality rates between fluids (or fluid protocols) in a sepsis resuscitation setting. There were 9,875 initial studies identified, of which 185 were suitable enough for full text-review. Following further exclusions, 14 randomised controlled trials were identified for analysis, involving 18,916 patients in total. The analysis was carried out in two forms, a 4- and 6- node network meta-analysis (NMA).  

The 6-node NMA compared:

• balanced crystalloids
• unbalanced crystalloids
• albumin
• low molecular weight HES
• high molecular weight HES
• gelatin

The 4-node NMA compared:

• crystalloids
• albumin
• HES
• gelatin

Results from the 4-node analysis suggested: 

• higher mortality with HES than crystalloids, with a high level of confidence (OR, 1.13 [95% CrI   0.99 to 1.30])
• Lower mortality with albumin than with crystalloids, with a moderate confidence (OR, 0.83 [95% CrI 0.65 to 1.04])
• Lower mortality with albumin than HES, with moderate confidence (OR 0.73 [95% CrI 0.56 to 0.95])

 [CrI; Credibility Interval]

The 6-node analysis suggested that for preventing mortality:

• Albumin is superior to saline, with moderate confidence (OR, 0.82 [CrI, 0.65 to 1.04])
• Balanced crystalloids are superior to high-molecular-weight HES with moderate confidence (OR, 0.82 [CrI 0.60 to 1.13])
• Balanced crystalloids are superior to low-molecular-weight HES with moderate confidence (OR, 0.75 [CrI 0.58 to 0.97])
• Albumin is superior to low-molecular-weight HES, with low confidence (OR, 0.79 [CrI 0.59 to 1.06])
• Balanced crystalloids are superior to saline, with low confidence (OR, 0.78 [CrI 0.58 to 1.05])

In their discussion, the authors suggest that the reduced mortality with balanced fluids may be a reflection of their more physiological properties, although it is not clear which element (pH, electrolyte composition or presence of a buffer) is responsible. The possibility of using “unbalanced” saline in primary resuscitation for sepsis is called into question by these results. Also discussed is the apparent variation in type of colloid, with albumin seemingly the most suitable in this setting. Gelatin solutions had less firm data for meaningful comparison.

Limitations are the relatively few studies selected for analysis compared to the number of trials performed, as well as the fact that it was impossible to account fully for the variation in colloid solvents – for example, albumin was dissolved in different crystalloids in different studies.