In this retrospective observational study, we identified three sNa trajectory patterns in patients with severe AKI utilitizing GBTM: AS, DS and ST groups. Patients whose serum sodium trajectories were described as ascending or descending were significantly associated with a higher risk of 30-day and 1-year mortalities compared to those with a stable sodium trajectory. This association persisted after adjusting for admission sNa and other potential confounding factors.
Dysnatremia, referring to hyponatremia or hypernatremia, is one of the most common electrolyte disorders in ICU. Numerous studies have shown that dysnatremia is independently associated with poor prognosis, even small changes in sodium concentration could significantly worsen the prognosis[8–11, 18, 19]. For instance, previous studies reported that both borderline dysnatremia on admission and sNa variations of ≥ 6 mEq/L are significantly associated with higher risks of mortality[8, 10, 18, 19]. Collectively, compared to the absolute sNa value at a specific time point, longitudinal sodium fluctuation provided more concise information on disease progression due to its close association to physiological responses, renal dysfunction, and clinical treatment. Yuya et al. conducted a secondary analysis of the AQUAMARINE study and found that sodium dipping, defined as sNa level declined below the baseline level within 48 hours, was associated with higher mortality of patients with acute heart failure after adjusting baseline sNa[20]. In contrast to the crude definitions of Yuya, Chewcharat used the GBTM algorithm to identify five distinct sNa trajectories based on longitudinal sNa levels. Compared with stable normonatremia, other sNa trajectory patterns were strongly associated with poor prognosis[14]. Consistent with prior studies, our results revealed that the trajectories of sNa correlated with the prognosis of patients with AKI in ICU, independent of baseline sodium levels, providing new insights into the connection between clinically common dysnatremia and patient prognosis.
Kidney is the main organ involved in water-electrolyte homeostasis and clinically, AKI often coexists with sodium disorders. Although it was documented that coefficient of sodium variation linearly associated with an increased risk of AKI[13], the causal relationship between AKI and dysnatremia remained unclear. In the cases of AKI, the prevalence of dysnatremia ranged from 22.5–24.6% and patients with dysnatremia had a higher risk of mortality[13, 21, 22]. Besides, Jonathan reported that it was the patients with trajectory described as uncorrected hypernatremia rather than fluctuating sodium who had the highest risk of mortality by retrospectively analyzing the sodium trajectories of 288 patients[23]. Probably due to sample sizes, its conclusion was not consistent with previous study[14]. There is a contradiction in whether correcting dysnatremia is beneficial to the prognosis. Restoration of initial dysnatremia appeared to benefit in-hospital survival for elderly patients[24]. On the contrary, no additional clinical benefit from correction of dysnatremia for patients undergoing continuous renal replacement therapy (CRRT) as indicated by an observational study[25]. One study even suggested that rapid correction of sodium could even be harmful[26]. In the present study, the GBTM is applied to determine the sNa trajectory patterns of AKI patients based on the patient's sNa values for 7 consecutive days after admission to the ICU. Furthermore, Kaplan-Meier curves and Cox regression models revealed that fluctuation of sNa was associated with the risk of mortality. Combined with our results, sodium fluctuation appeared to be a well-performed indicator of prognosis rather than a therapeutic target.
Since at least three published studies have used GBTM to identify serum trajectory patterns[14, 23, 26], it is necessary to note the differences between previous studies and ours. On the one hand, the populations enrolled varied: our study included patients diagnosed with AKI within 48 hours of ICU admission and previous studies focused on hospitalized patients with AKI, hospitalized patients and patients with heart failure, respectively. On the other hand, Xia determined sodium trajectories based on changes in sodium levels within 48 hours of admission, while Chavez and Chewcharat based on multiple in-hospital sodium levels without missing values considered. However, considering the effect of missing values and the possibility of excessive fluctuations in sNa in a short time, we utilized seven consecutive days of sNa to analyze sodium trajectory patterns.
Several limitations must be mentioned in the study. First, our inclusion of patients with AKI occurring within 48 hours of ICU admission would cause two problems. On the one hand, some patients would not have baseline serum creatinine data from 7 days earlier. On the other hand, patients with AKI occurring after 48 hours would be missed. Second, we used consecutive 1-week longitudinal sNa measurements to determine the sodium trajectories, which improved the reliability of the model. However, the sample sizes were reduced due to the exclusion of patients who did not have their sNa measured for 7 consecutive days, which may underestimate the impact of sNa trajectories on mortality in AKI patients. Third, serum sodium was not adjusted for serum glucose levels and sodium intake was not assessed in this study. Last but not least, due to the observational nature of the investigation, the causal relationship between dysnatremia and clinical outcomes could not be established.