Study population

This study was approved by the Acıbadem University Medical Research Assessment Council (ATADEK, 2016-18/1) on 10.11.2016. All data were fully anonymized without restriction after the ethics committee approval and ethics committee waived the requirement for informed consent. Data collection was started on 01.01.2018 and 2529 patients who were admitted to the tertiary intensive care unit (ICU) in Acıbadem International Hospital between January 1^st, 2011, and January 1^st, 2018 were retrospectively evaluated. Inclusion criteria were age>18 years, arterial blood gas samples and albumin levels at the ICU admission. Exclusion criteria were readmissions and missing outcomes (Fig 1).

Fig 1

Study flowchart.

Database

Patients demographics, blood gas and laboratory data were collected from the Acıbadem Health Group Database as from 01.01.2018. ABL 800 (Radiometer, Denmark, Copenhagen), which employs an ion-selective electrode, was used for blood gas analysis. At ICU admission, demographic data (age, sex), diagnosis (medical, elective and emergency surgeries), Acute Physiology And Chronic Health Evaluation II (APACHE II) score, pH, PaCO₂ (mmHg), Na (mmol L^-1), K (mmol L^-1), Ca (mmol L^-1), Cl (mmol L^-1), HCO₃ (mmol L^-1), SBE (mmol L^-1), SID_a (mmol L^-1), effective strong ion difference (SID_e) (mmol L^-1), strong ion gap (SIG) (mmol L^-1), BE_Cl (mmol L^-1), serum lactate (mmol L^-1), albumin (g L^-1) and outcomes were recorded.

Traditional approach

The effect of Cl is evaluated in accordance with accepted limits of Cl_obs. In this study, serum Cl_n ranges for Cl_obs were defined as 98–106 mmol L^-1 [6]. Cl_obs levels which are below 98 mmol L^-1 were defined as hypochloremia whereas Cl_obs levels which are above 106 mmol L^-1 were defined as hyperchloremia.

Chloride corrections

In this concept, it is claimed that Cl_obs should be corrected in accordance with serum Na changes. There are two approaches in literature;

Cl_def/exc

For the Cl_def/exc, accepted serum Na_n and Cl_n were 142 mmol L^-1 and 104–108 mmol L^-1 respectively [2] Firstly, Cl_corr is calculated as below;

Cl_corr = (Cl_obs x (Na_n / Na_obs))

If calculated Cl_corr levels are between 104–108 mmol L^-1, Cl_def/exc is equal to zero and it indicates normochloremia.

If calculated Cl_corr levels are below 104 mmol L^-1, Cldef/exc is calculated as below; Cl_def/exc = (104 mmol L^-1—Cl_corr)

Since Cl_def/exc will have positive value, it will indicate the alkalotic effect due to hypochloremia.

If calculated Cl_corr levels are above 108 mmol L^-1, Cl_def/exc is calculated as below; Cl_def/exc = (108 mmol L^-1—Cl_corr)

Since Cl_def/exc will have negative value, it will indicate the acidotic effect due to hyperchloremia.

Cl_mod

For the Cl_mod, accepted serum Na_n and Cl_n were 140 mmol L^-1 and 102 mmol L^-1 respectively [3]

Cl_corr is also calculated as the same equation as below;

Cl_corr = (Cl_obs x (Na_n / Na_obs))

However, if the calculated Cl_corr level is 102 mmol L^-1, Cl_mod is equal to zero and it indicates normochloremia.

If calculated Cl_corr levels are below 102 mmol L^-1, Cl_mod is calculated as below;

Cl_mod = (102 mmol L^-1—Cl_corr)

Since Cl_mod will have positive value, it will indicate the alkalotic effect due to hypochloremia.

If calculated Cl_corr levels are above 102 mmol L^-1, Cl_mod is calculated with the same formula as below;

Cl_mod = (102 mmol L^-1—Cl_corr)

Since Cl_mod will have negative value, it will indicate the acidotic effect due to hyperchloremia.

Cl/Na ratio

Normal limits for serum Cl/Na ratio are accepted as between 0.75–0.80 [5]. There are no accepted serum Na_n and Cl_n values in this concept.

If Cl/Na ratio is between 0.75–0.80, it is defined as normochloremia.

If Cl/Na ratio is below 0.75, it indicates the alkalotic effect due to hypochloremia.

If Cl/Na ratio is above 0.80, it indicates the acidotic effect due to hyperchloremia.

BE_Cl

In this concept, the difference between serum Na and Cl should be 32 mmol L^-1 [4]. BE_Cl is calculated as below;

BE_Cl = (Na—Cl—32 mmol L^-1)

If BE_Cl level is zero, it indicates normochloremia.

If BE_Cl levels are below zero, it indicates the acidotic effect due to hyperchloremia.

If BE_Cl levels are above zero, it indicates the alkalotic effect due to hypochloremia.

Other calculations

SID_a, SID_e and SIG, were calculated as follows: [11]

SID_a = ([Na]+[K]+[Ca])—([Cl]+[lactate])

SID_e = (2.46x10^-8x(P_aCO₂ / 10^-pH)) + (Albumin (g L^-1) x (0.123xpH-0.631))

SIG = SID_a—SID_e

The subgroup

To clearly demonstrate only the effects of Cl on the acid-base status, a subgroup that included patients with normal lactate (≤1.6 mmol L^-1), albumin (≥35 g L^-1, the lowest value of normal laboratory limits) and SIG (0–2 mmol L^-1) values was created (Fig 1) [3, 12].

Statistical analysis

All data were analysed by using SPSS version 27. The Kolmogorov smirnov test was used to detect normal distributions. The data were presented as percentages or medians and interquartile range (IQR). The effects of chloride on the acid-base status according to the different approaches were validated by the SBE and SID_a because they are parameters which indicate the total metabolic effects and the metabolic effects of electrolytes respectively. The agreement between each pair of approaches was analysed with the Kappa test. The Pearson correlation was used for all correlations. According to revised SBE formula [13], whose components are all electrolytes, albumin, lactate, SIG and PaCO2, for all patients, all Cl evaluation parameters (Cl_obs, Cl_def/exc, Cl_mod, BE_Cl, Cl/Na ratio), Na_obs, K, Ca, albumin, lactate, SIG and P_aCO₂ were added to the multivariate linear regression models to determine their relationship with the SBE. In the subgroup, the Bland & Altman test was used to determine the limits of agreement between SBE and each of Cl_mod, Cl_def/exc and BE_Cl. Acceptable agreement was a bias of up to ±1 mmol/L and the limits of agreement were less than ±1.96 mmol L^-1. A p-value <0.05 was considered statistically significant.

Definitions of hyperchloremic acidosis and hypochloremic alkalosis

Siggaard-Andersen emphasized that the degree of acidity was related to the amount of H⁺ ions after Arrhenius developed the dissociation theory [14]. Stewart and Kellum also mention that the unlimited source of H⁺ ions is water in the body [11, 15]. Under these circumstances, the amount of H⁺ ions in the plasma should be determined by water dissociation. However, terminologically, the use of hyperchloremia and hypochloremia to define Cl effect on acid-base status appears that only Cl level changes cause water dissociation and vice versa. Yet, the Stewart approach defines 3 independent variables to determine H⁺ ions in the plasma, one of which is SID_a [2, 11, 14]. In other words, the difference between strong cations and anions determines whether water will dissociate. Since the major cations and anions are Na and Cl in the plasma, water dissociation should depend on the Na-Cl difference, not only the Cl level. Already, Story et al. showed that hyperchloremic acidosis was a type of SID_a acidosis [16, 17]. Even so, Tani et al. had defined their groups in accordance with only Cl_obs levels, and the mean SID_a value of their normochloremic group was 33.9±3.5 although lactate and Na values were normal in this group [6]. Yet, this option can come true in only hyperchloremia, not normochloremia. When we divided our patients into three groups as such in Tani’s study, we also found the median (IQR) SID_a value for 1063 patients who accepted normochloremic was 36.3 (5.5). Hence, we think that using accepted Cl limits to define hyperchloremic acidosis or hypochloremic alkalosis is not adequate. Our study clearly showed that using only accepted Cl limits gave us different results when compared to other approaches. It had the weakest correlations and moreover, Cl_obs was not associated with the SBE (Tables 3–5). We believe that the reason for incompatible results in some chloride studies may be the use of the accepted Cl limits [6, 7, 18].

Chloride corrections and the Cl/Na ratio vs BE_Cl

The other approaches to evaluating the effect of Cl on the acid-base status are chloride corrections, the Cl/Na ratio and BE_Cl [2–5]. Since each of them is supposedly based on the Stewart approach, each should correlate with SID_a. In this study, we found that there were correlations between SID_a and each approach in the whole cohort and the subgroup (Table 3). However, the correlation of BE_Cl was stronger than those of Cl_def/exc, Cl_mod and the Cl/Na ratio. Furthermore, there was limited agreement among them, and we obtained different results for a single patient by using each of the approaches at the same time (Tables 1 and 4). We think that there may be several explanations of these results. First, the corrections state that Cl_obs should be corrected in accordance with Na because Na and Cl should be similarly diluted or concentrated based on the gain or loss of water [2]. This thesis may be valid in vitro. However, it is known that the distributions of Na and Cl among compartments in the body differ because of the Gibbs-Donnan effect, Hamburger shift, absorption mechanisms in the kidney and small and large intestines and the effects of some drugs, such as furosemide [1, 19–21]. Therefore, the Na and Cl concentrations may not increase or decrease in the same manner in different clinical situations. In fact, even if Na and Cl are similarly diluted or concentrated, the corresponding acidosis or alkalosis cannot be explained because SID will be constant. Furthermore, in the classification by Fencl, it is a paradox to use only Na levels to evaluate water excess or deficit while claiming that the dilution or concentration of Cl and Na are the same [2]. Second, the corrections are actually based on the Cl/Na ratio [2, 3]. The Cl/Na ratio is another usable parameter because it is based on the Na_obs and Cl_obs levels. Thus, Cl/Na ratio studies have consistent results as well [5, 9, 10]. However, in addition to using the Cl/Na ratio, correction approaches create a fictitious Cl level by using the accepted serum normal value ranges for Cl and Na. Yet, it does not make sense to correct Cl when a Cl_obs level exists. The Stewart Approach mentions neither limits for Na and Cl nor needed to correct serum Cl levels in accordance with Na changes [11, 15]. The only rule in this concept is the difference between strong cations and anions, which almost equals the Na-Cl difference defined as BE_Cl by O’Dell et al. [4, 11]. For this reason, BE_Cl should be the most valuable parameter for the evaluation of the effect of Cl. In this study, we found two more important pieces of evidence supporting the superiority of BE_Cl: a) the bias between BE_Cl and SBE in the subgroup that had normal metabolic values except Cl was less than that of the correction approach (Fig 2), and b) BE_Cl was significantly related to the SBE in all patients, whereas the correction and Cl/Na ratio approaches were not (Table 5). In other words, BE_Cl was the only parameter which determines Cl effect on SBE.

Additionally, these results compel us to discuss whether there are normal value ranges for Na and Cl or not. Although normal values of Na and Cl for physiologic events in the body exist, it appears that they are irrelevant while evaluating the acid-base status if BE_Cl is the most reliable approach to evaluate Cl effect.