Hyperkalemia in Two Patients with Diabetes Mellitus and Chronic Kidney Disease and the Role of Disrupted Internal Potassium Balance

To identify factors associated with the development of hyperkalemia in patients with chronic kidney disease (CKD), we analyzed conditions present during episodes of hyperkalemia in two patients with insulin-dependent diabetes mellitus who had elevated serum potassium concentration ([K]) in ≥ 20% of the blood samples during both the pre-hemodialysis and the hemodialysis period. In both patients, conditions causing derangements in the internal potassium balance (exchanges of potassium between the intracellular and the extracellular compartment), including severe hyperglycemia (serum glucose concentration ≥ 400 mg/dL) and catabolic illnesses, were present in ≥ 75% of the instances of simple hyperkalemia ([K] ≥ 5.1 mmol/L) and almost all of the instances of severe hyperkalemia ([K] ≥ 6.0 mmol/L) during both the pre-hemodialysis and the hemodialysis periods. Derangements of the internal potassium balance, many of which are potentially preventable, can be a major cause of hyperkalemia in patients with CKD before or after starting chronic hemodialysis. Careful analysis of the conditions associated with and potentially causing hyperkalemia in CKD patients is imperative for both treatment and prevention purposes. Categories: Endocrinology/Diabetes/Metabolism, Internal Medicine


Introduction
Hyperkalemia, even at modest levels, is associated with adverse outcomes, including shortened survival, in the general population [1].Evaluation of the pathogenetic mechanism(s) of each episode of hyperkalemia is required for both treatment and prevention.The concentration of potassium in serum ([K]) is determined by the interplay of two balances, the external balance (intake and output of potassium) and the internal balance (potassium exchanges between the intracellular and extracellular compartments).Either or both mechanisms may be at fault in each episode of hyperkalemia [2].
The frequency of hyperkalemia is high in patients with chronic kidney disease (CKD) [3] and even higher in those on chronic hemodialysis [4].Defects in both the external and internal potassium balance have been documented in CKD.Patients with CKD have limited capacity to excrete potassium loads [5].Advanced CKD leads also to defective uptake of potassium by cells as a consequence of uremia [6].We present two patients with insulin-dependent diabetes mellitus and CKD who had a high frequency of hyperkalemia (in ≥ 20% of the serum samples) in both the pre-hemodialysis (pre-HD) and the hemodialysis (HD) periods.The purpose of this report was to investigate mechanisms of hyperkalemia in these patients and to compare the pre-HD and HD periods.

Case Presentation Patients
Both patients were men with insulin-dependent diabetes mellitus.Table 1 shows age at initiation of dialysis, duration of follow-up pre-HD and during the HD period, and annual hospitalization rates, plus average days of hospitalization per annum.The table also shows the 95% confidence interval (CI) of the corresponding values for a control group comprised of all other diabetic men treated in the same dialysis unit over the same period [7].The control group was used only for comparison of morbidity with the two patients presented.Hospitalization was used as the measure of morbidity.The ages of the two patients at initiation of hemodialysis and the duration of follow-up in the pre-HD and HD periods were within the corresponding 95% CIs of the control group.The rates and lengths of hospitalizations exceeded the corresponding upper limits of the 95% CIs intervals of the control group in both the pre-HD and HD periods in both patients.The course of both patients during the pre-HD and HD periods was characterized by repeated episodes of severe hyperglycemia and episodes of catabolic illness.In the first patient, catabolic illnesses included left leg gangrene leading to high below-the-knee amputation with prolonged postoperative sepsis, sepsis secondary to an infected dialysis catheter, and sepsis associated with severe pneumonia.The second patient had repeated episodes of deep abscesses secondary to self-injection of illegal drugs, large muscle hematoma secondary to trauma, infectious endocarditis, and right femoral neck fracture with operative correction.Patient 1 died suddenly at home.Patient 2 expired two months after the repair of the right hip fracture after withdrawing all treatments.

Methods
The following additional information was collected for each patient in both the pre-HD and HD periods: (d) whether the laboratory tests were collected in an outpatient setting or during hospitalization, whether episodes of hyperglycemia were associated with ketoacidosis, and whether laboratory tests obtained after a test showing severe hyperglycemia were collected during treatment of this hyperglycemic episode with continuous insulin infusion.

Statistical methods
Continuous variables are presented as mean ± standard deviation.Initial statistical analysis addressed the pre-HD and HD periods combined.The relationship between parametric variables and [K] was evaluated by correlational analysis.For categorical variables, [K] values obtained while the variable was present were compared by unpaired student's t-test to [K] values obtained when the variable was absent.Hyperglycemia and the presence of catabolic illness were the two most prominent conditions associated with hyperkalemia in both patients.
[Glu] levels ≥ 400 mg/dL are reported as critical by the laboratory of this hospital and were considered "severe" hyperglycemia for this report.Serum samples with [Glu] < 400 mg/dL obtained in periods when catabolic illness was absent were labeled as control samples.In the pre-HD and HD periods separately, [K] levels were compared between hyperglycemic and control samples in two ways: In the first comparison, the control sample (Control Stepwise, multiple linear regression was performed in each patient to identify predictors of [K].A forward and backward stepping procedure was used with a P-value < 0.05 to enter and > 0.05 to remove a variable.The candidate variables included the period of observation (pre-HD vs. HD), the presence or absence of ketoacidosis, whether the blood sample was obtained during hospitalization or on an outpatient basis, whether the blood test was obtained during continuous infusion of insulin during treatment of severe hyperglycemia, presence or absence of catabolic illness, intake of drugs affecting potassium balance, and parametric variables, such as [Glu], SUN, tonicity, and [Na] Corr .All serum samples were included in this analysis.

Results
The first patient had, during the pre-HD period, 223 measurements of serum chemistries.Seventy-nine of these measurements (35.3%) exhibited hyperkalemia ([K] range 5. Tables 2-7 show comparisons of laboratory tests between episodes of hyperglycemia and control states (Tables 2-5) and between episodes of catabolic illness and control states (Tables 6, 7).The control states of absence of severe hyperglycemia and catabolic illness in the same patients (not in the control group of patients used for comparison of morbidity in Table 1) were defined in the Methods subsection.The bodies of Tables 2-7 show the comparison of the mean values, while the captions of the Tables show the number (percent) of the [K] measurements that were in the hyperkalemic range in each column.Tables 2, 3 compare biochemical values between the episodes of non-ketotic hyperglycemia and the two control states in Patients 1 and 2, respectively.Tables 4, 5 show the same values for episodes of diabetic ketoacidosis.Average [K] values were significantly higher at hyperglycemia than in either control stage in the pre-HD and HD periods in both patients.Substantially higher percentages of [K] levels were in the hyperkalemic range at hyperglycemia than in the control states.[K] values were significantly higher during acting catabolic illness than during the control periods in the pre-HD period in Patient 1 (Table 6) and during the HD period in Patient 2 (Table 7) .The slope of the regressions was exactly the same (0.002 mmol/L rise in [K] per 1 mg/dL rise in [Glu]) for all four regressions.Standard regression diagnostics showed that the assumptions of linearity, homoscedasticity, and normality were justified for all four regressions.

Pre-HD
Tables 8, 9 show the predictors of [K] by multiple linear regression.In Patient 1, high values of [Glu] and SUN, catabolic illness, and use of ACE inhibitors or ARBs were independent predictors of higher [K] values, while the pre-HD period, the period of continuous insulin infusion for treatment of severe hyperglycemia and periods of hospitalization were independent predictors of lower [K] values.For Patient 2, high values of [Glu] and SUN and catabolic illness were independent predictors of higher [K] values, while the pre-HD period, the period of continuous insulin infusion for treatment of severe hyperglycemia and use of sodium polystyrene sulfonate and diuretics were independent predictors of lower [K] values.When several other biochemical parameters were added to the list of candidate variables, multiple linear regression identified the following additional predictors of [K]: For Patient 1 [TCO 2 ] (p < 0.001) and anion gap (p < 0.001); and for Patient 2 [TCO 2 ] (p < 0.001) and [Na] Corr (p = 0.009).The coefficient assigned to [Glu] was 0.002 in all multiple linear regression models.
Figure 1 shows the percentages of hyperkalemic episodes associated with hyperglycemia, catabolic illness, and other causes.Pre-HD and HD periods were analyzed separately in each patient.Hyperglycemia and catabolic illness combined accounted for 76.2% to 90.0% of the hyperkalemic episodes.For severe hyperkalemia, hyperglycemia and catabolic illnesses accounted for 84.2% of the hyperkalemic episodes in the pre-HD period of Patient 1 and for 100% of the hyperkalemic episodes in the HD period of Patient 1 and the pre-HD and HD periods of Patient 2.

Discussion
The main finding of this report was that the great majority of the episodes of hyperkalemia in both patients were associated with severe hyperglycemia or active catabolic illness.These mechanisms of hyperkalemia were noted during both the pre-HD and HD periods.The hyperkalemic effects of hyperglycemia and catabolic illness are consequences of disturbances in the internal potassium balance.The exit of potassium from cells is reversible by administration of insulin in hyperglycemic hyperkalemia and irreversible when catabolic illness causes cell death.
The hyperglycemic egress of potassium from cells is caused by two interdependent mechanisms.The fundamental hyperkalemic mechanism is absence of insulin [11][12][13][14].Hyperglycemia secondary to insulin deficit causes also hypertonicity secondary to extracellular solute (glucose) gain [14].Hypertonicity causes exit of potassium from cells [15].Hypertonicity secondary to hyperglycemia is aggravated in patients with substantial renal function by excessive loss of water through osmotic diuresis [14,[16][17]].Katz's coefficient [9] provides an estimate of the fraction of hypertonicity that is due to extracellular glucose gain.This coefficient estimates that serum tonicity increases by 2.4 mmol/L for each 100 mg/dL rise in [Glu] [14].The corrected serum sodium concentration [10] uses Katz's coefficient to compute an estimate of the relative loss of water through osmotic diuresis [14].Osmotic diuresis causes, in addition to excessive water loss, large losses of potassium that counterbalance the exit of potassium from the cells.In the face of large potassium losses in patients with preserved renal function and severe hyperglycemia, [K] may be elevated, in the normal range, or even low [17].
Severe hyperglycemia developing in patients with end-stage renal disease (ESRD) is not associated with large water loss through osmotic diuresis.Consequently, the degree of hypertonicity in dialysis-associated hyperglycemia is predicted by Katz's formula with remarkable accuracy [18].Urinary potassium losses are also limited, and hyperkalemia is frequent in hyperglycemic episodes occurring in this setting [19][20].The degree of hyperglycemic hyperkalemia in dialysis patients is higher when ketoacidosis is present [19,21].Ketoacidosis contributes to the development of hyperglycemic hyperkalemia by affecting the function of several transport pathways in the cell membrane [22].Insulin is the only required treatment for both hyperglycemia and hyperkalemia in patients with ESRD, although additional measures may be necessary for extreme cases of hyperkalemia [19][20]23].
Hyperglycemia was frequent and caused hyperkalemia in both patients of this report during both the pre-HD and HD periods.We addressed the question whether the relationship between [Glu] and [K] differs between the pre-HD and HD periods.There is evidence suggesting that the regulation of internal potassium balance and the effects of insulin on certain aspects of transport across cell membranes differ between the pre-HD and HD periods; therefore, it is possible that the relation between [Glu] and [K] may also differ between these periods.A synopsis of this evidence is presented below.
Investigations of the disruption of cellular functions in uremia suggest that uremic abnormalities of the function of the sodium-potassium ATPase (Na/K-ATPase) of the cell membranes is the root of the uremic defect in cellular potassium uptake.Uremia (circulating uremic factor(s)) was shown to cause a decrease in the activity of the cellular membrane Na/K-ATPase first in red cells [24] and subsequently in several other cell types and tissues, including leukocytes, sarcolemma, and intestines [25][26].A class of digitalis-like compounds, labelled collectively as endogenous cardiotonic steroids (CTS), has been linked to the uremic inhibition of cell membrane Na/K-ATPase activity.The decrease in this activity leads to decreases in cytosolic potassium [27].CTS, which have pleiotropic actions, are either cardenolides (endogenous ouabain) or bufadienolides (telocinobufagin, marinobufagenin).In a recent report, marinobufagenin (MBG) was found to be an important inhibitor of the cell membrane Na/K ATPase in uremia [28].Plasma MBG levels were elevated in both patients with renal failure and experimental animals with renal failure following partial nephrectomy.The elevated plasma MBG levels were associated with decreases in cell membrane Na/K-ATPase activity, which returned toward normal levels after the addition of MBG-binding antibodies [28].Whether additional molecular mechanisms contribute also to the uremic defect of the internal potassium balance is not clear.
A hemodialysis dose leading to correction of clinical uremic manifestations also corrects the uremic defect in cellular potassium uptake [29].The effect of dialysis on plasma MBG levels has not been reported thus far, to our knowledge.Uremia also causes peripheral tissue resistance to the hypoglycemic effect of insulin [30], while the effect of insulin on cellular uptake of potassium is unchanged [31] or enhanced [32] in uremia.
No difference in the relationship between [K] and [Glu] was found between the pre-HD and HD periods in this report.All four linear regressions of [K] on [Glu] found the same regression coefficient (0.002).The regression equations compute that [K] increased by 0.2 mmol/L for each 100 mg/dL rise in [Glu] in both the pre-HD and HD periods in both patients.It is of note that, in addition to the simple regressions, all four multiple regressions assigned the same coefficient (0.002) to the relationship between [Glu] and [K].The effect of hyperglycemia on [K] was uniform in both patients throughout the observed course of CKD.This finding was robust.However, whether this finding is applicable in general to patients with diabetes mellitus and CKD will need studies in large cohorts of patients.Prevention of severe hyperglycemia in patients who exhibit a hyperkalemic effect of hyperglycemia similar to that found for the two patients in this report will also prevent, to a large extent, severe hyperkalemia.
The highest [K] and [Glu] values were observed in both patients during episodes of diabetic ketoacidosis.Although ketoacidosis was not identified as a predictor of [K], [TCO 2 ] and serum anion gap, both of which showed their largest deviations from normal values in episodes of ketoacidosis (Tables 2-5), had a significant effect on [K] by multivariable analysis.In addition, [K] values corresponding to the mean [Glu] values calculated from the linear regressions differed minimally from the corresponding actual mean [K] values for the episodes of non-ketotic hyperglycemia (Tables 2-3), but were lower, by 0.36 to 1.15 mmol/L, than the corresponding actual mean [K] values for the episodes of ketoacidosis (Tables 4-5).These findings are consistent with previous findings supporting an added hyperkalemic effect of ketoacidosis in CKD patients [19,21].
Although the magnitude of hyperkalemia was less in episodes of active catabolic illness than in hyperglycemic episodes, multivariate analysis identified catabolic illness as a predictor of hyperkalemia.Combined, hyperglycemia, and catabolic illness accounted for the majority of the episodes of hyperkalemia (Figure 1) and almost all the episodes of severe hyperkalemia in both patients.

Conclusions
In addition to bringing to the forefront derangements of internal potassium balance as a cause of hyperkalemia in CKD, the findings of this report also suggest that at least several of these derangements are preventable.The main limitation of the report is that it presents only two patients.The mechanisms of hyperkalemia in CKD patients are multiple and often involve the external potassium balance.Studies in large numbers of patients on chronic dialysis are needed to identify the principal causes of hyperkalemia in dialysis populations.The purpose of this report was to illustrate a method of detailed analysis of all the factors that can potentially cause hyperkalemia in individuals with CKD, including factors related to the internal potassium and external balance.Such an analysis should guide individualization of the preventive measures.

TABLE 2 : Serum concentrations at nonketotic hyperglycemia and control states. Patient 1
The highest [K] values were observed in episodes of diabetic ketoacidosis.

TABLE 8 : Predictors of serum potassium concentration. Multiple linear regression in Patient 1
ACE = angiotensin converting enzyme.ARB = angiotensin receptor blocker.