Lower estimated glomerular filtration rates in patients on long term lithium: a comparative study and a meta-analysis of literature

This was a comparative cross sectional study carried out by the University Medical Unit and the University Psychiatry Unit of the National Hospital of Sri Lanka. All consenting adult patients (18 years and above) who were on lithium and followed up at the University Psychiatry Unit as inpatients or outpatients were included in the study, and their data was obtained from the clinical records. The latest and previous creatinine values of patients were recorded, and eGFR values were calculated using the MDRD (Modification of Diet in Renal Disease) formula [7]. We also extracted data of patients with bipolar affective disorder who were not on lithium. These two groups were defined as below.

Group A: Patients with ICD-10 (International Classification of Diseases) clinical diagnosis of bipolar affective disorder who had been treated with lithium for a minimum period of one year. This group was further divided into two subgroups depending on whether the patients had diabetes or hypertension as a co-morbidity (subgroup A1: lithium users without diabetes or hypertension, subgroup A2: lithium users with diabetes, hypertension or both).

Group B: Patients with ICD-10 clinical diagnosis of bipolar affective disorder but not treated with lithium within the last 12 months. These patients were on treatment with other mood stabilizers.

Matching was done on a 1:1 basis for patients in subgroups A1 and A2 (group B was too small to carry out any comparisons). Patients from the University Medical Unit, who were admitted to wards or followed up in clinics due to a non-renal illness or for evaluation purposes (and therefore had a serum creatinine value recorded) were recruited as controls. The clinics of the University Medical Unit have a population of approximately 1500 patients followed up as outpatients. The selection of a control was done by checking each registered patient on a randomly selected clinic date according to the matching criteria detailed below.

Controls for subgroup A1: Patients without a) a psychiatric condition (and not on lithium), b) diabetes or hypertension and c) any other comorbidity known to effect the eGFR were matched to each of the A1 patients with respect to age and gender on a 1:1 basis.

Controls for subgroup A2: Patients without a psychiatric condition (and not on lithium) but having either a) diabetes, b) hypertension or c) both were matched for each of the group A2 patients (1:1 match) with regard to the age, gender and the type of co morbidity. For example, for a lithium-using 56 year old male with diabetes and hypertension, a non lithium-using 56 year old male with diabetes and hypertension was selected as the control.

Data pertaining to following aspects were collected from the recruits using a pre-tested interviewer administered data sheet: demography, past medical history, past psychiatric history, treatment history of lithium, complications of lithium therapy and serial creatinine values while on lithium.

The following patient categories were excluded from this study: a) patients on lithium diagnosed with renal disease or CKD due to another aetiology, b) non-consenting patients, c) patients with acute kidney injury, d) patients with an alternative cause for a high serum creatinine (e.g. myositis), e) patients with poor compliance with lithium and f) extremes of body mass index (18 < and > 30 kg/m²).

Ethical clearance for the study was granted by the Ethics Review Committee of the National Hospital of Sri Lanka.

eGFR values were compared between the groups using mean differences. Significance of mean differences in eGFR between groups was calculated with the Independent t test. Statistical significance was set at p < 0.05. Descriptive statistics were summarized into proportions and averages based on the scales of measurement. SPSS statistical software (version 15) was used for data analysis. Linear regression was used to correlate two continuous variables.

A recent meta-analysis has compared the eGFR values between lithium users and controls from studies published up to 2010 [8]. We built on that meta-analysis. We searched EMBASE, SCOPUS and PUBMED for publications between January 2011 – October 2013 which had the keywords “Lithium” and “glomerular filtration rate” or “GFR” or “eGFR” in the abstract. This time period was not covered by the previous meta-analysis. Three authors independently carried out the search and the data was filtered using Endnote X4 software (Thomson Reuters, Carlsbad, CA 92011, USA). Relevant studies published after 2010, and the data from our study were added to the data of studies identified in the previously published review and a new independent meta-analysis was carried out using the Revman (version 5) software [9].

Forty seven patients were recruited to group A and six patients to group B. Of the group A patients, 24 belonged to subgroup A1 and 23 to subgroup A2. Of the patients in subgroup A2, 11 had diabetes, six had hypertension and six patients had both co-morbidities. The characteristics of group A and its controls are given in Table 1. Since the numbers were small in group B, they were used for an overall comparison of eGFR only (Table 2).

Group A had a higher number of patients with hypothyroidism which is a known complication of long term lithium therapy (seven patients among lithium users vs. one patient in the control group). The mean duration of diabetes and hypertension were greater among controls than the cases. Comparing the two subgroups (A1, A2) of lithium users, those with comorbidities (A2) were older and had been on lithium for a longer duration (Table 1). None of the lithium users had a recorded history of any other lithium related renal complication such as diabetes insipidus, acute kidney injury or renal tubular acidosis.

The patients on lithium (Group A, n = 47) had a mean eGFR value of 79.71 ml/min/1.73 m² (SD ± 22.6, median: 79.9, interquartile range: 70.5 – 94.3). The age and gender matched controls (n =47) had a mean eGFR value of 89.31 ml/min/1.73 m²(SD ± 21.3, median: 90.6, interquartile range: 77.2 – 105.7). This difference was statistically significant (t- 2.12, df – 92, p = 0.04). Serial data on creatinine values were available for only 34 patients among the lithium users and the mean rate of decline of eGFR per year in that subgroup was -2.7 ml/min/1.73 m² (SD ± 9.8). The comparisons of mean eGFR of the sub groups are summarized in Table 2. Notably, impairment in eGFR remains statistically significant for the comparison between subgroup A1 and their controls. Correlation of eGFR values with duration of lithium therapy by linear regression (controlling for age and gender) showed a negative trend which was not statistically significant (beta = -0.07, B = -0.16, p = 0.66).

The previously published review by McKnight et al. [8] had considered six studies in their paper [10‐15]. We traced the original papers of these studies and re-entered the data for a new meta-analysis. Our search for the period from 2011 January to 2013 October recovered only one study with usable data [6]. There was another suitable study published in 2010 which had not been included in the earlier meta-analysis [16]. These data plus the results of our study were considered for the new meta-analysis. For five out of six studies of the earlier meta-analysis, the GFR values extracted by us were the same. However, in the remaining study (by Turan et al. [15]) we noted that the original authors had considered three groups of patients namely; lithium naïve patients, short term lithium users (less than three years) and long term lithium users (greater than 3 years). It seems the authors of the review had opted to include the GFR of short term lithium users for comparison. We felt that given the duration of lithium use in our study and other studies included in the meta-analysis, it was more appropriate to take the GFR values of the long term user group. This changed the direction of overall effect for that study compared to what was published in the previous meta-analysis [15] (See below). The results of our meta-analysis confirmed that lithium users had significantly lower GFR values compared to non-using controls; (9 studies, mean difference -10.3 ml/min, 95% CI: -15.13 to -5.55, 471 cases, 337 controls, p < 0.0001) (Figure 1). A summary of studies included in the meta-analysis is given in Table 3.

The numbers in our lithium using patient sample was limited by the fact that we excluded patients with poor compliance (patient reported compliance). We did not measure the serum lithium level of our recruited patients as a single current lithium level does not reflect compliance over many years. Unfortunately, facilities to measure serum lithium is not regularly available to patients free of charge and therefore such measurements had not been done for us to have an idea about patient compliance over time. The state-run clinics like ours provide free health care to patients, but investigations like serum lithium level are not available all the time. Most patients seeking health care in state-run cannot afford to get serum lithium levels tested from private sector laboratories. Several patients in the lithium-using category did not have periodic serial serum creatinine values in clinic records. Such patients were excluded from the analysis of the rate of decline in eGFR. Current guidelines suggest that eGFR alone is an inadequate measure to define at risk groups for CKD, and albuminuria also has to be taken in to account [19]. This measure was not compared between cases and controls in our study. Therefore, we have not classified patients and controls according to CKD risk categories or stages.