Nephrogenic diabetes insipidus is observed in 40–50 % of patients prescribed lithium. This reduced ability to concentrate the urine is correlated to the duration of lithium exposure (Bedford et al. 2008). However, in healthy volunteers, the ability to concentrate the urine was already reduced after just 1 month of treatment (Walker et al. 2005). Nephrogenic diabetes insipidus is characterised by a polyuro-polydipsic syndrome, which may be abundant—up to 10 l per day. Nephrogenic diabetes insipidus may occur less frequently with once-a-day dosing (Ljubicic et al. 2008). In patients, it usually develops early, after 8 weeks of treatment (Boton et al. 1987). When treatment is stopped, symptoms diminish slowly over 8 weeks (Bucht and Wahlin 1980). However, it can be irreversible, especially if lithium has been prescribed for more than 15 years (Bendz et al. 1996a).

Lithium-induced nephrogenic diabetes insipidus is caused by a negative regulation and a decrease of trafficking towards the membrane of aquaporin-2 channels (AQP2) in the principal cells of the collecting duct. AQP3 channels are also negatively regulated by lithium (Fig. 16.1) (Kwon et al. 2000). The ENaC channel is also dysregulated. Indeed, lithium enters into the principal cells of the collecting duct through the ENaC channel and is not able to exit out of the cell. Once in the cell, the lithium inhibits adenylate cyclase activity and cyclic adenosine monophosphate 3–5′(AMPc) production. Regulation of the adenylate cyclase also depends on glycogen synthase kinase 3β (GSK3β) (Kuure et al. 2007). Other key enzymes are regulated by lithium: the phosphorylation of Akt leads to the inhibition of GSK3β via a mechanism of phosphorylation (Nielsen et al. 2008). The inhibition of GSK3β is associated with an accumulation of β-catenin in the principal cells. This protein has a role in cellular adhesion and plays a role in the transcriptional coregulation of the genes of cellular growth. This can induce the high proliferation rate of principal cells observed with lithium (Christensen et al. 2006).

Medullary interstitial cells are also involved in lithium-induced nephrogenic diabetes insipidus. In these cells, lithium inhibits GSK3β and increases the expression of COX-2 and PGE-2, which also regulates the AQP2 (Rao et al. 2005).

In patients, Bedford and colleagues have shown a decrease of urinary excretion of AMPc and AQP2 that was correlated with the duration of lithium exposure (Bedford et al. 2008). The use of amiloride has been proposed for cases of severe nephrogenic diabetes insipidus: it inhibits the ENaC channel and thus the entry of lithium into the principal cells of the collecting duct (Batlle et al. 1985). In a cross-over study, Bedford and colleagues treated patients for 6 weeks and observed an increase in urinary osmolality (Bedford et al. 2008). In that study, lithemia remained stable, but kalemia and creatininemia should be monitored carefully. However, polyuro-polydipsic syndrome is usually well tolerated and does not require any treatment.

Even when serum lithium concentrations are maintained within the therapeutic range, the glomerular filtration rate (GFR) falls slightly in about 20 % of patients (Grandjean and Aubry 2009). In the early 1990s, it was reported that there was no evidence that patients taking lithium were at risk of progressive renal failure (Groleau 1994). Longitudinal studies showed a progressive decline in GFR over time that was considered not to exceed that seen with normal ageing (Gitlin 1999). However, most authors now disagree with that view (Grandjean and Aubry 2009). Nephrologists increasingly report patients with renal insufficiency and no other medical history than lithium therapy (Presne et al. 2003). In a laboratory database study, a very high percentage of lithium-treated outpatients had an eGFR below 60 mL/min/L.73 m², ranging from 39 % in the 20–39 year age group to 85 % in patients aged over 70 years (Bassilios et al. 2008). This was much higher than in patients from the general population. Recently, Minay and colleagues also compared eGFR in patients treated with lithium to a control group and showed that eGFR was below 60 mL/min/L.73 m in 17 % of patients, which was significantly more than in controls (13 %) (Minay et al. 2013). Therefore, the suggestion that only a few patients receiving long-term lithium are at increased risk of progressive renal insufficiency should be viewed with caution. It is now unequivocally established that long-term lithium administration may induce chronic tubulointerstitial nephropathy leading to renal failure, even in the absence of episodes of lithium intoxication (Hestbech et al. 1977; Markowitz et al. 2000).

Several studies have demonstrated the impact that time spent on lithium treatment has on the prevalence of lithium-induced nephropathy (Bendz 1983; Bendz et al. 1994, 2010; Presne et al. 2003; Bassilios et al. 2008). The maximal number of new cases of renal failure was observed after 16–20 years of treatment in a cohort study including 114 patients treated with lithium (Lepkifker et al. 2004). Presne and colleagues also showed a correlation between the estimated glomerular filtration rate and the duration of treatment with lithium (Presne et al. 2003). Since bipolar disorder often begins at a relatively young age, renal failure may develop by 40 years of age. Two multicentre studies have investigated the prevalence of lithium-induced nephropathy (Bendz 1983; Bendz et al. 1994) and showed that, after a mean treatment duration of 6.5 years, only 4 % of patients had elevated serum creatinine levels, whereas after 19 years of exposure, this proportion increased to 12 %. Recently, the same group established that the prevalence of chronic kidney disease (defined as plasma creatinine over 150 μmol/L) in the general lithium-treated population was about 1.2 % (excluding patients on renal replacement therapy) (Bendz et al. 2010). Time on lithium was the only identified risk factor in that study. Similarly, in the cohort described by Lepkifer and colleagues, mean duration of lithium treatment was 16.8 years and 21 % of patients had an abnormal creatininemia, defined by a creatininemia above 133 μmol/L (Lepkifker et al. 2004).

A recent meta-analysis of case-control studies compared the eGFR of lithium-treated patients and controls and showed a reduction in eGFR of 5 mL/min per year in lithium-treated patients. However, the very short follow-up periods of these case-control studies does not fully allow for thorough assessment of lithium nephrotoxicity, which is a much longer term concern (McKnight et al. 2012).

The mean annual loss of eGFR is slow, but lithium-induced nephropathy can progress to end-stage renal disease over several decades (Markowitz et al. 2000; Presne et al. 2003; Bendz et al. 2010). The prevalence of end-stage renal disease was sixfold higher in lithium-treated patients than in the general population in the study by Bendz and colleagues: 18 of 3,369 patients (0.5 %) reached end-stage renal disease (Bendz et al. 2010). Patients were on dialysis at a mean age of 46 years, after a mean duration of treatment of 23 years, and more than half of them had stopped lithium treatment for more than 10 years. Among French dialysis patients, the prevalence of lithium nephropathy was estimated at 0.22 % (Presne et al. 2003). In Australian and New Zealand registries, the prevalence of end-stage renal disease due to lithium nephropathy was estimated at between 0.2 % and 0.7 % in 2000–2003. In the French cohort, the only risk factor of end-stage renal disease was initial eGFR. The risk was similar between patients who had stopped lithium and those who had not. A delay of 20 years was observed between the beginning of treatment and dialysis (Presne et al. 2003). However, patients who received more than 750 mg/day were three times more likely to experience an annual eGFR decline above 2 mL/min than those who received a lower dose. Similarly, eGFR inversely correlated with the mean serum lithium concentration (Presne et al. 2003). Furthermore, in those patients who underwent renal biopsy in that study, the degree of interstitial fibrosis was related to the cumulative dose of lithium salt.

Lithium-induced nephropathy is an asymptomatic tubulointerstitial nephropathy, without hypertension or proteinuria. There is often an associated defect in the ability to concentrate urine. The main risk factor is the length of lithium exposure. Some comorbidities, such as diabetes, may also play a role in the progression of renal failure. Diagnosis is based on the association of long-term lithium treatment (more than 10 years) and a chronic tubulointerstitial nephropathy, without any other aetiology. Tubular cysts may be observed.

The presence of tubular cysts on imaging, magnetic resonance imaging or ultrasound, in normal-size kidneys, is highly characteristic of lithium toxicity (Alexander et al. 2008) and observed in 33–62 % of cases (Markowitz et al. 2000; Farres et al. 2003). Renal calcifications may be observed in cases of hyperparathyroidism. A renal ultrasound should be performed in patients with renal failure or those who have been treated with lithium for more than 10 years. Even if renal function is normal, the presence of characteristic lesions should lead to a reduction of lithium dose if possible. A renal biopsy may be indicated if the diagnosis is uncertain and will show a chronic tubulointerstitial nephropathy pattern. Interstitial fibrosis may be present early, even 5 years after initiation of treatment (Presne et al. 2003). There are usually no glomerular lesions. Clinical evolution is usually slow with a mean decrease of eGFR of 2.2 mL/min/year.

Long-term lithium treatment is also associated with hypercalcaemia and hyperparathyroidism. When examining the end point of isolated hypercalcaemia, the reported prevalence among those treated with lithium varies greatly, from 10 % up to 42 % (Bendz et al. 1996b; Presne et al. 2003; Grandjean and Aubry 2009). However, the prevalence of hyperparathyroidism in chronic lithium users (more than 10 years) has been estimated at 10–15 % in retrospective case series (Hundley et al. 2005). The prevalence of hyperparathyroidism is 7.5 times higher in patients with lithium for more than 15 years than it is in the general population (Bendz et al. 1996b). ‘False’ hypercalcaemia due to plasma volume depletion resulting from nephrogenic diabetes insipidus should be excluded in such individuals (Grunfeld and Rossier 2009). These data support a recommendation to routinely follow a serum calcium level in those patients treated with lithium therapy (Livingstone and Rampes 2006). Some cases of lithium-associated hyperparathyroidism are associated with nephrocalcinosis, nephrolithiasis and/or osteoporosis (Dwight et al. 2002; Awad et al. 2003; Hundley et al. 2005; Carchman et al. 2008).

A proposed mechanism is an alteration in the calcium-sensing receptor, resulting in a shift to the right of the calcium–parathyroid hormone (PTH) response curve (Livingstone and Rampes 2006). This then reduces suppression of serum PTH due to an increased threshold for calcium levels (Brown 1981; Riccardi and Gamba 1999). However, it is not entirely understood whether lithium causes hyperparathyroidism directly or somehow potentiates hyperparathyroidism in patients with early parathyroid dysregulation. One possible effect of lithium on parathyroid tissue may be promotion of the growth of pre-existing abnormal tissue (Saxe and Gibson 1991; Saxe et al. 1995).

Complete resolution of hyperparathyroidism may thus require total or partial parathyroidectomy (Bendz et al. 1996b; Awad et al. 2003). Ablation of a single parathyroid adenoma usually leads to normocalcaemia, even in patients who continue to receive lithium (Awad et al. 2003; Hundley et al. 2005; Grunfeld and Rossier 2009). In contrast, surgical treatment of multiglandular disease is technically difficult and hazardous (Nemeth et al. 1998). A conservative approach to the therapy of hypercalcaemia may also be appropriate. Cinacalcet hydrochloride is an allosteric activator of the calcium-sensing receptor present in chief cells of the parathyroid glands (Quarles et al. 2003; Peacock et al. 2005; Wuthrich et al. 2007). This calcimimetic action lowers the threshold for activation of the calcium-sensing receptor by extracellular calcium and decreased PTH secretion. Activation of the calcium-sensing receptor by calcimimetics could thus specifically antagonise the effects of lithium on the parathyroid glands (Sloand and Shelly 2006; Gregoor and de Jong 2007; Szalat et al. 2009).

Some acquired cystic kidney disease and other toxic tubulointerstitial nephropathies have been associated with an increased occurrence of renal and urothelial carcinomas, and some cases of renal tumours have been described in lithium-treated patients (Markowitz et al. 2000; Kjaersgaard et al. 2012). We recently studied a cohort of 170 lithium-treated patients and identified 14 patients (8.2 %) who developed renal solid tumours, including seven cases of malignant and seven of benign tumours (Zaidan et al. 2014a). The mean duration of lithium exposure at diagnosis was 21.4 ± 10.3 years. The renal cancers included three clear cell renal cell carcinomas (RCC), two papillary RCC, one hybrid tumour and one rare tumour (clear cell carcinoma with leiomyomatous stroma). The benign tumours included four oncocytomas, one mixed epithelial and stromal tumour and two angiomyolipomas. The percentage of renal tumours, particularly cancers and oncocytomas, was significantly higher in lithium-treated patients compared to 340 sex-, age- and eGFR-matched lithium-free patients. The standardised incidence ratio of renal cancer was also significantly higher in lithium-treated patients compared to the general population: 7.51 (95 % CI 1.51–21.95) and 13.69 (95 % CI 3.68–35.06) in men and women, respectively. These results indicated an increased risk of renal tumours in lithium-treated patients (Zaidan et al. 2014a). Collectively, we reported an association and not a causal effect (Licht et al. 2014; Zaidan et al. 2014b). Our work did not question the benefit of lithium in the management of bipolar disorders but should alert nephrologists and psychiatrists to the possibility of late tumour development, which should encourage regular renal screening of patients until further studies are conducted.

Accumulation of lithium in cells of the distal nephron and the early collecting duct via ENaC could account for the chronic nephrotoxic effects. Increased inhibitory phosphorylation of GSK3β leads to the activation of Wnt/β-catenin signalling, a critical pathway for progression of cystic kidney diseases (Sinha et al. 2005; Nielsen et al. 2008; Kjaersgaard et al. 2012). Altogether, lithium seems to interfere with multiple critical signalling pathways that regulate tubular cell proliferation, differentiation and apoptosis, any of which may explain the onset of tubular dilatations, cysts, subsequent renal damage and eventually tumours (Sinha et al. 2005; Nielsen et al. 2008; Kjaersgaard et al. 2012).