The main reason to undertake this study was the perceived need for better quantifying the presumptive long-term renal risks associated with lithium treatment, in terms that would make sense for a patient. In Sweden, healthcare legislation requires healthcare professionals to give individually adapted information about (among others) risks and benefits of available treatments, enabling shared decision-making. Lithium is perceived by a number of physicians and researchers as being underused (Zivanovic 2017), and this is a result of several factors: clinicians’ reluctance to initiate lithium treatment, treatment termination due to side effects (Öhlund et al. 2018), but also patients’ refusal to accept lithium. A recent survey exploring clinicians' attitudes toward lithium use in bipolar disorders (Hidalgo-Mazzei et al. 2023) reported that over 70% of respondents (886 clinicians in 43 countries) considered lithium as their first choice for maintenance medication in bipolar disorder. However, 55% of them expressed concerns about renal function alterations. The primary reasons for clinicians' reluctance to prescribe lithium were patients' negative beliefs or attitudes toward lithium, followed by concerns about long-term side effects and safety.

A noticeable discrepancy exists between the statements of surveyed physicians and real-world practices. Data from a large multinational bipolar cohort of 10,351 patients across North America, Europe and Australia revealed that lithium was prescribed to only 29% of the patients (Singh et al. 2023). These findings reinforce the concerns about the widespread under-prescription of lithium (Zivanovic 2017; Malhi et al. 2023). Unclear or difficult to comprehend information about potential risks may lead to an exaggerated negative perception of lithium, both among clinicians and patients. Meaningful patient information about the long-term renal effects of lithium should ideally include (among others): what is the risk for severe renal impairment for the actual patient and what is the time perspective considered? What are the possible consequences? What are the benefits of lithium? What are the alternatives?

In this respect, one of the study aims was to estimate cumulative incidence and lifetime risk for CKD4 + in lithium-treated patients in a way that would allow an easy-to-understand comparison with the general population. The lifetime risk is expressed in relation to a defined outcome-free start age and requires a follow-up time long enough to cover the age that may reasonably approximate life expectancy. In Sweden, the current life expectancy is nearly 85 years for women and just over 81 for men (Statistics Sweden 2022), hence 90 years of age was considered an appropriate lifetime horizon for estimation of lifetime risk. In that matter, the lifetime risk of CKD4 + was: 13.9%, 18.6%, and respectively 5.4% for Start age groups 55–64, 65–74, and ≥ 75 years.

These figures cannot be evaluated in isolation, what we want to know is the magnitude of the excess risk attributable to lithium. However, to our knowledge, data on lifetime risk for CKD in the general population is quite sparse and in bipolar patients not treated with lithium—not available at all. In a simulation study from the USA (Grams et al. 2013), the residual lifetime risk for CKD4 + for a lifetime horizon of 90 years was estimated at: 10.2–10.5% for white men of age 60–80 years and respectively 11.1–11.8% for white women of the same age. European data is available only from a prospective longitudinal study from Iceland (Inker et al. 2015). The estimates in the latter study were much lower (3.4–3.7% for men and 3.2–3.6% for women, aged 45–75 years). However, the results are not comparable, as the lifetime horizon considered in the Icelandic study was shorter (85 years) and the follow-up controls, upon which the CKD4 diagnosis was based, were scheduled at 3–7 years intervals, thus probably missing an important number of CKD4-cases. Interestingly, in the only group followed up until 90 years (patients aged 80), the lifetime risk was 7% for both women and men, close to the results in the American simulation study.

While direct comparisons with American figures (Grams et al. 2013) have limitations, this study serves as the sole available reference in the general population. Notably, within our study population, the lifetime risk of CKD4 + among those aged 55–75 exceeded that observed in the American study (Grams et al. 2013). This outcome was in line with expectations and aligns with prior research (Close et al. 2014; Shine et al. 2015; Van Alphen et al. 2021).

When assessing this excess risk, a few factors come into play. First, as detailed in the Methods section, some potential for overestimation persists. Second, it is important to acknowledge that a portion of the excess risk may be associated with other risk factors. Specifically, our CKD4 + group exhibited a significant burden of somatic diseases, as indicated in Table 2. Of the 103 patients who developed CKD4 + , only 19 (18.4%) had no recorded somatic comorbidities. This data implies that the increased incidence of CKD4 + should not be solely attributed to lithium; somatic comorbidities likely made a substantial contribution (Forty et al. 2014).

However, the lifetime risk for individuals starting lithium at ≥ 75 years was much lower. As renal impairment usually progresses slowly, often over many years, a possible explanation might be that these older individuals simply did not live enough to progress to CKD4 + , a large proportion of them having died of other causes, precluding Incident CKD4 + . As shown in Table 3, their cumulative incidence was 5.4% already at 10 years follow-up (highest among all Start age groups), but it didn’t increase thereafter. On the other hand, the Start age group ≥ 75 years is the smallest group, with the lowest number of individuals and outcomes, and the lowest precision of the estimates (large SD’s, as indicated in Table 3), thus estimate error may, in part, account for the low figure.

The available data did not allow to estimate the lifetime risk of CKD4 + in patients starting lithium earlier than 55 years (as their age at end of follow up was lower than 90 years). However, it showed that their risk for CKD4 + is essentially zero the first ten years after commencing lithium (see Fig. 4a–f and Table 3): at five years follow-up, their cumulative incidence for CKD4 + was zero, and at ten years it marginally increased (to 0.7%) only for the Start age group 35–44 years, and remained zero for the other two groups.

Our study does not provide specific guidance on managing patients with an accelerated decline in renal function, a matter of paramount clinical importance. The central question at hand is whether to continue or discontinue lithium treatment. If we acknowledge that lithium poses a risk for CKD, continuing its use could potentially worsen renal decline. On the other hand, discontinuing lithium may raise the risk of relapse (Kumar et al. 2023), which may have severe consequences, including the risk of suicide (Gitlin 2023), and offers no guarantees of renal function improvement. It is worth noting that other psychopharmaceuticals may also have adverse effects on the kidneys (Kessing et al. 2015; Bosi et al. 2023; Højlund et al. 2020).

While some evidence suggests that continuing lithium treatment may not necessarily lead to further deterioration of renal function toward ESRD (Pahwa et al. 2021; Kumar et al. 2023; Kessing et al. 2017; Pahwa and Singh 2022), cases of progression to ESRD have been reported (Gitlin 2023). Also, some evidence exists, albeit limited, that discontinuing lithium may improve renal function at least in some patients (Hoekstra et al. 2022). 

In summary, there is no definitive answer, no “one-size-fits-all”-recommendation, and managing patients with progressive renal function decline is arguably one of the most challenging aspects of the lithium treatment.

In our previous research (Golic et al. 2023), we found that the pre-treatment creatinine value, albeit within the reference interval, was a prognostic indicator for Incident CKD4 + over 10-years follow-up. The current results show that Start creatinine level remained prognostic even when the follow-up period was extended to more than 3 decades. Specifically, Start creatinine in the upper third of the reference range was strongly associated with a higher risk of CKD4 + compared to the lower third.

In the matched case–control study, as all our study participants were lithium-treated, we categorised the Time on Li and used the lowest exposure (Time on Li 1–5 years) as reference. We found an association between the length of lithium exposure and Incident CKD4 + , with larger OR for increased exposure (see Table 6). However, it is important to acknowledge that the use of OR as a measure of risk has its drawbacks. The only use of OR in medicine is as an approximation of risk, but the approximation is good only for rare diseases. What is a “rare disease” in this context is yet another question to which there is no clear answer. Opinions differ, some authors (Davies et al. 1998), consider that the disease rate should fall below 20%, others (Chen et al. 2010)—below 10%, for a reasonable “safe” use of OR. Regardless of the disease rate, the OR will always overestimate the risk ratio (RR) to some degree, but serious divergences between OR and RR occur only with large effects on groups at initial high risk (Davies et al. 1998). In order to avoid drawing exaggerated conclusions from studies using logistic regression, a method for interpreting OR was proposed (Chen et al. 2010), where OR below 1.5 indicates a weak association (small size effect) and higher than 5—strong association (large size effect). Hence, our findings suggest that lithium exposure of 20 years or more is associated with a substantial increase in the risk of CKD4 + , while the association was only moderate for lithium exposure of 5–20 years.

How do the present results compare to previous research? The association between the duration of lithium treatment and impairment of kidney function has been previously explored using “softer” outcomes. A Swedish study (Aiff et al. 2015) found a continuous increase in S-creatinine with the lithium treatment duration. A Dutch study (Van Alphen et al. 2021) found that the duration of lithium exposure was associated with the risk for CKD3 (although the OR suggests a small size effect). A more recent Swedish study (Fransson et al. 2022) found that bipolar and schizoaffective patients who have used lithium for more than 10 years had a steeper eGFR decline compared to both bipolar/schizoaffective patients with 0–10 years of lithium exposure and to a reference population. A recent Danish register study (Højlund et al. 2022) found that lithium was associated with increased risk for CKD, with stronger association for more than 10 years of use. These results, although not directly comparable with the present study, point in the same direction, namely that increased exposure to lithium is associated with decreased glomerular function and increased risk for CKD. However, not all research findings are consistent. A US study, involving 154 bipolar patients treated with lithium and followed for up to 9 years, found no association between the duration of lithium exposure and incident CKD stage 3 or higher (Pahwa et al. 2021).

It is worth noting that in all Start age groups, the risk of death from other causes was numerically larger than the risk for CKD4 + (see Fig. 4). Even in the youngest Start age group (18–35 years), the number of competing deaths outnumbered Incident CKD4 + by more than 4:1.

In this group of individuals receiving lithium, the risk of developing CKD4 + was primarily associated with older age. The age at which CKD4 + was diagnosed is as follows: 25 patients (24.3%) were 80 years or older, 40 patients (38.8%) were aged 70–79, 29 patients (28.2%) were aged 60–69, and 9 patients (8.7%) were aged 40–60. Notably, there were no cases among individuals under 40 years of age. The incidence pattern of CKD4 + observed in this lithium-treated population closely resembles that reported in a population-based study in the Netherlands (Blijderveen et al. 2014).

Over the past couple of decades, there has been a growing debate about the definition of CKD when age-independent criteria are applied. Critics argue that these criteria may erroneously label the natural, age-related decline in renal function, which typically begins around the age of 50, as a medical condition. Consequently, this could lead to the overdiagnosis of many asymptomatic elderly individuals, who might remain in CKD stage 3 throughout their lives.

Some experts (Delanaye et al. 2019) have proposed an age-adapted CKD definition, suggesting a pathological threshold of 45 ml/min/1.73 m2 for individuals aged 65 and older, and 75 ml/min/1.73 m2 for those under 65. This age-specific approach is seen as more medically and socially sensible. It advocates for greater attention to be directed toward younger patients with 'milder' CKD since they are at risk of disease progression and early mortality (Kula et al. 2023). By focusing on this younger demographic, based on the recommended thresholds, it becomes possible to proactively manage associated risk factors. This approach holds promise for potential developments in lithium monitoring guidelines as well.

To our knowledge, this is the first study to include a large number of unselected lithium-treated patients followed up to more than 35 years. The diagnosis of CKD4 + based on laboratory data rather than health charts ensured a high level of accuracy, as chart-diagnosis might be delayed or even missing in asymptomatic patients. The outcomes are expressed in easy-to-grasp concepts and the findings may have direct clinical applicability for patient communication and clinical decision making.

The study has a number of limitations. The risk of surveillance bias has been previously discussed (Kessing et al. 2015; Nielsen et al. 2018; Wiuff et al. 2023), whereby regular renal function monitoring in lithium patients is likely to result in a higher detection of asymptomatic CKD, compared to patients that are not monitored. This risk might be less important though when the outcome is as severe as CKD4 + .

The risk of overestimation of cumulative incidence and lifetime risk, due to possible uninformed censoring, has been mentioned before and must not be overlooked.

A number of assumptions were made in relation to parameter operationalization such as: one year without any S-Li was regarded as treatment discontinuation, and the presence of S-Li at least once a year was regarded as a treatment period. This may not necessarily reflect the reality, as some patients may have taken lithium without proper treatment monitoring. Patients' follow-up started with the first year of continuous treatment, according to our operational definitions. In reality, some patients may have had a longer lithium treatment than the one computed in our analysis, and this may have introduced a bias.

An important limitation of this study is the lack of complete data on somatic comorbidities and concurrent medications. The regression models were not adjusted for these factors, which are known to be associated with CKD in the general population (Low et al. 2015; Zeng et al. 2023) and in lithium-treated patients (Shine et al. 2015; Rej et al. 2014; Aiff et al. 2019). While it is unlikely that adjusting for somatic comorbidities would invalidate the results, such an adjustment would underscore the potential contribution of these risk factors more clearly.

The number of Incident CKD4 + was 103 in the whole material of 2381 patients. This allowed for multivariate analyses with categorised variables, observing the statistical rule of thumb of 10 outcomes per parameter. However, the Start age groups were comparatively small, and the number of Incident CKD4 + per group was low (in particular the youngest and the oldest Start age group), resulting in relatively low precision for the Cumulative Incidence estimates, as indicated by the comparatively large standard deviations presented in Table 3. Statistically meaningful sub-analyses (i.e., by sex) were not possible due to the low number of Incident CKD4 + among men (only 31).

The correlation between Time on Li and Incident CKD4 + was examined only within the lithium-treated population. Comparison was made between individuals with larger exposures and those with a reference exposure (Time on Li 1–5 years), but not with lithium-free individuals. We, therefore, did not explore whether the risk for CKD4 + is higher for our reference exposure (1–5 years) compared to lithium non-users.

The patients recruited in our study have been followed up and managed by Swedish physicians, acting in accordance with national and local monitoring guidelines. S-Li has been well kept within therapeutic limits, and the lithium treatment has been terminated in a number of patients. The local guidelines and routines, clinicians’ prescribing practice, patients’ medical literacy, compliance, and degree of involvement in the decision-making process are likely to influence the characteristics of the study population, the length of the treatment, the target serum lithium level and / or the outcomes. Hence, the results may only be generalizable to settings with similar lithium monitoring guidelines and recommendations, access to laboratory testing and socio-demographics.