A 20-year old man with bipolar disorder presented to an outside hospital with severe viral gastroenteritis. His labs were unremarkable except for a lithium level of 4.4 mM. The outside hospital lacked dialysis capabilities, so he was transported to the Genius General ICU. In the interim, he received aggressive IV crystalloid.
We were advised to place a hemodialysis catheter as soon as possible, in order to start dialysis without a moment's delay. However, upon arrival he looked great with excellent urine output and normal neurologic examination. He reported feeling “a thousand times better” then a few hours earlier. Rather than placing a hemodialysis catheter we checked a repeat lithium level, which was 3.4 mM. He was continued on intravenous fluids and recovered uneventfully.
If this patient had presented initially to Genius General Hospital, he would have been dialyzed. Presenting to a remote community hospital saved him from dialysis by forcing us to perform a therapeutic trial of rehydration while awaiting transportation.
The lithium level doesn't tell the whole story
In a world of uncertainty, we are naturally drawn towards data – numbers, tests, images, studies. However, it's unclear how much clinically useful information a lithium level actually provides us. Interpreting a lithium level is complicated by two main factors:
Factor #1: A single level doesn't reveal pharmacokinetics
Consider two patients with the same exact lithium level:
- The first patient is man on chronic lithium therapy who presents shortly after a massive lithium ingestion with a level of 4.4 mM. This lithium level is likely to increase soon, placing him at risk of deterioration.
- The second patient is described in the introductory case above, also with a lithium level of 4.4 mM. This patient was taking his usual dose of immediate-release lithium, with the last dose ingested yesterday. There is no reason to believe that his lithium level will increase further. With aggressive hydration, his lithium level will almost certainly decrease.
Although both patients have the same lithium level, the first patient is at substantially greater risk of deterioration and would be a better candidate for dialysis.
Factor #2: Serum levels don't predict with brain levels
The primary organ affected by lithium toxicity is the brain. The key parameter in determining toxicity seems to be lithium levels in the brain. Serum levels can be a poor predictor of brain levels for two reasons:
- Lithium takes time to cross the blood-brain barrier. Thus, a patient with acute ingestion may do fine despite a high serum lithium level. This occurs because the kidneys clear lithium from the blood before it has enough time to fully penetrate the brain.
- Even at steady-state, serum lithium level is a poor predictor of brain lithium levels (figure below, Riedl 1997). This variability makes it difficult to predict the clinical effect of any specific lithium level.
Recommendations for lithium intoxication vary widely
UpToDate and many other guidelines recommend dialysis for asymptomatic patients if the lithium is >4 mM, or >2.5 mM with renal dysfunction. The EXTRIP guideline recommends dialysis if the lithium level is >5 mM, or >4 mM with renal dysfunction.
Using any single lithium level to trigger dialysis is absurd. For example, imagine a patient with chronic asymptomatic intoxication, normal renal function, and an admission lithium level of 4.1 mM. Most guidelines recommend dialysis because the lithium level is above 4 mM. However, with hydration the lithium level will probably fall below 4 mM within a few hours. Why not treat conservatively and repeat the level? This is a common-sense approach used with other conditions potentially requiring dialysis, such as hyperkalemia (1).
Similarly, it may be premature to categorize renal dysfunction as “present” or “absent” based on admission laboratory values. Many patients present with pre-renal acute kidney injury that responds rapidly to fluid resuscitation.
There is no evidence to support dialysis for chronic asymptomatic hyperlithemia.
There is no high-level evidence to support dialysis in lithium intoxication. The largest study on this issue was Bailey 2000, which focused on eighteen patients for whom poison control recommended dialysis. Of these patients, only six actually underwent dialysis. No differences were observed when comparing patients who did or didn't undergo dialysis. The authors concluded that “the indications for hemodialysis in lithium poisoning should be reconsidered to include only the more severe cases.”
Likewise, there doesn't appear to be any evidence to support the concept of a lithium threshold that mandates dialysis. This explains why authors disagree about the lithium level that should trigger dialysis.
What is the theoretical rationale for dialysis in chronic, asymptomatic lithium intoxication?
A patient with chronic asymptomatic hyperlithemia is tolerating their current lithium level well. If they have adequate renal function, their lithium level is very likely to decrease over time with hydration and discontinuation of lithium intake. Why dialyze such a patient? It is impossible to improve a patient's condition if the patient is already asymptomatic.
The first rationale for dialysis is that perhaps the serum lithium hasn't had time to enter the brain yet. Maybe the serum lithium is poised just about to penetrate the brain and cause severe intoxication. This would imply that removal of lithium from the serum could avert entry into the brain, thereby avoiding severe poisoning. This argument makes sense in acute intoxication, but less so for chronic intoxication. Specifically, there are a few flaws with this rationale:
- For patients in whom chronic hyperlithemia is incidentally discovered, it seems unlikely that this is coincidentally right at the moment when disaster is about to strike.
- It seems unlikely that chronic hyperlithemia would rapidly progress from asymptomatic to severe toxicity. Thus, rather than prophylactically perform dialysis, it might be more sensible to follow the patient carefully and perform dialysis if clinical signs of toxicity should develop.
Given that the time course of lithium poisoning is relatively long compared to other poisons, evolving over days, it is not unreasonable that a conservative approach is initially adopted in some circumstances as long as frequent monitoring of renal function and serum lithium concentrations are performed so that treatment can be adjusted, if required. –Roberts 2014
The second rationale for dialysis seems to be avoidance of SILENT syndrome (syndrome of irreversible lithium-effectuated neurotoxicity). SILENT syndrome is a rare and poorly understood syndrome that may occur following lithium use, with persistent neurologic deficits including cerebellar dysfunction. The EXTRIP guideline speculates that more aggressive dialysis might reduce the risk of SILENT syndrome.
Dialysis is unlikely to reduce the risk of SILENT syndrome.
We certainly would like to avoid SILENT syndrome. Unfortunately, the most recent review of this syndrome concluded that there is no evidence that hemodialysis can prevent it (Niethammer 2007). These are some reasons why:
SILENT syndrome appears unrelated to the lithium level
Imagine that SILENT syndrome only occurred in patients with a high lithium level (i.e. >4 mM). In that case, it could make sense to dialyze patients to avoid exposure to such lithium levels. However, there is no clear relationship between SILENT syndrome and the lithium level. Adityanjee 2005 reported that SILENT syndrome could occur even at therapeutic lithium levels. Among reported cases of SILENT, the average lithium level was 2.3 mM, which is below even the lowest dialysis threshold.
Rapid reduction in lithium levels could increase the risk of SILENT syndrome
Swartz 1994 observed several patients with chronic lithium intoxication who seemed to deteriorate following withdrawal of lithium. To determine if this were common, they reviewed the literature to identify fifty patients with subacute or chronic lithium intoxication who were responsive at admission. Among these patients, the use of dialysis was associated with a higher rate of deterioration during hospitalization (85% vs. 46%, p=0.016) and persistent neurologic sequelae (62% vs. 22%, p=0.012). Such differences didn't seem to reflect baseline differences, since patients who underwent dialysis had less cognitive dysfunction upon admission (2).
While hyperlithemia can be toxic, its rapid correction can be more toxic. –Swartz et al. 1994
Other reports support this concept. Magarbane 2002 reported an association between dialysis and neurologic deterioration among a retrospective analysis of data from a single center. Patients with chronic lithium intoxication who deteriorated following reduction of lithium level have been described (Simard 1989, Swartz 1995). Similarly, some authors have reported a phenomenon of lithium withdrawal wherein an asymptomatic, stable patient on chronic lithium develops delirium after abruptly stopping lithium (Wilkinson 1979, Yuce 1979).
Deterioration related to lithium withdrawal might relate partially to osmotic shifts. Patients with hyperlithemia can have abnormal serum osmolality, either due to lithium itself or associated diabetes insidipus (Fabisiak 2002). Dialysis could trigger rapid shifts in lithium and sodium levels, promoting cerebral edema or demyelination (3).
Long-term implications for future treatment with lithium
There are some patients for whom lithium is a uniquely effective. In these patients, complete cessation of lithium is undesirable, as it may increase the risk of psychiatric relapse. Furthermore, ICU admission for dialysis could be traumatizing and might impair future adherence with lithium therapy.
For a patient who requires lithium, the most sensible approach to chronic asymptomatic hyperlithemia might be hospital admission for hydration and re-titration of lithium dose (4). For example, Swartz 1994 published a report of a woman with symptomatic chronic lithium intoxication with an initial lithium level of 2.4 mM. She was admitted and treated with additional lithium to prevent a rapid drop in her lithium level. Eventually her lithium dose was re-titrated and she was discharged on lithium in excellent condition.
Data describing the benefits, if any, of preemptive enhanced elimination in an asymptomatic patient with very high plasma lithium concentrations are extremely limited. – Baird-Gunning 2016
These guidelines suffer from a common problem: failure to admit uncertainty. There simply isn't enough evidence to make any recommendation about who to dialyze. Camouflaging uncertainty is always dangerous. In the past, I have used guidelines to determine who required dialysis. My patients would have been better served by a more nuanced evaluation and a thoughtful discussion with a toxicologist, rather than a paint-by-the-numbers approach.
Guidelines can also generate and perpetuate dogma. I've met some nephrologists who were entirely convinced that dialysis is mandatory for any lithium level above 4 mM. These physicians aren't stupid or lazy, they've merely internalized certain guidelines. Oversimplified guidelines can dumb down the level of thought involved in treatment decisions, by substituting uncertainty with a false sense of confidence.
- Patients with chronic, asymptomatic elevation of lithium levels are at little acute risk of immediate deterioration: their lithium level should fall over time, and they are tolerating their current lithium level well.
- There is no evidence to support the use of dialysis in stable patients with chronic asymptomatic hyperlithemia.
- Some reports suggest that abrupt reduction in lithium levels induced by dialysis may cause neurological deterioration.
- A sensible approach to chronic asymptomatic hyperlithemia in a patient with adequate renal function might be rehydration with close monitoring of lithium levels and clinical status.
- The Poison Review has written excellent posts on lithium intoxication here, here, and here. The most recent post makes an important point about the use of ultrasonography to perform goal-directed fluid resuscitation in patients with lithium toxicity. These patients may be profoundly volume depleted, so giving an arbitrary volume of fluid (e.g. the proverbial “two liters of saline”) may not be enough.
- (Editor) Here is the review article recently mentioned on The Poison Review: lithium-toxicity
- Admittedly, there are some situations in which a single potassium level might mandate dialysis (e.g. a patient with chronic dialysis-dependent renal failure and a markedly elevated potassium level). However, for most patients who aren't dialysis dependent, the usual approach is to start with a trial of medical therapy. Even if the potassium is scary high, medical therapy can be more effective and more rapid than dialytic therapy (keep in mind that placing a dialysis catheter and starting dialysis usually takes at least 2-3 hours, whereas medical therapy can work much faster). More discussion about this here.
- Of course, this is a retrospective correlational study. It is probable that patients who were selected for dialysis were sicker at baseline in some way (possibly not neurologic), and this could be a confounding factor related to their subsequent deterioration.
- Nobody really knows, though. It's also possible that the neurochemistry of the brain could adapt to lithium, and then withdrawal of lithium could lead to an imbalance. The entire topic is a bit of a mystery.
- Please note that these decisions are probably best made in consultation with the patient's psychiatrist. A psychiatrist may help to understand whether ongoing lithium therapy is critical for a specific patient, or whether there are safer alternative therapies. Regardless of what management decisions are made, it is important to consider both the short-term stabilization of the patient and longer-term plans regarding ongoing psychiatric care.
- IBCC chapter:Guide to APRV for COVID-19 - April 8, 2020
- PulmCrit Theoretical Post – The COVID Severity Index (CSI 1.0) - April 2, 2020
- PulmCrit wee – Why the SCCM/AARC/ASA/APSF/AACN/CHEST joint statement on split ventilators is wrong. - March 29, 2020