Long‐term adverse effects of paracetamol – a review

Abstract

Paracetamol (acetaminophen) is the most commonly used drug in the world, with a long record of use in acute and chronic pain. In recent years, the benefits of paracetamol use in chronic conditions has been questioned, notably in the areas of osteoarthritis and lower back pain. Over the same period, concerns over the long‐term adverse effects of paracetamol use have increased, initially in the field of hypertension, but more recently in other areas as well. The evidence base for the adverse effects of chronic paracetamol use consists of many cohort and observational studies, with few randomized controlled trials, many of which contradict each other, so these studies must be interpreted with caution. Nevertheless, there are some areas where the evidence for harm is more robust, and if a clinician is starting paracetamol with the expectation of chronic use it might be advisable to discuss these side effects with patients beforehand. In particular, an increased risk of gastrointestinal bleeding and a small (~4 mmHg) increase in systolic blood pressure are adverse effects for which the evidence is particularly strong, and which show a degree of dose dependence. As our estimation of the benefits decreases, an accurate assessment of the harms is ever more important. The present review summarizes the current evidence on the harms associated with chronic paracetamol use, focusing on cardiovascular disease, asthma and renal injury, and the effects of in utero exposure.

Introduction

http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=5239 (acetaminophen) was first synthesized in 1878 1, from its precursor http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=7402. Its use was not widespread initially, due to early reports of a link to methaemoglobinaemia 2, 3. After this association was discredited, it was marketed in the 1950s as a safer alternative to phenacetin, which by then had been found to be nephrotoxic and potentially carcinogenic 4. In the early 1980s, paracetamol overtook http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4139 as the most widely used over‐the‐counter (OTC) analgesic in the UK 5. It is now the most commonly used analgesic in the world 6, and the first step of the World Health Organization (WHO) analgesic ladder for the treatment of cancer pain 7.

Paracetamol is currently marketed as an analgesic and antipyretic, to be used for no more than 3 days without consulting a doctor 8. However, due in part to its inclusion in the WHO analgesic ladder, as well as decades of clinical experience, it is also prescribed in chronic conditions such as osteoarthritis and lower back pain. Recently, meta‐analyses of the randomized controlled trials (RCTs) covering these conditions have shown the effect sizes to be modest, although still statistically significant, compared with placebo (averaging a 4–5% reduction in pain) 9, 10, 11, 12. Despite this, paracetamol continues to be recommended as first‐line treatment in UK guidelines 13, and attempts to remove it as a recommendation from the UK's National Institute for Health and Care Excellence (NICE) guidance on osteoarthritis raised considerable concerns among medicines regulators and various specialist societies 14, particularly as this would leave opioids as the major alternative. Given the current opioid addiction epidemic ongoing in several US states 15, and a desire not to repeat this in the UK 16, 17, the introduction of opioids earlier in the pain management pathway is unlikely to be viewed favourably.

Paracetamol has less of an analgesic effect in chronic use than previously thought; there needs to be greater emphasis on accurately determining the harms of long‐term use at therapeutic doses. This helps clinicians to balance harms against likely benefits for individual patients and allows regulators to make recommendations on its availability in OTC preparations. The acute effects of paracetamol ingestion in overdose are well known 18. Harms with long‐term therapeutic use are less clear. Concerns have been raised over the effects on the cardiovascular, respiratory, renal, gastrointestinal and central nervous systems, as well as potential effects in the offspring of pregnant women ingesting paracetamol.

The present review summarizes our understanding of the evidence on the adverse effects of paracetamol in long‐term therapeutic use, informs clinicians of the risks and provides a clearer picture of the underpinning evidence base. This will, in turn, allow clinicians to discuss with their patients the relative benefits and harms of long‐term paracetamol use.

Mechanism of action

The mechanism of action of paracetamol is not completely understood but is likely to involve http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1376 (COX‐2) inhibition. Traditional nonsteroidal anti‐inflammatory drugs (NSAIDs) inhibit cyclooxygenase (COX) enzymes, preventing the metabolism of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2391 to http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=5245. COX enzymes also have a separate peroxidase function, and metabolize PGG₂ to http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4483, which, in turn, is converted to several different PGs by local tissues according to their individual needs 2, 19, 20. Unlike the closely related NSAIDs, paracetamol interferes with the peroxidase activity of COX isoenzymes, predominantly COX‐2, particularly when the cellular environment is low in arachidonic acid and peroxides 2, 19, 20. This explains paracetamol's apparent ‘central’ effect in earlier studies (as COX‐2 is constitutively expressed in neural tissue) 19, 21, and why it appears to be ineffective in inflamed tissues (where peroxide and arachidonic acid are abundant), seen in conditions such as rheumatoid arthritis. A proposed COX‐3 isoenzyme (an exon splice variant of COX‐1 seen in insects and rodents) has not been found in humans, and further studies suggest that paracetamol has no clinically significant effects on the COX‐1 exon splice variants found so far in humans 19, 21. Other possible mechanisms of action include the inhibition of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2364 reuptake (and subsequent cannabinoid receptor http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=56 stimulation) by paracetamol metabolite N‐arachidonoylphenolamine (AM404), which is produced by the conjugation of arachidonic acid and deacetylated paracetamol 22, and direct activation by this metabolite of the capsaicin receptor transient receptor potential cation channel subfamily V member 1 (http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=507) 23. Transient receptor potential cation channel, subfamily A, member 1 (http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=485) activation by paracetamol metabolites has also been suggested 24, 25.

Paracetamol is mostly metabolized by the formation of conjugates (with glucuronide and sulphate), and subsequently excreted in urine. In therapeutic dosing, around 10% of paracetamol is metabolized by http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=242 (CYP) enzymes to form http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6299 (NAPQI), which is subsequently conjugated with intracellular http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=6737, and ultimately excreted as cysteine and mercapturic acid conjugates. Less than 5% is excreted unchanged 26.

Search strategy

We conducted a literature search of PubMed, searching the years 1980 to 2016. An initial Pubmed review of ʻparacetamol [Title] OR acetaminophen [Title]ʼ with ʻside effects OR adverse effectsʼ revealed several key interest areas, which were subsequently searched for specifically as follows: we combined ʻparacetamol [Title] OR acetaminophen [Title]ʼ with: ʻhypertension OR blood pressureʼ; ʻmyocardial infarction OR cardiac OR cardiovascularʼ; ʻstroke OR CVA OR cerebrovascular accidentʼ; ʻliver OR hepatic OR transaminase OR aminotransferaseʼ; ʻgastrointestinal OR bleeding OR anaemiaʼ; ʻrenal OR kidney OR CKD OR chronic kidney diseaseʼ; ʻrespiratory OR asthma OR chestʼ; ʻreproductive OR maternal OR ADHD OR attention deficitʼ. Papers were selected uisng the following criteria: (i) human subjects; and (ii) meta‐analyses, reviews, RCTs, prospective studies and cohort studies. English language was not included as a filter but this would not have excluded any papers from review. Titles and abstracts were then reviewed, and relevant articles reviewed in full. Key papers identified in references were also reviewed by the authors, when considered relevant (see Figure 1 for our search strategy).

Figure 1.

Outline of search strategy

Cardiovascular disease

Studies examining the effect of paracetamol on the incidence of cardiovascular disease are relatively sparse when compared to those on NSAIDs 27. Early studies focused on hypertension (which we have reviewed previously 28), owing to the known association of NSAIDs with hypertension, and the similar mechanism of action of paracetamol 29. One such study was a placebo‐controlled crossover study of 20 treated hypertensive patients, in whom a 4 mmHg rise in blood pressure (BP) was found when paracetamol was administered 30. Given that a 2 mmHg rise in systolic BP is associated with a 7% increase in the risk of ischaemic heart disease and a 10% increased risk of stroke 31, this apparently small increase in BP could have serious population‐based consequences.

However, observational and interventional studies examining the effect of paracetamol on hypertension have produced conflicting results 28. To date, most 32, 33, 34, but not all 35, 36, observational studies suggest that long‐term paracetamol use increases the risk of developing hypertension. The Nurses' Health Study II, which included 80 020 participants, found that regular NSAID or paracetamol use was associated with an increased risk of developing hypertension 33: the relative risk (RR) of developing hypertension on NSAIDS was 1.86 [95% confidence interval (CI) 1.51 to 2.28) and on paracetamol was 2.00 (95% CI 1.52 to 2.62). It also seems that there is some evidence for a dose–response relationship between daily paracetamol dose and the risk of incident hypertension. This was observed not only in the Nurses’ Health Studies I and II 37, but also by Roberts et al. 6 for overall cardiovascular risk in their systematic review of paracetamol‐related adverse effects.

By contrast, a retrospective observational study by Dawson et al. 36, with propensity matching, found no impact of paracetamol on BP in a cohort of 2754 participants with treated hypertension. Although observational studies may find an association between paracetamol use and hypertension, underlying confounders (such as chronic inflammatory conditions) need to be considered. Unfortunately, to date, interventional studies examining the impact of paracetamol on BP have been limited by study design and small sample size. One recent study, by Sudano et al. 38, randomized 33 patients with established coronary artery disease to paracetamol 1 g three times per day or placebo in a double‐blinded crossover study. Two weeks of treatment with paracetamol significantly increased mean systolic ambulatory BP (from 122 ± 12 mmHg to 125 ± 12 mmHg; P = 0.02) and diastolic ambulatory BP (from 73 ± 7 mmHg to 75 ± 8 mmHg; P = 0.02). Although this difference is unlikely to have a significant effect on an individual patient's cardiovascular outcomes, it may explain the finding that self‐reported frequent paracetamol use in women is associated with an increase in cardiovascular events similar to that seen with frequent NSAID use 27. Fulton et al. 39 showed no increased risk of myocardial infarction or stroke in a hypertensive cohort of 4000 subjects, and no change in BP, which suggests that any increase in risk may be driven by BP alone. Further research in this area is clearly required, and there is currently a suitably powered double‐blind, placebo‐controlled, crossover trial from our centre examining the effects of 2 weeks of paracetamol use on BP in hypertensive patients (https://clinicaltrials.gov/ct2/show/NCT01997112), which should report soon.

Respiratory effects

After aspirin was recognized to cause the rare but serious complication of Reye's syndrome, its use was banned in children under 12 years of age 40, 41. As aspirin use as an antipyretic waned in developed countries and paracetamol use became more common 42, concerns over paracetamol's association with asthma were raised 43. Observational and cross‐sectional studies demonstrated a connection between paracetamol use and asthma diagnoses or exacerbations 44, 45, 46, 47, 48, 49, 50, 51. However, as for BP, almost all of these studies suffer from confounding by indication: recurrent symptomatic respiratory infections and febrile illnesses are more common in asthmatic patients and contribute to the onset of asthma in childhood 52, 53, 54. In some studies, an increase in the risk/odds for developing asthma with increasing paracetamol use becomes nonsignificant when adjusted for recurrent respiratory tract infection 55, 56, 57, although this is not universal 50. Meta‐analyses of these observational studies tend to show only a small effect [e.g. odds ratio (OR) 1.15 for use in infancy], and suffer from considerable heterogeneity 44, 52.

A link between paracetamol use and asthma is biologically plausible. Paracetamol metabolism involves the antioxidant glutathione, which is depleted when large doses of paracetamol are taken. There are papers describing glutathione depletion at therapeutic doses of paracetamol 58, 59, and increased oxidative stress could contribute towards either the development of asthma or inflammatory exacerbations in asthmatics 60, 61, 62, 63, 64. Glutathione depletion may also change T helper (Th) physiology towards a Th2 phenotype, which is associated with atopic disease 46. Paracetamol may also cause an imbalance in lipoxygenase activity, brought about by COX inhibition, resulting in increased http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=272 and decreased http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1883 production 60, 61, 65. This latter mechanism has support from studies carried out in patients with aspirin‐associated asthma, in which decreases in forced expiratory volume in 1 s (FEV₁) following paracetamol administration were observed 66. In one such study, 34% of aspirin‐sensitive participants showed cross‐reactivity to paracetamol 67, and in patients with aspirin‐associated asthma it is recommended to use the lowest effective dose of paracetamol for analgesia 68, 69. It should be noted that a later double‐blind RCT in non‐aspirin‐sensitive subjects (n = 85) taking paracetamol 1 g twice daily or placebo for 12 weeks showed no differences in bronchial hyperresponsiveness 70.

A few researchers have attempted to study the effects of paracetamol on asthmatic patients (adult and paediatric) in RCTs. Ioannides et al. 70 randomized adults with mild‐to‐moderate asthma to placebo or paracetamol 4 g d^–1 for 12 weeks, then submitted them to a http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=7438 challenge test. Airway hyperresponsiveness was similar in both groups (amount of methacholine required to reduce FEV₁ by 20%, paracetamol group – placebo was −0.48 mg ml^–1, 95% CI –1.28 to – 0.32) but this study was notably underpowered (n = 94; recommended sample size 650) 70. More recently, Sheehan et al. 71 conducted an RCT in which they randomized children aged 0.5–5 years to receive either paracetamol or http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2713 for analgesia/antipyresis over the following 48 weeks. Participants received a mean of 5.5 doses (range 1–15), with no between‐group differences. The RR for asthma exacerbations was similar between the two groups, and there were no significant differences in secondary outcomes (asthma‐controlled days, unscheduled care, use of rescue medication), indicating that, at least in mild‐to‐moderately asthmatic children, paracetamol was as safe to use as ibuprofen 71.

Overall, there is evidence of a weak association of paracetamol use with asthma, but causation cannot be established. RCTs are limited, but seem to provide reassurance that paracetamol is safe to use in patients with established asthma 70, 71, 72.

Gastrointestinal (GI) effects

The acute effects on the liver of paracetamol in overdose have been well documented 73, 74, 75. However, the effect of chronic therapeutic‐dose paracetamol use on the liver and GI system in general is less clear. Concerns are generally focused around GI blood loss and chronic hepatotoxicity.

GI bleeding

Paracetamol has long been considered the ‘safe’ analgesic alternative to NSAIDs in patients prone to GI bleeding. Indeed, in studies of the analgesic effects of NSAIDS it is commonly used as a comparator, owing to the ethical issues of withholding analgesia 76, 77, 78. There is some evidence to support the safety of paracetamol. Examining adverse events reported in the Spanish drug monitoring system, Carvajal et al. 79 found that paracetamol use was associated with nausea (3.3% of all reported adverse events) and dyspepsia (4.2%), but not GI bleeding. Furthermore, a meta‐analysis of individual patient data from three case–control studies, looking at the risk of GI bleeding with individual NSAIDs, included paracetamol as a comparator and found no increased risk of GI bleeding with increasing daily doses of paracetamol 80.

However, recent epidemiological studies have identified a potential increased risk of upper GI bleeding with doses of paracetamol ≥2–3 g d^–1. In 2001, a case–control study was conducted using the UK's General Practice Research Database (GPRD) 81. Adults aged 40–79 years with no history of prior GI disease or alcohol misuse (n = 13 605) were followed up between 1993 and 1998. The incidence of upper GI complications was documented, as was the prescription of paracetamol and potentially confounding medications. Compared with nonusers of paracetamol, users of ≤2 g d^–1 did not have a significant increase in GI complications. However, use of >2 g d^–1 had an adjusted RR (95% CI) of 3.6 (2.6 to 5.1). When this analysis was confined to those patients with no prior NSAID prescription or antecedents of GI disorders (e.g. dyspepsia), the adjusted RR was 5.7 (2.0 to 16.4). When combined with NSAID, the risk increased to 13.2 (9.2 to 18.9), indicating a substantial interaction. It is important to recognize the potential influence of channelling bias in this instance (NSAIDS are not prescribed to those at risk of upper GI bleeding unless necessary, so high‐risk patients may be disproportionately prescribed paracetamol). The authors tried to compensate for this by excluding a history of Mallory–Weiss tear, cancer, oesophageal varices, coagulopathy or alcohol‐related disease, and adjusting the RR for age, smoking, upper GI risk factors and concomitant medications, but this (they admitted) cannot exclude all bias. Additionally, the study was of prescriptions, not ‘real‐world’ use. The authors had no data on OTC use of paracetamol by patients, and the daily dose was calculated from prescription frequencies, both of which have the potential to confound the results (although would not explain the apparent dose–response relationship found). The same group later published a follow‐up examination of the link between GI complications and paracetamol in the GPRD 82, and found a pooled RR of 1.3 (1.1 to 1.5). Furthermore, in users of ≥2 g d^–1, the RR was 3.6 (2.6 5.1). They did not detect evidence of heterogeneity or publication bias.

In 2008, a study of GI‐related hospitalisation in elderly patients in Quebec examined the effect of NSAIDs, paracetamol ≤3 g d^–1, paracetamol >3 g d^–1, and proton pump inhibitors (PPI) 83. Using paracetamol ≤3 g d^–1 as the reference population, they found hazard ratios (HRs) for GI‐related hospitalization of 1.2 (95% CI 1.03 to 1.40) for paracetamol >3 g d^–1, 1.63 (95% CI 1.44 to 1.85) for NSAID, and 2.55 (95% CI 1.98 to 3.28) for combined usage. With the use of a PPI, the hazards became nonsignificant except in the combined usage group [HR 2.15 (95% CI 1.35 to 3.40)]. These data would suggest that elderly patients taking paracetamol with or without concomitant NSAID are at risk of GI‐related hospitalization. The authors hypothesize that the additional, weak, nonspecific COX inhibition from paracetamol could have an additive effect to that of NSAIDs, creating an increased risk of gastric mucosal injury when used together.

Early RCTs in this area appeared to give reassuring results. One crossover study examining the effects of 7 days of paracetamol, http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4795 or placebo on endoscopic appearances found no acute effects of paracetamol on the GI mucosa 84. However, more recent RCTs have been less reassuring. In 2011, Doherty et al. 76 examined the effects of paracetamol (3 g d^–1), ibuprofen (1200 mg d^–1) and a combination of the two (ibuprofen 600 mg/paracetamol 1.5 g daily, or twice this dose) for chronic knee pain in a parallel‐group RCT of 892 patients. Although the study was powered to detect differences in analgesic effect [a 5.5‐point reduction in the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain scale], they also collected data on adverse events. After 13 weeks, they examined the proportion of the groups that had a decrease in haemoglobin of ≥1 g dl^–1. This was 19.6% in the ibuprofen group, 20.3% for paracetamol, 24.1% for the low‐dose combination and 38.4% for the high‐dose combination, which was significantly different to the other three groups. As there was also a small but significant drop in platelet count, and an increase in mean cell volume, the authors suggested that the haemoglobin decrease was likely to be due to occult GI blood loss. They concluded that paracetamol 3 g d^–1 and ibuprofen 1200 mg d^–1 were associated with similar amounts of occult blood loss, and that there was an additive effect in the higher‐dose combination. More recently, in 2016, the authors of the PERFORM trial (Prevention of cerebrovascular and cardiovascular Events of ischaemic origin with teRutroban in patients with a history oF ischaemic strOke or tRansient ischaeMic attack) examined the effects of paracetamol and ibuprofen on cardiovascular effects and bleeding using a nested case–control study within their cohort of 19 120 participants with recent ischaemic stroke 85. A total of 800 cases were paired with 1600 controls, and the incidence of bleeding (in general, but including intracerebral haemorrhage or intraocular bleed) was recorded. They found that the use of ≥3 g d^–1 was associated with bleeding events [OR 3.72 (95% CI 1.58 to 8.75)]. Doses of ≤3 g d^–1 were not associated with significant risk, but the trend test was significant (P = 0.02), indicating a dose–response relationship.

Based on these data, it seems that when taken regularly at doses of >2–3 g d^–1 (i.e. at daily doses normally seen in chronic use), there is a significant risk of GI bleeding with paracetamol. The dose–response relationship seen in some of the studies would indicate that something in the mechanism of action of paracetamol can cause GI bleeding as an adverse effect, and that this effect is additive when combined with NSAIDs.

Hepatotoxicity

Over the past few decades, there have been several case reports and small studies implying a connection between the ingestion of therapeutic doses of paracetamol and liver injury 86. It has been known for many years that therapeutic paracetamol use (≤4 g d^–1) has been associated with subclinical rises in liver injury markers 74. However, transient rises in alanine aminotransferase (ALT) can be secondary to many factors, such as exercise, vitamin intake, congestive heart failure, diabetes and medications such as aspirin, heparins and statins 87, 88. Whether such an enzyme rise results in clinically significant liver injury is less clear. Heard et al. 89 looked at this issue with healthy volunteers in an RCT of long‐term paracetamol ingestion (dose 4 g d^–1). They found that ~50% of the paracetamol group experienced no ALT rise, ~25% had a transient rise, which was gone by day 16, and ~25% had ALT normalize by day 40 89. These findings are consistent with those from Dart and Bailey's review of observational data in >40 000 patients, showing a low incidence of transaminitis (0.4–1.0%) and no progression to hepatotoxicity 90.

Of those case reports of liver injury in patients taking therapeutic doses, additional factors such as alcohol abuse, nutritional deficiency or concurrent febrile illness are usually present 91. As the toxic metabolite of paracetamol NAPQI is produced via CYP metabolism (predominantly the http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1330 isoenzyme), clinicians have hypothesized that induction of these enzymes through alcohol misuse might predispose patients to liver injury. This appeared to be supported by animal studies showing that CYP2E1 was induced by ethanol in rodents, and that levels of NAPQI and hepatotoxicity were increased when paracetamol was administered 86. However, researchers have failed to replicate this finding in humans, and have found evidence of the opposite: CYP2E1 appears to increase only modestly with short‐term alcohol use, reversing soon after abstinence 86, and one examination of cirrhotic livers found them to have 59% less CYP2E1 than control samples 92. In addition, NAPQI levels are not increased in chronic alcoholics taking paracetamol and, although the drug's half‐life is prolonged in chronic liver disease, this does not significantly affect metabolism or lead to NAPQI accumulation/hepatotoxicity 86, 93. Similarly, taking CYP‐inducing medication (such as rifampicin and CYP‐inducing anticonvulsants) does not seem to lead to an increased production of NAPQI when paracetamol is taken at therapeutic doses 94. This has led some researchers to hypothesize that glutathione depletion may be the causal factor in those few cases where therapeutic‐dose paracetamol has resulted in liver injury 91. Glutathione must be >70% deplete before NAPQI starts to accumulate, but in a starvation state (such as that seen in some alcoholics) this could occur. Indeed, it is known that chronic alcohol misuse is associated with glutathione deficiency 95. Despite these concerns, owing to its lack of a direct effect on coagulation (although there is evidence that a dose of 4 g day^–1 taken for 2 weeks raises the international normalized ratio of patients taking warfarin by ~0.8 96) and (apparent) GI safety profile, paracetamol remains the first‐line analgesic of choice for patient with chronic liver disease 97. There does not seem to be evidence for therapeutic paracetamol treatment causing hepatotoxicity, either in healthy individuals or chronic liver disease patients, with the exception of those in a poor nutritional state 98, 99.

Hepatotoxicity in children

Children metabolize paracetamol differently to adults 100, and there is some concern that they may also suffer as a result of ingestion of therapeutic doses of paracetamol. This prompted Lavonas et al. 87 to perform a systematic review in 2010, examining 62 studies and >32 000 children receiving therapeutic‐dose paracetamol (≤75 mg kg^–1 d^–1, up to a maximum of 4 g d^–1) for an average of 3–5 days. The range of settings (inpatient, outpatient, primary care, developed and developing world) and indications for paracetamol (infective illness, postoperative pain) was comprehensive. In their analysis, no child showed symptoms of liver disease, and only 10 showed any hepatic adverse events at all (incidence 0.031%, 95% CI 0.015 to 0.057%). They concluded that short‐term, therapeutic dose paracetamol is not associated with significant hepatotoxicity.

Renal effects

Acute kidney injury is said to occur in 1–2% of patients in paracetamol overdose 101 and most commonly occurs in the setting of severe paracetamol‐induced hepatotoxicity 102. Renal biopsy, while not often performed, shows evidence of acute tubular necrosis, particularly of the proximal tubule 101. While the explanation for hepatotoxicity is well known 18, the causes of renal toxicity are less clear: possible reasons include the local generation of NAPQI or other toxic metabolites from paracetamol by CYP or COX enzymes 103. Administration of N‐acetylcysteine has no effect on peak creatinine concentration, suggesting that the depletion of glutathione stores is not the sole cause of renal toxicity 104, 105. The clinical outcome from paracetamol‐induced nephrotoxicity in the absence of concomitant liver failure is good, with only 1% of patients needing temporizing dialysis and most patients returning to baseline renal function within 1 month 101.

Analgesic nephropathy is characterized by interstitial nephritis and progressive reduction in renal size due to repeated episodes of papillary necrosis. The association between the analgesic phenacetin and nephropathy was first described in 1953 106, and by the 1970s analgesic nephropathy was reported to be responsible for at least 10–20% of cases of chronic renal failure in the UK and Australia 106. Despite phenacetin's withdrawal from sale in the 1980s, analgesic nephropathy has not been eradicated, suggesting that other agents may also be responsible 107, 108. As the major active metabolite of phenacetin is paracetamol 109, some questioned whether chronic paracetamol use might also cause chronic kidney disease. In 1994, Perneger et al. 110 studied 716 subjects with end‐stage renal disease (ESRD) and found that this was associated with an increase in paracetamol use in a dose‐dependent fashion, with ~10% of the overall incidence of ESRD attributable to paracetamol use. The study unfortunately failed to adjust for possible previous use of phenacetin and NSAIDs, bringing its results into question. A large review in 2000, requested by the regulatory authorities of Germany, Switzerland and Austria, examined all published data on nonphenacetin analgesic nephropathy 111. Overall, its findings were that there was insufficient evidence to claim that nonphenacetin‐containing analgesics were causally associated with nephropathy, suggesting that further research was required 111.

Pregnancy

Paracetamol is administered to pregnant women as an antipyretic agent and for the management of mild‐to‐moderate pain. The presumed safety of this agent has resulted in paracetamol becoming one of the most common prescriptions in pregnancy: ~50–60% of pregnant women in North and Western Europe self‐report using this medication 112. Its popularity is mainly due to the recommendation of paracetamol over other analgesics, with NSAIDs having a less favourable risk profile in pregnant women, and use of aspirin limited due to concerns over its effect on the fetus 113, 114.

In recent years, the safety of paracetamol in pregnancy has come under increasing scrutiny. Paracetamol and its metabolites cross the placenta 115 and undergo different pharmacokinetic/pharmacodynamic processes in neonates than in adults; an immature glucuronide conjugation system makes the sulphation pathway the major route of metabolism in neonates 116. Paracetamol has been postulated to cause a diverse range of embryo–fetal and neonatal adverse effects, dependent on dose, duration of treatment and the trimester of exposure. However, large cohort studies have not found an association between maternal paracetamol use in the first trimester and either adverse pregnancy outcomes or congenital malformations 117, 118. Nevertheless, there is some evidence of increased risk with paracetamol use in pregnancy and neurodevelopmental disorders, respiratory illness and reproductive toxicity.

Neurodevelopmental effects

The association between paracetamol exposure in utero and the risk of long‐term neurological disorders has been the focus of several controversial pharmaco‐epidemiological studies. Brandlistuen et al. 119 suggested that maternal paracetamol use for >28 days during pregnancy was associated with problems in gross motor development, communication, externalizing and internalizing behaviour, and higher activity levels, when compared with controls. These data were obtained from a Norwegian sibling‐controlled study (n = 2919) and based on parental reports of child behaviour at 18 months and 36 months. Notably, the group also reviewed ibuprofen exposure, to control for possible confounders arising from paracetamol indication, and concluded that ibuprofen exposure was not associated with adverse neurodevelopmental outcomes. There was no relationship between trimester of exposure to paracetamol and any of the above outcomes.

Maternal paracetamol use was later linked to a neurodevelopmental clinical outcome in the Danish National Cohort Study. Liew et al. 120 suggested that maternal paracetamol use during pregnancy was associated with a higher risk of receiving a hospital diagnosis of hyperkinetic disorder (HR 1.37, 95% CI 1.19 to 1.59), use of attention‐deficit/hyperactivity disorder (ADHD) medications (HR 1.29, 95% CI 1.15 to 1.44) or having ADHD‐like behaviours at age 7 years (HR 1.13, 95% CI 1.01 to 1.27). The strengths of this flagship study lie in its large sample size (n = 64 322) and adjustment for a large number of potential confounders. Notably, these associations were stronger with increased frequency of paracetamol use and were not confounded by maternal inflammation or infection during pregnancy. Using hospital outcome coding data in the same patient cohort, the group later identified an association between prenatal paracetamol use and an increased risk of autistic spectrum disorder (ASD) accompanied by hyperkinetic symptoms (HR 1.51, 95% CI 1.19 to 1.92), but not with other ASD cases 121; other studies have suggested an association with ASD symptoms in male offspring only, with associations dependent on the frequency of exposure 122.

The mechanism by which paracetamol and its metabolites may affect neurological development is poorly understood. Animal studies have reported behavioural and cognitive changes in mice given paracetamol during neonatal brain development – specifically, locomotor activity and attainment of spatial learning 123. Levels of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=4872 (BDNF) in the neonatal brain were affected (significantly increased in the frontal, and decreased in the parietal, cortices), postulating that this may be the mechanism of action. The role of BDNF in development and brain maturation has been extensively reviewed elsewhere 124.

In conclusion, on the basis of these studies, only weak associations between paracetamol exposure and neurodevelopmental issues have been identified, and no causal link can be inferred. The epidemiological studies that support a link are subject to confounding by unmeasured environmental factors, recall bias, diagnostic inaccuracy (most rely on coding data or parental recall for their outcomes) and differences in drop‐out rates. Notably, few studies confirm the effect of duration and timing of paracetamol exposure, details that are critical in the assessment of toxicological risk in pregnancy.

Asthma

The potential mechanisms by which paracetamol may contribute to the development/exacerbation of asthma have been described earlier. How paracetamol exposure in utero could cause asthma is less clear, unless glutathione levels are lowered sufficiently in the fetus to affect lung development. Some support for maternal intake of paracetamol affecting offspring comes from mouse studies, where adult mice exposed to paracetamol in utero underwent an allergic airway challenge 125. Increased airway infiltration by leukocytes (notably eosinophils) was observed, suggesting an increased susceptibility to asthma, but this finding has not been consistently reproduced 126.

The Avon Longitudinal Study of Parents and Children was one of the first epidemiological studies to examine the causal link between paracetamol exposure during pregnancy and childhood asthma 127. Frequent paracetamol use in late pregnancy (20–32 weeks) was associated with an increased risk of wheezing in the offspring at 30–42 months (adjusted OR 2.10, 95% CI 1.30 to 3.41), particularly if wheezing started before 6 months (termed ‘persistent wheezers’ – OR 2.34, 95% CI 1.24 to 4.40). Two further cohort studies suggested that paracetamol use during any time of pregnancy was associated with a small increased risk of asthma or bronchitis among children at 18 months (RR 1.17, 95% CI 1.13 to 1.23) and 7 years (RR 1.15, 95% CI 1.02 to 1.29) 128, 129. Interestingly, maternal pain showed a positive association with asthma development without the use of paracetamol 129.

However, maternal infections, including respiratory infections, have already been associated with an increase in childhood asthma 130, 131. Paracetamol use may simply be a surrogate for these disease states. Notably, maternal paracetamol use for non‐infectious disorders revealed no increased risk of wheezing in children 132. Further studies expanded on this theme of confounding by paracetamol indication and have highlighted that the increased risk of asthma diagnosis in children exposed to paracetamol prenatally (unadjusted OR 1.36, 95% CI 1.14 to 1.61) drops significantly (OR 1.26, 95% CI 1.02 to 1.58) when adjusted for potential confounders 133. For an in‐depth review of paracetamol exposure and asthma in children, and the issue of confounding, see elsewhere 60. Further clarification of this issue will be difficult, as RCTs would be both unethical and impractical 52.

Endocrine and reproductive toxicity

The incidence of cryptorchidism is reportedly increasing, which is particularly concerning given its association with early adulthood disorders such as low sperm count and testicular germ cell cancers 134. When considered together, these conditions represent a testicular dysgenesis syndrome, a disorder related to androgen disruption during the fetal programming window 135. Experimental data have shown reduced testicular PG and http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=2858 production and reduced ano–genital distance (a marker for androgen action) in rats prenatally exposed to paracetamol 136. Prenatal paracetamol exposure is likely to result in a reduction in key steroidogenic enzymes (http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1358, http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1361), implicated in the reduced fetal plasma testosterone (45% reduction; P = 0.025) and seminal vesicle weight (18% reduction; P = 0.005) 137. These changes were noted in castrated host mice bearing human fetal testis xenografts following exposure to therapeutic doses of paracetamol for 7 days. Notably, however, exposure for 1 day had no effect 137. Another recent study has linked reduced germ cell development in human fetal testes and ovary xenografts when exposed to paracetamol; this effect was linked to PGE₂‐mediated alterations in epigenetic regulatory genes, indicating that the effect of paracetamol on the fetus may affect the genetics of subsequent generations 138.

Several clinical studies associate paracetamol exposure during pregnancy with increased occurrence of cryptorchidism, particularly when used in for >2 weeks in the second trimester 136, 139, 140. Few of these studies considered indication for paracetamol use in their analyses, and, latterly, reanalysis of these data sets showed slightly lower HRs for paracetamol exposure during weeks 8–14 among women who did not report an illness that would trigger weak analgesic use 141. This is an interesting paradoxical observation, given that this time frame represents the human fetal programming window, disruption of which has previously been linked to reduced male infant ano–genital distance 135, 142. However, we should also note that several large cohort studies have not identified any association between paracetamol and cryptorchidism 117, 143, 144, 145. Indeed, the use of paracetamol may decrease the risk of selected congenital abnormalities when used for febrile illness 144.

The continuing search for evidence that paracetamol causes harm in pregnancy clearly highlights the difficulty in implying causation from pharmaco‐epidemiological studies. Extrapolation of preclinical toxicology data to humans may suggest associations with asthma, ADHD and androgen disruption but the small associations seen in clinical cohort studies may be explained by various confounders and biases inherent in the study designs. Confidently teasing apart these issues would require RCTs, which would be difficult to perform ethically in pregnant populations. Carefully designed, long‐term, sibling‐ and sex‐matched cohort studies are more ethically acceptable, and would further our understanding of the risks. While the evidence base is uncertain, care should be taken to avoid raising poorly founded concerns among pregnant women because of the risk of switching to other analgesic/antipyretic drugs with less favourable risk profiles 113. Untreated febrile illness is associated with severe harm to both mother and child, posing a far greater risk than that postulated for paracetamol exposure 130, 146, 147, 148. Practical advice would be to avoid the protracted use of paracetamol for nonfebrile illness, a view shared by many study authors 137, 149.

Discussion

Clearly, there remains considerable uncertainty regarding the chronic adverse effects of paracetamol use. The evidence base in each of the above sections relies mostly on observational and cohort studies, and so is prone to inherent biases. The positive associations found in these studies are generally weak, and often contradictory. Few RCTs have been performed but, when undertaken, usually give reassuring results. Further studies are required in many areas, but RCTs may be difficult to perform, either because they would need to be very large to detect the modest increases in risk seen in the observational studies, or because of the significant ethical issues of using placebo in patients in pain, as well as of conducting trials in children and pregnant women.

The two areas in which the evidence is most convincing are hypertension and GI bleeding. A small BP rise of 4 mmHg would be clinically important at the population level, and the outcome of ongoing RCTs should clarify the reliability of this estimate. This may be particularly important in patients with angina or pre‐existing hypertension. The fairly consistent evidence for GI bleeding associated with paracetamol use, along with its additive effect when combined with NSAIDs, may be less well known but similarly important. When considering prescribing paracetamol in the chronic setting it would seem wise to consider these adverse effects, based on current data, and discuss them with the patient.

Whether paracetamol use in the chronic setting should be restricted is doubtful, given that the alternatives are NSAIDs and opioids. Indeed, in patients intolerant of NSAIDs, their next option would be opioid medication, which comes with risks of addiction, drowsiness and fatal accidental overdose.

In summary, the average therapeutic effect for chronic pain syndromes is small, but there is accumulating evidence of clinically significant adverse effects in chronic use. Despite this, for patients who derive clear symptomatic benefit, or only take occasional therapeutic doses, the risks are probably very small. For this reason, paracetamol can be seen as the ‘least‐worst’ option – which probably means that it will remain, for now at least, the first‐line analgesic of choice.

Conclusion

The present review is designed to provide an objective summary of the evidence base for chronic adverse effects of paracetamol use. We hope that by highlighting the key epidemiological studies, RCTs, meta‐analyses and reviews, we have provided a valuable summary of knowledge in this field. We hope this work will help clinicians and their patients to make an evidence‐based, informed decision regarding their chronic pain management, based on the likelihood of clinically relevant adverse effects.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 150, and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 151, 152, 153.

Competing Interests

There are no competing interests to declare.

McCrae, J. C. , Morrison, E. E. , MacIntyre, I. M. , Dear, J. W. , and Webb, D. J. (2018) Long‐term adverse effects of paracetamol – a review. Br J Clin Pharmacol, 84: 2218–2230. 10.1111/bcp.13656.