Chronic pain in critical care survivors: a narrative review

Summary

Chronic pain is an important problem after critical care admission. Estimates of the prevalence of chronic pain in the year after discharge range from 14% to 77% depending on the type of cohort, the tool used to measure pain, and the time point when pain was assessed. The majority of data available come from studies using health-related quality of life tools, although some have included pain-specific tools. Nociceptive, neuropathic, and nociplastic pain can occur in critical care survivors, but limited information about the aetiology, body site, and temporal trajectory of pain is currently available. Older age, pre-existing pain, and medical co-morbidity have been associated with pain after critical care admission. No trials were identified of interventions to target chronic pain in survivors specifically. Larger studies, using pain-specific tools, over an extended follow-up period are required to confirm the prevalence, identify risk factors, explore any association between acute and chronic pain in this setting, determine the underlying pathological mechanisms, and inform the development of future analgesic interventions.

Over the last two decades, mortality during admission to intensive care has decreased.1, 2, 3 As a result, the challenges of ‘ICU survivorship’ have become an increasingly important focus.⁴ With between 216 and 2353 admissions to intensive care per 100 000 population in Europe and North America,⁵ and more than 270 000 UK critical care admissions in 2015–6,⁶ the long-term sequelae of ICU admission represent a significant health burden. Many long-term survivors of critical care admission display physical, mental, and cognitive impairments, now recognised as post-intensive care syndrome (PICS).⁷ Although there is no official definition of PICS, it is generally recognised as including new or worsening dysfunction in one or more of the following domains:

• (i)Physical function: including pulmonary or neuromuscular impairment, and reduced ability to walk or carry out activities of daily living
• (ii)Cognitive function: impairment in executive function, memory, mental processing, and visual–spatial perception
• (iii)Psychological function: depression, anxiety, and post-traumatic stress disorder (PTSD)

There is increasing evidence that chronic pain in ICU survivors may be an important mediator for the physical, cognitive, and psychological symptoms and signs associated with PICS, yet until recently the role of chronic pain in this cohort has received little attention. This article aims to provide a narrative review of the evidence for the prevalence, characteristics, aetiology, and therapeutic interventions for chronic pain after ICU admission, and to summarise future research priorities in this condition.

Methods

This narrative review consists of two components. Firstly, a literature search was conducted including PubMed, Medline, and Embase databases using the following study terms: ‘chronic pain’ or ‘quality of life’ AND ‘intensive care’ or ‘critical care’ or ‘survivors’. Only English language publications involving studies with adult humans were included. Studies conducted since January 1, 2000 were included to reflect contemporary critical care practice, but where these studies referenced older studies, these were also assessed for relevance (date of search: October 11, 2018). This was not intended to be a full systematic review addressing a specific question, as the aim was to provide a broad description of the topic and to highlight important areas for future research.

In an effort to increase the data available for this review, and to act as an illustration of data that may be available for secondary analysis, principal investigators of 25 recent UK-based critical care studies (published in the last 10 years), which included long-term follow-up and the collection of pain symptoms as part of a EuroQol-5D⁸ (EQ-5D) instrument, were contacted by the senior author. Whilst we acknowledge that there are likely differences globally, UK-based studies were selected, as it was anticipated that patients included would have access to broadly similar treatment pathways within the National Health Service after discharge (although wide regional variation is acknowledged), for example, access to general practitioners, pharmacies, chronic pain clinics, and emergency services for management of chronic pain. Principal investigators (PIs) were invited to submit ‘pain/discomfort’ EQ-5D data from baseline and any other subsequent time points. Data reported using the 5L format of the EQ-5D were converted to the 3L format as follows: ‘no pain or discomfort’ remained the same, ‘slight’ and ‘moderate’ pain or discomfort were converted to ‘moderate’ pain, and ‘severe’ and ‘extreme’ pain converted to ‘extreme’ pain. As this was not a systematic search for data, a formal meta-analysis was not performed, but data were summarised in a proportional format for prevalence of symptoms for each study cohort.

Prevalence

Chronic pain after intensive care admission was first described in cohort studies examining health-related quality of life (HRQoL) in the months and years after ICU discharge. Whilst the majority of these studies did not collect pain-specific outcome data, pain was reported as a commonly problematic HRQoL domain9, 10, 11, 12, 13, 14, 15, 16, 17, 18 with tools, such as the EQ-5D⁸ and the 36-Item Short Form Health Survey (SF-36).¹⁹ A limited number of studies have been conducted with the primary aim of exploring chronic pain in ICU survivors.20, 21, 22, 23, 24, 25, 26 One such study²⁴ has recently suggested the use of the term ‘chronic ICU-related pain’ (CIRP) to describe pain that persists for at least 3 months post-ICU discharge, in keeping with the International Association for the Study of Pain (IASP) definition of chronic pain,²⁷ and this will be used as the definition to discuss prevalence in this review.

Estimates of the prevalence of chronic pain in ICU survivors range from 14% to 77% depending on the population, the time point when outcomes were evaluated post-discharge, and the type of tool used.11, 14, 16, 17, 20, 22, 23, 24, 25, 28, 29, 30, 31, 32 In common with other long-term outcomes in ICU survivors, it is important to appreciate the premorbid state of the patient, so that the impact of the intensive care admission itself on the development, or mediation, of chronic pain can be elucidated.

Comparison with the premorbid state is challenging, as ICU admission is often not planned, thereby preventing prospective pre-admission data collection. As a consequence, the reporting of the pre-admission presence or absence of pain relies either on the retrospective recall by the patient, or on the proxy reporting by relatives or friends. Both methods may be subject to bias for many reasons, and it is not clear which is most reliable.33, 34 Studies that use these methods to specifically define ‘new onset’ chronic pain after ICU admission report prevalence of between 22% and 33%.22, 24

Chronic pain is also a condition that affects a significant proportion of the general population³⁵; therefore, to infer a greater prevalence in ICU survivors, comparison with a control group where other risk factors, such as age and co-morbidity profile, are adjusted for is required. Comparison with appropriate control groups has been performed in several studies examining the long-term consequences of ICU admission, as normative data sets exist for the SF-36 and EQ-5D, but in studies where pain was the main outcome of interest, none have attempted to directly compare the prevalence of pain to matched control groups.16, 21, 22, 24, 25, 26 Data from simple ‘population normal’ studies are of arguable value, as the co-morbidity burden, which predisposes to emergency admission to critical care, is substantial, rendering the critical care population not ‘normal’. Thus, estimates from simple case control studies are unlikely to be very accurate.

With the preceding caveat in mind, in line with other HRQoL domains, pain does appear to be significantly more prevalent in the post-ICU population than in normative data sets.¹³ Using the EQ-6D in a Dutch population, Timmers and colleagues¹⁴ identified that pain and discomfort were the most frequently reported domain, with 57% of patients reporting problems compared with 34% in the general population. However, Orwelius and colleagues³⁰ showed that the effect of ICU admission on the proportion of patients reporting pain (44.3% of the post-ICU population vs 25.7% of those that were not admitted to the ICU) was mitigated once the number of pre-existing co-morbidities was accounted for in the reference population. A systematic review of HRQoL after ICU discharge, including 48 studies,¹⁸ identified a significant decrement in pain domains compared with population norms when using the EQ-5D at 12 months (20 studies). However, when the SF-36 was used (30 studies), this decrement was not significant, and it is unclear whether the different results can be accounted for by the choice of outcome measure or other study-specific characteristics.

Core outcome sets have been defined for research in rehabilitation of ICU survivors and acute respiratory failure survivors specifically.36, 37 Although these recommendations stress the importance of including measures for recording pain, no validated pain measures, other than a VAS, are specifically recommended.36, 37 This may be attributable to the fact that so few studies exist using pain-specific tools common in chronic pain research in ICU survivor populations. It is generally considered that the use of the VAS as a measure of pain intensity fails to evaluate the multidimensional complexity of a condition, such as chronic pain.³⁸

In studies where pain has been at the centre of the hypotheses, the Brief Pain Inventory (BPI)³⁹ has been used most frequently, and one study each reported the German Pain Questionnaire⁴⁰ and the Pain Self-Efficacy Questionnaire.⁴¹ Needham and colleagues³⁷ reflect that pain-specific measures, in addition to core HRQoL measures, should be included in the future research agenda.³⁷ Such research will undoubtedly require joint working between clinical and basic science researchers from both chronic pain and intensive care.

Pain intensity

Pain intensity has long been recognised as a core outcome in clinical research,⁴² and it is advised that severity is recalled over different periods of time and with respect to different functional domains, for example, with the Von Korff and colleagues⁴³ pain grading or the chronic pain grade questionnaire⁴⁴; however, this level of detail has rarely been reported in studies involving ICU survivors.

One study that did use the Von Korff and colleagues⁴³ grading revealed that more than 50% of patients with CIRP reported high disabling pain with moderate (21.2%) to severe (36.4%) limitation in daily living.²⁴ Several studies do report some other measures of pain intensity. Moderate-to-severe pain, measured on the BPI, was reported in 31% of 295 medical and surgical ICU patients at 3 months post-discharge, which increased to 35% at 1 yr.²⁵ Marx and colleagues²¹ also used the BPI, and reported 44% of sepsis patients scoring pain intensity of 0–3/10 and 56% 4–10/10. From the EQ-5D, it is possible to determine whether patients report pain as ‘slight’, ‘moderate’, ‘severe’, or ‘extreme’ if the 5L version is used (or ‘no’, ‘moderate’, or ‘extreme’ pain if the 3L version is used). Although most studies do not describe this level of detail, Granja and colleagues⁴⁵ identified 33% of their cohort as reporting moderate and 7% extreme pain/discomfort on EQ-5D.

From the SF-36, it is also possible to determine a numerical ‘bodily pain’ 0–100 score, where higher values indicate a more favourable status. Although this cannot be taken as a proxy for pain intensity, it could provide quantitative information about pain impact. Herridge and colleagues⁴⁶ reported a median bodily pain SF-36 score of 62/100 and 74/100 at 1 and 5 yr, respectively. A pooled estimate of acute respiratory distress syndrome (ARDS) studies reporting the SF-36 calculated a decline in bodily pain score over 1–4 yr post-discharge of 20.⁴⁷ It is not clear to what extent such changes in a numerical HRQoL bodily pain subdomain infer a clinically meaningful change in ICU survivors.

To increase our understanding of pain prevalence and severity in ICU survivors across a range of cohorts, we have collated EQ-5D scores from more than 12 000 patients in recently published UK-based studies (Table 1; Fig. 1). This was to illustrate data relating to pain or discomfort that were collected, but may not have been published in the original reports. Of the 25 studies identified, the investigators were able to contribute data from 17 (68% response rate). The data indicate that at least moderate pain or discomfort is common during the first year after ICU admission across many different patient cohorts, and that a variable proportion report extreme symptoms. The Optimisation of Cardiovascular Management to Improve Surgical Outcome (OPTIMISE) study is illustrative of a study that recruited only surgical patients, yet it does not appear to stand out as having a very different proportion of patients with pain after discharge.⁴⁸

Table 1.

Summary of studies contributing data to the evaluation of reported pain symptoms shown in Figure 1. All studies were conducted in the UK, and recruited patients were from units in the UK or The Republic of Ireland. CCU, critical care unit; COPD, chronic obstructive pulmonary disease; DOH, Department of Health; EME, Efficacy and Mechanism Evaluation programme; ESICM, European Society of Intensive Care Medicine; GCS, Glasgow Coma Scale; HFOV, high-frequency oscillatory ventilation; HRB, Health Research Board; HSC R&DD, Health and Social Care Research & Development Board; ICHT, Imperial College Healthcare NHS Trust; ICIS, Intensive Care Society of Ireland; ICNARC, Intensive Care National Audit & Research Centre; MRC, Medical Research Council; NHS, National Health Service; NIHR BRC, NIHR Biomedical Research Centre; NIHR-CS, NIHR Clinician Scientist Award; NIHR-HTA, National Institute of Health Research Health Technology Assessment programme; OPTIMISE, Optimisation of Peri-operative Cardiovascular Management to Improve Surgical Outcome; RfPB, Research for Patient Benefit programme; R&D, Research and Development; TBI, traumatic brain injury

Study	Funding	Population	Intervention	Comparator	Outcome	Methods
Harvey and colleagues49 (2016) (CALORIES)	NIHR-HTA	2400 adult patients (329 surgical); unplanned CCU admissions	5 days early parenteral nutritional support, n=1200	5 days early enteral nutritional support n=1200	No significant difference in 30 day all-cause mortality	Multicentre RCT
Mouncey and colleagues50 (2015) (ProMISe)	NIHR-HTA	1260 adult patients (no surgical patients); septic shock	Early goal-directed therapy, n=630	Usual resuscitation, n=630	No significant difference in 90 day all-cause mortality	Multicentre RCT
Pearse and colleagues48 (2014) (OPTIMISE)	NIHR-CS	734 adult patients (all surgical); >50 yr; high risk undergoing major gastrointestinal surgery	Cardiac-output-guided haemodynamic therapy algorithm for i.v. fluid and inotrope (dopexamine) infusion, n=368	Usual care, n=366	No significant difference in 30 day mortality, complications	Multicentre RCT; updated systematic review and meta-analysis, including RCT data
Harrison and colleagues51 (2013) (RAIN)	NIHR-HTA	3210 adult patients; admitted to the CCU after actual/suspected TBI; GCS score of <15	N/A	Comparison of type of critical care unit where care was delivered	Results suggest neuro-CCU management may be cost-effective	Multicentre, prospective cohort study
Wildman and colleagues52 (2009) (CAOS)	MRC Health Services Research Fellowship; NHS R&D grant	832 adult (no surgical patients); >45 yr old; breathlessness, respiratory failure, or change in mental status attributable to an exacerbation of COPD, asthma, or both	N/A	N/A	Pre-CCU data can aid prediction of 180 day mortality; high burden of symptoms at 180 days post-admission	Multicentre, prospective observational cohort study
Griffiths and colleagues53 (2013)	DOH; Oxford Health Charity; NIHR BRC (Imperial and Oxford)	293 patients (173 medical, 89 surgical, 30 trauma, and 1 unknown); >16 yr	N/A	N/A	12 months after discharge negative impact on employment and negative impact on family income	Multicentre, prospective cohort study
Ji and colleagues54 (2017) (PARAMEDIC-1)	NIHR-HTA	1035 adult patients (no surgical patients); out-of-hospital cardiac arrest	LUCAS mechanical chest compression device, n=658	Manual chest compressions, n=377	No clinically important differences in long-term outcomes	Multicentre, cluster randomised trial
Lall and colleagues55 (2015) (OSCAR)	NIHR-HTA	795 adult patients (108 surgical); mechanically ventilated patients with ARDS	HFOV (Novalung R100 ventilator) until the start of weaning, n=398	Conventional ventilation, n=397	HFOV had no effect on 30 day mortality and no economic advantage	Multicentre RCT
Perkins and colleagues56 (2018) (BREATHE)	NIHR-HTA	364 adult patients (no surgical patients); mechanically ventilated and failed spontaneous breathing trial failed	Protocolised weaning via early extubation to non-invasive ventilation, n=182	Protocolised standard weaning, n=182	Early extubation to non-invasive ventilation (intervention group) did not shorten time to liberation from any ventilation	Multicentre, randomised, allocation concealed, open-label clinical trial
Page and colleagues57 (2017) (MoDUS)	NIHR	142 adult patients (no surgical patients); mechanically ventilated	Simvastatin 80 mg daily (<28 days), n=71	Placebo daily, n=71	No difference in duration of delirium and coma	Multicentre RCT
Pattison and colleagues58 (2018)	Supported by The Royal Marsden BRC infrastructure	50 adult cancer patients	Diary administration	N/A	Diaries can be a means of emotional support for patients post-ICU discharge	Single centre; mixed (qualitative and quantitative); observational
Pattison and colleagues59 (2015)	ESICM, European Critical Care Research Network, Edwards Nursing Science Award	77 adult cancer patients; critical care stay >48 h	N/A	N/A	Patient shaped by ongoing illness and treatment	Single centre; mixed (qualitative and quantitative); observational
Page and colleagues60 (2013) (HOPE-ICU)	NIHR	141 adult patients (50 surgical); mechanically ventilated	Haloperidol, irrespective of coma or delirium status, n=71	Placebo: saline IV TDS 0.9%, n=71	Haloperidol did not modify duration of delirium in critically ill patients	Single-centre RCT
Batterham and colleagues61 (2014) (PIX)	NIHR RfPB	59 adult patients (8 surgical); mechanically ventilated	Supervised, hospital-based aerobic training intervention, n=29	Usual care, n=30	Accelerated short-term natural recovery process in intervention group; feasible	Multicentre, exploratory parallel-group controlled trial
McAuley and colleagues62 (2014) (HARP-2)	NIHR-EME, HRB, HSC R&DD, ICSI, REVIVE	540 adult patients; ARDS	Simvastatin 80 mg daily (maximum of 28 days), n=259	Placebo once daily, n=281	Simvastatin did not improve clinical outcomes in ARDS	Multicentre RCT
Hatch and colleagues63 (2017) (ICON)	Intensive Care Foundation; Bupa Foundation	Pain data provided for 7325 adult patients; medical and surgical	Questionnaire follow-up for responses at either 3, 12, or 24 months	N/A	Questionnaire burden; had no effect on return rates; affected answers to the same questionnaire (EQ-5D-3L)	Multicentre, parallel-group RCT
Wright and colleagues64 (2018) (EPICC)	NIHR RfPB	308 adult patients (148 surgical); invasive or non-invasive ventilation	Target of 90 min of physical rehabilitation per weekday, n=150	Target of 30 min of physical rehabilitation per weekday, n=158	ICU-based physical rehabilitation did not improve physical outcomes at 6 months	Multicentre, parallel-group RCT

CALORIES, Trial of the Route of Early Nutritional Support in Critically Ill Adults; ProMISe, Protocolised Management in Sepsis; RAIN, Risk Adjustment in Neurocritical Care; CAOS, COPD and Asthma Outcome Study; PARAMEDIC-1, pre-hospital randomised assessment of a mechanical compression device in cardiac arrest; OSCAR, OSCillation in ARDS; BREATHE, Extubation to Non Invasive Ventilation for difficult to wean patients; MoDUS, Modifying Delirium Using Simvastatin; HOPE-ICU, HalOPEridol Effectiveness in ICU delirium; PIX, Post Intensive care eXercise; HARP-2, Hydroxymethylglutaryl-CoA reductase inhibition with simvastatin in Acute lung injury to Reduce Pulmonary dysfunction; ICON, Intensive Care Outcomes Network; EPICC, Intensive versus standard physical rehabilitation therapy in the critically ill.

Fig 1.

Proportion of subjects reporting no, moderate, and extreme pain or discomfort after ICU admission. *Indicates that the authors did not have access to the number of subjects alive at the time point. The detailed reasons for missing data were also not known to the authors. BL, baseline; DC, at discharge; EGDT, early goal-directed therapy; EN, enteral nutrition; HFOV, high-frequency oscillatory ventilation; Int, intervention; IV, invasive ventilation; LUCAS, LUCAS Chest Compression System; M, month; NIV, non-invasive ventilation; PN, parenteral nutrition; W, week. Full references for included studies are indicated in Table 1.

Temporal trajectory of pain

The trajectory of chronic pain after discharge is also unclear. Although one meta-analysis reported improvement in pain by 7 months when described using the SF-36, it also indicated that there was no significant improvement in pain at 12 months when EQ-5D results were analysed.¹¹ Similar EQ-5D results were reported in the subpopulation analysis of patients with ARDS.⁴⁷ In a more recent systematic review, little change in SF-36 bodily pain scores was identified after 9 months, yet on EQ-5D the proportion of patients reporting pain increased until 1 yr, after which no data were reported.¹⁸ Hayhurst and colleagues²⁵ demonstrated a reasonably stable proportion of patients reporting pain, and consistent intensity and interference scores over the first year post-discharge using the BPI.

A growing number of studies with longer follow-up time (>1 yr) after discharge are being reported. Cuthbertson and colleagues¹³ demonstrated a trajectory using SF-36 bodily pain scores, where pain initially improves at 1 yr, but then deteriorates again at 5 yr. However, at 5 yr, Herridge and colleagues⁴⁶ reported a slightly improved median (SF-36) pain score compared with 1 yr. One study in a surgical population reported an incidence of chronic pain of 57% as long as 8 yr after discharge, but did not report whether this was significantly higher than the control population.¹⁴

Pain trajectory may well be population specific, and cardiac ICU admission in particular appears to be unique. A multicentre cohort study of cardiac ICU patients reported an incidence of chronic pain of 40% at 3 months that decreased to 9.5% by 2 yr.²³

Pain aetiology

The cause of chronic pain in ICU survivors is largely inferred from studies examining the sites of pain and risk factors associated with the onset of pain. There is evidence for the presence of nociceptive, neuropathic, and nociplastic pain in ICU survivors. Nociplastic pain is a new mechanistic descriptor adopted by the IASP to define pain that ‘arises from altered nociception despite no clear evidence of actual or threatened tissue damage causing the activation of peripheral nociceptors or evidence for disease or lesion of the somatosensory system causing the pain’.⁶⁵ This type of pain would include those reporting more non-specific, diffuse pain with no obvious cause for the activation of nociceptors or neuropathy.

Potential sources of acute pain are frequent in the ICU, and it is hypothesised that sustained, under-treated acute pain may contribute to a transition to chronic pain, similar to the process recognised in chronic post-surgical pain.⁶⁶ One study has shown that those with CIRP reported higher pain intensity and distress scores when recalling their ICU experience,⁶⁷ but, as yet, there is little evidence linking acute objective pain scores during admission to the development of chronic pain post-discharge. The ability to link acute and chronic pain in the ICU may be in part hampered by an infrequent or inappropriate assessment of pain during ICU admission,⁶⁸ although it is hoped that updated guidelines⁶⁹ and prioritising acute pain assessment in the critically ill may improve this.

In a single-centre general ICU questionnaire-based study, Battle and colleagues²² reported that the most frequently reported site of pain at 6 months post-discharge was the shoulder (22%), indicating a higher prevalence of chronic shoulder pain than that identified in the general UK population (11.7%).⁷⁰ Limitation of shoulder movement was also commented on by Herridge and colleagues⁴⁶ in their study of long-term follow-up of patients with ARDS. Recently, Gustafson and colleagues¹¹⁶ sought to identify the prevalence and risk factors for shoulder impairment post-ICU admission. More than 50% reported shoulder pain at 6 months, and 95% had reduced range of movement.

The other body sites highlighted as painful by Battle and colleagues²² included the lower limb (9%), lumbar spine (9%), cervical spine (6%), upper limbs (6%), abdomen (4%), and pelvis (3%).

Sepsis was also found to associate with the development of chronic pain, and it was suggested that increased time on inotropic support could lead to later instigation of rehabilitation. The inflammatory state associated with sepsis also mediates neurological damage, as demonstrated in critical illness polyneuropathy (CIP).71, 72 Many of the small molecules important in inflammation, including neuropeptides and cytokines, are pro-nociceptive. They could lead to the development of chronic pain through altering membrane excitability, descending nociceptive control, and synaptic plasticity.66, 73 ICU-related weakness secondary to immobility, CIP, and critical illness myopathy (CIM) can result in rapid deconditioning and potentially joint-related pain⁷⁴ and contractures.22, 75

The hypothesised microcirculatory and metabolic derangement identified in CIP can affect sensory and motor transmission.⁷⁶ Sensory impairment, such as numbness and tingling in glove and stocking distributions, has been reported after ICU discharge and suggest a neuropathic origin.76, 77, 78 Abnormal nerve conduction studies (NCS) have been reported up to 5 yr after ICU admission.79, 80 Sensory symptoms and signs can also be identified in the absence of abnormal neurophysiological testing, suggesting small fibre impairment, and have been shown to persist for up to 2 yr.⁷⁸

Recently, three studies have sought to find evidence of structural abnormalities of small nerve fibre architecture through the assessment of intra-epidermal nerve fibre density (IENFD) on lower limb skin biopsy in ICU patients. Latronico and colleagues⁸¹ described reduced IENFD in 14 patients a month after admission, despite a range of NCS findings and not all exhibiting features of CIP or CIM. Skorna and colleagues⁸² performed serial skin biopsies in patients with normal NCS on ICU admission, and also showed a reduction in IENFD. However, only one study performing biopsies a significant time after discharge has been published⁸³ and included only four patients at four months. It could not show a significant reduction in IENFD when compared with normative data, despite patients reporting paraesthetic symptoms. All of these studies were small, had a relatively short follow uptime and collected only limited patient-reported outcome measures with respect to neuropathic symptoms.

Although IENFD provides an objective structural measure of small fibre pathology, in some chronic pain conditions it has been shown to correlate poorly with patient-reported symptoms.⁸⁴ Quantitative sensory testing (QST) is commonly used in neuropathic pain to provide an assessment of small fibre function,85, 86 and in some conditions can be correlated to IENFD.⁸⁷ In a recent study, Baumbach and colleagues⁸⁸ sought to determine evidence of small nerve fibre dysfunction using QST in 84 critical care survivors at 6 months post-discharge. Small fibre dysfunction was detected in 42.5% of patients, and was associated with significantly higher pain intensity, increased pain-related disability, and decreased HRQoL.

These studies suggest a role of small fibre dysfunction in CIRP, but larger studies using a more complete battery of tests and patient-reported outcomes for neuropathy and associated neuropathic pain are needed.

It is not clear if ICU survivors report multiple sites of pain. Generalised or multisite chronic pain is a common phenomenon observed in certain chronic pain populations,89, 90 and suggests a central augmentation of sensory processing.

Potential risk factors

The presence of pre-existing painful conditions has been associated with the development of new onset chronic pain after ICU admission.23, 24 This reflects evidence from general chronic pain populations that show high proportions of coexisting chronic pain conditions,⁹¹ but findings within ICU survivors are not consistent.²² The contribution of the number of pre-admission co-morbidities to long-term HRQoL after ICU admission is well documented,15, 30, 92 and Griffith and colleagues¹⁷ were able to demonstrate that pain specifically was associated with more medical co-morbidity. There is conflicting evidence to associate traditional measures of illness severity in ICU patients, such as number of days of mechanical ventilation20, 22 and length of stay on ICU14, 22, 23 with chronic pain. Acute Physiology and Chronic Health Evaluation score has not been associated with chronic pain,¹⁷ whilst an increased length of hospital stay has.20, 22

Several features of PICS,⁷ such as depression, anxiety, PTSD, and cognitive impairment, have well-recognised co-prevalence with chronic pain.91, 93, 94, 95 The relationship between pain and psychological distress is, however, bidirectional, which makes determining causality complex. ICU trauma patients with chronic pain report more symptoms of PTSD, depression, and anxiety,³¹ and patients with chronic pain after cardiac surgery self-report lower mental health scores in the SF-36 a year after discharge.⁹⁶ However, relationships between those experiencing chronic pain and adverse psychological outcomes after ICU admission are yet to be properly explored, making it difficult to tease out the influence of the ICU environment, the cause of ICU admission, psychological factors, and pain on each other.

Battle and colleagues²² were able to demonstrate on multivariate analysis that older age was associated with CIRP at 6 months. Although the influence of age on functional outcomes after ICU has been explored, pain, as assessed by HRQoL measures, was previously shown not to be influenced by age.¹¹ This is in contrast to the association demonstrated between younger age and chronic postoperative pain,⁹⁷ which has also been identified in cardiac post-surgical patients.²³ A frequently cited hypothesis for this is that there is an age-dependent decline in peripheral nociceptive function; however, in ICU, other factors associated with age may be more important, such as reduced muscle mass or poorer nutritional status.17, 97

In recent years, an emphasis has been placed on the role of opioid exposure as a risk factor for the development of CIRP. In a general ICU population, exposure to opioids during admission was shown not to be significantly associated with the development of pain or the severity of pain interference.²⁵ There is evidence from surgical populations that pre-surgical opioid use is associated with an increased risk of postoperative opioid use and misuse,⁹⁸ and a large retrospective epidemiological study in Canada identified pre-admission opioid use (in 6.2% of admissions) was a significant predictor of opioid use at 1 yr post-ICU discharge.⁹⁹ A further study showed that 5% of the elderly population admitted to the ICU was taking chronic opioids and that 63% continued to use them post-discharge. Those most likely to continue opioids were medical patients who were using fentanyl and benzodiazepines at admission.¹⁰⁰

Specific ICU populations

Longitudinal cohort studies to identify risk factors for CIRP have indicated that a diagnosis of sepsis is associated with the development of chronic pain²² and more severe symptoms,¹⁶ although the underlying mechanisms remain unclear. However, a recent case-control study, comparing more than 200 sepsis and non-sepsis patients, failed to demonstrate a significant difference in prevalence of new-onset chronic pain at 6 months post-discharge,²² and in another cohort, an admission diagnosis of sepsis was not shown to confer higher risk than trauma for the development of chronic pain.²⁹ Marx and colleagues²¹ sought to characterise chronic pain in survivors of sepsis, and showed relatively high pain interference scores for interference with work (mean: 5.1/10) and general activities (mean: 4.2/10). Although it might be assumed that surgical patients would be at higher risk of CIRP as a result of an acutely painful intervention, Soliman and colleagues¹⁶ demonstrated that medical patients reported the lowest HRQoL at 1 yr, and that the greatest contributors to this decrement were pain and discomfort. However, this study did not control for pre-existing co-morbidities.

Survivors of ARDS are often studied as a discrete cohort for long-term follow-up. In an ARDS-specific meta-analysis of HRQoL (13 cohorts),⁴⁷ patients reported more pain than the general population, and showed that pain continues to deteriorate at least to 12 months post-discharge. However, a more recent study, reported by Herridge and colleagues⁴⁶ showed an improvement in pain between 1 and 5 yr.

The systematic review by Dowdy and colleagues¹¹ identified that trauma patients reported more severe pain and discomfort in the EQ-5D compared with other ICU survivors, and Timmers and colleagues¹⁴ identified that 60% of trauma patients reported pain at a median of 8 yr follow-up, the highest prevalence of all subgroups of a Dutch surgical ICU population. Pavoni and colleagues¹⁰¹ also reported a high prevalence (79%) of chronic pain 1 yr after ICU admission for severe burns.

Persistent postoperative pain after cardiac surgery, commonly involving critical care admission, has been well investigated and is comprehensively discussed elsewhere.¹⁰² A recent systematic review¹⁰³ highlighted that a pooled estimate of 37% of patients reported persistent pain at 6 months and 17% at least 2 yr after cardiac surgery, that neuropathic pain was common, and that more than half of those with pain experienced moderate-to-severe symptoms. An association between length of stay in the ICU and the severity of persistent pain after cardiac surgery has also been demonstrated, suggesting that either length of stay was a proxy for disease complexity or severity, or that ICU-specific features in the patient journey have an important influence on pain outcome.

Impact of chronic pain

The impact of chronic pain after ICU has mostly been explored through the use of measures of functional impairment and healthcare utilisation. Hayhurst and colleagues²⁵ reported a high proportion of patients describing pain-related functional interference with daily activities at 3 and 12 months (59% and 62%), but interference was generally mild (median of 2/10), lower than that reported by Marx and colleagues²¹ (mean interference with daily activities: 5.1/10). Thirty-two per cent of those reporting pain at 6 months in the UK-based Battle and colleagues²² study reported seeking the advice of a healthcare professional to manage their pain, and 45% of participants in a Dutch cohort study had sought access to a multidisciplinary programme to address their functional needs, including pain by 3 months.¹⁰⁴

Interventions to manage chronic pain

To our knowledge, no RCTs have been conducted with the specific aim of reducing chronic pain after ICU admission. Theoretically, interventions could be targeted during acute admission to modify potential risk factors, such as acute pain, or during interaction with healthcare professionals as part of post-ICU follow-up. The evidence for early and post-discharge rehabilitation interventions on long-term outcomes has been subject to a review of systematic reviews.¹⁰⁵ This identified that patients may derive short-term benefits from early interventions, and that there was insufficient evidence to support post-discharge interventions; however, pain was not mentioned as a specific outcome in this review. Frustratingly, although the majority of the original studies included collected pain data in the form of HRQoL indices, very few specifically reported pain subdomains or perform statistical analysis on the data. Walsh and colleagues¹⁰⁶ were an exception to this, but showed no significant difference in pain scores up to 1 yr between ICU subjects that underwent an enriched, individualised exercise and nutritional rehabilitation programme, and those following usual care. The lack of availability of pain-specific data may change, as the recommendations for core outcome sets to include pain measures are integrated into new study protocols.36, 37

The development of interventions targeted at pain is likely limited by the lack of evidence for the pathophysiological mechanisms involved and as yet conflicting evidence for modifiable risk factors associated with the development of pain specific to this population. Transferring any post-discharge intervention into clinical practice is also likely to be impeded by the inconsistent provision of ICU follow-up services,¹⁰⁷ despite the National Institute for Health and Care Excellence (NICE) recommendation¹⁰⁸ and the fact that pain specialists are rarely included in such services.

Future research priorities

In common with all research investigating the long-term consequences of ICU admission, the loss to follow-up attributable to mortality and disengagement with outpatient services can impact on response rates, meaning that very large initial sample sizes may be required to detect statistically meaningful effects. There is also likely to be a bias in those that do respond, dependent on the research question. Similarly, the challenge of collecting pre-admission pain-related information, potentially both patient-reported and objective measures, such as QST, must be overcome or accounted for, and the use of patient recall or proxy reporting of chronic pain investigated.

Data from very-long-term follow-up are emerging, but are still lacking, especially with respect to validated pain measures. If cancer survivorship is used as a comparison, long-term follow-up studies commonly start at 5 yr. Ideally, longitudinal cohorts should incorporate multi-aetiology and multicentre designs to allow for assessment of the impact of disease- and service-specific factors on the development of chronic pain. Certain ICU cohorts, such as those with traumatic brain injury or cognitive impairment, are often excluded from long-term follow-up studies, but chronic pain may well be particularly prevalent in those with neurological pathology. Innovative studies to include such populations are required to estimate the prevalence and identify risk factors to inform the design of interventions.

The accurate and interpretable recording of multisite pain is challenging, yet is likely to be important in this population. Novel web-based or electronic forms of recording painful sites may be useful to assess the spatial burden of pain and aid the determination of pain aetiology. Although pain measures, such as VAS, for pain intensity have been recommended for clinical research in ICU survivors,³⁶ studies, including other validated pain measures, commonplace in the chronic pain research arena, to assess pain characteristics and impact could provide valuable information about pain aetiology and functional impairment, and reflect more comprehensively the complex condition that is chronic pain. The reporting of subdomains of commonly collected HRQoL data relating to pain could also improve our understanding of pain without increasing the data collection burden. Concluding which measures are most relevant in this unique cohort will require further investigation and a consensus agreement across specialists in both the pain and critical care fields.

Chronic pain has the potential to mediate all areas of PICS and the complex interplay between the cognition, psychological, and social sequelae, and physical impairment, in this cohort. This needs to be explored further in order to provide patients with a holistic approach to rehabilitation. An awkward confounder is the quantity and effectiveness, or otherwise of pain medication and services that patients may have accessed; being free of pain, but taking large doses of analgesics is not the same thing as being genuinely free of pain, but may result in the same score being entered into a questionnaire or scale.

There is currently limited evidence linking acute and chronic pain in ICU survivors. It is recognised that good acute pain assessment and management are associated with improved acute outcomes,109, 110, 111, 112 yet several studies have identified that acute pain management in ICU remains poor in many parts of the world.68, 113, 114, 115 A coordinated approach in response to new acute pain management guidelines⁶⁹ could alter the trajectory of chronic pain in this population.

Larger deep profiling studies incorporating subjective and objective measures in the same subjects are also needed to provide information on pain-generating mechanisms contributing to symptoms in the ICU survivors. It is likely that the mechanisms are heterogeneous given the diversity of the population, admitting diagnoses, and the different courses of ICU treatment. Methods for identifying those at risk of developing chronic pain and stratifying this heterogeneity could lead to clinically relevant tools to improve pain management during acute and chronic phases of the patient pathway. Chronic pain in the ICU is a common and debilitating problem, yet no interventions have yet been developed or subjected to clinical trial. It is anticipated that, as our understanding of the pathological mechanisms and risk factors improves, such interventions can be designed and integrated into rehabilitation programmes.

Authors' contributions

Literature search: HIK.

Drafting manuscript: HIK, HL.

Conducted the search of studies that contributed data to the sub-analysis, contacted the study teams, and devised the idea: AC, SJB.

Extracted and collated data from information sent by study teams: AC.

Editing and contributing to the final version of the manuscript: all authors.

The collaborators listed extracted, and sent summary data, from their original studies to AC and SJB for inclusion.

Handling editor: J.G. Hardman

Editorial decision: 19 March 2019

Declaration of interest

The authors declare that they have no conflicts of interest.

Funding

Data represented as the HARP-2 study generated by UK Efficacy and Mechanism Evaluation (EME) Programme (EME 05:01 Simvastatin to reduce pulmonary dysfunction in patients with acute respiratory distress syndrome: the HARP-2 RCT), an MRC and NIHR partnership (08/99/08). The EME Programme is a collaboration between the MRC and NIHR, with contributions from the Chief Scientist Office in Scotland, the National Institute for Social Care and Health Research in Wales, and the Health and Social Care (HSC) Research and Development Division, Public Health Agency for Northern Ireland; by a Health Research Award (HRA_POR-2010-131) from the Health Research Board (HRB), Dublin; and with contribution from the HSC Research and Development Division, Public Health Agency for Northern Ireland, the Intensive Care Society of Ireland, and REVIVE. It contains information licensed under the Non-Commercial Government Licence v2.0.