Adverse cognitive effects of glucocorticoids: A systematic review of the literature

Objectives: Glucocorticoids as a drug class are widely used in the treatment of many conditions including more recently as one of the mainstay treatments for the SARS-CoV-2 infection. The physiological adverse effects are well described. However, less is known and understood about the potentially deleterious neuro-cognitive effects of this class of medication. Methods: We carried out a systematic review of the literature using two separate search strategies. The first focussed on the rates of reporting of adverse cognitive effects of glucocorticoid use in randomised controlled trials. The second looked at those studies focussing directly on adverse cognitive effects associated with the use of glucocorticoids. MEDLINE, Embase and Cochrane Library was searched for randomised controlled trials utilising glucocorticoids as a part of a treatment regimen. Additionally, these databases were also used to search for articles looking directly at the adverse cognitive effects of glucocorticoids. Results: Of the forty-three RCTs included as a part of the first search strategy, only one (2.3%) included specific documentation pertaining to cognitive side effects. As a part of the twenty studies included in the second search strategy, eleven of the included studies (55%) were able to demonstrate a correlation between glucocorticoid use and decreased cognition. Most studies within this strategy showed that GCs predominately affected hippocampus-dependent functions such as memory, while sparing executive function and attention. Conclusions: Overall, the data reporting of adverse clinical effects of glucocorticoid use is poor in recent RCTs. Given the demonstrable effect on predominately hippocampal-dependent cognitive functions evident within the literature, more thorough documentation is needed within clinical research to fully appreciate the potentially widespread nature of these effects.


Introduction
Glucocorticoids (GCs) as a drug class are widely used as treatments for many conditions including autoimmune diseases, malignancy, allergic conditions, and more recently as a treatment for the SARS-CoV-2 infection [1,2].As a result of their broad utility and potency, they are now the most widely prescribed anti-inflammatory and immunosuppressant medication worldwide [3].The physiological effects of GCs are well understood.Additionally, the physiological effects of sustained exposure to high doses of GCs are also well known (Cushing's syndrome, Abbreviations: GC, Glucocorticoid; RCT, Randomised controlled trial; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; NOS, Newcastle-Ottawa Scale; QoL, Quality of life; OPE, Oral prednisolone equivalent; PROs, Patient-reported outcomes; SF-36, 36-Item Short Form Survey; FACIT-F, Functional Assessment of Chronic Illness Therapy -Fatigue; HAQ, Health Assessment Questionnaire; HAQ-DI, Health Assessment Questionnaire -Disability Index; FACT-BMT, Functional Assessment of Cancer Therapy -Bone Marrow Transplantation; FACT-G, Functional Assessment of Cancer Therapy -General; EQ-5D-5L, EuroQol 5 Dimensions, 5 Levels; GHS, Global Health Status; PostopQRS, Postoperative Quality of Recovery Scale; EoE-QoL-A, Eosinophilic Oesophagitis Quality of Life Scale for Adults; EQ-5D, EuroQol 5 Dimensions; MG-QOL15, Myasthenia Gravis Quality of Life 15-Item Scale; EORTC QLQ-C30, European Organisation for Research and Treatment of Cancer Core Quality of Life Questionnaire; EORTC QLQ-MY24, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire 24-Item Myeloma-Specific Module; MTX, Methotrexate; CHOP, Cyclophosphamide, Doxorubicin, Vincristine, Prednisolone; HCQ, Hydroxychloroquine; PO, Oral; S/C, Subcutaneous; IV, Intravenous; BD, Twice daily; TDS, Three times daily.hypertension, osteoporosis, diabetes mellitus and immunosuppression) [4].While these physiological phenomena are well elucidated in the literature, less is known about the neuro-cognitive effects of this class of medication, particularly within the human brain and its effects on memory [5][6][7].
Varney et al. first drew attention to the presence of severe and persistent cognitive disturbances observed in patients exposed to supraphysiological GC therapy [8].This phenomenon, sometimes referred to in the literature as "steroid dementia" superficially resembles Alzheimer's disease with observed impairment in delayed verbal recall, encoding deficits and difficulty with stepwise thinking [8][9][10][11][12][13].While the cognitive effects of supra-physiological GC exposure are still being elucidated, current evidence suggests that the impairments observed are mainly due to the effect GCs have on the hippocampus and prefrontal cortex [5,9,14].Hippocampal atrophy, reduced dendrite length and alterations in neurotransmitter receptor populations in the hippocampus have all been observed in functional imaging, animal, and human studies, supporting the evidence for the potentially deleterious effects of GCs on the hippocampus in chronically elevated doses [3,[14][15][16].
Reporting of adverse effects in clinical trials is instrumental in ensuring medical treatments are safe, well tolerated and ultimately beneficial for the patient.Although the adverse effect profile of GCs has been well reported in clinical trials, the rates of reporting of the cognitive effects of this class of medication is not widespread [17,18].Despite the documented association between memory and GCs within the literature, clinical trials using GCs as a part of a treatment regimen continue to omit measures of cognitive function or memory as a part of the possible list of adverse events.
The first aim of this study was to examine the rates of adverse cognitive effects of GCs reported within randomised controlled trials (RCTs) that used GCs as a part of a treatment regimen.The second aim was to examine those studies investigating the cognitive effects of GCs directly to better understand the rates of these adverse events using a systematic review methodology.

Methods
This study was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement methodology [19].

Search strategy
The electronic databases MEDLINE, EMBASE and Cochrane Library were searched from inception until February 2022 using two separate search strategies.The first strategy was aimed at capturing RCTs that list GCs as a part of a treatment regimen to assess the rates of adverse cognitive effects documented.The second strategy was aimed at capturing studies investigating GCs and their effects on cognition and memory.The two search strategies can be viewed in Supplementary Fig. 1.Bibliographies from individual included studies and review articles were searched manually to identify additional relevant papers.All references were exported to EndNote (EndNote X9.3.3) for screening and to remove duplicates.Title, abstract and full-text screening was undertaken by a single reviewer (LS).As a part of the first search strategy, RCTs were included if they utilised oral or intravenous glucocorticoids as a part of a treatment arm; involved adult participants; were published in the last five years; were published in English; had a full text article available; and reported adverse events in both comparator groups.Articles identified in the first search strategy were excluded if they did not document a clear dose of GC, or the route of administration was any other than intravenous or oral.If multiple reports described the same trial, the most recent full-text publication was included.For the second search strategy, primary research articles were included if their aim was to investigate the effect of GCs on cognition in humans.For both search strategies, articles were excluded if there was no full text manuscript available, the article was not published in English, and the study was not conducted on human adults.

Quality assessment
The quality of all included RCTs was assessed through the use of an 8item modified-Jadad scale which assessed reported randomisation, blinding, withdrawals, inclusion/exclusion criteria, adverse events and statistical analysis [20].This scale has a maximum score of 8 points, with high quality studies scoring between 5 and 8 points.Quality of nonrandomised studies was assessed using the Newcastle-Ottawa Scale (NOS), a bias assessment tool for observational studies recommended by the Cochrane Collaboration [21].The total NOS scale was 9 for casecontrol and cohort studies and 10 for cross-sectional studies (selection, comparability, and outcome/exposure).Studies with total score ≤ were considered low-quality, and those with scores ≥ 5 were considered high-quality [21,22].All quality assessment was completed by a single reviewer (LS) with any uncertainty resolved through discussion with a second reviewer (PR).

Data extraction
A pre-determined data form was used to independently extract data from the full-text studies.Data extracted included publication details (author, year, country), disease state, sample characteristics (sample size, mean age and % of male participants) and study details (study type, intervention, comparator, equivalent oral dose of prednisolone, measure of quality of life (QoL) or cognition, documentation of adverse cognitive effects and length of follow up).

Search results and study characteristics
After removal of duplicate articles, a total of 654 and 212 studies were identified for the first and second search strategies respectively.After title and abstract screening, 77 and 28 full-text studies were assessed for eligibility for the first and second search strategies respectively.A total of 63 studies were included in this review from both the first and second search strategies (First strategy total = 43; Second strategy total = 20).The PRISMA diagrams demonstrating the process of the literature search and study selection is shown in Figs.1A and 1B.
For the second strategy, a total of 20 studies published between and 2021 analysing a total of 11,636 patients were included.Mean age ranged from 22.2 to 75.6 years with the majority being males (51.1 %).

Quality assessment
Results from the quality assessments of the included studies are shown in Supplementary Tables 1 and 2. Overall, for the studies included as a part of the first search strategy, the modified-Jadad score indicated high quality (a score of 5-8) for all studies.
The quality assessment for studies included as a part of the second search strategy was divided into RCTs and non-randomised studies.For the RCTs included, the modified-Jadad score indicated that 6 of the 12 studies (50 %) were of high quality.Overall, the non-randomised studies included were all high quality according to the NOS (scoring ≥5 stars total).

Intervention characteristics
In the studies identified as a part of the first search strategy, there was a high degree of variability in terms of the type of GCs used as well as the dosing regimens.To simplify the comparisons between studies, all GC doses were converted to a daily oral prednisolone equivalent (OPE) dose.
In the studies identified as a part of the second search strategy, there was also a high degree of variability in terms of the type/s and regimens of GCs used.These regimens were also converted to an OPE dose for ease of comparison.The daily dose of GCs in these studies ranged from 2.5 mg to 625 mg OPE.Three of the twenty studies (15 %) identified in this group either did not report the doses of glucocorticoids used, or the OPE was not calculable from the data available in the manuscript [64,66,74,75].

Principal findings
This large systematic review included two separate, but related search strategies encompassing RCTs utilising GCs as a part of a treatment regimen in addition to studies aimed at investigating the cognitive effects of GCs directly.
The first search strategy included 13,310 patients in forty-three studies with a daily dose of GC between 1.66 and 625 mg OPE.Only one of the forty-three studies included specific documentation of pertaining to cognitive adverse effects.In addition to this, twenty-four of the forty-three studies included documentation of umbrella terms potentially relating to cognitive and nervous system adverse effects.Measures of QoL were included in fifteen of forty-three studies, of which only FACT-BMT, EORTC QLQ-C30 and PostOpQRS contained a component or subsection pertaining to cognition or memory.Overall, the studies utilising QoL tools that contained a cognitive component demonstrated a neutral effect of GCs on cognition with both Royse et al.The second search strategy included 11,636 patients contained within twenty studies with a daily GC dose between 2.5 and 625 mg OPE.Several different study types were included within the second search strategy including cross-sectional studies, RCTs, cohort and casecontrol studies.The domains of cognition tested as a part of the cognitive testing included hippocampal-dependent and independent memory, attention, and executive function.Within the eleven studies able to demonstrate a relationship between GC exposure and decreased cognitive function, four looked at the chronic effects of GCs and the remaining seven focussed on the acute and subacute effects.
Within the studies focussing on the acute and subacute effects of GC exposure, hippocampal-dependent functions, as evidenced by disruption to delayed verbal recall and visuospatial ability were predominately affected [7,10,13,69,80].Interestingly, these effects were observed to be independent of patient's attention [12,72].
In addition to the acute and sub-acute effects of GCs, the long-term effects of GCs on cognition and memory were explored in several papers included in this study.Of those studies investigating the cognitive effects of chronic GC exposure, hippocampal-dependent memory was the predominant cognitive impairment observed [11,17,64,65].Unfortunately, the doses within these studies differed significantly, making more in-depth comparison difficult.However, Werumeus et al. (2015) noted that both "low" and "high" physiological doses of GCs administered over the course of 20 weeks were correlated with hippocampaldependent memory deficits in as many as one third of study participants [17].

Clinical implications
Despite evidence within the literature pointing towards GCs having a demonstrable effect on cognition with both short-and long-term use, this effect is not well reported within recent clinical trials.Many trials within the literature fail to document any measure of patient QoL.Those that do, utilise measures that do not include subsections relating to cognition or memory.As a result of this poor reporting, there is no way of quantifying the magnitude or severity of the effect that GCs are likely having on patients.
Additionally, those studies that do investigate the cognitive effects of GCs directly utilise tools that are inconsistent.While there is a vast menagerie of cognitive testing tools available with varying ease of use, applicability and reliability, the inconsistencies within the reporting of such studies further compounds the difficulty in generalising and comparing their results.

Strengths and limitations
The strengths of this review include two robust search strategies encompassing a large, diverse, and contemporary patient cohort; high quality studies; strong data with regards to steroid dosing in large clinical trials and thorough understanding of the measures of QoL and cognition used in contemporary clinical trials.
However, this study was limited by a high degree of heterogeneity in terms of the studies included, particularly with regards to the disease states, treatment regimens, mean OPE dose and measures of QoL and cognition.Because of the high degree of heterogeneity between studies with respect to direct and indirect measures of cognition it may be that impairments on cognition may have been missed owing to less sensitive tools utilised by some studies.Additionally, the high degree of heterogeneity makes comparing the results of these studies difficult.Another possible limitation of this study was the inclusion of articles investigating the effects of GCs in treating acute illnesses including septic shock, meningitis and COVID-19.It is possible that these conditions may cause temporary neurocognitive deterioration independent of the presence of GCs [81].A challenge in future studies looking at the effects of GCs on cognition will be elucidate which effects are due to the underlying illness and which are due to the presence of GCs.However, inclusion of these studies emphasises the paucity of reporting in terms of neurocognitive adverse events more broadly within contemporary RCTs.In addition to this, some of the RCTs included in this study contained treatment regimens with similar GC dosing across both treatment and comparator groups [32,45,48].In studies with similar GC exposure across both treatment groups, interpreting which adverse neurocognitive effects are due to GCs themselves and which are due to underlying disease or other elements of the treatment regimen becomes difficult.However, even in the articles that did include differing GC regiments across treatment groups, there was a distinct lack of QoL and cognitive measures tested making direct comparisons of GC effect and cognition in non-dedicated studies difficult.
This review excluded routes of administration of GCs other than IV or PO.This was done with the hope of being able to elucidate a dose-response relationship from the available data.However, due to the broad range of dosing regimens and methods, this was not possible.

Future directions
As a part of future research, more targeted studies are needed with regards to the adverse cognitive effects of GCs.An important part of this work could be to develop an objective and validated tool for detecting cognitive impairment in the context of GC usage.In addition, mechanisms through which GCs act both in the short and long term to produce cognitive impairment could be better elucidated to better understand how and why this occurs as well as which patient populations are at particular risk for this adverse effect.Finally, a dose-response relationship, if one exists, would be a useful focus for this work for clinicians to better understand the risks of GCs when they are prescribed as a part of a treatment regimen.
Once an objective and validated tool for detecting the adverse cognitive effects of GC use has been created, it would then be able to be used in clinical trials employing GCs as a part of a treatment regimen.This would then allow the clinical community to have a better grasp of the size of the problem when it comes to the cognitive effects of GCs.Additionally, it would allow for a more uniform method of reporting in clinical trials.

Conclusion
While GCs are an essential part of many treatment regimens, more work is needed in this field to elucidate the potential mechanisms involved in their effect on cognition.Additionally, more thorough, and consistent documentation of cognitive adverse effects is needed within clinical research to fully appreciate the potentially widespread nature of these effects.

Special recognition
Special recognition and appreciation must go to the co-author Philip C. Robinson; MBChB, PhD FRACP a,d who helped to conceptualise, formulate and edit the final manuscript for this paper prior to his unexpected passing.Dr Robinson's generosity, kindness and genuine desire to help others achieve, were atributes which medical students and physician-researchers everywhere should seek to mirror.Thank you Dr Robinson.

Fig. 1A .
Fig. 1A.PRISMA flow diagram for studies included as a part of the first search strategy.

Fig. 1B .
Fig. 1B.PRISMA flow diagram for studies included as a part of the second search strategy.
Royse et al. (2017) and Jagasia et al. (2019) both demonstrating nonsignificant changes cognitive sub-scores of the FACT-BMT and Post-OpQRS measures.However, Pidala et al. (2020) demonstrated a potentially adverse effect of GCs on cognition as demonstrated by a significantly lower mean FACT-BMT score in those patients exposed to GCs as a part of a treatment regimen.
demonstrated a reduction in tests of hippocampal-dependent memory by as much as 27.36 %.Additionally, Frol et al. (2013) demonstrated a significant difference in the number of patients with an impaired global clinical rating (GCR) of cognition when treated with GCs (59 % vs 12 %).Finally, Ancelin et al. (2012) demonstrated a 1.8-fold increased risk of decline in frontal executive function in women using GCs over a period of 7 years.
(2017)  andJagasia et al. (2019) failing to show a significant change in the FACT-BMT and PostOpQRS scores respectively.None of the studies contained within the first search strategy utilised a direct measure of cognition.

Table 1
Baseline characteristics of studies included as a part of the first search strategy.
(continued on next page) L.Spannenburg and H. Reed

Table 2
Baseline characteristics of studies included as a part of the second search strategy.
Brunner et al. (2005) was able to demonstrate a 10.3 % decline in the hippocampal-dependent Rey Auditory Verbal Learning Test scores (RAVLT) in those patients treated with GCs compared to placebo.Domes et al. (2005) was able to show impaired performance on the Phone Numbers subtest of the LGT-3 whereas Wolkowitz et al. (1990) was only able to demonstrate a slightly increased rate of errors of intrusion in a paragraph recall test in those patients treated with GCs.