Side effects of corticosteroids in patients with advanced cancer: a systematic review

Corticosteroids are used in the treatment of a wide variety of diseases. In the field of palliative care, corticosteroids are frequently prescribed to alleviate distressing symptoms in patients with advanced cancer, including anorexia, fatigue, and pain [1‐3]. However, corticosteroids are associated with several side effects: susceptibility to infection, hyperglycemia, insulin resistance, proximal myopathy and catabolic effects, skin changes, and adrenal insufficiency [4]. Patients with advanced cancer are particularly vulnerable to adverse events (AEs) associated with corticosteroid use. Treatment decisions should be decided based on a risk-benefit balance. It is therefore important to assess both effectiveness and AEs associated with corticosteroid use in people with advanced cancer.

Randomized controlled trials and consecutive cohort studies are both study designs that collect data prospectively and can contribute data to the incidence of an event because they have clear numerators and denominators. Given that randomized controlled trials (RCTs) and consecutive cohort studies report AEs, both can provide useful data [5, 6]. Equally, both may suffer from varying quality of AE reporting [7].

In order to manage symptoms without harms, it is important to know the risk-benefit balance of corticosteroid use for patients with advanced cancer. However, no systematic review of AE reporting in clinical studies of corticosteroids in patients with advanced cancer has been reported to date. The purpose of this systematic review was to identify the AE reporting in RCTs and consecutive cohort studies of corticosteroids in patients with advanced cancer.

We conducted a systematic review of the literatures using online databases.

Inclusion criteria: (1) patients with life-limiting advanced cancer; (2) corticosteroids prescribed for symptom control (such as pain, fatigue, anorexia, nausea, well-being, and headache attributed to metastases); (3) adult participants; (4) randomized controlled trial (placebo-controlled or comparison with other drugs) or a consecutive, prospective cohort study; (5) use of any of the following corticosteroids: dexamethasone, methylprednisolone, prednisone, prednisolone, or betamethasone; (6) oral or intravenous (systemic) administration route; (7) written in English language.

Exclusion criteria: (1) corticosteroid administration by inhalation; (2) other chronic progressive diseases for which corticosteroid therapy is indicated, including ulcerative colitis and rheumatoid arthritis; (3) cancer patients undergoing chemotherapy and receiving steroids only for chemotherapy-related nausea; (4) steroids prescribed as chemotherapy (for example, in leukemia); (5) reviews or summary studies; (6) child and adolescent cancer patients; (7) case reports and retrospective studies; and (8) studies with no corticosteroid administration.

The relevant literature was extracted from PubMed, Medline, SCOPUS, Cochrane reviews, and CINAHL online databases. Search filters were created by the authors with the assistance of an experienced librarian. The details of the search filters are shown in Supplementary Material 1. Following PRISMA guidelines, two review authors (Y.H. and H.M.) independently selected studies that met the criteria. Disagreements were resolved by discussion with all authors. The systematic review protocol was registered on PROSPERO (CRD42015025813) and is available [8]. Selected studies were assessed for reporting of AEs based on CONSORT extension for harms [9]. Risk biases of RCTs were also assessed according to Cochrane guidelines [10].

All participants in the selected RCTs were patients with advanced cancer. The most frequently reported primary outcome was nausea and vomiting (8), followed by pain (5), loss of appetite (3), quality of life (3), and fatigue (2). Types of corticosteroids included dexamethasone (13), methylprednisolone (9), and prednisolone (4), and one study included no information on the types of corticosteroids used. Doses of corticosteroids varied widely, and duration of corticosteroid exposure ranged from 1 to 250 days.

In the included 12 consecutive cohort studies, (Supplementary Material 3), primary outcomes included pain (3), anorexia (1), fatigue (1), spinal compression (1), and bowel obstruction (2). In four studies, the priorities of outcomes were not specific, and participants were taking corticosteroids to relieve several symptoms. Dexamethasone was the most commonly used corticosteroid in the consecutive cohort studies (7), followed by methylprednisolone (1). Four studies included participants who were taking different types of corticosteroids. Doses of corticosteroids and the length of the study period differed widely across the selected studies.

Among the studies included in this review, many failed to report sufficient information about AEs associated with corticosteroid. An extension of the CONSORT checklist is available to report AEs in RCTs [9] and was developed following calls to improve the quality of AE reporting among RCTs [5, 6]. In this review, two RCTs and one consecutive cohort study in which a validated AEs assessment tool was used were published after developing CONSORT extension. However, the remaining studies used no validated tools to assess AEs. As previous reviews have been suggested, [7, 12, 13] more promotion to improve quality of AE reporting in clinical should be needed.

Corticosteroids are commonly used in clinical practice to treat a variety of conditions. Clinicians should therefore be aware of both the benefits and harms of corticosteroid. Patients of deteriorating condition may be more susceptible to toxicities arising from pharmacological intervention. To avoid the occurrence of harmful events by the proper use of pharmaceutical products, clinical researchers should be more aware of the importance of reporting AEs.

Sixteen RCTs (59%) and seven consecutive cohort studies (58%) did not report the method used for assessing AEs. Assessment of the side effects of corticosteroids without specific measurements may lead to under-estimation of the actual harm caused by corticosteroids, which can present with a wide variety of symptoms. Furthermore, the use of different measurement tools may limit the ability to analyze differences in AEs between studies. PRO-CTCAE [14], developed to assess cancer patients’ subjective side effects, should be considered as a standardized tool for assessment and minimizing the risk of under-estimation of AEs of corticosteroids.

Nevertheless, there was no significant difference in the proportion of AEs between corticosteroid and placebo groups in recent RCTs in which validated AE assessment tools were used [15, 16]. This may be explained by low statistical power related to small numbers of participants and short study periods. Restrictive inclusion/exclusion criteria may also be related to the low rates of AEs observed in the RCTs. In terms of the timing of AE assessment, 15 RCTs (56%) and 11 consecutive cohort studies (92%) did not report the timing of AE assessment. Corticosteroids present not only acute side effects but also long-term side effects [17]. It is important that consecutive cohort studies of larger numbers of participants taking corticosteroids for longer periods of time with high-quality AE assessment are reported.

RCTs provide the highest level of evidence of treatment effectiveness. However, most RCTs are designed to assess benefit, capturing AEs or harms collaterally [18]. While measuring effectiveness of treatment is essential, assessing harms is also important. The CONSORT extension for harms was developed in 2004 [9]. This guideline provides lists of recommendations for AE reporting in clinical trials. However, recent reviews show that the quality of reporting has not improved since the publication of the CONSORT extension for harms in 2004 [7, 12, 13]. Authors insist more effort is needed to improve AE reporting in clinical trials.

The outcomes of consecutive cohort studies, in which the criteria of participants were less restrictive, may better reflect clinical practice than the RCTs. However, it is often difficult to differentiate delirium due to corticosteroid from the natural disease progression in patients with terminal cancer [19]. For example, delirium may be caused by several factors such as opioids, infectious diseases, cerebral metastases, and organ failure, as well as by corticosteroids. The vulnerable populations included in the studies may be less likely to show resolution of risk factors for delirium. Nevertheless, a previous cohort study showed that approximately 50% of delirium symptoms might be reversible, even in patients with advanced cancer [20]. The cause of AEs in advanced cancer patients receiving corticosteroids should therefore be carefully assessed with validated tools.