A systematic review assessing the under-representation of elderly adults in COVID-19 trials

In this review, we included all randomized clinical trials (RCTs) registered on preprint repositories and/or published between December 2019 and May 22nd 2020. Participants in the trials were adult patients (age ≥ 18 years) with COVID-19. The review was performed in accordance with PRISMA guidelines [12]. Clinical trials comparing any pharmacologic agents and devices to another intervention or control in patients with COVID-19 were included. COVID-19 diagnosis was considered as defined in individual trials.

The following interventions were considered for inclusion: drug therapy/prophylaxis, biological therapy (e.g., immunoglobulins, convalescent plasma), device (e.g., continuous positive airway pressure, oxygen therapy). Phase I trials and studies evaluating diagnostic tests were excluded, as were studies including only pediatric patients. We also excluded trials assessing Chinese traditional medicine, as these are less relevant for other countries. For comparison, we also searched for large observational cohorts including COVID-19 patients. We planned to extract data from cohorts including over 1,000 patients and reporting outcomes. We aimed to compare the age of included patients to those included in RCTs.

We searched for full-text RCTs in PubMed, leading internal medicine and infectious diseases journals (see list below), and preprint repositories (including medRxiv at https://​www.​medrxiv.​org/​; arxiv at https://​arxiv.​org/​; and bioRxiv at https://​www.​biorxiv.​org/​). Search terms used for PubMed search were “COVID19 OR COVID-19 OR SARS-CoV-2 OR 2019-nCoV OR SARS2”, combined with the Cochrane filter for RCTs [13]. For preprint repositories, the search term “COVID-19” was combined with “randomized”.

Journal sites searched included those of The New England Journal of Medicine (https://​www.​nejm.​org/​coronavirus), The Lancet (https://​www.​thelancet.​com/​coronavirus), JAMA (https://​jamanetwork.​com/​journals/​jama/​pages/​coronavirus-alert), Annals of Internal Medicine (https://​annals.​org/​aim/​pages/​coronavirus-content), the journals of the Infectious Diseases Society of America (IDSA) (https://​www.​idsociety.​org/​public-health/​COVID-19-Resource-Center/​), Emerging Infectious Diseases journal (at https://​wwwnc.​cdc.​gov/​eid/​), and Clinical Microbiology and Infection (https://​www.​clinicalmicrobio​logyandinfection​.​com/​). We hand searched all study titles published in “COVID-19 resource centers” of each of the above journals for relevant RCTs or observational studies. We also searched for unpublished RCTs in the clinicaltrials.​gov registry, using their link to listed clinical studies related to the coronavirus disease (COVID-19).

An additional search for observational studies was performed using the term “COVID19 OR COVID-19 OR SARS-CoV-2 OR 2019-nCoV OR SARS2”, combined with ‘observational’ OR ‘cohort’ OR ‘prospective’ OR ‘retrospective’. We also reviewed observational data from the ISARIC database [6]. No language restrictions were applied to any of the searches. Two reviewers independently conducted the search and applied inclusion criteria (either VP, NT, or YLW). Any discrepancies were resolved by a third reviewer (DY). The titles and as needed, abstracts or full texts, of the studies were each reviewed by the authors for their relevance. Specifically for observational studies, we also applied a criterion of sample size of over 1000 patients.

Two reviewers independently extracted the following data. For full-text RCTs: publication status and site, setting (hospital vs. other), purpose (prophylaxis vs. treatment), inclusion and exclusion criteria with emphasis on criteria that may eventually lead to exclusion of elderly: any upper age limit, comorbidities, or exclusion for medications/polypharmacy or due to mental or cognitive disorders. Similarly, consent options, potentially a factor limiting elderly patients’ participation, number with limited life expectancy or do-not-resuscitate orders, were also documented. Additional data extracted included interventions and outcomes, including report of outcomes for age-specific subgroups. Studies were also analyzed according to their primary endpoint, i.e., clinical (mortality, time to clinical improvement or clinical improvement rate, duration of invasive mechanical ventilation, hospitalization, transfer to/from intensive care unit (ICU), discharge to long-term-care facilities (LTCF) or rehabilitation, adverse events (AEs), including serious AEs and QTc > 500 milliseconds) or virologic outcomes (duration of viral detection in clinical samples). For observational studies, we extracted data regarding patients’ age and mortality.

Inclusion and exclusion criteria were analyzed, with emphasis on age and other confounding factors cited above. The percentage of trials performing age-adjusted analyses was also reported. Combined mean age for all RCTs was calculated. For studies reporting age as median, the median was considered equivalent to the mean, as suggested by the Cochrane handbook [13]. Mean participant ages were compared between RCTs and large observational studies. For RCTs, risk of bias was assessed using the domains recommended by the Cochrane handbook. These were graded as low, high or unknown risk of bias, according to the Cochrane handbook’s criteria [13]. For observational studies we used NIH Study Quality Assessment Tool for observational cohort and cross-sectional studies (https://​www.​nhlbi.​nih.​gov/​health-topics/​study-quality-assessment-tools). Two authors performed the quality assessment independently (DY and NT).

The search yielded 2191 RCTs; after applying eligibility criteria, 12 full-text trials assessing various treatment options for COVID-19 were included (Fig. 1). Risk of bias assessment of these trials is detailed in Supplementary Table 1. All trials but one were judged low risk of bias for allocation generation; eight for allocation concealment; and four were double blind studies. (For full data see Supplementary Table 1). All trials addressed drug therapy, of which eight were performed in China. Table 1 summarizes characteristics of included trials. Two were published in PubMed [14, 15], four at journal websites [16‐19] and six in medRxiv [19‐24]. Four evaluated lopinavir/ritonavir-based therapy [14, 17, 19, 24], three chloroquine or hydroxychloroquine therapy [15, 20, 23], two favipiravir or baloxavir [21‐23], two remdesivir [16, 18] and one α-lipoic-acid [25]. The mean age of patients included in RCTs was 56.3 years. No trial specifically targeted the elderly. An upper age limit was reported in three studies (25%). Chen et al. reported 60/240 patients aged ≥65 years (25%) [20], Beigel 382/1063 (36%) patients aged ≥65 years [18], and Zhong nine patients between 60 and 70 and two > 70 years among a total of 17 patients (53%) [25]. Three studies performed age-specific subgroup analyses, among which one for patients ≥65 [23] (Table 1). Eight, six and five trials had exclusion criteria of liver function abnormalities or severe liver disease [14, 16, 19‐24], severe kidney disease [16, 19, 20, 22‐24], and heart disease/arrhythmia/prolonged QT, respectively [17, 19, 20, 23, 24]. No study excluded patients due to polypharmacy but three excluded patients taking medication potentially interacting with lopinavir/ritonavir, ribavirin, interferon or arbidol [14, 17, 24]. Four trials excluded patients due to the presence of a mental or cognitive criteria: “a mental illness affecting therapeutic compliance” [24], “unable to cooperate with investigators due to cognitive impairment or poor mental status” [23], “inability to comprehend and to follow all required study procedures” [17, 18]. One study used an exclusion criterion of patients who would probably not cooperate or finish the study: “base on the researcher’s judgment, there are other factors that may cause the subject to be forced to terminate the study midway, such as other serious diseases, serious laboratory examination abnormalities, other factors that affect the safety of the subject or study data and blood sample collection” [22]. Five trials excluded patients for expected survival time under 48 h or for critical illness [20‐22, 24, 25]. Only six trials allowed consent to be provided by a legal guardian [14‐18, 21], and three reported allowing a family member to provide consent [6, 15, 17]. Primary outcomes varied: only seven RCTs chose a clinical outcome (one reported 28-day mortality and six clinical improvement or recovery); five RCTs reported virological outcomes (time to or rate of negative polymerase chain reaction (PCR)). None of these studies reported on discharge to rehabilitation or long-term care facilities (LTCF) or other clinical outcomes relevant for the geriatric population, such as cognitive or functional decline.

Our clinicaltrials.​gov search yielded 1651 registered studies addressing COVID-19; among these, 870 interventional studies included only adults, and 650 were RCTs evaluating therapeutic and prophylactic interventions for COVID-19. Two hundred RCTs (200/650, 31%) had an upper age limit: 56 used a cutoff of 65 years, 153 of 80 years, and 47 of 99 years.

We reviewed 881 titles from PubMed searching for observational studies including over 1000 patients and reporting clinical outcomes (Fig. 2). We found two such cohorts, from which we extracted data on age and mortality: Guan et al. (China) described 1099 COVID-19 patients hospitalized in 552 hospitals [26]. Median age was 47 (IQR 35–58). Of the entire cohort, only 15.1% were 65 or older. When stratified according to disease severity, patients ≥65 years old made up 27% of severe patients but only 12.9% of non-severe patients. Overall mortality was 1.4%; it was not reported by age group. Richardson et al. (USA) reported the outcomes of 5700 patients hospitalized with COVID-19 in the New York City area [27]. Here the median age was 63 (IQR 52–75, range 0–107); 36% were ≥ 70 years old. Overall mortality was 9.7% (2634 were discharged or had died at the study’s end of follow-up). In the 60–69, 70–79, 80–89 and ≥ 90-year age groups, mortality was 6.4, 12.6, 24.9 and 28.6%, respectively. Mortality rates for those receiving mechanical ventilation in the 18-to-65 and > 65-year age groups were 76.4 and 97.2%, respectively. In the same age groups, mortality rates for those not requiring mechanical ventilation were 19.8 and 26.6%, respectively. Quality rating of these two studies according to NIH quality assessment tool was judged as good quality for Richardson et al. [27] and fair quality for Guan et al. [26] (Supplemental Table 2).

As of May 6th 2020, the ISARIC database contained information on 20,276 COVID-19 patients from 35 countries [6]. Median age was 72 (range 0–104 years). Here, 53% of the patients were ≥ 70 years old. Overall mortality was 26.6%, yet in the 60–69 and ≥ 70-year age groups, it was 21 and 39%, respectively.

Multiple data sources were comprehensively reviewed in this search, using a broad search term of all COVID-19 associated studies. However, for the search of observational studies a significant hand search was required, which may have resulted in omissions. In addition, the search of leading journals was based on the journals’ COVID-19 resource centers, without a specific search term used by this review authors.

Elderly persons are underrepresented in COVID-19 RCTs, although they are the demographic most frequently and severely affected by the disease. Clinical research including the elderly has never been easy; nevertheless, future trials will need to address this vulnerable and oft-forgotten population, particularly when these individuals are regularly receiving off-label therapies anyway. Both interventional RCTs, including elderly patients, are needed; as well as observational studies including both older and non-older patients, the latter to clarify the special characteristics of older patients with COVID-19.