key: cord-0742317-mapvcpd8
authors: Taquet, Maxime; Geddes, John R; Husain, Masud; Luciano, Sierra; Harrison, Paul J
title: Neuropsychiatric disorders and COVID-19 – Authors' reply
date: 2021-06-17
journal: Lancet Psychiatry
DOI: 10.1016/s2215-0366(21)00202-9
sha: c22f5530051df59abf95de365921918a8a78708b
doc_id: 742317
cord_uid: mapvcpd8

nan

We thank Elizabeth Charlton and colleagues, Josef Finsterer, and Ella Burchill and colleagues for their comments on our Article in The Lancet Psychiatry. 1 Charlton and colleagues raise several interesting points. Regarding posttraumatic stress disorder (PTSD), we did not explore this specific diagnosis, although we did in an earlier Article. 2 We have now done so, extending the window for the index event to April 20, 2021. The risk of a first diagnosis of PTSD within 6 months of a COVID-19 diagnosis was 0·58% (95% CI 0·50-0·67 Using the same matched cohorts of patients with COVID-19 and with influenza diagnosed between Jan 20, 2020, and April 20, 2021, we also investigated the incidence and HRs for subtypes of dementia (table). The majority of diagnoses were of unspecified dementia but the relative increase was broadly similar across categories. We did not exclude people with a history of mild cognitive impairment or delirium, and therefore some patients diagnosed with dementia might have been in this high-risk or prodromal group, as we noted in the Discussion of our Article. 1 We have no data as to which of the COVID-19 cases had been asymptomatic, but we assume that this group is substantially underrepresented in our dataset because there is a bias towards symptomatic people presenting for testing (especially early in the pandemic), and because we used the U07.1 ICD-10 code to define cases, which refers to a confirmed diagnosis. Asymptomatic COVID-19 might well be associated with lower rates of subsequent psychiatric or neurological disorder, and our results should be interpreted with this important possibility in mind. We agree that asking about COVID-19 should become a routine item in medical history questionnaires. The idea of reverse redeployment will be attractive to mental health professionals but we suspect rather less so to our general medical colleagues.

Josef Finsterer commented on the overlap between the influenza and respiratory infections cohorts. We agree that we could have made them mutually exclusive; however, we chose not to do this to enable the respiratory infection cohort to be sufficiently large to enable all the COVID-19 cases to be included after propensity score matching. Our study was observational, and we did not attempt to list or explore all the potential mechanisms that might be involved. For instance, we did not investigate the list of putatively neurotoxic compounds that some Incidence within 6 months after COVID-19

Incidence within 6 months after influenza Hazard ratio p value

Dementia in other diseases classified elsewhere (F02)

Data in parentheses are 95% CIs. Dementia subtypes are presented with their ICD-10 codes. The sum of incidences exceeds the total incidence of dementia because the same patient might be diagnosed with one subtype (eg, unspecified dementia) and then another (eg, Alzheimer's disease) within the follow-up period. No data can be shown for frontotemporal dementia and Lewy body dementia because they occurred in fewer than ten patients in each cohort (which is the minimum number to be returned by TriNetX to safeguard patients' anonymity). relapse rates between participants receiving LAIs and those receiving oral antipsychotics-an outcome that is arguably more relevant to patients than the composite primary outcome of hospitalisation or relapse. Second, Kishimoto and colleagues assessed the risk of bias of the randomised trials, but it was not clear for which outcome this assessment was made, and they did not make an overall risk-of-bias judgment for each trial. They judged 24 (75%) of the 32 trials to be at high risk of bias for at least one domain and 30 (94%) were judged as either at high risk of bias in at least one domain or at unclear risk of bias for multiple domains. Using the judgments by Kishimoto and colleagues, and following guidance in the Cochrane Handbook for Systematic Reviews of Interventions for reaching an overall risk-of-bias judgment, 4 30 (94%) of the included randomised trials should potentially be judged as having an overall high risk of bias.

Third, the authors did not assess the certainty of the evidence for any outcomes. On the basis of the randomised trials alone and assuming that their risk-of-bias assessment applied to the outcome of relapse, the certainty of the evidence for that outcome using the GRADE framework 5 should arguably be rated as very low, due to downgrading for risk of bias, heterogeneity (I²=54·5), and indirectness, since the antipsychotics differed between the LAI and oral antipsychotic groups in most included studies. A similar rating would probably apply to the primary efficacy outcomes. The certainty of the evidence from non-randomised studies is, if possible, likely to be rated even lower.

Taken together, the evidence presented by Kishimoto and colleagues 1 does not appear to support conclusions regarding whether outcomes in schizophrenia would improve by increased use of LAIs or not. I declare no competing interests. Cochrane Denmark is funded by the Danish Government.

In their systematic review of longacting injectable (LAI) versus oral antipsychotics, Taishiro Kishimoto and colleagues included randomised trials, cohort studies, and prepost studies. 1 On the basis of their findings, which included a lower risk of hospitalisation or relapse (their primary outcome) with LAIs than with oral antipsychotics, they suggested that increased use of LAIs could improve outcomes in schizophrenia. I would like to comment on some methodological issues.

First, because randomised trials are more likely to provide unbiased estimates of the effects of interventions than are other study designs, they should generally be preferred over non-randomised studies, such as some of the studies included by Kishimoto and colleagues, when synthesising the evidence to guide appropriate patient care. 2 The validity of non-randomised studies, and hence syntheses including them, such as those done in this review, is inherently threatened by biases, especially confounding and selection bias. 2, 3 On the basis of the randomised trials alone (27 studies, 7407 participants), the authors found no significant difference in patients might have received, since a comprehensive pharmacoepidemiological assessment was beyond the scope of the study. Similarly, we could have put various diagnostic combinations together, but we chose to present Guillain-Barré syndrome separately because of the previous suggestions of a specific association with COVID-19. 3 We agree that undiagnosed COVID-19 in the control cohorts will have occurred, and we mention this and its implications in the Discussion of our Article. 1 To expect control cohorts in a real-world electronic health records study to be based on systematic negative PCR test data would be unrealistic. Finally, we acknowledge that we did not attempt to include every neurological syndrome. We have subsequently reported on cerebral venous thrombosis 4 and will be studying headache and some of the other diagnoses in future analyses.

Ella Burchill and colleagues rightly draw attention to the salient finding regarding myoneural junction and muscle disorders. They are correct that most diagnoses in this category were to G72.8 rather than to myasthenia gravis or other specific diagnoses, and we agree that critical illness-associated neuropathy and myopathy are indeed plausible explanations. We also agree that neuromuscular complications of COVID-19 merit attention both for research and rehabilitation.

SL is an employee of TriNetX. All other authors declare no competing interests.

harrison@psych.ox.ac.uk Department of Psychiatry (MT, JRG, PJH) and Nuffield Department of Clinical Neurosciences (MH)

6-month neurological and psychiatric outcomes in 236 379 survivors of COVID-19: a retrospective cohort study using electronic health records

Bidirectional associations between COVID-19 and psychiatric disorder: retrospective cohort studies of 62 354 COVID-19 cases in the USA

Incidence, clinical characteristics, risk factors and outcomes of Guillain-Barré syndrome in patients with COVID-19

Cerebral venous thrombosis and portal vein thrombosis: a retrospective cohort study of 537,913 COVID-19 cases