Psychological distress as a risk factor for the efficacy of chemotherapy in advanced gastric cancer patients.
Academic Article
Overview
abstract
BACKGROUND: Type 2 diabetes (T2D) is a potential risk factor for poor outcomes associated with COVID-19, as are associated conditions including obesity, chronic kidney disease, and socioeconomic risk factors. T2D providers during the pandemic are thus caring for an especially vulnerable population. Published literature as well as online media, including patient discussion forums, raise questions about T2D management during the pandemic. For example, does glycemic control mitigate COVID-19 risk? Do specific diabetes drugs confer benefit or risk in the setting of SARS-CoV-2 exposure? OBJECTIVES: Aim 1 was intended to understand issues with lack of data completeness that might affect the executability of the later aims. We used Medicare data from 2013 to 2016 as a gold standard and assessed the sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) of New York City's INSIGHT Clinical Research Network (CRN) data from those years in identifying hospitalizations and diabetes drug use among Medicare patients with T2D. Aim 2 was to conduct a cohort study to test the hypothesis that elevated baseline glycated hemoglobin (HbA1c) is not independently associated with the primary outcome (hospitalization with COVID-19). The intent of aim 2 was to inform management of T2D during the COVID-19 pandemic by assessing whether relatively intensive T2D management, with the goal of maintaining low average blood glucose levels, is necessary or helpful in reducing risk of hospitalization with COVID-19. Aim 3 was to conduct a comparative cohort study to test the null hypothesis that none of the antidiabetic drug classes of metformin, sodium-glucose cotransporter-2 (SGLT2) inhibitors, dipeptidyl peptidase 4 (DPP-4) inhibitors, glucagon-like peptide 1 (GLP-1) receptor agonists, and sulfonylureas affect hospitalization rates or hospital deaths for COVID-19. The intent of aim 3 was to inform the management of T2D during the COVID-19 pandemic by assessing whether use of some antidiabetic drugs vs others could increase or lower the risk of hospitalization with COVID-19. METHODS: For aim 1, data from the INSIGHT CRN were randomly divided into training and test sets. In the training set, least absolute shrinkage and selection operator regression was used to develop, from candidate variables in the data set, models to predict high completeness of the data for hospitalization and antidiabetic drug use. For hospitalization, completeness means the probability that any hospitalization of a patient identified through Medicare claims data is captured in the INSIGHT CRN records of hospital admissions; for antidiabetic drug use, completeness means the probability that use of a given antidiabetic drug class shown by the Medicare claims data is also captured in the INSIGHT CRN electronic prescribing data. For aim 2, HbA1c was modeled as a continuous variable in the primary analysis, and in secondary analyses, patients in the CRN training data set were divided into 4 exposure groups based on their most recent HbA1c measurements of 5.7% to 6.4%, 6.5 to 7.4%, 7.5 to 9.9%, and 10% or higher. Time to COVID-19 hospitalization was modeled as a function of baseline HbA1c with adjustment for covariates, and in secondary analysis was compared in the 4 groups in a propensity score-weighted Cox proportional hazards model adjusting for potential confounders. Data were available and were analyzed for patients up until June 15, 2020. In-hospital death was examined as a secondary outcome. For aim 3, adult patients with prevalent use of metformin, sulfonylureas, DPP-4 inhibitors, SGLT2 inhibitors, or GLP-1 receptor agonists before the index date of March 15, 2020, were included. Prevalent use of each of these medications was examined as a potential risk factor for COVID-19 hospitalization in propensity score-weighted Cox proportional hazards models adjusting for covariates. Patients were followed until June 15, 2020. RESULTS: For aim 1, patients were divided into deciles of data completeness (sensitivity) for hospitalization—that is, the probability that a given hospitalization documented in Medicare claims data would be captured in the INSIGHT CRN data, ranging from 22% in the lowest decile to 71% in the highest decile. Positive predictive value for metformin use ranged from 95% to 98%, meaning that if metformin prescriptions were present in the INSIGHT CRN data, there was a high probability of ongoing metformin use documented in the Medicare administrative claims data, whereas NPV for metformin use ranged from 62% at the lowest decile of data completeness to 82% in the highest decile. The statistics for other antidiabetic drugs were similar. For aim 2, of 4192 patients who met the inclusion criteria, 1083 patients had a baseline HbA1c of 5.7% to 6.4%, 1319 patients had an HbA1c of 6.5% to 7.4%, 1388 patients had an HbA1c of 7.5% to 9.9%, and 402 patients had an HbA1c of 10% or higher. As a linear variable, HbA1c was associated with higher rates of COVID-19 hospitalization (adjusted hazard ratio [aHR], 1.14 [95% CI, 1.05-1.23]) and of death from COVID-19 (aHR, 1.18 [95% CI, 1.05-1.33]), where the hazard ratio reflects the increase in hazard associated with an 1% absolute increase in HbA1c. For aim 3, of the 2982 patients who met all inclusion criteria, 2199 were exposed to metformin, 782 to sulfonylurea, 1047 to a DPP-4 inhibitor, 326 to a GLP-1 receptor agonist, and 275 to an SGLT2 inhibitor. The unadjusted rates of COVID-19 hospitalization were 0.28, 0.50, 0.41, 0.52, and 0.28 per 1000 person-days for patients with baseline prescriptions for metformin, sulfonylureas, DPP-4 inhibitors, SGLT2 inhibitors, and GLP-1 receptor agonists, respectively. After adjustment for covariates, no antidiabetic drug classes were associated with statistically significant differences in COVID-19 hospitalization. CONCLUSIONS: Rates of missing data on hospitalizations and antidiabetic drug exposure in the INSIGHT CRN data could be substantially improved by restriction to patients with predicted low levels of missing data (ie, high levels of data completeness). Patients with diabetes with high baseline HbA1c had a greater risk of COVID-19 hospitalization. Preventive efforts for COVID-19 should allocate additional resources to patients with diabetes with poor glucose control. All antidiabetic drug classes had similar adjusted hazard ratios for COVID-19 hospitalization. LIMITATIONS: As documented in aim 1, rates of missing data on important outcomes (ie, hospitalization) and exposures (ie, antidiabetic drugs) in the CRN were substantial in the lower deciles of data completeness. This research also has the typical limitations of observational research, including the potential for unmeasured and/or uncontrolled confounding.
