Substance use disorders (SUDs) are common: ~10.8% of adults in the US have had a problem with drug use1, 2. SUDs are often associated with multiple comorbid conditions that are known risk factors for severe outcomes of COVID-19 infection, including cardiovascular, cerebrovascular, immune, hematological, pulmonary, metabolic, oncological, hepatic, renal, infectious, neurological and psychiatric diseases3-11. Additionally, studies from the early pandemic showed that patients with SUDs – including alcohol use disorder, cannabis use disorder, cocaine use disorder, opioid use disorder, and tobacco use disorder – were at increased risk for COVID-19 infection and associated severe outcomes, especially among African Americans6.

In the US, three vaccines have been approved since December 2020: two mRNA vaccines developed by Pfizer-BioNTech and Moderna, and an adenovirus vaccine by Johnson & Johnson. Clinical trial data showed an efficacy of 95% for the Pfizer-BioNTech12, 94.1% for the Moderna13 and 66.3% for the Johnson & Johnson vaccine14 in preventing COVID-19 infection. Clinical trials for COVID-19 vaccines did not explicitly include SUD patients, though they did include – for example, in the clinical trial for Pfizer-BioNTech vaccine – participants with a range of other diseases, including cancers, cardiovascular diseases, human immunodeficiency virus (HIV) infection, and renal diseases12. Currently, there are no systematic studies examining the real-world effectiveness of COVID-19 vaccines in populations with various SUDs. Vaccines are very effective, but breakthrough infections have been recorded15-18, highlighting the need to identify populations that might be most vulnerable, as we have entered a worrisome new phase of the pandemic.

Studies have shown that individuals with compromised immune function, such as organ transplant recipients and cancer patients, have limited rates of SARS-CoV-2 IgG seroconversion19-23. Drugs and alcohol affect immune function, which is likely to contribute to the higher risk for infections in individuals with SUDs3, 6, 24, 25. Thus, we hypothesized that individuals with SUDs could be at increased risk for vaccine breakthrough COVID-19 infection.

In this study, we estimated the risk for breakthrough COVID-19 infection among vaccinated patients with various SUDs compared to matched vaccinated individuals without SUDs. We also examined how the rate of breakthrough cases changed between December 2020 and August 2021.

METHODS Study population

We used the TriNetX Analytics network platform26, which allows access to de-identified data of 84.5 million unique patients from 63 health care organizations in the US, among whom 15 million (age ≥12 years) had medical encounter(s) with health care organizations since December 1, 2020.

The study population comprised 579,372 individuals who fulfilled the following inclusion criteria: a) they had medical encounter(s) with health care organizations since December 1, 2020; b) they had documented evidence of full vaccination in electronic health records (i.e., they had received a second dose of Pfizer-BioNTech or Moderna vaccine, or a single dose of Johnson & Johnson vaccine) between December 1, 2020 and August 14, 2021; and c) they had not contracted COVID-19 infection prior to vaccination.

The fully vaccinated study population included 30,183 patients with SUD and 549,189 patients without SUD. Among the fully vaccinated population with SUD, 7,802 patients had a diagnosis of alcohol use disorder, 2,058 of cannabis use disorder, 1,011 of cocaine use disorder, 2,379 of opioid use disorder, and 21,941 of tobacco use disorder.

TriNetX Analytics provides web-based real-time secure access to patient electronic health records from hospitals, primary care and specialty treatment providers, covering diverse geographic locations, age groups, ethnic groups, and income levels. Though the data are de-identified, end-users can use the platform built-in functions working on patient-level data for cohort selection and matching, analyzing incidence and prevalence of events in a cohort, and comparing characteristics and outcomes between matched cohorts. Multiple studies have used TriNetX to study risk, disparity, sequelae, temporal trends, clinical characteristics, and outcomes of COVID-19 infection27-30.

The status of COVID-19 infection was based on the ICD-10 diagnosis code of “COVID-19” (U07.1) or lab-test confirmed presence of “SARS coronavirus 2 and related RNA” (TNX:LAB:9088). The status of full vaccination was based on the Current Procedural Terminology (CPT) relevant codes for Pfizer-BioNTech (0002A), Moderna (0012A) and Johnson & Johnson (0031A) vaccines.

The status of SUD was based on the ICD-10 diagnosis code of “mental and behavioural disorders due to psychoactive substance use” (F10-F19). The status of alcohol use disorder was based on the ICD-10 diagnosis code of “alcohol related disorders” (F10); that of cannabis use disorder on the code of “cannabis related disorders” (F12); that of cocaine use disorder on the code of “cocaine related disorders” (F14); that of opioid use disorder on the code of “opioid related disorders” (F11); and that of tobacco use disorder on the code of “nicotine dependence” (F17). Other subtypes of SUD, such as methamphetamine use disorder, were not examined due to their small sample sizes.

For breakthrough outcome measures, the status of hospitalization was based on the CPT code “hospital inpatient services” (013659), while the status of death was based on the vital status code “deceased” that TriNetX regularly imports from the Social Security Death index.

Procedures

We tested whether fully vaccinated SUD patients had higher risk for breakthrough infection than non-SUD patients. Separate analyses were performed for alcohol use disorder, cannabis use disorder, cocaine use disorder, opioid use disorder, and tobacco use disorder.

The cohorts of SUD and non-SUD patients were created by propensity score matching for demographics (age, gender, ethnicity); adverse socioeconomic determinants of health (including “problems related to education and literacy”, “problems related to employment and unemployment”, “occupational exposure to risk factors”, and “problems related to housing and economic circumstances”, according to the ICD-10); lifetime comorbidities (hypertension, heart diseases, cerebrovascular diseases, obesity, type 2 diabetes, cancers, chronic respiratory diseases, chronic kidney diseases, liver diseases, blood diseases and disorders involving immune mechanisms, HIV infection, dementia, depression, and psychotic disorders), and vaccine types (Pfizer, Moderna and Johnson & Johnson).

The TriNetX built-in propensity score matching function was used (1:1 matching using a nearest neighbor greedy matching algorithm with a caliper of 0.25 times the standard deviation). The outcome was COVID-19 infection at least 14 days after patients received the second dose of Pfizer-BioNTech or Moderna vaccine or a single dose of Johnson & Johnson vaccine. Kaplan-Meier analysis was performed to estimate the probability of breakthrough infection from day 14 after full vaccination to August 28, 2021. Comparisons between cohorts were made using a log-rank test (a built-in function in TriNetX). The hazard ratio (HR) was used to describe the relative risk of breakthrough infection based on comparison of time to event rates, and was calculated using a proportional hazard model (a built-in function in TriNetX). The proportional hazard assumption was tested using the generalized Schoenfeld approach.

We tested whether fully vaccinated patients who received Pfizer-BioNTech vaccine had a different risk of developing breakthrough COVID-19 infection compared with a matched cohort of patients who received Moderna vaccine. Johnson & Johnson vaccine was not examined due to small sample size. The Pfizer and Moderna cohorts were propensity-score matched for demographics, adverse socioeconomic determinants of health, and comorbid medical conditions. Kaplan-Meier analysis was used to estimate the probability of breakthrough infection from day 14 after full vaccination to August 28, 2021. Separate analyses were performed for SUD, SUD subtypes, and non-SUD individuals. HR was calculated to compare the relative risk of breakthrough infection in two matched cohorts.

We explored how the rates of breakthrough infection in fully vaccinated SUD and non-SUD populations, measured by cases/person-day for each month, evolved between December 2020 and August 2021. TriNetX built-in functions were used for calculating proportion rates.

We tested whether fully vaccinated patients with breakthrough infection had different risk for hospitalization and death compared with a matched cohort without breakthrough infection. Breakthrough and non-breakthrough cohorts were propensity-score matched for demographics, adverse socioeconomic determinants of health, comorbid medical conditions, and vaccine types. For the breakthrough cohort, overall risks of hospitalization and death were calculated from the day of infection to August 28, 2021. For the non-breakthrough cohort, overall risks of hospitalization and death were calculated from day 14 after full vaccination to August 28, 2021. Relative risk (RR) was used to compare matched cohorts. Separate analyses were performed for SUD and non-SUD populations.

We investigated how the risks for breakthrough infection in fully vaccinated patients differed by age, gender and ethnicity. The case cohort comprised fully vaccinated patients with one of the following demographic factors: female, older (age ≥65 years), or African American. The comparison cohort comprised matched vaccinated SUD patients with one of the following corresponding factors: male, younger (age <65 years), or Caucasian. Two cohorts were propensity-score matched on other demographics, adverse socioeconomic determinants of health, comorbid medical conditions, and vaccine types. Kaplan-Meier analysis was performed to estimate the probability of breakthrough infection from day 14 after full vaccination to August 28, 2021 in matched cohorts. HR was used to compare the relative risk of breakthrough infection between matched cohorts. Separate analyses were done for SUD, non-SUD and each SUD subtype.

We examined how the timing of recent medical encounters for SUD diagnosis was associated with the risk of breakthrough infection among fully vaccinated SUD patients. Four cohorts of SUD patients were used: a) “Ever” (all SUD patients, irrespective of when they had a medical encounter for their diagnosis, thus including both recovered patients and those with active SUD); b) “February 2019” (patients who had a medical encounter for their SUD diagnosis after February 2019); c) “February 2020” (patients who had a medical encounter for their SUD diagnosis during the pandemic, i.e. any time after February 2020); and d) “December 2020” (patients who had a medical encounter for their SUD diagnosis after the COVID-19 vaccine was approved, thus most likely having a currently active SUD). The “Ever” group was used as the reference one to which the risk of breakthrough infection in the other groups was compared. Separated analyses were conducted for each SUD subtype.

Statistical tests were conducted with significance set at p<0.05 (two sided) using R, version 3.6.3.

RESULTS

The demographic characteristics of the fully vaccinated patients and the sample sizes as a function of SUD subtype are shown in Table 1. Among vaccinated SUD patients, 75.6% received Pfizer-BioNTech, 21.1% Moderna, and 3.3% Johnson & Johnson vaccine. Among vaccinated non-SUD population, 88.2% received Pfizer-BioNTech, 10.6% Moderna and 1.2% Johnson & Johnson vaccine.

Table 1. Characteristics of substance use disorder (SUD) and non-SUD vaccinated populations AUD CUD CocaineUD OUD TUD SUD Non-SUD Total number of patients 7,802 2,058 1,011 2,379 21,941 30,184 549,189 Age (years, mean±SD)* 59.3±14.4 47.9±16.3 55.1±12.2 59.1±14.2 59.6±13.5 59.3±14.4 54.7±19.8 Gender (% male)* 61.8 60.1 61.5 45.7 50.3 51.4 43.1 Ethnicity (%) White 69.0 57.7 41.9 62.8 62.1 63.2 63.4 African American* 21.9 33.4 50.1 29.6 28.5 26.2 14.3 Hispanic/Latino 5.0 4.6 5.0 3.2 3.8 4.3 12.3 Asian 1.2 1.1 1.0 1.0 2.2 2.0 8.6 Unknown 7.4 7.4 6.7 6.0 6.9 7.2 12.6

Adverse socioeconomic determinants of health (%)*

10.8 18.7 22.6 14.1 7.8 7.9 1.2 Lifetime medical conditions (%) Hypertension* 63.3 50.8 66.8 67.2 62.9 61.6 22.8 Heart diseases* 19.6 17.0 24.2 21.0 21.5 20.1 5.3 Cerebrovascular diseases* 15.0 13.0 19.6 15.0 13.3 13.2 3.6 Obesity* 27.7 31.1 33.4 36.6 31.2 30.4 9.3 Type 2 diabetes* 21.6 19.9 28.9 30.7 25.7 24.6 8.4 Cancers* 48.8 40.9 46.6 44.5 45.2 44.9 16.2 Chronic respiratory diseases* 30.1 35.1 44.8 39.7 38.7 34.7 7.6 Chronic kidney diseases* 11.9 11.5 18.8 15.7 10.8 11.3 3.5 Liver diseases* 26.3 18.0 29.2 21.9 15.4 16.9 3.2 Blood diseases and disorders involving immune mechanisms* 41.1 40.0 50.1 49.8 34.3 35.6 10.5 HIV infection* 3.3 8.4 12.5 7.1 3.1 3.2 0.3 Dementia* 2.2 0.9 1.6 2.2 1.2 1.4 0.5 Major depression* 37.0 51.8 52.3 48.0 29.2 30.9 6.0 Psychotic disorders* 4.7 12.9 16.9 6.3 3.5 3.6 0.3 Lifetime organ transplants (%)* 3.9 3.7 3.8 3.4 1.8 2.6 0.7 * Significant difference between SUD and non-SUD populations, p<0.001. AUD – alcohol use disorder, CUD – cannabis use disorder, CocaineUD – cocaine use disorder, OUD – opioid use disorder, TUD – tobacco use disorder

Patients with SUD were older (mean age: 59.3±14.4 years) than those without SUD (54.7± 19.8 years). There were more men in the SUD population (51.4% vs. 43.1%), and the percentage of African Americans was higher in the SUD (26.2%) than in the non-SUD (14.3%) sample. The prevalence of adverse socioeconomic determinants of health was also higher in the SUD population than in patients without SUD (7.9% vs. 1.2%). Vaccinated patients with SUD had a higher lifetime prevalence of all comorbidities, as well as of transplants (all p<0.001).

Among the vaccinated population, the risk of breakthrough infection ranged from 6.8% for tobacco use disorder to 7.8% for cannabis use disorder, all significantly higher than the 3.6% in the non-SUD population (p<0.001). The HRs between SUD and non-SUD cohorts after propensity score matching for demographics (age, gender, ethnicity) and vaccine types remained significantly higher for all SUD subtypes except for tobacco use disorder, being highest for cocaine use disorder and cannabis use disorder (HR=1.17, 95% CI: 1.01-1.35 for alcohol; HR=1.92, 95% CI: 1.39-2.66 for cannabis; HR=2.06, 95% CI: 1.30-3.25 for cocaine; and HR=1.31, 95% CI: 1.00-1.71 for opioids) (see Table 2).

Table 2. Risk of breakthrough COVID-19 infection in propensity-score matched (demographics and vaccine types) substance use disorder (SUD) and non-SUD populations Cohort Patients in cohort Risk in cohort Risk in matched non-SUD cohort Hazard ratio (95% CI) AUD 7,802 7.2% 3.7% 1.17 (1.01-1.35) CUD 2,055 7.8% 2.3% 1.92 (1.39-2.66) CocaineUD 1,011 7.7% 2.4% 2.06 (1.30-3.25) OUD 2,379 7.1% 3.2% 1.31 (1.00-1.71) TUD 21,935 6.8% 3.9% 1.06 (0.98-1.15) AUD – alcohol use disorder, CUD – cannabis use disorder, CocaineUD – cocaine use disorder, OUD – opioid use disorder, TUD – tobacco use disorder

After controlling for adverse socioeconomic determinants of health and comorbid medical conditions, the risk for breakthrough infection no longer differed in SUD compared to non-SUD cohorts, except for patients with cannabis use disorder, who remained at significantly increased risk (HR=1.55, 95% CI: 1.22-1.99) (see Table 3).

Table 3. Risk of breakthrough COVID-19 infection in propensity-score matched (adverse socioeconomic determinants of health and comorbid medical conditions, in addition to demographics and vaccine types) substance use disorder (SUD) and non-SUD populations Cohort Patients in cohort Risk in cohort Risk in matched non-SUD cohort Hazard ratio (95% CI) AUD 7,754 7.2% 6.9% 1.09 (0.96-1.22) CUD 2,032 7.8% 5.4% 1.55 (1.22-1.99) CocaineUD 991 7.7% 7.5% 1.15 (0.83-1.58) OUD 2,360 7.0% 7.6% 0.94 (0.76-1.16) TUD 21,757 6.8% 6.8% 1.03 (0.96-1.11) AUD – alcohol use disorder, CUD – cannabis use disorder, CocaineUD – cocaine use disorder, OUD – opioid use disorder, TUD – tobacco use disorder

Among SUD and non-SUD populations, the risk for breakthrough infection was higher in individuals who received the Pfizer than the Moderna vaccine, after matching for demographics, adverse socioeconomic determinants of health, and comorbid medical conditions (HR in SUD cohort: 1.49, 95% CI: 1.31-1.69; HR in non-SUD cohort: 1.45, 95% CI: 1.38-1.53). The same trend was observed in SUD subtypes (see Table 4).

Table 4. Risk of breakthrough COVID-19 infection in propensity-score matched (demographics, adverse socioeconomic determinants of health, and comorbid medical conditions) substance use disorder (SUD) and non-SUD populations receiving Pfizer and Moderna vaccine Cohort Risk in patients receiving Pfizer Risk in patients receiving Moderna Hazard ratio (95% CI) SUD 8.7% 6.3% 1.49 (1.31-1.69) AUD 8.9% 7.1% 1.41 (1.10-1.80) CUD 8.2% 7.3% 1.16 (0.68-1.97) CocaineUD 7.3% 4.9% 2.78 (1.08-7.16) OUD 9.7% 6.6% 1.56 (1.01-2.42) TUD 9.0% 5.8% 1.69 (1.46-1.97) Non-SUD 5.4% 4.7% 1.45 (1.38-1.53) AUD – alcohol use disorder, CUD – cannabis use disorder, CocaineUD – cocaine use disorder, OUD – opioid use disorder, TUD – tobacco use disorder

The rate of breakthrough infection in the SUD population steadily increased from 0 cases/person-day in January 2021 to 0.001 cases/person-day in June 2021 to 0.0025 cases/person-day in August 2021 (2.5 times faster than in June 2021). A similar trend was observed in the non-SUD population: the rate of breakthrough infection steadily increased from 0 cases/person-day in January 2021 to 0.0009 cases/person-day in June 2021, and then reached 0.0049 cases/person-day in August 2021 (5.4 times faster than in June 2021) (see Figure 1).

Time trend of incidence rates (cases/person-day) of breakthrough COVID-19 infection in patients with and without substance use disorder (SUD)

Within the SUD population, the overall risk for hospitalization was 22.5% in the breakthrough cohort compared to 1.6% in the matched non-breakthrough cohort (RR=14.4, 95% CI: 10.19-20.42). The overall risk for death was 1.7% in the breakthrough cohort, compared to 0.5% in the matched non-breakthrough cohort (RR=3.5, 95% CI: 1.74-7.05).

Within the non-SUD population, the overall risk for hospitalization was 17.5% in the breakthrough cohort compared to 0.5% in the matched non-breakthrough cohort (RR=34.2, 95% CI: 28.05-41.67). The overall risk for death was 1.1% in the breakthrough cohort compared to 0.2% in the matched non-breakthrough cohort (RR=6.0, 95% CI: 4.20-8.66).

No significant age, gender and ethnic disparities of breakthrough infections were observed in SUD patients after matching for other demographics, adverse socioeconomic determinants of health, comorbid medical conditions and vaccine types, except for patients with cannabis use disorder, among whom African Americans had higher risk than matched Caucasians (HR=1.63, 95% CI: 1.06-2.51). Among vaccinated non-SUD population, older individuals (age ≥65 years) were more likely to have breakthrough infections than younger patients after matching for gender, ethnicity, adverse socioeconomic determinants of health, and comorbid medical conditions (HR=1.08, 95% CI: 1.04-1.13); women had lower risk than matched men (HR=0.87, 95% CI: 0.84-0.90); and African Americans had higher risk than matched Caucasians (HR=1.12, 95% CI: 1.07-1.18) (see Figure 2).

Hazard ratios of breakthrough COVID-19 infection in fully vaccinated substance use disorder (SUD) and non-SUD populations: female vs. male; older (age ≥65 years) vs. younger (age <65 years); African American vs. Caucasian. Two demographic-stratified cohorts were propensity-score matched based on other demographics (age, gender, ethnicity), adverse socioeconomic determinants of health, comorbid medical conditions, and vaccine types. AUD – alcohol use disorder, CUD – cannabis use disorder, CocaineUD – cocaine use disorder, OUD – opioid use disorder, TUD – tobacco use disorder

Within the SUD population, the risk for breakthrough infection was higher for patients who had recent medical encounters for their SUD diagnosis, ranging from 7.0% in the “Ever” group to 10.5% in the “December 2020” group (p<0.001 between these two groups). The same trends were observed for SUD subtypes (see Table 5).

Table 5. Risk of breakthrough COVID-19 infection among fully vaccinated substance use disorder (SUD) patients who had medical encounters for their diagnosis at different time cutoffs Medical encounter for SUD Patients on cohort Patients with infection Risk of infection p SUD Ever 30,183 2,113 7.0% Ref. Feb. 2019 4,185 366 8.7% 0.003 Feb. 2020 13,621 1,181 8.7% <0.001 Dec. 2020 9,041 946 10.5% <0.001 AUD Ever 7,802 563 7.2% Ref. Feb. 2019 4,185 366 8.7% 0.003 Feb. 2020 2,959 294 9.9% <0.001 Dec. 2020 1,858 222 11.9% <0.001 CUD Ever