Infections as a Risk Factor for and Prognostic Factor After Substance-Induced Psychoses

We included all people born in Denmark since 1981. For the analyses concerning conversion from substance-induced psychosis, we included only the subpopulation diagnosed with substance-induced psychosis according to the definition below.

Information on substance-induced psychosis was obtained from the Psychiatric Central Research Register as ICD-10 codes F1x.5 (16). We also conducted analyses on the following specific types of substance-induced psychosis: alcohol (F10.5), opioids (F11.5), cannabis (F12.5), sedatives (F13.5), cocaine (F14.5), amphetamines (F15.5), hallucinogens (F16.5), volatile solvents (F18.5), and multiple drug use or other substances (F19.5).

Infections were identified through diagnostic codes in the National Patient Register (17), as listed in Table S1 in the online supplement. Pregnancy-related infections were not analyzed separately but were included in the overall definition of “any infection.” All other sites of infection were analyzed separately as well as being included in the “any infection” category.

Information on diagnosis of schizophrenia was obtained from the Psychiatric Central Research Register as ICD-10 codes F20.x. Since this information was used only in the subanalysis of people previously diagnosed with substance-induced psychosis, and because substance-induced psychoses were only made into separate diagnostic entities in ICD-10, we did not need to include information on schizophrenia using ICD-8 codes.

Alcohol or substance use disorders were identified by linking three registers. In the Psychiatric Central Research Register and the National Patient Register, the following diagnostic codes were used to indicate alcohol or substance use disorders: ICD-8 codes 291, 303, and 571.0; ICD-10 codes F1x (except F1x0.5), E52, G31.2, G62.1, G72.1, I42.6, K29.2, K70, K86.0, O35.4, Y57.3, Z50.2, Z71.4, and Z72.1. This was supplemented with information from the National Prescription Registry (18) regarding receipt of medication for alcohol or substance use disorders, identified as the following ATC codes: N07BB01, N07BB02, N07BB03, N07BC01, N07BC51, N07BC02, or N07BC03. We also included information on parental substance use disorder, using the same information. Parental psychosis was defined in the Psychiatric Central Research Register as ICD-8 codes 295, 297, 298.29, 299.39, 298.99, 299.05, 299.09, 301.09, and 301.29 as well as ICD-10 F2x. Other parental psychiatric disorders were identified as all diagnoses in ICD-8 chapters 290–315 and the IDC-10 F chapter that were not part of either the substance use disorder or the psychotic disorder definitions. Highest parental level of education was identified in the Statistics Denmark registers (19). All parental variables were constructed separately for fathers and mothers and included an “unknown” category for cases in which the respective parent was not registered or for which the outcome of interest, e.g., education, was not available for the parent.

The first part of the study concerned the role of infections in the development of substance-induced psychosis. For these analyses, we conducted Cox regression with time-varying covariates, in which people were followed from birth until substance-induced psychosis, death, migration, or end of registers (April 10, 2017). Data on infections were entered into the analyses as time-varying covariates. Analyses in model 1 were adjusted for sex, year of birth (continuous variable), and calendar year (time-varying continuous variable). Analyses in model 2 were further adjusted for alcohol or substance use disorders as a time-varying covariate, as well as parental education, parental alcohol or substance use disorders, parental psychotic disorder, and other parental psychiatric disorders. The analyses were conducted both for any substance-induced psychosis and for each specific type of substance-induced psychosis.

The second part of the study concerned the role of infections in the conversion from substance-induced psychosis to schizophrenia. For this analysis, we conducted Cox regression with time-varying covariates, but included only those individuals from the first subanalysis who had developed substance-induced psychosis and followed them from this point until incident diagnosis of schizophrenia, death, migration, or end of registers. People with schizophrenia diagnosed prior to the first diagnosis of substance-induced psychosis were excluded from the analyses. Dataon infections were entered into the analyses as time-varying covariates, both in the period before incident substance-induced psychosis and following the diagnosis of substance-induced psychosis. Analyses in model 1 were adjusted for sex, year of birth (continuous variable), and calendar year (time-varying continuous variable). Analyses in model 2 were further adjusted for infections requiring care in the secondary health care sector prior to the first diagnosis of substance-induced psychosis. Analyses in model 3 were further adjusted for alcohol or substance use disorders as a time-varying covariate, as well as parental education, parental alcohol or substance use disorders, parental psychotic disorder, and other parental psychiatric disorders.

Figure 1 shows the cumulative incidence rates of substance-induced psychosis according to presence of infections. For people with no infections treated in the secondary health care sector, the cumulative incidence of substance-induced psychosis by age 36 was 0.4%, compared with 0.6% for individuals with at least one such infection. The highest cumulative incidence rate was found after hepatitis, with a cumulative incidence rate by age 36 of 3.4%.

Table 2 lists the relative contributions, as hazard ratios, of infections to the risk of being diagnosed with substance-induced psychosis. In model 1, all types of infections were associated with an increased risk of substance-induced psychosis. The largest increase in risk, at nearly eightfold, was observed for hepatitis. Sepsis, skin infections, urogenital infections, and the category of other infections were all associated with more than a 50% increase in risk of substance-induced psychosis compared with no such infections. In model 2, all associations were attenuated, but only CNS infections were no longer statistically significantly associated with the risk of substance-induced psychosis. In this model, hepatitis was still associated with more than a threefold increase in substance-induced psychosis, whereas all other types of infections had hazard ratios between 1 and 2.

Figure 2 shows the association between any infection and substance-induced psychosis as a function of time since the first severe infection, indicating that the risk of substance-induced psychosis was doubled the first 2 years after a severe infection. Even 20 years after the severe infection, the risk of substance-induced psychosis was still increased (hazard ratio=1.28, 95% CI=1.16–1.42).

Tables S2–S8 in the online supplement list hazard ratios for the associations between infections and psychosis induced, respectively, by alcohol, opioids, cannabis, cocaine, amphetamines, hallucinogens, and mixed or other substances. There were too few cases of sedative- and volatile-substance-induced psychosis for the statistical models to converge.

Infections generally increased the risk of all types of substance-induced psychosis except hallucinogen-induced psychosis. There were, however, differences in which type of infection predicted types of induced psychoses. For instance, hepatitis was associated with an increased risk of cocaine- or amphetamine-induced psychosis as well as psychosis induced by mixed or other substances, but not cannabis-induced psychosis. Hepatitis was associated with an increased risk of alcohol- or hallucinogen-induced psychosis in model 1 but not in the fully adjusted model 2. Sepsis was associated with an increased risk of all types of substance-induced psychosis except those induced by alcohol, cocaine, or hallucinogens. Gastrointestinal infections were associated with most types of substance-induced psychosis in model 1, but after full adjustment in model 2, they were associated only with cocaine- and possibly opioid-induced psychoses. Skin infections were associated with all types of substance-induced psychosis except hallucinogen-induced psychosis. Respiratory infections were associated with psychosis induced by opioids, cocaine, and mixed or other substances. Urogenital infections were associated with psychosis induced by opioids, cannabis, and mixed or other substances. CNS infections were positively associated only with opioid-induced psychosis but displayed a protective effect against psychosis induced by the category of mixed or other substances, the only such negative association observed in our study. Finally, other infections increased the risk of all types of substance-induced psychosis except those induced by opioids or hallucinogens.

Given the generally broad confidence intervals caused by the relatively infrequent outcomes of specific types of substance-induced psychosis, the absolute strengths of the observed associations are somewhat uncertain.

Of the 3,618 people with substance-induced psychosis identified in the previous analysis, 351 were excluded for prior diagnoses of schizophrenia, leaving 3,267 for the analyses regarding conversion from substance-induced psychosis to schizophrenia. These people were followed for 15,793 person-years, during which time 813 conversions to schizophrenia occurred. For people with no infections treated in the secondary health care sector, the cumulative incidence rate of schizophrenia 20 years after substance-induced psychosis was 37.3%. For those with such infections, the cumulative incidence rate was 33.7%. The highest cumulative incidence rate was observed for people with hepatitis diagnosed after substance-induced psychosis. In this group, after 20 years, 57.5% had converted to schizophrenia.

Table 3 lists the relative contributions, as hazard ratios, of infections both preceding and following substance-induced psychosis to the risk of converting from a diagnosis of substance-induced psychosis to one of schizophrenia. Overall, neither was associated with an increased risk of conversion to schizophrenia. When specific sites of infection were analyzed, only hepatitis was associated with an increased risk of conversion. There was very little change between results from the three models with different variables used for adjustment.

To our knowledge, this is the first study to investigate the association between infections and substance-induced psychosis. Our overall findings, however, mirror previous findings on the association between infections and schizophrenia, including previous observations that the link is particularly strong for hepatitis (6, 8–10, 20). This supports the hypothesis of an immune-related component not just for schizophrenia but for psychosis in general (10, 20). The biological mechanisms through which infections would increase the risk of psychosis, including substance-induced psychosis, are still poorly understood. Infections can affect brain functioning directly by entering the CNS or by inducing peripheral inflammation, making the blood-brain barrier more permeable to detrimental substances in the blood, such as cytokines and autoantibodies (10).

Of course, substance-induced psychoses would never manifest in the absence of use of substances. This raises the possibility that the associations we observe with infections may simply be associations between infections and substance use, which in some people then develop into substance-induced psychosis. For this reason, we adjusted our analyses for a range of potential confounders, including substance use disorders. While these adjustments attenuated many associations to a certain degree, most of the associations remained both strong and statistically significant. However, especially in the case of hepatitis, some residual confounding regarding severe substance use disorder may still be present. Hepatitis is linked to polydrug use, and this may not have been completely accounted for even in our fully adjusted models. Consequently, we may have overestimated the true association between hepatitis and substance-induced psychosis.

Noncausal mechanisms for our findings should also be considered. Infections may be considered a marker of social exclusion and poor general health. This may point to the link being one of shared underlying risk factors rather than one of causality. While this is likely to explain part of our results, it should be noted that the observed associations were particularly strong during the first 2 years after an infection. This does point toward the infection being involved etiologically in the development of substance-induced psychosis.

While hepatitis was the strongest overall risk factor for substance-induced psychosis, further analyses suggested that this type of infection was linked primarily to psychoses induced by cocaine, amphetamines, or mixed or other substances. Substance users have been well documented to have an increased risk of hepatitis (21), and this may explain why, for instance, the association between hepatitis and alcohol-induced psychosis disappeared after adjustment for, among other confounders, alcohol and substance use disorders. However, in light of this adjustment, it does seem to be the case that hepatitis increases vulnerability to psychotic states from these two classes of substances, irrespective of whether the person might have acquired hepatitis through substance use. Likewise, the act of injecting substances, rather than the substances themselves, may also play a role, as associations remained significant for substances with a potential for injection.

Sepsis was associated with all types of substance-induced psychosis except those induced by alcohol, cocaine, or hallucinogens. These latter three types are all relatively infrequent (at least in younger populations), and these absent associations may indeed reflect inadequate statistical power. Hence, it may actually be the case that sepsis, a severe, systemic infectious state, increases vulnerability to the psychotogenic properties of all substances. Sepsis has certainly been shown to affect the blood-brain barrier, and it can have long-term neurological and cognitive sequelae (22, 23). Similarly, skin infections were also associated with nearly all individual types of substance-induced psychosis.

Gastrointestinal infections were broadly associated with most types of substance-induced psychosis, but most of these associations disappeared when we controlled for, among other covariates, alcohol and substance use disorders. Why gastrointestinal infections would be particularly subject to such confounding is not entirely clear.

The reasons for the selective associations we observed for respiratory, urogenital, and CNS infections are not readily evident and may require further studies to be fully understood.

There was no difference in the rate of conversion to schizophrenia depending on whether the substance-induced psychosis was preceded by a severe infection. Similarly, most infections after the substance-induced psychosis did not increase the risk of subsequent conversion to schizophrenia. However, hepatitis following substance-induced psychosis nearly doubled the risk of conversion to schizophrenia, with no attenuation of the association even in the fully adjusted model. To our knowledge, no previous studies have examined whether infections increase the risk of conversion from substance-induced psychosis to schizophrenia. However, the finding that hepatitis is the strongest candidate for development of schizophrenia mirrors a previous study (6), and also mirrors the finding that hepatitis was the strongest risk factor for substance-induced psychosis in the first part of the present study.

Our findings lend strong support to the hypothesis that there is an immunological component to psychosis. This support is particularly strong given that we observed associations between infection and psychosis even in a set of disorders in which the etiology may have seemed straightforward in the sense that the disorders were induced by substances such as cannabis or amphetamines. If the exact mechanisms underlying the psychotogenic properties of infections or the immune response can be identified, this is likely to lead to improvements in treatment for psychotic disorders. A further hope is that it may even be possible to use this knowledge for primary prevention of psychosis.

Our findings may also hold direct relevance in terms of secondary and tertiary prevention of both substance-induced psychosis and later conversion to schizophrenia. The fact that among people with hepatitis, more than 3% will have developed a substance-induced psychosis by age 35 holds some clinical potential. If substance use can be prevented in this group, it may reduce their risk of psychosis. Similarly, if substance use cannot be prevented, it may be worth following these individuals closely in order to prevent substance-induced psychosis or at least to catch the disorder at an early stage. Since hepatitis is also linked to injection drug use, there may also be an argument for reducing the infection rate in substance users by supplying injection drug users with clean syringes. The positive health impacts of such an intervention would naturally go far beyond those related to substance-induced psychosis (24).

The use of nationwide Danish treatment registers is a strength of our study, as they allow us to identify incident cases without the risk of selection bias that occurs if individuals have to consent to being part of a study. It also allows a prospective design, which would probably not be feasible for studies approaching individual patients for consent to participate in a study when the outcome is so rare that large numbers of study subjects are needed.

Our study design also has limitations, however. We relied exclusively on treatment registers of the secondary health care sector for infections, substance-induced psychosis, and conversions to schizophrenia. While most cases of schizophrenia will probably eventually present in the secondary health care sector, the same is not necessarily true for either infections or substance-induced psychosis. Many infections do not require hospitalization, meaning that we only identified the most severe infections. If other types of infections—for example, those that go untreated or are treated in primary care—also increase the risk of substance-induced psychosis, this would have led to our estimates being at least slightly conservative. It is also possible that some proportion of substance-induced psychoses do not lead to contacts with the secondary psychiatric health care system because of their often transient nature. Again, this would likely cause our estimates to be slightly conservative.

A further limitation lies in the relatively young age of our study cohort. They were born in 1981 or later and were followed until April 2017, so the oldest individuals in the cohort only reached age 36 during the study’s time frame. This was a necessary choice because substance-induced psychoses were not a separate diagnostic entity in the ICD-8 system, which was used in Denmark until 1994. Consequently, it is unclear whether our results can be generalized to people getting severe infections in their 40s and beyond. Some of the individuals in our study may develop substance-induced psychosis or schizophrenia after the final date of our follow-up, which would again likely indicate that our results may be slightly conservative.