Genetic prediction of the causal relationship between schizophrenia and tumors: a Mendelian randomized study

Background Patients with schizophrenia are at a higher risk of developing cancer. However, the causal relationship between schizophrenia and different tumor types remains unclear. Methods Using a two-sample, two-way Mendelian randomization method, we used publicly available genome-wide association analysis (GWAS) aggregate data to study the causal relationship between schizophrenia and different cancer risk factors. These tumors included lung adenocarcinoma, lung squamous cell carcinoma, small-cell lung cancer, gastric cancer, alcohol-related hepatocellular cancer, tumors involving the lungs, breast, thyroid gland, pancreas, prostate, ovaries and cervix, endometrium, colon and colorectum, and bladder. We used the inverse variance weighting (IVW) method to determine the causal relationship between schizophrenia and different tumor risk factors. In addition, we conducted a sensitivity test to evaluate the effectiveness of the causality. Results After adjusting for heterogeneity, evidence of a causal relationship between schizophrenia and lung cancer risk was observed (odds ratio [OR]=1.001, 95% confidence interval [CI], 1.000–1.001; P=0.0155). In the sensitivity analysis, the causal effect of schizophrenia on the risk of lung cancer was consistent in both direction and degree. However, no evidence of causality or reverse causality between schizophrenia and other tumors was found. Conclusion This study elucidated a causal relationship between the genetic predictors of schizophrenia and the risk of lung cancer, thereby providing a basis for the prevention, pathogenesis, and treatment of schizophrenia in patients with lung cancer.


Introduction
Schizophrenia is a serious mental illness that affects the thinking, behavior, and emotions of patients (1).Recent studies have demonstrated a complex relationship between schizophrenia and tumor incidence (2).
Symptoms of schizophrenia include hallucinations, delusions, and confusion (3,4), whereas tumors are pathological changes caused by excessive cell growth.Although they seem unrelated, the relationship between them may be more complicated than anticipated.However, whether there is a causal relationship between schizophrenia and tumors remains controversial.Some observational studies have suggested a causal relationship between schizophrenia and tumor risk (5), whereas others have not (6).In addition, the incidence of tumors in patients with schizophrenia differs from that in the general population.Patients with schizophrenia may have a lower cancer incidence (7); however, this does not imply people with schizophrenia will not develop tumors but that the risk is relatively low.In contrast, a significant difference was observed in the overall incidence of cancer among people diagnosed with mental disorders, and tumors are often diagnosed at an advanced stage in patients with mental disorders (8).Patients with schizophrenia may be at increased risk of developing certain types of tumors following antipsychotic treatment (9), which could be attributed to lifestyle, drug therapy, and genetic factors.
Although the relationship between schizophrenia and tumors is not fully understood, this possible link is of considerable importance for the research and treatment of schizophrenia.Understanding this relationship may help us better understand the pathological mechanisms of schizophrenia and provide new implications for future treatment strategies.The relationship between schizophrenia and tumors is complex and delicate (2).Further research is needed to explore and understand this relationship to provide a scientific basis for the prevention and treatment of schizophrenia and tumors.
When observational studies explore the causal association between schizophrenia and cancer, the level of evidence for causal inference is often low due to confounding factors such as smoking, body mass index, socioeconomic status (SES), antipsychotic drugs, and genetics (10)(11)(12).Table 1 summarized the common confounding factors affecting schizophrenia and tumors.Among the diverse epidemiological research methods, randomized controlled trials are the strongest research design in clinical trials because of their randomness, control, and blindness.Randomized controlled trials are the "gold standard" to verify causality (27).However, because the research and implementation cost too much time, money and manpower; Strict inclusion criteria may lead to a decrease in the representativeness of the test population to the target population.The standard interventions used are not completely consistent with clinical practice; Limited sample size and short follow-up lead to inadequate detection of rare adverse  events.These limitations pose challenges in extrapolating the findings of Randomized Controlled Trial (RCT) to practical clinical applications.In addition, for rare and life-threatening diseases that lack effective treatments, routine RCTS may be difficult to implement, require high time costs, or may raise ethical concerns.Therefore, randomized controlled trials face various challenges in clinical practice (28)(29)(30).Mendelian randomization (MR) is a statistical method for evaluating etiological inferences in epidemiological studies.MR is based on genome-wide sequencing data and uses genetic variations with strong correlations to exposure factors as tool variables to assess causal relationships between exposure factors and clinical outcomes (31).During meiosis, alleles are randomly separated; therefore, MR can reduce deviations caused by confounding factors (32).
In this study, a large-scale genome-wide association analysis (GWAS) dataset was used to conduct a two-sample MR study to explore the causal relationship between schizophrenia and the risk of multiple cancers.These tumors included lung adenocarcinoma, lung squamous cell carcinoma, small-cell lung cancer, gastric cancer, alcohol-related hepatocellular cancer, tumors involving the lungs, breast, thyroid gland, pancreas, prostate, ovaries and cervix, endometrium, colon and colorectum, and bladder.This study will help reveal the causal relationship between schizophrenia and cancer, which may aid in prevention and treatment.
Because the data in this study were obtained from a public database, ethical approval from the participants was not required.

P value and F-statistics calculation
We used F-statistics to evaluate the statistical strength.The Fstatistics are equal to ((N-k-1)/k) × (R2/(1-R2)), where N and k represent the sample size and number of SNP, respectively (34).An F-statistic < 10 indicates that the genetic variation used is a weak tool variable that may contribute to a bias.We determined the variance in phenotype explained by each instrument by R2: where EAF is the effect allele frequency, b is the effect quantity, N is the sample size, and Se (b) is the standard error of the effect quantity.We used R2 and F-statistics to evaluate the statistical power of each tool variable.Supplementary Table S1 shows the number of MR statistical powers and valid tool variables for each pair.

Statistical analysis
The causal relationship between exposure (schizophrenia) and outcome (tumor) was determined using two-sample Mendelian randomization (35).Figure 1 illustrates the study process.The inverse variance weighting (IVW) method was used to estimate the causal relationship between exposure and outcome (36).This method ignores the existence of intercept terms and fits the reciprocal of the outcome variance as a weight; therefore, it can reliably estimate the impact of exposure on the outcome under the condition of heterogeneity between tool variables (37).The results are presented as odds ratios (OR) and 95% confidence intervals (CI).IVW is the most effective MR method; however, it assumes that all tool variables are valid.If the horizontal multiplicity is not zero, a bias exists.The weighted median method used most SNPs to determine causality (38).Therefore, we conducted a sensitivity analysis to ensure the accuracy of the causal effect results, including the leave-one-out sensitivity test, heterogeneity test, and pleiotropy test.The main purpose of the leave-one-out sensitivity test was to calculate the MR results of the remaining instrumental variable (IV) after eliminating it.If there is a substantial difference between the estimated MR and the summary results of other IVs after excluding a certain IV, the MR results are sensitive to the IV.The main purpose of the heterogeneity test was to examine the differences between the different IVs.Moreover, if there are considerable differences between different IVs, the heterogeneity of these IVs is large.The multiplicity test mainly tests whether there is horizontal multiplicity in multiple IV.The MR-Egger regression intercept is commonly used to express horizontal multiplicity if the intercept is not zero.In addition, the MR-PRESSO package is a commonly used R packet for testing horizontal multiplicity.MR analysis was performed using the R package "Two Sample MR" (version 0.5.7) in R (version 4.3.1).

Causal relationship between tumor and schizophrenia
To analyze the causal effects of different tumors on the risk of schizophrenia, we first identified SNPs that are closely related to different tumors.Among the 16 tumor types, only 6 had a strong correlation SNP, 6 had no strong correlation SNP, and 4 had <5 SNP.Therefore, we could not conduct an MR analysis of these 10 tumors.The IVW-P value of prostate cancer was 0.008 in the reverse MR analysis; however, the beta value of the MR-Egger was not consistent with the direction of IVW in further MR analysis, and the result was invalid.Therefore, we found no effect of tumors on the risk of schizophrenia, indicating that almost all tumors had no significant effect on the risk of schizophrenia (P > 0.05; Table 3).

Discussion
To date, the possibility of a causal relationship between schizophrenia and tumor incidence remains controversial.Cohort studies by some scholars have shown that there is no significant difference in the incidence rates of colorectal cancer, breast cancer, lung cancer, and all common cancers between patients with schizophrenia and patients without schizophrenia (39,40).However, some scholars have also found in cohort studies that, under the influence of different confounding factors, the incidence rates of schizophrenia and different cancers in different groups have completely different effects.For example, under the influence of the confounding factor of smoking, the incidence of lung cancer in male patients with schizophrenia is reduced (41), while the incidence of lung cancer and breast cancer in women is increased (42).This shows that confounding factors do have a significant impact between the two, and the way of impact is uncertain.The relationship between schizophrenia and cancer is controversial, possibly owing to research design and confounding factors such as smoking or diet, antipsychotics, and different cancer screening and treatment methods (11,12).Moreover, cancer is a geriatric disease, and the life expectancy of patients with schizophrenia is shortened by 10-25 years (43).The advantage of MR is to control confounding factors, so that we can get more reliable and accurate results between exposure and outcome at the genetic level.This study examined the causal relationship between genetically predicted schizophrenia and the risk of various cancers, including lung adenocarcinoma, lung squamous cell carcinoma, small-cell lung cancer, gastric cancer, alcohol-related hepatocellular cancer, tumors involving the lungs, breast, thyroid gland, pancreas, prostate, ovaries and cervix, endometrium, colon and colorectum, and bladder.We found suggestive evidence (OR=1.001,95% CI:1.000-1.001,P=0.0155) of genetic predictions of causality between schizophrenia and lung cancer; however, reverse causality could not be determined.This study demonstrated that schizophrenia is associated with an increased risk of lung cancer.A literature search revealed that the association between schizophrenia and lung cancer involves different mechanisms.Shaw et al. (44) showed that antipsychotic drugs fluspirilene, penfluridol, and pimozide induce apoptosis and inhibit metastasis through p53, STAT3, STAT5, protein phosphatase 2A, cholesterol homeostasis, integrin, autophagy, USP1, wnt/b-catenin signal transduction and DNA repair in vivo and in vitro.In addition, penfluridol and pimozide have synergistic effects with existing chemotherapeutic drugs such as cisplatin.Accumulating evidence suggests that antipsychotics are potential anticancer agents.Zuber et al. have shown that the relationship between schizophrenia and lung cancer may be related to common genetic risk factors; smoking is closely related to schizophrenia and lung cancer, and the variation in nicotinic acetylcholine receptors may lead to this overlap (43).In addition, the anti-schizophrenic drug penfluridol can inhibit the growth and metastasis of various G0/G1 stagnant lung cancer cells by increasing the expression of p21/p27 and reducing the expression of the cyclin-CDK complex.Penfluridol is an i n e x p e n s i v e d r u g ap p r o v e d b y t h e F o od a n d D r u g Administration.This study demonstrated the potential of antischizophrenic drugs in the treatment of lung cancer.This study had several advantages.First, we used two-sample MR analysis to minimize the deviation caused by confounding factors.Second, we performed an MR analysis of schizophrenia and 16 types of cancer; these 16 types of tumors are common and representative in the clinic.In addition, the database sources for each pair of exposures and results did not overlap and were of European origin, ensuring the basic requirements for the MR analysis of the two samples.Fourth, the exposed F-statistics were all greater than 10, indicating that there was no weak tool deviation.Fifth, to ensure the validity of the causal effect results, we conducted a sensitivity analysis, including the leave-one-out sensitivity, heterogeneity, and pleiotropy tests.
Despite these advantages, our study had certain limitations.First, we used only GWAS data on PGC schizophrenia; therefore, the results may be inadequate.In addition, because the two samples of the MR analysis required ethnic consistency, our study focused on the European population and did not cover the Asian and African populations, which may have led to incomplete results.In addition, population stratification may interfere with the causal relationship between schizophrenia and tumors (45).Although the population studied was European, the internal structure of the population and potential sex factors were not considered.Sex differences in the incidence of schizophrenia have been reported (46), and sex differences between schizophrenia and lung cancer have gained considerable attention.Furthermore, on reversing the MR analysis, we found that the number of strong correlations SNP was relatively small, resulting in invalid results that limited our ability to draw real causality conclusions.In addition, the OR value is 1.001, indicating that there is a potential correlation between schizophrenia and lung cancer.Schizophrenia is a risk factor for lung cancer, but the possible risk factors are not very obvious.It is necessary to further explore the relationship between schizophrenia and lung cancer in the future.This result also suggests that we need to pay more attention to the incidence of lung cancer in patients with schizophrenia in the future.Finally, our study aimed to determine whether there is a causal relationship between schizophrenia and lung cancer; however, the underlying mechanism has not yet been studied.Future investigations analyzing the possible mechanisms increasing the risk of lung cancer in patients with schizophrenia are warranted.

Conclusion
A two-sample MR analysis revealed a potential causal relationship between the genetic prediction of schizophrenia and lung cancer using schizophrenia as an exposure factor and 16 cancer types as outcomes.However, the causal relationship between lung cancer and schizophrenia has not been determined, and the causal effect and reverse causal relationship of schizophrenia on other types of cancer have not been determined.This demonstrates the importance of preventing schizophrenia in the prevention and treatment of lung cancer.Future large-scale prospective studies and more in-depth mechanism research are required to validate the study findings.

FIGURE 1 Two
FIGURE 1Two-sample Mendelian randomized study design for schizophrenia and different cancers.Solid blue lines indicate the association between tool variables (SNP) and exposure (schizophrenia) and between exposure and outcome (different types of tumors).The red solid line indicates the correlation of reverse causality.The dotted line with crossover indicates that this association conforms to two basic assumptions of Mendelian randomization: 1. Genetic variation (SNP) is independent of the confounding factor between exposure and outcome 2. Genetic variation affects the results only through exposure.

FIGURE 3 MR
FIGURE 3MR leave-one-out sensitivity analysis for Schizophrenia on Lung cancer.

TABLE 1
Common confounding factors affecting schizophrenia and tumors.

TABLE 2
Results of MR analysis of different tumors in schizophrenia.

TABLE 3
Results of MR analysis of tumor in schizophrenia.