Association between health insurance status and malignant glioma

Abstract

Background

Prior studies have suggested an association between patient socioeconomic status and brain tumors. In the present study we attempt to indirectly validate the findings, using health insurance status as a proxy for socioeconomic status.

Methods

There are 2 types of health insurance in Germany: statutory and private. Owing to regulations, low- and middle-income residents are typically statutory insured, whereas high-income residents have the option of choosing a private insurance. We compared the frequencies of privately insured patients suffering from malignant neoplasms of the brain with the corresponding frequencies among other neurosurgical patients at our hospital and among the German population. To correct for age, sex, and distance from the hospital, we included these variables as predictors in logistic and binomial regression.

Results

A significant association (odds ratio [OR] = 1.59, CI = 1.45-1.74, P < .001) between health insurance status and brain tumors was found. The association is independent of patients’ sex or age. Whereas privately insured patients generally tend to come from farther away, such a relationship was not observed for patients suffering from brain tumors. Comparing the out of house and in-house brain tumor patients showed no selection bias on our side.

Conclusion

Previous studies have found that people with a higher income, level of education, or socioeconomic status are more likely to suffer from malignant brain tumors. Our findings are in line with these studies. Although the reason behind the association remains unclear, the probability that our results are due to some random effect in the data is extremely low.

Malignant gliomas are a major brain tumor type of unknown cause, originating from brain tissue and showing a highly infiltrative growth pattern. The most malignant variant, glioblastoma, amounts to 47.7% of malignant glioma and its incidence ranges around 3.2 new cases per 100 000 citizens per year in Western countries.¹ The life expectancy of patients suffering from malignant glioma is still disastrous, with a median overall survival of about 14.6 months.²

Despite significant research efforts, the cell of origin for glioma development remains elusive. The genesis of malignant gliomas is thought to involve multiple sequential and/or cumulative genetic mutations resulting from intrinsic and environmental factors.³ Known risk factors are rare hereditary syndromes (such as the Hippel-Lindau and Li-Fraumeni syndromes or neurofibromatosis) or ionizing radiation.^3–6 Next to genetic predisposition, other risk factors for the development of malignant glioma are under discussion, such as sunlight exposure and vitamin D intake.^7–9 Furthermore, (low-grade) glioma might share similar risk factors with, for example, multiple sclerosis.^8,9 Like multiple sclerosis, glioma show a heterogeneous geographical distribution and higher incidences in higher-developed countries.¹⁰ A large cohort study on more than 4 million patients included in the Swedish National Cancer Register identified higher socioeconomic status with better education and high income as a significant risk factor,¹¹ with the incidence risk ratio greater than or equal to 1.14. Earlier studies had also found an association between higher socioeconomic status and various tumors of the nervous system.^12,13

Germany has one of world’s oldest national social health insurance systems originating in the 19th century. There are 2 types of health insurance in Germany: statutory and private (“gesetzliche” and “private Krankenversicherung” in German). For employees with an income below a certain threshold (60 750 € in 2019) and for jobless individuals, statutory health insurance is mandatory. This insurance already covers all necessary medical and dental treatments, and dependent family members can be included in it. The monthly instalments are a fixed fraction of the salary and are deducted directly from income. The self-employed, certain public servants, and employees earning more than the income threshold can choose between private health insurance with a risk-based insurance premium or voluntary statutory health insurance. Depending on a person’s age, health, and family situation, private health insurance may be cheaper than the statutory one. Nevertheless, it usually offers additional benefits, such as treatment by the head of the department or single rooms. Doctors and hospitals also have an incentive to treat privately insured patients because private health insurance can be charged up to 3.5 times more for the same treatment. Privately and statutory insured patients both may freely choose their doctors and hospitals. Statutory-insured residents may buy a supplementary private insurance in addition the statutory health insurance to obtain some of the benefits. Therefore, the insurance status (statutory vs private health insurance) can be considered as a surrogate marker for a person’s socioeconomic situation.

Yet, a potential correlation between socioeconomic or insurance status and frequency of malignant glioma has not been analyzed for the German population. Moreover, some potential confounders have not yet been taken into account. Social differences might not be as pronounced as in other countries and Germany’s 2-tiered health insurance system is considerably different from that of other countries. The aim of the present study is to epidemiologically analyze a potential association between health insurance status and the frequency of malignant glioma based on data from a large German neuro-oncological center. We investigated (1) whether the insurance status of patients suffering from malignant brain tumor differs from the insurance status of remaining neurosurgical patients or of the general German population; (2) whether the association between insurance status and brain tumor is consistent between the sexes; and (3) whether the differences can be explained by hidden factors, such as higher age or higher mobility of privately insured patients. With regards to the special German health insurance system, we hope to provide insight regarding risk factors of glioma development, confirm previously reported findings about an association between socioeconomic status and glioma genesis, and contribute to the discussion about the underlying mechanisms.

Methods

Ethical Statement

We performed the present analysis in accordance with the Declaration of Helsinki, and the study was approved by the local research ethics committee and institutional review board (internal study number: 2019–545).

Study Design, Inclusion, and Exclusion Criteria

This study was conceived as a retrospective cohort study to assess whether private insurance is statistically related to the frequency of malignant brain tumors. Therefore, we considered all patients treated at our department of neurosurgery and Center of Neuro-Oncology, at a quaternary university hospital, in the last 20 years (January 1, 1999, through December 31, 2018). Children younger than 18 years and patients not residing in Germany were excluded to avoid a possible selection bias for our department, we also compared the insurance status of all neurosurgical patients treated in our department with the insurance status of all residents of Germany.

Data Sources

Data concerning patients at our hospital were retrieved from the hospital electronic medical record database. Data regarding the general population and its health insurance distribution were compiled from the data available from the Association of Private Health Insurers (Verband der Privaten Krankenversicherung e.V.)¹⁴ and from the German Federal Statistical Office (DeStatis).¹⁵

Study Size

The study time period of 20 years was chosen based on our experience of having surgically treated, on average, 10 new patients with a malignant brain tumor each month, leading to some 2400 distinct inpatients. This number of cases should have allowed the detection of an odds ratio (OR) of 1.14 or more at a significance level of .05 with power greater than 0.9. Under rare disease assumption, this OR corresponds to the lowest incidence risk ratio in the Swedish study.¹¹

Outcome Measures, Potential Confounders, and Definitions

Primary variables of interest were patient diagnosis at discharge (nominal: malignant neoplasm of the brain or other) and their insurance status (nominal: private or statutory). We considered each patient as being privately insured if he or she had a full or supplementary private health insurance at the time of admission and compared the insurance status in our hospital with the insurance status of the German population. Therefore, the total number of privately and statutory insured patients were the variables of interest for the adult population of Germany in the year 2011. The numbers of fully and supplementary privately insured individuals, separately for men, women, and children, and separately for inpatient and outpatient supplementary insurances, were taken from publicly available data sources.^15,16 The number of statutory-only insured patients, with regard to inpatient and outpatient insurance, was computed as the difference between the total number of residents of the corresponding sex, as provided in the data.

We identified and defined primary brain tumors by their International Classification of Diseases, ninth (ICD-9) or tenth revision (ICD-10) code, as recorded in the patient’s electronic medical record. In ICD-9 the World Health Organization’s encodes “Malignant neoplasm of brain” with the code 191, whereas in the newer ICD-10, C71 encodes “malignant neoplasms of brain.” For patients with multiple diagnoses, we considered them to have had a malignant neoplasm of the brain if at least one of the diagnoses matched that description. Each patient was counted only once in each group, outpatient and inpatient, at the time of her or his first admission.

We considered the following secondary (control) variables as potential confounders: sex (nominal: female or male), admission year (numeric), age at admission to hospital (numeric), case type (nominal: inpatient or outpatient), the zip code of the patient’s address (nominal), and the respective distance between place of residence and our hospital (numeric). The distance to our hospital was approximated as the straight-line distance between the area with the patient’s zip code and the area with the zip code of our hospital. We used geographic coordinates of the zip code areas from the General German Cyclist Club.¹⁷

We considered the distance between the patient’s residential zip code and the zip code of the hospital as a measure for patient mobility to be advised and treated in our department.

Bias

We considered the following potential sources of bias:

1. Sex discrepancy, in insurance status and disease status;

2. Age, as a possible common factor for both insurance status and disease;

3. Location, because the catchment area of our hospital includes a regional capital, where, because of the presence of many state officials, lawyers, consultants, etc, the privately insured segment of the population might be overrepresented;

4. The distance of patients’ residence to the treatment center. Recent studies^18,19 suggested that human mobility might be related to the socioeconomic status. Consequently, patients more likely to be privately insured might be more mobile or more likely to afford care while away from home;

5. The combination of patient’s age and his or her residence’s distance, because mobility might vary differently with age for people with a higher and a lower socioeconomic status;

6. Selection bias: the possibility that privately insured patients may be treated surgically even in cases when the intervention would have otherwise been considered forlorn, unlikely to benefit the patient. Such an approach would convert more privately insured outpatients into inpatients than it would for statutory insured patients;

7. Inverse causation: the possibility that people diagnosed with malignant neoplasms of the brain, being aware of the seriousness of their situation, actively apply for private health insurance in the hope of receiving better care.

Statistical Analysis

Simple, uncorrected ORs were computed from contingency tables and tested using the Fisher exact test. For comparison of 2 ORs we used the Z test on their logarithms. To correct for potential confounders, we performed regression analysis and included them as predictors. Depending on the outcome variable, logistic, binomial, or linear regression, was used. Nominal (binary) response variables were modeled by logistic regression, and numeric response variables by linear regression. Fractions were modeled by binomial regression, with the denominator as the number of “trials” and the numerator as the number of “successes.” Variables were chosen based on the question we wanted to answer, and interaction terms were taken into account. To control for possible associations/collinearities between predictor variables, we performed multiple simpler analyses before the final, multivariable analysis. Because the predictors have shown only low levels of correlation, we included them all in the final analysis, as well as the interactions between the insurance status and the remaining predictor variables. A summary of the regression analyses is given in Table 1. For linear regression, the significances of the individual regression coefficients were derived from their t values, which were computed as the ratio of the coefficient and the corresponding standard error. For logistic and binomial regression, we used z values for determining the significance of individual coefficients. Distance to the hospital had a highly skewed, left-leaning distribution, so, to minimize the effect of outliers, we stratified it into deciles (ie, whether distance falls into the lowest 10%, the next 10%). Stratified distances were treated as numeric values. All computations were performed in R, version 3.5.2 “Eggshell Igloo.” ²⁰ A significance level of .05 was used.

Table 1.

Regression Results. Interactions Between Variables Are Denoted by × Sign

Outcome	Binomial regression
	Privately insured neurosurgical patients (proportion of all neurosurgical patients)
Predictor	Odds ratio	95% CI	P
Age	1.014	(1.012 to 1.016)	< 10–43
Diagnosis = C71/191	1.710	(1.173 to 2.493)	.005
Age × diagnosis	0.998	(0.992 to 1.004)	.55
Outcome	Binomial regression
	Privately insured neurosurgical patients (proportion of all neurosurgical patients)
Predictor	Odds ratio	95% CI	P
Distance decile	1.042	(1.031 to 1.053)	< 10–12
Diagnosis = C71/191	1.724	(1.374 to 2.164)	< 10–5
Distance decile × diagnosis	0.972	(0.940 to 1.007)	.11
Outcome	Binomial regression
	C71/191 cases at our department (proportion of population)
Predictor	Odds ratio	95% CI	P
Sex = male	1.547	(1.410 to 1.698)	< 10–19
Insurance = private	1.799	(1.561 to 2.073)	< 10–15
Sex × insurance	0.780	(0.648 to 0.939)	.009
Outcome	Linear regression
	Median distance from hospital (C71/191 patients only)
Predictor	β	95% CI	P
Age	–.263	(–0.382 to –0.144)	< 10–4
Insurance = private	1.965	(–7.8 to 11.73)	.69
Age × insurance	.003	(–0.167 to 0.175)	.97
Outcome	Logistic regression
	Diagnosis = C71/191
Predictor	Odds ratio	95% CI	P
Age	0.994	(0.992 to 0.998)	< .001
Sex = male	1.256	(1.140 to 1.384)	< 10–5
Distance decile	1.071	(1.053 to 1.090)	< 10–14
Insurance = private	2.348	(1.481 to 3.722)	< .001
Age × insurance	0.994	(0.988 to 1.001)	.08
Sex × insurance	1.102	(0.905 to 1.342)	.33
Distance decile × insurance	0.972	(0.938 to 1.006)	.11

Results

Analyzed Cohorts

Our primary cohort consisted of 30 119 neurosurgical inpatients from the years 1999 through 2018. International patients (N = 237) and patients with a missing diagnosis (N = 24, 4 of them privately insured) were excluded from further analysis. Within the cohort there were 2446 patients treated for a malignant neoplasm of the brain. In the reference year (2011) the German adult population was 67.1 million, with 11.7 million (17.4%) having private health insurance covering inpatient treatment. Detailed demographics are presented in Table 2.

Table 2.

Population and Patient Demographics

	Total		Men		Women
Population
No.	67 058 600		32 396 400	(48.3%)	34 662 200	(51.7%)
Privately insured	11 671 600	(17.4%)	6 323 400	(9.4%)	5 348 200	(8%)
Cohort
No.	30 119		15 805	(52.5%)	14 313	(47.5%)
International patients (excluded)	237	(0.8%)	129	(0.4%)	108	(0.4%)
No diagnosis (excluded)	24	(0.08%)	13	(0.04%)	11	(0.04%)
Remaining No.	29 858	(99.1%)	15 663	(52%)	14 194	(47.1%)
Age, mean (SD), y	58.5	(16.3)	58.3	(16.1)	58.7	(16.5)
Age, minimum to maximum, y	18-99		18-98		18-99
Privately insured	5651	(18.9%)	2920	(9.8%)	2730	(9.1%)
Diagnosis: C71/191
No.	2446		1422	(58.1%)	1024	(41.9%)
Age, mean (SD), y	56.9	(15.7)	56.6	(15.4)	57.3	(16.2)
Age, minimum to maximum	18-90		18-87		18-90
Privately insured	614	(25.1%)	361	(14.8%)	253	(10.3%)

Insurance Status and Glioma Frequency

Of the 2446 distinct inpatients treated for a malignant neoplasm of the brain at our department, 614 (25.1%) had private health insurance. In the same period, we treated 27 412 patients for other diagnoses, 5037 (18.4%) of whom were privately insured. This fraction is close to that of privately insured individuals in the German general population (17.4%). The difference between the insurance status of our patients with a malignant brain tumor and the remaining neurosurgical patients was significant (Fisher test on contingency table; OR = 1.49, CI = 1.35-1.64, P < .001). The difference between the former and the German population was also significant (OR = 1.59, CI = 1.45-1.74, P < .001).

Although in Germany men are more likely than women to have private health insurance (19.5% vs 15.4%), and although malignant brain tumors affect men more often than women (see later in this article), men and women suffering from the disease both were more likely to be privately insured than the corresponding group in the general population. Thus, of 1422 male patients suffering from a malignant brain tumor, 361 (25.4%) had private insurance, compared to 6.3 million out of 32.4 million adult males in Germany (OR = 1.4, CI = 1.24-1.58, P < .001). In female patients suffering from malignant brain tumors, the effect was even more pronounced: 253 out of 1024 (24.7%) were privately insured, compared to 5.3 million out of 34.7 million in the population (OR = 1.8, CI = 1.55-2.08), P < .001). The difference in the ORs was significant (Z test on log-OR, P < .01).

Glioma risk generally increases with age. Also, with advancing age, people are more likely to have private health insurance. This could because income tends to increase with age, which would allow more people to afford private insurance later in life, or because people with lower socioeconomic status tend to die earlier. To exclude possible bias, we modeled the fraction of privately insured patients as a function of their age and neurosurgical diagnosis (“malignant brain tumor” vs “other”). Among our neurosurgical patients, age and brain tumor diagnosis both are significantly associated with insurance status (binomial regression, OR =1.01, CI = 1.01-1.0), P < .001 and OR = 1.71, CI = 1.17-2.4), P < .01, respectively). There is no interaction between the 2 terms (OR = 1, CI = 0.99-1, P = .55) Fig. 1.

Fig. 1.

Fraction of privately insured patients, as a function of the patient’s age and the neurosurgical diagnosis. With increasing age, the fraction of privately insured patients also increases, for all neurosurgical patients (binomial regression, P < .001). However, patients suffering from a malignant neoplasm of the brain (solid line) are more likely to be privately insured than patients suffering from other neurosurgical diagnoses (dashed), at all ages (P = .005).

Insurance Status and Mobility

Distance between the patient’s residential zip code and the zip code of the hospital was higher in privately insured neurosurgical patients than in statutory insured patients (binomial regression, fraction privately insured vs diagnosis and distance decile, OR = 1.04, CI = 1.03-1.0, P < .001). Therefore, privately patients arrived from farther away and we more mobile to be advised and treated in our department. There is, however, no interaction between distance and diagnosis when the diagnosis is dichotomized into “malignant brain tumor” or “other” (OR = 0.97, CI = 0.94-1.01, P = .11) Fig. 2.

Fig. 2.

Fraction of privately insured patients, as a function of the patient’s residence’s distance from the hospital and the neurosurgical diagnosis. The fraction of privately insured neurosurgical patients increases with the patient’s residence’s distance from the hospital, suggesting that privately insured patients are more mobile, irrespective of diagnosis (binomial regression, P < .001). Still, patients suffering from a malignant brain tumor (solid line) are more likely (P < .001) to be privately insured than patients suffering from other neurosurgical diagnoses (dashed). To correct for the highly skewed distribution of distances (most patients come from the vicinity and only a minority from far away), the distances have been grouped into deciles and patients counted for each decile.

Although our patients’ mobility (ie, the distance between their place of residence and our hospital) decreased with their age by 0.26 km/year (linear regression, median distance vs age, CI = 0.14-0.38, P < .001), we could not find any difference in decline of mobility for privately vs statutory insured patients (CI = –7.8 to 12), P = .96) Fig. 3.

Fig. 3.

Linear relationship between the patient’s age and the median distance from the hospital at that age, for privately and statutory insured patients. Patient’s mobility, measured as the median distance from the hospital for each age, decreases with increasing age by 0.26 km (linear regression, P < .001), irrespective of insurance status (P = .97). There is no difference in mobility between privately and statutory insured patients (P = .69).

Possible Selection Bias

Owing to the earlier mentioned incentive to hospitals to treat privately insured patients, we had to consider the possibility that privately insured patients are accepted for treatment even in cases when a statutory patient would have been turned down, such as when the treatment was unlikely to benefit the patient. That would constitute a selection bias. To exclude this possibility, we checked whether private insurance for out-of-house services appear more often among the patients who were accepted for treatment and those who never became our inpatients. From the complete list of our brain tumor patients, including outpatients and inpatients, we compiled a list of patients who were registered as outpatients but never became inpatients. This list of only-out patients was manually reviewed independently by 2 authors (I.F. and M.A.K.) to identify patients who came only for a second opinion or whose diagnosis of malignant brain tumor could not be confirmed. These patients were excluded from the list. In case of a disagreement between the list reviewers, a consensus solution was sought. After the review, 105 outpatients who never became inpatients remained on the list. We could not find evidence that the frequencies of private insurance differed between these patients and our inpatients (OR = 2, CI = 0.75-7.7, P = .24).

Multivariable Logistic Regression

Finally, we performed a multivariable logistic regression to adjust for the effects of age and other covariates. We modeled the probability that an inpatient is treated for a malignant brain tumor, as opposed to other neurosurgical diagnosis, as a function of his or her insurance status, age, sex, and distance from the hospital. All of the listed variables showed to be significant predictors, but the insurance status did not interact with any other variable. The results of this and other regressions are summarized in Table 1. Although the risk of brain tumor occurrence increases with advancing age, at least until age 70 (Fig. 4), the risks of other neurosurgical diseases increase faster. This is reflected in the OR of less than 1 for age.

Fig. 4.

Relative number of brain tumor patients, stratified by age and health insurance status. The number of patients treated for malignant glioma at our hospital, relative to the German population of corresponding age, rises until approximately age 70. This is equally true both for privately (solid line) and statutory-insured patients. The smooth curves through the points were calculated using fourth-degree polynomial regression. The curves are similar in form but much lower than reported incidence²¹ because only a fraction of the German population falls within our catchment area.

Discussion

The main results of this study are (1) there is a significant difference between the insurance status of patients suffering from malignant brain tumor on the one side and the remaining neurosurgical patients or compared to general German population on the other; (2) compared to the German population, a higher proportion of privately insured patients among glioblastoma patients of both sexes was observed; and (3) the differences could not be explained by a higher mobility of privately insured patients or by age as the hidden factor governing the fraction of privately insured individuals.

We cannot explain the statistical relationship between private insurance and a higher frequency rate of glioma. However, some potential biases, such as age, sex, a higher mobility of citizens with higher socioeconomic status, and a selection bias, could be excluded in our present analysis.

The potential bias of inverse causation was considered highly unlikely based on insurance policy grounds. First, acquiring full private health insurance is an option for only a small fraction of the population who meets the income and employment status criteria. For the remaining population, only supplementary private insurance is an option. However, health insurance companies are hardly willing to insure patients known to suffer from severe diseases, or, at best, will do so at a very high premium, which would likely offset any benefits from the insurance. Also, new contracts typically exclude coverage of any conditions that might appear shortly after acquiring the insurance (the abstention period). To get private insurance at reasonable rates, patients would have to apply for it well before seeking medical help. But symptoms that are serious enough to prompt patients to seek private health insurance are also serious enough to force them to seek immediate medical attention. We consider it highly unlikely that patients suffering from such symptoms would decide to wait for the abstention period of the new insurance to pass. For that reason, we are confident that no, or, at worst, very few patients switched from statutory to private insurance after falling ill with a brain tumor. Transitions in the opposite direction—privately insured patients switching to statutory insurance—are theoretically possible, but, if they happened before the diagnosis, they would actually strengthen our findings. It would mean that there were actually even more privately insured people falling ill with a brain tumor than recorded. Transitions after diagnosis would not affect the findings.

A correlation of higher malignant glioma incidence has been described in several recent studies. Our present results are in line with previous studies: In the previously mentioned large Swedish cohort study, women and men with 3 or more years of university education, intermediate and high nonmanual employees, and high income in men had significant higher incidence rate ratios for glioma, whereas never-married men or men never having been in cohabiting relationships had a significantly lower risk for developing a malignant glioma.¹¹ Another Swedish study previously showed an association between disposable income and an increased risk for glioma independent of marital status.²² Increasing socioeconomic status was significantly correlated with an increasing incidence of glioblastoma in an analysis of the data of the Surveillance, Epidemiology, and End Results Program.²³ However, a relationship between self-reported household income and malignant glioma incidence was not observed in a smaller study.¹³ In that study, patients without a relationship at the time of tumor diagnosis had a lower risk for developing a malignant glioma.

A significant variation in incidence of glioma by ancestry, race, ethnicity, or country of origin has been reported.²⁴ For historical reasons, these data are normally not collected in Germany, so we could not include them in our analysis. However, Germany is ethnically more homogeneous than countries with a colonial past and until 2000, it had a quite restrictive immigration policy. In 2011, foreigners constituted about 8.6% of the German population, most of them (79%) from another European country. Therefore, we hypothesize that ethnicity difference cannot account for the observed results in our study.

The German microcensus asks about the populations’ health insurance status only every 4 years. Therefore, the insurance status data for the year 2011, as the closest to the midpoint of the study (2009), was taken as representative for the whole population under study. Although this might not be entirely correct, available data suggest this is likely erring on the safe side: For the period 1999 to 2018, the German population was the lowest in 2011, whereas the number of privately insured individuals rose almost continuously and was already above the average in 2011. The true fraction of privately insured people in the population is thus likely lower than assumed, and the ORs higher than reported in this study.

This analysis includes only a single-center cohort. It would be of interest to examine whether socioeconomic status could be identified as a risk factor for malignant glioma in a nationwide analysis. Moreover, it would be interesting to compare malignant glioma incidence in Western and Eastern Germany for several reasons: (1) Owing to the country’s separation until 1989, citizens in former Western and former Eastern Germany might have been exposed to different risk factors, (2) the economic power of former Western and former Eastern Germany has not been aligned within the last 30 years, and (3) it is unclear whether the lower socioeconomic status of most residents in former Eastern Germany and a lower frequency of privately insured patients translates into a lower incidence of malignant glioma. Further, multicentric studies are required to analyze the influence of these factors.

We assume a yet unknown confounder, which is a risk factor for (a later) glioma development, and that might be influenced by the socioeconomic status. However, we can only speculate about this confounder. Genetic aberrations and infectious diseases, such as human papilloma virus infections, are established causes in some cancer types; accordingly, cytomegalovirus infection (CMV) has been assumed to play an oncomodulatory role and subsequently to correlate with glioma genesis.^25–28 Socioeconomic situation is well known to correlate with CMV infection rate and time point. CMV infections in the early childhood were hypothesized to be protective against malignant glioma development, while CMV infections in the later childhood or adulthood seemed to be a risk factor for glioblastoma. However, striking evidence is yet lacking. Next to infectious causes, lifestyle factors correlate with residents’ socioeconomic position and might be risk factors for glioma genesis. For example, higher-income parts of the population are more likely to travel by plane and to distant destinations, which exposes them to increased radiation, lower oxygen levels, and different environmental hazards. However, flight-induced radiation and lower oxygen levels should also affect flight personnel, but no increased incidence of brain tumors has been reported for that group. Early use of mobile phones, when they were expensive and transmitted at higher power levels, also cannot be completely excluded, but this is still an unlikely cause. Radio waves are nonionizing radiation, and a mechanism that could cause malignant brain tumors, but not other neurosurgical diseases, has not been convincingly put forward. Moreover, such a mechanism would also have to affect people living or working in the vicinity of radio transmitters, including the earlier mentioned flight personnel, but there is currently no scientific evidence supporting this. It is generally assumed that people with a higher socioeconomic status have a healthier lifestyle, which is reflected, among other things, in their higher life expectancy. In fact, lower socioeconomic status is linked to a lower frequency of physical activities and consumption of healthy food and higher rates of obesity and cigarette consumption.^29–32 We cannot exclude, however, that this otherwise healthier lifestyle includes risk factors specific for the development of malignant glioma. Again, whether this hypothetical factor is related to eating habits, physical activity, or attitudes toward literature and fine arts is pure speculation.

Limitations

We are aware of the following limitations of our study: (1) This study evaluated only a single-center cohort and thus might not be representative for the whole German population, much less for the whole world. (2) Glioma patients might not be representative of all neurosurgical patients. It is conceivable that our department has a significantly better reputation regarding the treatment of malignant brain tumors than for other neurosurgical diagnoses and that this reputation is better known to higher-educated, better informed and, probably, privately insured patients. Also, benefits of being treated at a top-level center are more pronounced for neuro-oncological patients than for other neurosurgical patients, and patients with higher socioeconomic status might be more aware of this. In combination, this could lead to a selection bias from the patients’ side. (3) Insurance status is not a perfect marker for socioeconomic status: Self-employed people may choose private insurance regardless of their income, and high-income employees may voluntarily opt for a statutory insurance. Therefore, the conjecture that our study supports previous findings might not be correct. Whether people who have high socioeconomic status while being insured on statutory insurance have the same risk as privately insured or as statutory-insured people is an interesting question that we could not answer. (4) Although we think we did not confuse the cause and the effect, that is, that patients suffering from malignant brain tumors obtain private insurance because of the disease, we could not statistically disprove it. Our reasoning is based on our clinical knowledge and the knowledge of the insurance system, as described in the “Methods” section earlier. (5) We did not differentiate between glioblastoma and other malignant gliomas. It is possible that the 2 exhibit a different association with socioeconomic status. (6) Lacking data on patients’ race and ethnicity, we could not investigate whether they are associated with socioeconomic status or brain tumors. Available data also did not allow us to control for other confirmed or hypothesized risk factors, such as exposure to sunlight or ionizing radiation, hereditary factors, or vitamin D. Moreover, we are not able to analyze that a proven healthier lifestyle with a higher rate of physical activities and more frequent consumption of healthier food might a confounding factor related to the higher rate of malignant glioma. (7) The present study analyzed a population consisting of patients who were treated in our neuro-oncological center and therefore already have a certain pathology. Hence, the frequency of malignant gliomas does not necessarily represent the incidence of glioma in the catchment area of our hospital.

Conclusion

Taking health insurance status in Germany as a surrogate marker for patients’ socioeconomic position, our findings are in line with previous findings. The socioeconomic position in Germany with its comparatively less pronounced social differences is positively associated with malignant neoplasms of the brain. We could not observe a similar association for other neurosurgical diagnoses at our hospital. The mechanism behind this effect is unclear and requires further investigation. From a practical perspective, health insurance status, being a binary variable, is a very simple and straightforward criterion for identifying individuals with an increased risk of developing brain tumors, both in clinical practice and at an epidemiological level. However, we aim to recall that a higher socioeconomic position is a risk factor for glioma genesis and to invigorate the discussion about the yet unclear underlying mechanisms. Whether and which policy decisions need to be taken will have to be decided based on cost-benefits deliberations.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Conflict of interest statement.

None declared.