Age at diagnosis and sex interact to modify primary malignant glioma incidence and survival

Although the focus of most etiologic studies of glioma is on their potential molecular determinants, there is a renewed interest in the role of demographic variables either as causal factors or their indicators. A similar analytic trend is seen in studies of glioma survival time. In a study of 294 886 primary malignant glioma cases, published in this issue of Neuro-Oncology, Wang et al¹ have made a substantial contribution by describing the interactions between age and sex on glioma incidence and survival. Of particular interest, are their age- and sex-specific analyses by glioma histologic subtype (anaplastic astrocytoma, anaplastic oligodendroglioma, diffuse astrocytoma, glioblastoma, and oligodendroglioma). Overall, they found that male glioma subtype age-adjusted incidence rates were higher than were those among females (Figure 1, Supplementary Table 3). Glioblastoma patients diagnosed between ages 40 and 49 years, showed the strongest male-to-female age-adjusted incidence rate ratios (AIRR) (AIRR: 1.79, 95% CI: 1.74-1.85, Figure 2, Supplementary Table 3). Male-to-female age-specific glioma hazard ratios (HRs) were also generally elevated (Figure 3, Supplementary Table 4), although those for glioblastoma patients, except for patients diagnosed between ages 20 and 29 years (HR: 1.21, 95% CI: 1.08-1.35), were close to 1.00, indicating minimal age- and sex-specific survival differences.

The age-sex interaction in glioma incidence and survival, documented by the authors, may be used to generate, or evaluate etiologic or treatment hypotheses as well as clinical guidelines. However, there are 2 methodological issues that should be addressed to allow proper interpretation of their findings. First, the benefit of the large sample, on which this study is based, is that it reduces random variability. Nonetheless, a large sample, with high statistical power, may also produce statistically significant P-values and correspondingly small confidence intervals that are substantively meaningless.² Thus, when interpreting results of analyses from studies using large samples, the magnitudes of the effect measures (eg, AIRR, HR) are more illuminating than are the sizes of the P-values or confidence intervals. This issue is particularly important in demographic studies, given, as the authors note, that the age and sex of glioma patients are indicators of many processes that are not known or, even if known, not measured. For example, in the abstract, the authors state that the overall male-to-female incidence rate ratio was lowest for children diagnosed between ages 0 and 9 years (AIRR:1.04, 95% CI:1.01-1.07, P = .003). Similarly, they write that females diagnosed between ages 0 and 9 years had worse survival than males (HR: 0.93, 95% CI: 0.87-0.99, P = .032). Although statistically significant, both effect measures were close to 1.00, suggesting the absence of male-to-female differences in both childhood glioma incidence rates and survival. The proximity of these childhood effect measures to the null may simply reflect the absence of factors causing sex differences among children.

The second methodological issue rests on the heterogeneity of the age- and sex-specific distributions among histologic subtypes.³ As the authors note, age and sex interactions vary by glioma subtype. The pattern of the summary or overall incidence rates (Figure 1, Supplementary Table 1), based on combining all subtypes, primarily reflects the larger numbers of astrocytoma and glioblastoma cases, concealing the unique age-specific incidence rate patterns of oligodendroglioma. These subtype-specific differences may result from differences in astrocytoma and oligodendroglioma etiology. The male-to-female incidence rate ratios for glioblastoma peaked at ages 40-49 years at diagnosis (Figure 2, Supplementary Table 3). Again, this pattern was concealed by the overall incidence rate ratios (Figure 2, Supplementary Table 1). Patterns of age-sex astrocytoma- and oligodendroglioma-specific male-to-female hazard ratios also differed (Figure 3, Supplementary Table 4).

While these methodological issues suggest slightly different interpretations of the findings than those of the authors, they do not alter the strength or novelty of their work. Histology-specific graphs and tables were presented in detail and will provide a foundation for subsequent research. Nonetheless, an example of a problem that may arise from thinking that small P-values are meaningful, independent of their associated effect measures, is that a clinician reading the paper may base treatment of a 60- to 69-year-old glioblastoma patient on their sex, thinking there is a survival difference by sex when there is probably not (HR: 1.07, 95% CI: 1.04-1.09, P < .001, Figure 3, Supplementary Table 4). Failure to consider histology-specific male-to-female incidence rate ratio patterns (Figure 2, Supplementary Table 3) when developing etiologic hypotheses may also limit the scope of subsequent research.

A potential motor of the age-sex interactions, not noted by the authors, is the effect of the microbiome on glioma risk. The microbiome consists of an extensive and diverse collection of bacteria, archaea, fungi, protozoa, and viruses. Most of these microorganisms colonize the gastrointestinal tract and are referred to as gut microbiota. The complex bidirectional interaction between the gut microbiota, their metabolites, and the brain, is known as the gut-brain axis. Mehrian-Shai et al⁴ hypothesize that alterations of gut microbiota that alter the brain immune system and induce inflammation may contribute to gliomagenesis. Whereas Dehhaghi et al⁵ point out that the complexity of the brain-gut interaction may either induce or prevent brain tumors. While antibiotics are administered to deplete pathogenic bacteria, they can modify other bacterial subsets thereby altering the microbiome and thus affecting the brain. D’Alessandro et al⁶ found that treating glioma-bearing mice with antibiotics altered their intestinal microflora and promoted tumor growth. In contrast, Hu et al⁷ observed that the antibiotic Clofoctol suppressed glioma stem cell proliferation.

In a recent paper, Zhang et al⁸ describe an age-sex interaction in microbiome composition. They found that females have more gut microbiota species known to benefit host metabolism than do males. Furthermore, they noted that differences in microbiome composition between the sexes were restricted to 26- to 50-year-olds. There were no substantive differences in this composition among younger or older people. In the context of Wang et al’s findings of a strong glioblastoma male-to-female incidence rate ratio among 40- to 49-year-olds, the age- and sex-specific differences in the microbiome suggest the possible role of the gut-brain axis in glioblastoma etiology and warrant investigation.

Funding

None.