IgE, allergy, and risk of glioma: Update from the San Francisco Bay Area Adult Glioma Study in the Temozolomide era

Authors


Abstract

The consistently observed inverse relationship of allergic conditions with glioma risk and our previous demonstration that immunoglobulin E (IgE) levels also were lower in glioma patients than controls suggest that atopic allergy may be related to a mechanism that inhibits or prevents glioma. We sought to extend these results with a new and larger series of patients (n = 535 with questionnaire data; 393 with IgE measures) and controls (n = 532 with questionnaire data; 470 with IgE measures). As expected, glioma cases were less likely than controls to report history of allergies [among self-reported cases, Odds ratios (OR) = 0.59, 95% confidence interval (CI): 0.41–0.85]. IgE levels also were lower in glioma cases versus controls (OR per unit log IgE = 0.89, 95% CI (0.82–0.98). However, this inverse relationship was only apparent among cases receiving temozolomide, a treatment which became part of the “standard of care” for glioblastoma patients during the study period. Among patients receiving temozolomide, IgE levels in cases whose blood samples were obtained within 30 days of diagnosis were slightly higher than controls, whereas IgE levels in cases whose blood sample was obtained >60 days after diagnosis were significantly lower than controls (OR = 0.80; 95% CI: 0.71–0.89). Thus, although our results robustly confirm the inverse association between allergy and glioma, the results for IgE are affected by temozolomide treatments which may have influenced IgE levels. These results have implications for the study of immunologic factors in glioma as well as for immunotherapy protocols for treating glioma. © 2009 UICC

Approximately, 14,000 patients are diagnosed with a glioma each year in the United States. The etiology of adult glioma is largely unknown and is thought to be multifactorial; various genetic, infectious and immunological factors have been implicated.1 Recent epidemiological studies have reported that adults with glioma are 1.5- to 4-fold less likely than controls to report a variety of allergies,2–6 which ranks the lack of allergies among the most consistent risk factors for glioma reported to date. In addition, we previously reported an inverse relationship between immunoglobulin E (IgE), a biomarker for atopic allergy, and glioma risk.7 We found poor concordance between self-reported allergy and IgE levels, and the strongest IgE–glioma association was observed among the least prevalent allergen—food IgE.7 In addition, glioma patients who had elevated levels of IgE had approximately 8 months longer survival than individuals with lower or undetectable levels,8 demonstrating potential clinical significance.

The goal of this current study was to utilize new population-based cases and controls to confirm our previous reports of an inverse relationship among self-reported allergy, IgE levels and glioma risk. However, during the new subject recruitment period, the standard of care for glioblastoma changed to include treatment with temozolomide (Temodar),9 a chemotherapy agent. Therefore, in addition to replicating our previous studies, the current analysis provides the first opportunity to evaluate the relationship among self-reported allergies, IgE levels and glioma risk in patients treated with temozolomide.

Material and methods

Subjects, interviews and specimen collection

Histologically confirmed gliomas (International Classification of Diseases for Oncology, morphology codes 9380–9481) diagnosed from November 2001 to September 2004 were identified using the Northern California Rapid Case Ascertainment program and included in this study. Eligible cases were aged 20 or older, had pathologically confirmed glioma and resided in the six county region of San Francisco (SF) Bay Area, (Alameda, Contra Costa, Marin, San Mateo, San Francisco and Santa Clara). Controls aged 20 years or older from the same residential area as cases were identified using random digit dialing and were frequency matched to cases based on age, gender and ethnicity. The University of California San Francisco Committee on Human Research approved the methods for this study (IRB approval H6539-04956-21A). We call this ascertainment series, “Series 3” to distinguish it from the previous recruitments, Series 1 (1992–1994) and Series 2 (1997–2000).

In-person interviews with cases (or their proxies) and controls lasted approximately 2 hr and used a structured questionnaire and show cards. Subjects were offered a brief telephone interview if they declined the full in-person interview. The allergy history assessment questions were very similar to the ones we used for our Series 2 glioma study.5

Detailed information regarding history of allergies was collected in tabular form on eight questionnaire pages. Data were collected by asking “Have you ever had reactions to” the following allergens: house dust, mold or mildew, pollens, poison oak/ivy, stinging or biting insects, eggs, dairy products, shellfish, wheat, peanuts or peanut butter, other nuts, soy, alcohol, coffee, other foods, toiletry items (soaps/detergents), cosmetics, deodorant (perfumes/colognes/aftershave), cats, dogs, other animals, prescription and nonprescription drugs, tobacco smoke and wool. Additional spaces were included in the questionnaire for “other” items that the patient identified as allergens but were not specifically asked for by name by the interviewer. Interviewers prompted subjects with show cards for each of the general allergen categories. For a “yes” response, the interviewer then asked whether the allergens produced any of the following symptoms also listed on show cards (runny nose, burning/watery eyes, sneezing/congestion, wheezing/asthma, rash/hives, itching, swelling/inflammation, nausea/vomiting, diarrhea, headaches, anaphylactic shock, other: specify).

The questionnaire also asked extensive information about family and personal medical history including asthma and eczema, demography, drugs used, and other personal information including smoking and diet.

Blood and sera were collected either at the time of interview or at a later time. Participants were asked on a separate blood draw questionnaire at that time about currently used medications and chemo- and radiation therapy. For analytical purposes, medications were classified into 17 categories (Supporting Information Table 1).

Table I. Description of Participants: Age, Gender, Ethnicity, Education, Income, Smoking History and Histology, San Francisco Bay Area Adult Glioma Study (2001–2005)
 Glioma casesControls All (n = 532)Chi-square/t test p value1
All (n = 535)Self-reporting (n = 406)Proxy-reporting (n = 129)
  • 1

    Testing for difference between cases and controls. Chi-square used for categorical, and t-test for continuous variables.

Mean Age ± SE54.6 ± 0.750.9 ± 0.766.2 ± 1.153.8 ± 0.70.421
% White767675750.683
% Male565852520.151
% College Graduate515438590.007
Mean Education (yr) ± SE14.9 ± 0.215.3 ± 0.213.7 ± 0.315.7 ± 0.1<0.001
Household Income (USD/yr) (%)
 ≤$29,999201730180.537
 $30,000–49,99916132415 
 $50,000–69,99913141017 
 $70,000–99,99916181218 
 $100,000+31351932 
Smoking History (%)
 Never Smoked475037480.699
 Past Smoker39384440 
 Current Smoker13121912 
Histology (%)
 Glioblastoma625585NA 
 Anaplastic Astrocytoma12148NA 
 Astrocytoma791NA 
 Anaplastic Oligodendroglioma442NA 
 Oligodendroglioma562NA 
 Oligoastrocytoma120NA 
 Ependymoma110NA 
 Juvenile Pilocytic Astrocytoma220NA 
 Medulloblastoma222NA 
 Other451NA 
 Astrocytoma NOS111NA 

Treatment information (such as, temozolomide, other chemotherapy, biopsy vs. resection, radiation therapy) for the brain tumor was obtained through medical record abstraction and SEER registry data. Additional information on drug treatments was obtained with a short questionnaire asking for a list of current medications that were being given at the time of blood draw.

We attempted to obtain pathological specimens for all glioma cases which were then reviewed and classified by an academic neuropathologist. To date, 484/535 cases have been reviewed, including 365 of 396 with IgE measurements. Kenneth Aldape (MD Anderson, n = 318) or Tarik Tihan (UCSF, n = 43), or both (n = 4) reviewed these cases.

IgE measurements

IgE levels were assessed using a standardized clinical instrument designed for this purpose: Pharmacia Diagnostics UniCAP fluorescent “sandwich” assay.10 Total food and respiratory allergens were measured. We compare some of the IgE measurements from an earlier series to the current series in the “Results” section of this paper to help to clarify the basis of differences observed; for details about Series 2 subjects and methods see the earlier papers.5, 7 IgE levels were determined at only a single time point for all SF Bay Area patients.

Longitudinal (repeated) IgE measurements

We examined the stability of IgE measurements in a single patient at six time points over 1 month. This patient was not a part of the SF Bay Area Glioma Study. This patient who has been previously reported by Heimberger et al.,11 had completed a 6-week daily temozolomide course with radiation therapy, a 6-week EGFRvIII vaccine course without temozolomide, and three prior 1-month cycles with 5 days on temozolomide followed by 23 days off. On the 23rd day (when blood cell counts recovered), the EGFRvIII vaccine was administered. Blood for the current study was drawn during the 16th treatment cycle of sequential temozolomide and EGFRvIII vaccine. The patient did not have a history of atopic allergy prior to surgery.

Statistical methods

Odds ratios (OR) for cases versus controls reporting a history of allergy were estimated with logistic regression, controlling for age, gender and ethnicity (white/nonwhite), education (college education) and smoking history. OR were estimated for all cases versus controls, self-reporting cases versus controls, and proxy-reported cases versus controls. OR also were computed for having any versus no history of allergy, by numbers of allergies reported (none, 1–3 and 4 or more), and by route/source of exposure, respiratory or food. Descriptive statistics and OR were computed with SAS.12

IgE quantities were compared in cases versus controls based on the following categories: (i) for total IgE, IgE levels > 100 kU/L are clinically “elevated”, 25–100 kU/L “borderline” and < 25 kU/L “normal”; (ii) for food and respiratory IgE, < 0.35 kU/L are termed “nonelevated” and ≥ 0.35 kU/L, “elevated.” Continuous measures were determined by measuring fluorescence against the standard curve with known quantity inputs; these data also were log transformed to improve normality. Measures of IgE that fell under the lower limit were adjusted to 0.35 kU/L. Quartiles of IgE levels based on those of controls also were created.

Control for potential bias and confounders

A common difficulty for retrospective interview studies is the potential for reporting bias; i.e., patients with disease might be more or less likely to recall, to fabricate or to be prompted by interviewers to supply information that might have contributed to the etiology of their disease. This seems unlikely to happen for history of allergies which would not be commonly thought to affect glioma, or if they were, might be thought to be positively associated rather than negatively. Another bias could come from an estimate of lifetime incidence of allergy that may be biased by the age of the individual; i.e., older individuals might provide less accurate information. Also, proxies might be less likely to know of and report temporally distant or minor allergies. To help control for these potential biases, OR were adjusted for age, gender, and ethnicity, and results emphasize associations for self-reported cases. OR were also adjusted by education level attained (college degree/no college degree). Finally, to control for potential confounders, OR were computed stratifying by tumor histopathology (glioblastoma or other glioma histologies), self/proxy status and temozolomide treatment status, where indicated.

Results

Of the 745 eligible cases, full interviews were obtained for 535 cases (72%) and abbreviated interviews for 27 (4%). Of the remaining eligible cases, there were language problems (n = 3), the subject's physician refused contact with the subject (n = 4), the subject or a proxy could not be located (n = 15), the subject or proxy refused (n = 97), or they were not reported through Rapid Case Ascertainment (n = 64). Of the 10,952 phone numbers dialed to obtain controls, 6% (n = 600) identified an eligible control, 2,477 refused information before or after study introduction (23%) and the remaining numbers (71%) either had no response after 10 calls (25.7%) or did not identify an eligible subject due to language barrier (5.8%), business/fax/modem line (9.1%), line out of service (12.2%), etc. Of eligible controls, 92% (n = 565) agreed to participate either in the full (n = 532) or an abbreviated (n = 33) interview. For the current analysis only subjects completing the full interview were included leaving 535 cases along with 532 controls (Table I). For 129 of the 535 enrolled cases (24.2%), questionnaire data was reported by a proxy. Additional information on the distribution of ethnic groups, sex, age, education, income, smoking status, proxy status and histopathological diagnosis for cases and controls are presented in Table I.

Cases and controls did not differ significantly by age, percent white, sex, total years of education, or smoking history. Controls were more likely to have a college degree, (p = 0.007, χ2). Within the cases, the proxy-reporting cases were significantly older than self-reporting cases (66 vs. 51 years, respectively, p < 0.001), had a lower percentage of college graduates (38% vs. 54%, respectively, p = 0.001), lower reported income (p < 0.001), significantly different smoking status (p = 0.019), and a higher percentage of glioblastoma multiforme diagnoses (p < 0.001). Proxy- and self-reporting cases did not differ by ethnicity or sex.

Allergy and glioma

Glioma patients reported significantly fewer allergies than controls when pooled [OR = 0.50, 95% confidence intervals (CI): 0.36–0.70, Table II] or separated by proxy or self-reporting patients (OR = 0.28, 95% CI: 0.17–0.46 and OR =0.59, 95% CI: 0.41–0.85, respectively, Table II). Adjustments were made for age, gender and ethnicity (white/nonwhite), college education and smoking history, which did not substantively change any allergy or IgE OR. ORs were 0.58 for subjects reporting 1–3 allergies and 0.39 for subjects reporting 4 or more allergies, suggestive of a dose–response. There was a significant deficit of history of respiratory but not food allergies in cases compared with controls (Table II).

Table II. Case–Control Odds Ratios for History of Allergy, San Francisco Bay Area Adult Glioma Study (2001–2005)
 ControlsAll casesSelf-report casesProxy-report cases
n%n%OR1 (95% CI)p valuen%OR1 (95% CI)p valuen%OR1 (95% CI)p value
  • 1

    Models were adjusted for age, gender, ethnicity, college education and smoking history.

Number of Allergies
 No Allergies Reported6713123231.00 80201.00 43331.00 
 ≥1 Allergy46587408770.50 (0.36–0.70)<0.001322800.59 (0.41–0.85)<0.00186670.28 (0.17–0.46)<0.001
Number of Allergies
 No Allergies Reported6713123231.00 8019.91.00 4333.31.00 
 1–3 Allergies26650273510.58 (0.41–0.82)0.00220651.20.66 (0.45–0.96)0.0296751.90.36 (0.22–0.61)<0.001
 4+ Allergies19937135250.39 (0.27–0.57)<0.00111628.90.50 (0.33–0.75)0.0011914.70.15 (0.08–0.29)<0.001
Respiratory Allergies
 No Respiratory Allergies19938265511.00 18044.81.00 8565.91.00 
 ≥1 Respiratory Allergy32362258490.63 (0.49–0.81)<0.00121553.50.73 (0.56–0.96)0.0254333.30.35 (0.23–0.54)<0.001
Food Allergies
 No Food Allergies40777424801.00 31778.91.00 10782.91.00 
 ≥1 Food Allergy12123106200.88 (0.65–1.18)0.408420.90.90 (0.66–1.24)0.5382217.10.77 (0.45–1.31)0.333

Serum IgE and glioma

Total IgE levels were available for 393 cases and 470 controls. The mean log transformed serum IgE levels were lower for cases (3.26 ± 0.08) than for controls (3.49 ± 0.07). The case–control OR was 0.89 (95% CI: 0.82–0.98) for each unit of increase in log IgE. When clinical categorical IgE levels were considered, IgE levels were lower among cases than controls, but failed to reveal a dose–response. Categorical food and respiratory IgE yielded OR lower than 1 (OR = 0.58 and 0.80, respectively), but in either case the 95% CI included 1 and were not significant (Table III).

Table III. Case–Control Odds Ratios for IgE Levels, San Francisco Bay Area Adult Glioma Study (2001–2005)
 ControlsAll cases
nmean ± SENmean ± SEOR1 (95% CI)p value
  • 1

    Models were adjusted for age, gender, ethnicity, college education and smoking history.

  • 2

    Ln(IgE) was expressed as a continuous variable.

  • *

    p value for overall trend test.

Ln (Total IgE)24703.49 ± 0.693933.26 ± 0.0790.89 (0.82–0.98)0.01
 n%n%OR1 (95% CI)p value
Total IgE
 Normal, <2519942197501.00 
 Boderline, 25–10016535111280.66 (0.48–0.90)0.05
 Elevated, >1001062385220.79 (0.55–1.12)0.18
Respiratory IgE
 Nonelevated27358244621.00 
 Elevated19742148380.80 (0.60–1.06)0.12
Food IgE
 Nonelevated43192372951.00 
 Elevated3981950.58 (0.33–1.03)0.06
IgE Quartiles
 Quartile 111725123311.000.07*
 Quartile 21182590230.69 (0.47–1.01)0.05
 Quartile 31182589230.68 (0.47–1.00)0.05
 Quartile 41172591230.70 (0.48–1.03)0.07

We noted that in the current series, time from blood draw to laboratory processing was on average 1 day faster than the prior series (median 1 day between draw and freezing, as opposed to 2 days in prior series). This time is not different between cases and controls, but we tested whether it would have any difference to measured IgE levels. Sera from three healthy volunteers were drawn into multiple red-top tubes and stored at room temperature for 0, 1, 2, 3 and 4 days, then processed and frozen. IgEs were measured from the frozen sera; in all cases measurements did not decrease over time and IgE levels scored within 5% of the original “0 time” measurement, indicating that there is no detectable degradation of IgE in whole coagulated blood over the variable time frames that samples are processed. One of the three volunteers displayed “elevated” IgE, one high “borderline” and the third, “normal.”

Because we were able to ascertain and interview cases in the current series an average of 43 days faster than the previous series (the average time from diagnosis to blood draw was 69 days in Series 3 vs. 112 days in Series 2),7 we also considered whether bloods drawn closer to the time of diagnosis had higher IgE levels, more similar to control levels. For patients taking temozolomide, there was an inverse relationship between IgE levels and time between diagnosis and blood draw; for cases not treated with temozolomide, the IgE levels were highest within 30 days of diagnosis but there was no additional decrease in levels > 60 days from diagnosis (Fig. 1).

Figure 1.

Relationship of IgE level in blood among cases and the length of time between diagnosis and blood draw. Series 3 patients, (a) no temozolomide treatment and (b) on temozolomide therapy. Mean and standard error are shown. Using a general linear model, the trend of IgE was not significantly different among the three categories in no temozolomide treatment, (a, p = 0.15), but significant among the temozolomide treated (b, p = 0.001).

Glioma medications

Most medications were prescribed to too few patients to substantially bias the case–control OR (see Supporting Information Table 1). However, large numbers of cases took temozolomide and dexamethasone. Those who were prescribed temozolomide (63% of our patient population) demonstrated significantly lower IgE levels; least square mean (LS mean) of IgE for temozolomide patients was 3.14 (SE = 0.13) compared with those not taking temozolomide (LS mean = 3.61, p = 0.006). Only the temozolomide-treated cases had lower IgE levels compared with controls, and this relationship held true for both glioblastoma and lower grade gliomas (Table IV). Patients taking temozolomide were more likely to be self reporting and younger, but were not significantly different with regards to gender, ethnicity, or income (Table V). Our study ascertainment period intersected closely with the introduction of temozolomide, as only 14% of glioblastoma cases were on the drug early in this case series, and 76% near the end (Table V). IgE levels of people taking this drug were substantially lower with increasing time since diagnosis (Fig. 1). Cases taking dexamethosome at the time of blood draw had a ln IgE mean of 3.32, and when temozolomide-treated patients were removed, 3.47. For temozolomide-treated patients not on dexamethosone (only 33 patients), the mean ln IgE was 3.08. We did not detect significant effects of other medications, or other therapies on IgE levels (including surgery and radiation) when statistically adjusted for temozolomide treatment (data not shown).

Table IV. Case–Control Odds Ratios for IgE Levels by Temodar Use and GBM Status, San Francisco Bay Area Adult Glioma Study (2001–2005)
 ControlsAll casesGBMNon-GBM
n%n%OR1 (95% CI)p valuen%OR1 (95% CI)p valuen%OR1 (95% CI)p value
  • 1

    Models were adjusted for age, gender, ethnicity, college education, and smoking history.

  • 2

    Includes cases with unknown Temodar status (35 total).

All Cases2
 Total IgE
  Normal, <2519942197501.00 116521.00 81481.00 
  Borderline, 25 to 10016535111280.66 (0.48–0.90)0.0156250.56 (0.38–0.83)<0.0155330.82 (0.54–1.25)0.36
  Elevated, >1001062385220.79 (0.55–1.12)0.1852230.81 (0.54–1.23)0.3333200.70 (0.43–1.16)0.17
No Temodar Use
 Total IgE
  Normal, <251994262461.00 21451.00 41471.00 
  Borderline, 25 to 1001653540300.77 (0.49–1.22)0.2714300.81 (0.39–1.67)0.5726300.77 (0.45–1.34)0.36
  Elevated, >1001062332240.93 (0.57–1.53)0.7812261.11 (0.51–2.38)0.8020230.83 (0.45–1.53)0.55
Used Temodar
 Total IgE              
  Normal, <2519942121541.00 88561.00 33491.00 
  Borderline, 25 to 1001653560270.57 (0.39–0.83)<0.0135220.46 (0.30–0.72)0.0025370.90 (0.50–1.61)0.72
  Elevated, >1001062343190.62 (0.40–0.96)0.0334220.69 (0.43–1.11)0.139130.45 (0.20–1.00)0.05
Table V. Temodar Use by Various Characteristics and By Histologic Type (Excluding Those with Unknown Temodar Use) N = 479, San Francisco Bay Area Adult Glioma Study (2001–2005)
 GBMNon-GBM
Used temodarNo temodarUsed temodarNo temodar
nrow %nrow %nrow %nrow %
  1. Bold percentages indicate that the chi-square, p value is less than 0.05.

Ethnicity
 NonWhite4163243721413059
 White14562883853417759
Age
 <505276162448417059
 50–594668223214471653
 60–69586926319471053
 70+303848623211179
Gender
 Female7155584528355365
 Male11568543246465454
Proxy Status
 Self report15476482471439557
 Proxy report323364673201280
Blood Collection Status
 No blood collected272965717262074
 Blood collected15977472367448756
Total IgE
 IgE missing293165697262074
 Normal IgE8881211933454155
 Borderline IgE3571142925492651
 Elevated IgE347412269312069
Diagnosis Year
 Dx Year 20012141286229571
 Dx Year 20024960324022314969
 Dx Year 20036759464132483552
 Dx Year 20046876222418501850
Allergy History
 No Reported Allergy History4859334111421558
 Reported Allergy History13864793663419259
Education
 Not a College Graduate8957674333405060
 College Graduate9768453241425758

Cohort effects

Besides the introduction of temozolomide, another difference between our two series pertains to the birth cohort of participants. We note that the greatest difference between cases and controls in Series 2 is the age 53–65 central age tertile, OR = 0.58 (95% CI: 0.45–0.76, Supporting Information Table 2). The median age of these controls is age 59, which corresponds to a median birthdate of 1939. For Series 3, the corresponding cohort tertile shifts 9 years forward with a median birthdate of 1948. Individuals within this same age tertile in Series 3, who were not treated with temozolomide, did not demonstrate reduced IgE (OR = 0.94, 95% CI:0.70–1.26, Supporting Information Table 2).

Seasonality

We analyzed whether there were differences in seasonality of blood draws. Because of recruiting schedules, slightly more controls were ascertained in January–April time frame. Pollen levels are highest in our study area (San Francisco Bay Area) in the months of March–June. However, when we included the season of blood draw variable in the case–control model, ORs were not changed (data not shown). We also considered the question, “Have you had an allergic reaction in the past month?” An answer “yes” to this question was not predictive for high IgE levels among controls (data not shown).

Repeated measurements

We performed repeat measurements for a single patient who was not part of the SF Bay Area Glioma Study. IgE levels in this patient remained nearly constant at 30 kU/L, which is a clinical “low borderline” level for allergy. This patient did not have any history of allergy, and the IgE were not reactive to respiratory (Phadiotop) or food (Fx5) allergens (data not shown). Given that IgE half life in serum is only 36 hr, this indicated that IgE production is not sensitive to the hematotoxic or hemato-recovery cycles induced by the temozolomide treatments. The steady state of IgE is a contrast to varied T-regulatory cell, CD8+ cell counts and other immune cell counts measured with the same blood samples (Supporting Information Table 4 and Ref. 11).

Discussion

This manuscript describes a case–control replication study in which, due to a change in the standard of care, the majority of cases received an immune-modulating factor: temozolomide. The inverse association of self-reported allergy and glioma (including a dose–response) here was nearly exactly the same as our previous series, indicating a robust association.5, 7 IgE levels were also inversely related to glioma although far more weakly than we previously observed and most of the inverse relationship was confined to analyses including only cases who had received temozolomide as a treatment. This inverse association may also be impacted by length of temozolomide treatment (Fig. 1). Despite this introduction of temozolomide as the standard of care, the prevalence of elevated IgE levels among cases was paradoxically increased in the current series (Series 3), reducing IgE case–control differences observed previously.7 Self-reported allergies were more frequent in the current series among cases (20% in Series 3 vs. 13% in Series 2), suggesting some potential cohort differences between the series, which may also impact IgE levels. Both new treatments and a potential new cohort of individuals in Series 3 have then complicated this replication study with regard to IgE as a variable impacting glioma status.

During the early stages of glioma treatment, temozolomide is given daily, typically over the same time course as radiation treatments. Following initial treatment, temozolomide is prescribed in 5-day “on” and 23 day “off” recovery periods. It seems to be during this subsequent period that IgE levels were lower and case–control OR become significant (Fig. 1, Supporting Information Table 3). As these were cross-sectional data, the possibility remains that the lower levels were because of other unmeasured factors. A similar loss of immune function parameters (CD4 T-helper cells) was found in a study of melanoma patients treated with temozolomide over a longer period.13 CD4 cells were not significantly lost after the first treatment cycle but became more severe with subsequent cycles,13 mirroring the lymphopenia effect noticed in glioblastoma patients9 and the fall in IgE shown here (Fig. 1). Whether temozolomide may actively suppress IgE levels cannot be answered with the current data series and requires a study of repeat measurements during therapy. One patient with repeat measurements here indicated that IgE levels are stable over a 1-month period during later stages of treatment, and are not sensitive to changes in T-regulatory cells among other cell types (Supporting Information Table 4). This patient does not inform what happens to IgE during initial therapy, which is a question that will require additional studies.

We did not detect case–control differences in IgE levels with cases not treated with temozolomide, who comprised 37% of the patient cohort. No significant differences in gender, income, or ethnicity were found among patients who did not take temozolomide compared with those who did (Table V). However, there was a strong inverse association of temozolomide therapy and age—older individuals were less likely to be prescribed temozolomide. In our previous series, the relationship between IgE and case–control status was weakest in the oldest tertile (Supporting Information Table 2). Assuming a similar profile for the current Series 3, we would expect less of an IgE difference between older cases and controls, which constitute the bulk of the patients not treated with temozolomide. This factor, combined with unknown clinical considerations directing the decision not to treat with temozolomide, make it difficult to compare the temozolomide treated and untreated patients. Like our previous study, we did not detect significant effects from dexamethosone on IgE levels, nor other medications outside of the cytotoxic chemotherapeutics.

A remaining possible explanation on IgE differences between the two cohorts may be that the patients themselves are fundamentally different in Series 3, having grown up in the post-World War (WW) II era. Interestingly, the prevalence of childhood allergies is quite different between these groups, reported at 13% in Series 2 and 20% in Series 3 (p = 0.02). This corresponds to the rise of suburbanization in the United States with concomitant rise in the prevalence of allergies.14–16 Indeed, the average of 50 years of age for those with IgE measurements in Series 3 corresponds to birthdates between 1950 and 1954, which puts the bulk of participants at a post-WWII birth when modern trends including reduced exposure to microbes and endotoxins, intestinal parasites and increased use in antibiotics, affected the prevalence of allergies and potentially affected the allergic axis in the interaction with brain tumors. The increases in allergies are attributed to the “hygiene hypothesis” which posits that the immune system has not had a normal early modulation from frequent infections, making it vulnerable to overreact to other environmental antigens.17 This enhanced predisposition to overreact is sometimes referred to as the “missing immune deviation” hypothesis, since humans are born with a strong Th2-allergic-phenotype which is poised to develop balanced immunity in response to frequent infections early in life.18 If the immune deviation to a balanced Th1/Th2 immune system does not occur early in life, then lifetime risk of allergic disease is increased. Our control population had the same IgE distribution in the current series as the prior series,7 indicating that profound population shifts have not taken place; however, our cases had higher IgE levels in the current series. It may be possible that these cases are reacting to their tumor or to temozolomide treatment (initially) with higher IgE levels because of an enhanced predisposition to allergies due to their birth in the modern era. Further studies on a possible cohort effect are warranted.

For the current series we obtained sera from a higher proportion of cases than for our prior series—and ascertained them somewhat closer to diagnosis. We were also able to transport the blood on average 1 day faster (1 vs. 2 days) from the field to the laboratory. We assessed whether time since diagnosis and transit time for blood was related to IgE levels, with no significant relationship found (when excluding patients who took temozolomide). We also performed a reconstruction experiment in the laboratory to assess whether IgE levels are affected by storing whole blood at room temperature. IgE levels did not perceptibly change during a 4-day period that blood was stored at room temperature. We also note that the improvement in transport time affected both cases and controls equally, so is unlikely to play a role in case–control differences. It is, therefore, highly unlikely that our improvement of sampling techniques would explain any differences between Series 2 and 3.

Although we have emphasized the differences between the series, it is also important to note that many results were similar. Respiratory IgE had similar OR in both series and reported respiratory allergy OR were also similar and significant. In both series, we observed much stronger OR among proxy- compared with self-reported individuals; in both series this is likely the result of reporting bias. Proxies may not know about allergies, amplifying the OR; therefore self-report OR are likely to be closer to the truth (Table II). Case–control ORs for self-reported food allergies also did not differ between the series; however, one key difference was found in the relationship of food IgE between the two series (OR = 0.12 vs. 0.88, respectively, Series 2 and 3). The Series 2 result may be a false positive finding because of small numbers. The relationship of reported respiratory allergens is more robust; it exhibited virtually the same relationship in both reports. This finding should refocus attention on allergens and allergic responses thatenter via the nasal route, and away from the digestive route,which was not confirmed in the current analysis (Series 3). Interestingly, cytokines and other peptides administered intranasally can enter the brain directly,19 presenting a possible direct effect on intracranial immune responses from respiratory allergy pathophysiology.

In sum, our current results on reported allergy and glioma are quite similar to our and other's previous reports. Although we still observed overall that cases had lower levels of IgE than controls, the effect was only apparent in the temozolomide treated patients, and our cross-sectional analysis suggests that the lower levels of IgE in the patients could possibly be because of temozolomide treatment. To address differences between the series to the extent possible, we considered recruitment and sampling changes and differences in the population ascertained, including the later birth cohort of the Series 3 population. Our efforts do not fully explain the differences between the series, and cast some uncertainty on the capacity of IgE as an adequate biomarker to illuminate an immunologic mechanism that suppresses glioma, at least in the current temozolomide era. Less clear is the reason for overall higher IgE levels in glioma patients compared with the previous series, which occurred despite the temozolomide IgE suppression that may take place over the course of treatment. This IgE suppression mirrors the loss of T-helper cell function as studied in a temozolomide-treated melanoma cohort13 and serves as a warning to those developing immune-mediated glioma treatments. Definitive evaluation of the role of IgE as a biomarker of glioma risk awaits results from a large cohort study that collects serum before diagnosis. Additionally, the complete effects of temozolomide and other therapies on IgE await the analysis of serial samples obtained before and throughout glioma therapy.

Acknowledgements

The authors thank the Northern California Cancer Center for glioma patient case finding and to the Pathology Departments of Alexian Hospital, Alta Bates Medical Center, Brookside, California Pacific Medical Center, DR Pinole, Eden Hospital, El Camino Hospital, Good Samaritan, Highland Hospital, John Muir, Kaiser Redwood City, Kaiser San Francisco, Kaiser Santa Teresa, Los Gatos Hospital, Los Medanos Hospital, Marin General, Merrithew, Mills Peninsula Hospital, Mt. Diablo Hospital, Mt. Zion Medical Center, Naval Hospital, O'Connor Hospital, Ralph K Davies Medical Center, Saint Louise, San Francisco General, San Jose, San Leandro, San Mateo County, San Ramon Valley, Santa Clara Valley, Sequoia, Seton Medical Center, St. Francis, St. Lukes, St. Rose, Stanford, Summit, UC San Francisco, Valley Livermore, Veterans Palo Alto, Veterans SF and Washington Hospital for providing tumor specimens for review. J.L.W. is a Scholar of the Leukemia and Lymphoma Society of America.

Ancillary