Chlamydia pneumoniae Infection and Risk of Lung Cancer

There is accumulating evidence that infection with Chlamydia pneumoniae may be associated with an increased risk of lung cancer (1-6). Prospective and retrospective studies have reported 50% to 100% increased risks of lung cancer among those with elevated IgA antibody titers to this organism. Chronic inflammation and its sequelae may be part of the causal pathway.

We conducted a nested case-control study to further investigate this association while controlling for major lung cancer risk factors. Additional objectives were to identify possible modifying factors of the C. pneumoniae-lung cancer association and to describe histology-specific and site-specific associations. We used IgA antibody titers to C. pneumoniae as our exposure measure because IgA antibodies have a short half-life and thus have been proposed as a serologic marker of chronic or persistent C. pneumoniae infection (7, 8). To better understand the predictors of IgA antibody detection, we also examined the demographic, lifestyle, dietary, and medical correlates of IgA titers ≥16 among controls.

The present report is based on a case-control study nested within the β-Carotene and Retinol Efficacy Trial (CARET), a randomized, double-blind, placebo-controlled prevention trial that evaluated whether supplementation with β-carotene and vitamin A would reduce the risk of lung cancer among two high-risk groups: heavy smokers and persons with occupational exposure to asbestos. The design and findings of the study have been published elsewhere (9-11). Briefly, CARET was initiated with a pilot phase in 1985 and expanded 10-fold at six study centers throughout the United States (the “efficacy” study). CARET randomized 14,254 heavy smokers and 4,060 asbestos-exposed workers. Eligibility criteria for heavy smokers included being between ages 50 and 69 years and a current or former smoker (quit within the previous 6 years) with ≥20 pack-years of cigarette smoking history. Eligible asbestos-exposed participants were men ages 45 to 69 years, who were current or former cigarette smokers who quit within the previous 15 years. In addition, they had to have occupational exposure to asbestos beginning at least 15 years prior to recruitment into the trial and either 5 years of work in a high-risk trade or chest X-ray findings consistent with asbestos exposure. Subjects were randomized to receive a standard daily regimen of either 30 mg β-carotene plus 25,000 IU of retinyl palmitate or placebo. In January 1996, the active intervention was terminated prematurely because the group receiving retinyl palmitate and β-carotene had a 28% excess lung cancer incidence and 17% excess overall mortality rate compared with the group receiving a placebo.

At baseline, information on demographic factors, general health, health history, smoking history, and usual diet was collected. Participants identified the age when they started smoking, any period(s) of time when they did not smoke cigarettes and their duration(s), and average number of cigarettes they smoked. Pack-years of smoking were calculated as average number of cigarettes smoked per day divided by 20 multiplied by the number of years smoked. To assess usual dietary intake over the preceding year, we used a 110-item semiquantitative food frequency questionnaire that was developed at the Fred Hutchinson Cancer Research Center and is similar to the National Cancer Institute/Block Food Frequency Questionnaire (12).

Blood was collected from pilot participants every year and from efficacy participants at their first visit and every 2 years after randomization. Serum was separated by centrifugation and stored for a maximum of 2 weeks in brown glass freezer vials at −20°C prior to long-term storage at the −70°C at CARET Serum Bank (10).

Sources of initial reports of outcomes were received from participants or their next of kin at yearly or biannual contacts as well as from cancer registries, state boards of health, and the National Death Index. The CARET Internal End Points Committee reviewed all initial reports of lung cancer and obtained clinical records and pathology reports from the diagnosing physician/hospital. The committee used this information to determine the primary site, histology, date of histologic diagnosis, and relationship of the cancer to death if the patient had died (9).

In the present study, we selected 520 of the 836 confirmed lung cancer cases diagnosed through June 2000. To maximize the power to detect associations in important subgroups (i.e., women and those diagnosed at younger ages), we selected all women (n = 264) and all men ages <60 years at diagnosis (n = 84). About 30% of men over age 60 years (n = 172) were randomly selected after stratifying on 5-year age groups. One control subject without lung cancer was selected for each case from among those who were alive and being followed at the time of the case's diagnosis. The control subject was matched to the case for year of randomization, age at randomization (within 5-year age groups), gender, pilot or efficacy study, exposure cohort (heavy smoker or asbestos–exposed), and smoking status (current or former at baseline). Nine case-control pairs were omitted for the following reasons: insufficient blood sample available (n = 4), sample lost (n = 2), only a postdiagnosis blood sample available (n = 2), and missing information on smoking frequency or duration (n = 1). Furthermore, three pairs were concordant on all factors and were dropped from conditional logistic regression models, leaving 508 pairs. Participants with implausibly high or low energy intakes (<800 or >5,000 kcal/d for men and <600 or >4,000 kcal/d for women; n = 128; ref. 13) were excluded from the analyses on diet.

Serum collected at baseline or, for those without baseline serum available (n = 70), at the earliest time available prior to lung cancer diagnosis in the cases and a matched time in the controls was used to evaluate C. pneumoniae IgA and IgG antibody titers using the microimmunofluorescence test (14). Purified C. pneumoniae elementary bodies (strain AR39, Washington Research Foundation, Seattle, WA) were used as the antigen. Details of the test technique are described elsewhere (14). Only even fluorescence of the elementary bodies in the antigen dots was considered to be positive. After screening at a dilution of 1:16, the highest dilution demonstrating definite fluorescence was recorded as the titer (expressed as a reciprocal of the serum dilution). The antibody determinations of each case and matched control were tested simultaneously in the same titration series (batch) in a blinded fashion. By convention, C. pneumoniae titers (IgA or IgG) ≥16 were considered seropositive, which is consistent with the cutoff used in other studies (1, 2, 5, 6).

We used conditional logistic regression, which takes into account the matched design, to estimate the odds ratios (OR) and 95% confidence intervals (95% CI) of lung cancer in relation to IgA and IgG antibody titers. The other characteristics listed in Table 1 were evaluated for confounding and were included in models if they changed the β coefficient for dichotomous IgA by >10%. Ultimately, all models were adjusted for education (college graduate or not), body mass index (continuous), years smoked (log transformed), and cigarettes smoked per day (continuous). Although we were concerned that storage time could potentially affect antibody titers, because participants were matched on randomization year, they were also effectively matched on the time between blood draw and microimmunofluorescence test.

The purpose of our main study was to determine whether IgA titers ≥16 were associated with an increased risk of lung cancer (identified as the “case-control study”). In secondary analyses, we evaluated lung cancer risk by categorical IgA titer level and also conducted analyses with IgG antibody titers. The association between C. pneumoniae and lung cancer risk may differ depending on the presence of other factors (age, sex, intervention arm, etc.); consequently, we evaluated the presence and magnitude of multiplicative interactions by conducting analyses stratified on the characteristics of interest. In these analyses, former smokers were defined as individuals who had quit at least 1 year before baseline to reduce misclassification of those who only quit briefly at, or only shortly before, baseline (15). To examine whether C. pneumoniae may be associated with certain types (e.g., histologic type and lobe of origin) of lung cancer, we also constructed polychotomous regression models, which were adjusted for the confounding factors as well as the matching factors.

We used unconditional logistic regression to identify correlates of IgA seropositivity among controls (identified as the “correlates of IgA seropositivity study”). In these analyses, seropositivity was our outcome variable. Models were adjusted for age, sex, race, pilot or efficacy study, and time between blood draw and microimmunofluorescence test.

Table 1 gives distributions of various characteristics among cases and controls. Cases were matched to controls on 5-year age group, sex, exposure cohort, pilot or efficacy study, and smoking status at baseline and thus did not vary by these characteristics (Table 1). The median age of cases and controls was 59 years, and about half were women. Those who developed lung cancer were less likely to be college graduates and more likely to have been randomized to the intervention, to have a body mass index <25 kg/m², and to have a history of chronic bronchitis or emphysema. As expected, those who developed lung cancer had a heavier smoking history (cigarettes per day, years smoked, and pack-years). Cases tended to have a higher energy intake and percentage energy from fat but to consume fewer servings of fruits and vegetables compared with controls (data not shown). The most common histologic type was adenocarcinoma (26%) followed by squamous cell carcinoma (20%) and small cell carcinoma (18%); 51% of cancers arose in the upper lobe of the lung.

Table 2 provides the distributions of IgA and IgG antibody titers among cases and controls and their corresponding adjusted ORs and 95% CIs. Compared with controls, cases were more likely to have IgA titers ≥16 (55.4% versus 51.3%) and ≥256 (5.1% versus 2.5%). Those with antibody titers ≥16 had 1.2 times the risk of lung cancer (95% CI, 0.9-1.6) compared with those with lower titers. There was a significant trend (P = 0.007) of increasing ORs with increasing IgA titers primarily due to an OR of 2.8 (95% CI, 1.1-6.7) associated with titers ≥256. Because a history of chronic bronchitis or emphysema could be a confounder or in the causal pathway, we conducted analyses both with (OR, 1.3; 95% CI, 1.0-1.7), and without adjustment for it (Table 2). For IgG, the proportions of cases and controls with titers ≥16 were similar (63.8% and 64.2%, respectively), although cases were more likely to have IgG titers ≥256 than controls (21.7% versus 16.1%; Table 2). To determine the association between high IgG titers and lung cancer risk, independent of IgA titers, we also conducted analyses adjusting for IgA titers. In these analyses, results were attenuated and there was no longer evidence for a stronger association with high titers (OR, 1.1 for IgG ≥256 versus IgG <16).

Table 3 gives ORs for the association between IgA ≥16 and lung cancer risk in specific subsets of the cohort. Lung cancer risks associated with IgA ≥16 were greater for men in the asbestos-exposed worker cohort, individuals enrolled in the pilot study, and former smokers who had quit at least 1 year before baseline, although only the smoking status-IgA interaction approached statistical significance (P for interaction = 0.08). To be more certain that we had excluded those with subclinical lung cancer at the time blood was collected, we progressively eliminated individuals who were diagnosed within 5 years of blood sampling. There was little change in the ORs, suggesting that IgA titers measured at baseline represented pre–lung cancer levels. IgA seropositivity was more strongly associated with squamous cell carcinomas. There was also a suggestion, based on small numbers, of increased risk for cancers occurring in the main lobe (OR, 2.5; 95% CI, 0.6-10.1) and middle lobe (OR, 1.9; 95% CI, 0.9-4.1) of the lung.

Table 4 gives the prevalence of IgA seropositivity among controls along with ORs and 95% CIs. Seropositivity increased somewhat with age (Table 4). Men in the smoker cohort were more likely to be seropositive than women or than men in the asbestos-exposed worker cohort. Those of “other” or unknown race were more likely to be seropositive than Caucasians. Seropositivity was lower among those who were obese and who had self-reported health that was not excellent. There was a suggestion of lower seropositivity with increasing time since quitting but little pattern of seropositivity with any other cigarette smoking variables or any dietary variables (data not shown).

At least six other studies (1-6) have investigated the association between IgA antibodies to C. pneumoniae and lung cancer risk, two of which were prospective in design (2, 6). Despite differences in populations and study designs, ORs observed in the prospective studies were similar and ranged between 1.6 and 2.2 after adjustment for smoking. The four case-control studies (1, 3-5) also observed ORs of similar, or slightly greater, magnitude. In studies that used more than one cutoff for IgA titers, stronger ORs were observed for higher IgA titers (e.g., ≥64; refs. 2-4).

In comparison, our results for IgA seropositivity were generally weaker and compatible with there being no association. As expected, we observed no association with IgG seropositivity. ORs were statistically significantly elevated for IgA titers ≥256, suggesting that higher titers may be a better predictor of lung cancer risk than lower antibody titers. However, because we did not design the study to investigate this categorization, it needs replication in other studies. Furthermore, it is unclear biologically how higher titers may be related to severity or chronicity of infection.

In analyses limited to controls, we did not observe any differences in IgA seropositivity by smoking behaviors, except perhaps time since quitting, contrary to previous study findings (16, 17). This may be due to homogeneity of study participants in terms of their smoking, misclassification of smoking or C. pneumoniae seropositivity, or because there truly is no association.

Of the three studies that investigated histology-specific results (2, 5, 6), one study reported a stronger association with squamous cell and small cell carcinomas combined (6), whereas the others did not observe a statistically significant difference by histologic type (2, 5). We observed a stronger association for squamous cell and to a lesser extent, for small cell carcinomas and adenocarcinomas. Several studies have observed a stronger association between C. pneumoniae infection and lung cancer among those diagnosed at younger ages (<60 years; refs. 1, 5, 6); we found little support for that notion.

Prospective studies such as ours should minimize any influence of disease itself on titer levels, reduce selection bias, and generally include better adjustment of potential confounding factors. In case-control studies where blood was sampled after diagnosis (1, 3-5), it is possible that lung cancer caused an elevation in C. pneumoniae titers. It is therefore unclear whether lung cancer cases were more susceptible to chronic infection with C. pneumoniae or whether chronic infection contributed to an increased risk of lung cancer. Serum was collected before diagnosis in our study. When we successively excluded cases that had blood drawn 1 to 5 years before diagnosis, results remained the same, suggesting that C. pneumoniae infection preceded disease onset.

Another issue in case-control studies is the short survival of persons diagnosed with lung cancer. To the extent that C. pneumoniae infection is related to length of survival after lung cancer diagnosis, results of studies that exclude those who died shortly after diagnosis, or were too sick to participate, may be biased. However, we are aware of no data that suggest such a relationship.

Because both age and smoking are strongly associated with lung cancer and possibly positively associated with C. pneumoniae IgA and IgG titers (16, 17), they are potentially important confounders. Inadequate control for these factors could result in a bias away from the null. In several case-control studies that found particularly strong associations (3, 4), controls were more likely to be younger and former or never smokers than cases, and investigators did not control for any measures of smoking. In our study, confounding due to cigarette smoking was addressed both in the design phase by restricting the study to heavy smokers (see eligibility criteria in Materials and Methods) and by matching on smoking status at baseline, and in the analysis phase by adjusting for daily number of cigarettes smoked and years of smoking. However, because there was little association between cigarette smoking and IgA seropositivity in our study, there was likely little confounding due to cigarette smoking.

Definition of “chronic” infection varied in each of the studies. For example, one study used a combined measure of IgA ≥16 and immune complexes (6), whereas others used IgA ≥64 (4) and/or IgG ≥512 (1, 3). Furthermore, it was unclear if the criteria were chosen a priori or based on the data. Data-driven determination of cut points could result in a bias away from the null. Conversely, nondifferential misclassification of IgA antibody titers could result in a bias toward the null. We can use information on the reliability of C. pneumoniae antibodies to determine how misclassification of C. pneumoniae titers might attenuate relative risks. For example, if the true relative risk for the association between lung cancer and IgA titers ≥16 were 2.0, reliability of the magnitude observed in our reliability study (κ = 0.39; ref. 18) would result in an attenuated relative risk of at most

A relationship between C. pneumoniae infection and lung cancer is biologically plausible. Although it is not known how infection with C. pneumoniae might induce or lead to the progression of lung cancer, chronic inflammation may be involved. Agents that cause inflammation, such as infectious agents, may cause prolonged irritation, resulting in cell death and increased mitotic activity. The subsequent cell division that occurs during repair of the damaged tissue may increase the risk of cancer at the affected site (20). For example, multiple studies have linked chronic infection with Helicobacter pylori to an increased risk of gastric adenocarcinoma (21-23). C. pneumoniae may play a similar role in lung cancer development. C. pneumoniae stimulates the release of inflammatory mediators such as tumor necrosis factor-α, interleukin-1β, and interleukin-8 (24). Chronic infection generally, and interleukin-8 specifically, may cause genetic damage. Interleukin-8 also acts as a promoter of tumor growth for human non-small cell lung carcinoma through its angiogenic properties. C. pneumoniae can impair or even block apoptosis of infected cells (25) by induction of interleukin-10 (26), resulting in chronic infection and an increased risk of malignant transformation of infected cells.

In summary, we found a more moderate overall association between lung cancer risk and elevated IgA titers than most previous studies. However, a key limitation of all such studies is the relatively modest reliability of the microimmunofluorescence test on which they are based (18) and the resulting conservative bias. Thus, to better understand the role of C. pneumoniae in lung cancer carcinogenesis, it is important that we identify measures of chronic C. pneumoniae infection that are valid and reliable as well as being inexpensive and minimally invasive to obtain. Due to the intracellular nature of C. pneumoniae (which makes it difficult to culture), the inaccessibility of lung tissue for sampling, and the lack of a gold standard for identification of chronic infection, development and evaluation of new measures and their role in lung cancer is challenging. Nevertheless, considering the public health importance of lung cancer, such efforts could be richly rewarded.