Recent publications have reported the technical and clinical validation of EarlyCDT-Lung, an autoantibody test which detected elevated autoantibodies in 40% of lung cancers at diagnosis. This manuscript reports the results of EarlyCDT-Lung run on four new (postvalidation) data sets. The following four cohorts of patients (n = 574) with newly diagnosed lung cancer were identified: group 1 (n = 122), 100% small cell lung cancer (SCLC); group 2 (n = 249), 97% non-small cell lung cancer (NSCLC); group 3 (n = 122), 100% NSCLC; group 4 (n = 81), 62% NSCLC. Serum samples were obtained after diagnosis, prior to any anticancer treatment. Autoantibody levels were measured against a panel of six tumor-related antigens (p53, NY-ESO-1, CAGE, GBU4–5, Annexin 1, and SOX2) in the EarlyCDT-Lung panel and previously established cutoffs applied. In groups 2, 3, and 4, patients were individually matched by gender, age, and smoking history to a control individual with no history of malignant disease. Assay sensitivity was tested in relation to cancer type and stage, and in the matched normals to demographic variables. The autoantibody panel showed sensitivity/specificity of 57%/n.d (not done) for SCLC in group 1, 34%/87% for NSCLC in group 2, 31% and 84% for NSCLC in group 3, and 35%/89% for NSCLC and 43%/89% for SCLC in group 4. There was no significant difference in positivity of EarlyCDT-Lung and different lung cancer stages. These studies confirm the value of an autoantibody assay, EarlyCDT-Lung, as an aid to detecting lung cancer in patients at high risk of the disease. Cancer Prev Res; 4(7); 1126–34. ©2011 AACR.
Recent publications have reported on the technical and clinical validation of an autoantibody assay for lung cancer, EarlyCDT-Lung (1, 2). The clinical manuscript reported that these immunobiomarkers detected both non-small cell (NSCLC) and Small Cell Lung Cancer (SCLC), and that there was no significant difference between different lung cancer stages, indicating that the antigens included identified early- as well as late-stage disease. As such, EarlyCDT-Lung was reported to offer a diagnostic tool and a potential system for monitoring patients at high risk of lung cancer.
The need for an aid to detect lung cancer early is undisputed. Lung cancer is the worldwide leading cause of cancer-related mortality (3). Outcomes are substantially better with early, localized disease compared with locally advanced and metastatic disease, with 5-year survival rates of 53%, 23.7%, and 3.5%, respectively (4). A recent review of SCLC, previously regarded as a disease for which the primary treatment was systemic chemotherapy, has reported excellent survival for early, localized disease that has been resected with or without adjuvant chemotherapy (5). Lim and colleagues reported a 5-year survival rate of 52% for stage 1 without adjuvant chemotherapy (6), whereas Brock and colleagues reported a survival rate of 58% overall for stage 1, rising to 87.5% for stage-1 patients who had surgery followed by platinum-based adjuvant chemotherapy (7). There is, therefore, increasing evidence that early-stage disease treated by surgery with or without (neo)adjuvant chemotherapy can have substantially better 5-year survival rates than late-stage disease.
Ongoing clinical trials are investigating the use of spiral computed tomography (CT) in “at-risk” individuals (8–17). One of the major problems with CT is the high rate of false positives (as high as 50% in a prevalence round; ref. 10), which dictates that many individuals require follow-up examinations and a substantial proportion of individuals undergo unnecessary thoracotomy (18). A recent manuscript by the Lung Screening Study reported that up to 7% of patients who were screened by CT underwent some level of invasive procedure (19). This suggests that a test with a higher specificity than CT that can identify high-risk individuals with early-stage disease would be a valuable aid to the early detection of lung cancer.
This article reports the results of EarlyCDT-Lung in 4 new (postvalidation) data sets from individuals in the United States, Canada, and the United Kingdom involving measurement of these immunobiomarkers in the serum of patients with newly diagnosed lung cancer (prior to any treatment) and matched controls.
Findings from 4 separate groups of patients with newly diagnosed lung cancer are reported. Blood samples were obtained after diagnosis but prior to receiving any anticancer treatment. In 3 of the 4 groups (groups 2–4), patients with lung cancer were, as far as possible, individually matched by gender, age, and smoking history to control individuals with no previous history of malignant disease. These controls were taken from the normal population. Blood samples from more than 5,000 individuals were collected and were used to match with the individual cancer patients. Matching was conducted on the basis of basic demographics but without any knowledge of autoantibody data. The demographic characteristics of the control versus the study population are given in the Appendix.
Group 1 comprised 122 patients with SCLC presenting to a single center in the United Kingdom. Baseline patient characteristics are shown in Table 1. Samples from this group were run on the EarlyCDT-Lung test without matched controls as the aim was to provide further confirmation of the sensitivity of the test for SCLC in a larger group of patients. The validation data set contained 73 SCLC samples (2). Group 2 comprised 249 patients with lung cancer collected in multiple European centers. The lung cancer patients were matched for age, sex, and smoking history with samples from normal populations in Europe (n = 237) and the United States (n = 246; ref. Table 1). The normal controls do not exactly match the number of lung cancer patients, because after the studies were run it was noted that 15 of the controls had been included in other postvalidation studies reported in this article: the authors felt that any individual control sample should not be included more than once. Group 3 comprised 122 patients with lung cancer treated at a single center in Vancouver, Canada, who were matched to 114 control samples from high-risk individuals who did not have lung cancer (Appendix; Table 1). The 122 patients with lung cancer included 3 individuals who were initially designated as controls but were found to have developed lung cancer in the follow-up period. These 3 were, therefore, included in the cancer group for the sensitivity and specificity analysis. It should be noted, however, that it was only after the laboratory data had been transferred to our collaborators in Vancouver that the clinical data were made available to the laboratory researchers. Group 4 comprised 81 patients who were also matched to controls based on age, sex, and smoking history. One of the primary reasons for including the matched normals in groups 2 to 4 was to provide further confirmation of the specificity of the EarlyCDT-Lung test in high-risk individuals.
Tumor pathologic information was available for the patients with lung cancer, including TNM (Tumor, Node, Metastasis) staging, tumor-type SCLC or NSCLC, and NSCLC subtype histology (Table 2). Because this was not a CT screening trial, no CT data are available for these patients. In the clinical validation manuscript (2), early-stage disease included stage-1 or -2 NSCLC and limited disease of SCLC, and the same definition was used when analyzing these 4 new data sets to assess the sensitivity of EarlyCDT-Lung for early- and late-stage disease.
Autoantibody positivity by stage of disease and histologic subtype was not reported in the clinical validation manuscript. However, with significantly greater numbers of lung cancers, these data were analyzed by combining the 4 postvalidation data sets and the validation data set described by Boyle and colleagues (2).
Serum samples were evaluated in the EarlyCDT-Lung assay for autoantibodies against p53, NY-ESO-1, CAGE, GBU4–5 (also known as FLI3072 or TDRD12), Annexin 1, and SOX2, as previously reported (1, 2). For each group, samples from patients with cancers, matched normals, and control sera for the assay were interspersed: samples were assayed in an order so that any batch effects would be spread over all sample types. The laboratory staff, performing the assay, were blinded to the disease state of individual samples. In group 3, the samples were run, and the assay results returned to the clinician supplying the samples before any clinical data were released.
Autoantibody levels were determined by a quality-controlled, semiautomated indirect ELISA in which samples were allowed to react with a titration series of antigen concentrations. All liquid-handling steps were carried out by using an automated system. Briefly, purified recombinant antigens were diluted to provide a semilog titration series for each antigen from 160 to 1.6 nmol/L. Control antigens consisting of the purified BirA or NusA tags were also included to allow subtraction of the signal because of nonspecific binding to bacterial contaminants. Antigen dilutions were adsorbed to the surface of microtiter plate wells in phosphate buffer at room temperature. After washing in phosphate-buffered saline containing 0.1% Tween 20, pH 7.6, microtiter plates were blocked with a gelatine-based blocking buffer. Serum samples (diluted 1 in 110 in a blocking buffer) were then added to the plates and allowed to incubate at room temperature with shaking for 90 minutes. Following incubation, plates were washed, and horseradish peroxidase-conjugated rabbit anti-human IgG (Dako) was added. After a 60-minute incubation, the plates were washed and 3,3′,5,5′-tetramethylbenzidine was added. Color formation was allowed to proceed for 15 minutes before the optical density of each well was determined spectrophotometrically at 650 nm. The assay included a calibration system which utilized fluids drained from pleural or peritoneal cavities of patients with lung cancer, producing calibrated reference units (1).
All assays were conducted as 2 replicates and the mean value taken as the overall assay measurement. Samples were judged to be positive if they fulfilled 2 criteria, that is, they showed a dose–response to the antigen titration series and the measured autoantibody signal to 1 or more of the antigens in the EarlyCDT-Lung assay was above the cutoff set for that antigen in the commercial assay.
The initial data analysis to determine whether the sample was positive or negative was carried out in a completely automated system in which the sample list and raw plate data were kept separate until a final merge. The assay results (positive or negative) were then added to the different data sets with the clinical data and the sensitivity and specificity calculated.
For the statistical analysis, positivity rates were analyzed as 2 × r contingency tables by using standard χ2 tests with the respective degrees of freedom. For the forest plots, CIs for sensitivity were derived under a binomial assumption.
In the validation studies (2), the cutoffs for the autoantibody assays to the 6 antigens in the commercial EarlyCDT-Lung assay had been set to achieve a specificity of 90% in the matched control groups, to produce a test that could be used for early detection in a high-risk population and that would be health economically viable. To accomplish this, a Monte Carlo direct search method (20) was applied to find an optimized set of antigen-specific cutoffs yielding the maximum sensitivity for the fixed specificity of 90%. In these new studies, no further optimization was carried out and the commercial cutoffs were applied, providing further confirmation of the clinical utility of the commercial cutoffs.
The sensitivity and specificity of the EarlyCDT-Lung assay in each of the 4 groups, broken down by tumor type (NSCLC and SCLC), are shown in Table 3. For comparison, the sensitivity and specificity reported for the panel of the same 6 antigens in the Clinical Validation manuscript (2) are also included in Table 3. These show that the results for the 4 new data sets, by using the commercial assay cutoffs (i.e., not optimized for each individual data set), fall within the 95% CIs of the validation data. The one exception was the specificity for group 3 where the matched normal controls had a lower than expected specificity; however, these individuals had almost double the mean pack-years compared with the validation population (45.2 compared with 26, respectively), making them a much higher risk for cancer development.
Combining all data sets where all 6 antigens were measured (Table 3) gave 1,077 patients with lung cancer plus 1,296 matched controls. The sensitivity/specificity of the EarlyCDT-Lung was 38%/88% overall, with 34%/88% for NSCLC and 50%/88% for SCLC.
In this study, positive predictive values (PPV) for EarlyCDT-Lung, along with prevalence-based accuracy values for an assumed lung cancer prevalence of 1.5% would be 4.5% and 87.0%, respectively. At a lung cancer prevalence of 2.0%, PPV would be 6.0% with an accuracy of 86.8%, and at 2.7% prevalence, PPV would be 8.0% with 86.4% accuracy.
Effect of patient and disease characteristics on autoantibody assay sensitivity and specificity
Antigen positivity by histologic subtype for the panel and also for each of the antigens is shown in Tables 4 and 5. There was a higher sensitivity for SCLC compared with NSCLC (P ≤ 0.001) but no difference in sensitivity between the subtypes of NSCLC (P = 0.35). The results by tumor staging according to the International Association for the Study of Lung Cancer (IASLC, 7th edition) are shown in Tables 6–9. When the stage of disease was looked at within NSCLC (I–IV) and SCLC (limited and extensive disease), there was no significant difference (P = 0.54 and P = 0.78, respectively). For the 4 postvalidation data sets, the sensitivity of EarlyCDT-Lung for early- and late-stage disease is shown in Figure 1.
Irrespective of cancer type, early detection improves prognosis by allowing earlier treatment before the cancer spreads. The National Lung Screening Trial has shown that early screening, in the form of low-dose CT scans, can decrease lung cancer mortality by 20%, which highlights the value of early screening (21). However, the high proportion of noncancerous changes detected on chest CT, and the additional expensive diagnostic procedures needed, makes a strong case for a simple biomarker test that can be used as a diagnostic tool.
This report further confirms that EarlyCDT-Lung is a validated assay for the detection of autoantibodies to selected cancer-associated antigens in the peripheral blood. The study also confirms that the assay, by using the previously validated cutoffs, gives a sensitivity up to 40% for an overall lung cancer population. In patients with lung cancer, NSCLC typically accounts for 80% to 87% of cases and SCLC accounts for 13% to 20% of all cases, the exact proportions depending on a variety of factors such as the proportion of smokers versus former smokers and the level of smoking history. A further important point is that because the cutoffs used are those previously defined, they were not optimized for any of the 4 data sets. This provides further prospective confirmation of the reproducibility and clinical utility of the test.
For all 4 study groups, the sensitivity of the test by type of lung cancer (i.e., NSCLC and SCLC) was within the 95% CI of the validation study results (Table 3). The validation data set contained 73 SCLC samples. Although this was more than 13% of the validation group (2), and therefore greater than the percentage of lung cancers which are small cell according to the Seer database, it was felt that a larger data set was warranted to more accurately assess the sensitivity of the 6-antigen EarlyCDT-Lung test in SCLC. The data also confirm that the test detects early-stage cancer (stage I/II NSCLC plus limited SCLC) as well as it detects late-stage disease (stage III/IV NSCLC plus extensive SCLC; Fig. 1). This is particularly important if these immunobiomarkers are to act effectively as an aid to early detection. The presence of such a signal in early-stage disease is precisely what would be expected of an in vivo amplification signal such as the humoral immune response. This is in contrast to cancer-associated antigens, which are markers of tumor burden and not useful for the early detection or screening of breast (22, 23) or colorectal cancers (24, 25).
Previous publications (1, 2, 26–36) have highlighted the potential value of a panel of autoantibodies for the early detection of cancer. This study shows the sensitivity of both the overall panel and each individual autoantibody assay (Tables 4–9), and in doing so highlights the benefit of measuring autoantibodies to a panel of cancer-associated antigens compared with only 1 autoantibody assay. Tables 4–9 highlight that the panel as currently presented has a higher sensitivity for SCLC than NSCLC. They also highlight that individual autoantibody assays have different percentage sensitivity for different subtypes of lung cancer. As more assays are run and the number of patients with lung cancers increases, it may be possible to give an estimate of what subtype of lung cancer a patient is most likely to have, based on the pattern of autoantibody results.
Although it may be argued that if the control samples used were not matched to the patient samples by time in storage, this could lead to differences in antibody levels between the groups. The controls, we describe here, were collected around the same time as the cancer cases (started in 2007 or 2008, depending on sample sets, and finished in 2010). In addition, our sample stability studies over 2 to 3 years do not indicate any decreases in signal when the blood samples are properly stored (unpublished data).
Individual autoantibodies such as p53 autoantibodies have been detected prior to diagnosis of lung cancer in smokers with chronic obstructive pulmonary disease (37) or in patients with asbestosis (38). In the latter publication, the average lead time (time from first positive sample to diagnosis) was 3.5 years (range 1–12 years). Similar publications on other single autoantibodies (39–41) also indicate the induction of autoantibodies happening relatively early in the process of carcinogenesis. Autoantibodies to a panel of cancer-associated antigetns have been reported up to 5 years before screening CT scans (32) in lung cancer and up to 4 years before screening mammography detected breast cancers in young women at increased risk (31, 33). A recent presentation on SCLC has shown that EarlyCDT-Lung was positive in prediagnostic samples between 1 and 49 months prior to diagnosis of SCLC (42).
The study also confirms that the test has good specificity. In groups 2 to 4, matched normals were run and the specificity lay within the previously reported 95% CI of the validation data (Table 3). In group 3, the specificity was 84%, which was just below the lower margin of the 95% CI. In a group of high-risk smokers or ex-smokers, there will always be some individuals who are harboring an occult lung cancer. The specificity will vary somewhat if the risk profile of a group were to be higher or lower than the validation group. The matched normals in group 3 had almost double the mean pack-years compared with the validation population (45.2 compared with 26, respectively) or the matched normals in groups 2 and 4 (20.3 and 20.4 pack-years, respectively), and it is therefore not surprising that the specificity was slightly lower in this group.
Other researchers have developed risk models based on demographics from large population-based studies (43). This approach may be useful for the initial identification of a cohort at high risk of lung cancer over a defined period but does not allow repeated reassessment of the risk as many of the demographic factors in the models do not change significantly over the time period. The integration of immunobiomarkers in the blood with established demographic models may provide additive information and also provide a dynamic system for monitoring whether an individual at high risk seems to be developing a lung cancer.
In summary, these studies confirm the findings of the assay validation study (2) that EarlyCDT-Lung can detect up to 40% of lung cancers and that these immunobiomarkers detect early-stage disease as well as they detect late-stage disease. Furthermore, the pattern of autoantibody results varies between individuals and in future may provide an estimate as to what subtype of lung cancer an individual has developed. The study also confirmed that the specificity of the test is good, which is a prerequisite for it to be useful as an aid to early detection. The robust specificity of the EarlyCDT-Lung test indicates that it should make a major contribution to the diagnosis and monitoring of lung cancer patients.
It would also be important to examine the validity and utility of this test in populations with noncancer pulmonary pathologies (e.g., Chronic obstructive pulmonary disease and pneumonia). We have prospectively gathered information on concomitant benign autoimmune diseases, but not on other disorders. Data on benign lung conditions have been published in our previous validation paper which included 63 patients with benign lung conditions (2). The specificity of EarlyCDT-Lung was 89% for this group.
We understand and acknowledge that no cancer marker is 100% tumor-site specific and that some false-positives for lung cancer may in fact have another type of cancer. In this respect, we have preliminary data that show that the core antigens (e.g., p53 and NY-ESO-1) are also elevated in other types of cancer, such as breast or ovarian cancer. Nonetheless, in the population we are targeting, the principal demographic risk is that of lung cancer (around 2 per 100) whereas, for example, the risk of ovarian cancer is an order of magnitude lower. For this reason, we anticipate that the proportion of patients with a non–lung derived cancer will be very small. Furthermore, patients with a positive test but no detectable lung cancer should check with their physician that they have had any screening tests for other cancers (as advised by the American Cancer Society).
This study has shown that the EarlyCDT-Lung antibody panel has clinical utility for detecting lung cancer in clinical samples. There are ongoing studies testing the sensitivity and specificity of EarlyCDT-Lung in prediagnostic samples to fully assess the utility of the panel in monitoring asymptomatic patients for lung cancer. Future work is already ongoing to look for ways to increase the sensitivity and/or specificity. This includes investigating new antigens that are additive to the current panel and also looking at using not only cutoffs for each assay based on a high-risk control population but also assessing sequential changes in an individual's results or profile compared with their own baseline test results. In addition, combining these immunobiomarkers with demographic risk models (41) to assess if they are additive is ongoing.
Demographic characteristics of the control versus the study population
A total of 574 lung cancer sera (402 were from patients with NSCLC, 156 with SCLC, and 16 of unknown histology) were compared directly with 802 normal sera, which were analyzed as controls. Samples were obtained, with full informed consent, at different sites.
Group 1 comprised 122 patients with SCLC presenting to a single center in the United Kingdom. There were 68 males and 54 females, and the median age was 65 years (range 43–86). Group 2 comprised 249 patients with lung cancer collected in multiple European centers. The lung cancer patients were matched for age, sex, and smoking history with samples from normal populations in Europe (n = 237) and the United States (n = 246). In group 2, there were 201 males and 48 females. Controls for group 2 were selected from a prospective collection of blood samples taken from a larger sample set of normal populations in the Midlands of England and the Midwest of America. Controls for patients in group 2 were matched on the basis of gender and age (+/− 4 years). As all subjects in this group were smokers, pack-year matching was attempted but a tight match was prohibited by lack of information. The normal controls do not exactly match the number of lung cancer patients, because after the studies were run it was noted that 15 of the controls had been included in other postvalidation studies reported in this article: the authors felt that any individual control sample should not be included more than once. The median age (range) of the lung cancer patients and controls was 62 (23–82) and 62 (23–82) years, respectively.
Group 3 (n = 240) comprised 120 patients with lung cancer treated at a single center in Vancouver and Canada, who had been matched to 120 control samples from high-risk individuals who did not have lung cancer. The gender distribution was female (n = 63 and 69), male (n = 48 and 51), and unknown (n = 9 and 0) for cancers and controls, respectively. The median age (range) was 69 years (±10) for cancer patients and 62 years (±6) for controls. Pack-years smoked were 39±24 for the cancers and 45±16 for the controls. EarlyCDT-Lung results were available on 236 of the 240 samples which were returned to the Vancouver center blind of any clinical data. The mean follow-up on these patients was 57±13 months. There were initially 119 patients who had lung cancer and 117 controls with EarlyCDT-Lung results. Three controls with EarlyCDT-Lung results were diagnosed with lung cancer during the follow-up period (1 male and 2 female ex-smokers 5, 30, and 40 months after the blood sample had been taken): these were placed in the cancer group for the sensitivity and specificity analysis. This gave 122 with cancer and 114 controls. Group 4 comprised 81 patients (43 males and 38 females) who were also matched to controls based on age, sex, and smoking history. One of the reasons for including the matched normals in groups 2 to 4 was to provide further confirmation of the specificity of the EarlyCDT-Lung test in high-risk individuals.
Disclosure of Potential Conflicts of Interest
Caroline J. Chapman and John F.R. Robertson are consultants to Oncimmune Ltd., a University of Nottingham spinout company. JFRR is also a company shareholder.
The research reported was supported by Oncimmune Ltd. and the University of Nottingham.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The article was drafted by the authors. However, the authors acknowledge and thank Sandra Cuscó, PhD, of Complete Medical Communications, who assisted in formatting the manuscript and submitting it, which was funded by Oncimmune Ltd.
- Received November 11, 2010.
- Revision received March 14, 2011.
- Accepted March 31, 2011.
- ©2011 American Association for Cancer Research.