Predictive value of image cytometry for diagnosis of lung cancer in heavy smokers

Study population

In total, 2,889 smokers interested in the RIDTELC Lung Study in the period from May 2000 to April 2002 contacted the study administrator by telephone after announcement of the study in the local newspapers and on local radio stations. Heavy smokers with a cumulative cigarette consumption of ≥30 pack-yrs aged 50–74 yrs were eligible for the study. After excluding individuals either out of the age and exposition range or with a history of cancer in the last 5 yrs, a total of 2,480 voluntary suitable participants were recruited from the general population in the Ruhr industrial area around Bochum (n = 2,422) and a smaller group from the urban region of Copenhagen (Denmark) (n = 58). After giving written informed consent, each participant completed a questionnaire providing demographic data, pertinent medical history, smoking history, socio-economic information and occupational exposure history to harmful agents. The list of occupational exposures, adopted from Jöckel et al. 16, included 28 occupational groups and asked for educational levels. Special emphasis was given to asbestos exposure and working in the mining industry.

After lung function testing, sputum was collected following induction with hypertonic saline solution, and a digital chest radiograph investigation was performed. All data were collected by an independent study monitor and archived on a password-protected PC with access for the study monitor only. The study design was approved by the ethics committee of the Augusta Teaching Hospital (Bochum, Germany).

Chest radiograph screening

All participants had posterior–anterior and lateral digital chest roentgenograms (Thoravision with Selen-detector-head, super 50 CPD generator, 50 kW, 150 kV, frame 43×49 cm; Phillips, Aachen, Germany), which were read by a radiologist. The results of these radiographs were interpreted as: 1) no abnormality detected; 2) abnormalities not suggestive of malignant tumour; or 3) suspected malignant tumour. All abnormal chest films were re-examined by a pulmonologist. All participants with suspected malignant tumours were called for computer tomography (CT) examination.

Sputum collection

Under the supervision of a respiratory therapist, all participants underwent sputum induction by inhalation of aerosolised 3% saline with the addition of salbutamol (25 mg into 20 mL saline) for 20 min using an ultrasonic nebuliser (Aneasthoschall 880; HICO, Cologne, Germany) or a jet nebuliser (Pari Sole, Starnberg, Germany). A group of 67 (0.3%) participants who did not produce sufficient sputum after induction were asked to collect additional spontaneous sputum on the following 3 days. Sputum was collected in 50-mL labelled centrifuge tubes, filled with 25 mL of saccomanno preservative, with the addition of 0.1% dithiothreitol (DTT).

Slide preparation

The liquefied sputum specimens in the saccomanno–DTT solution were centrifuged at 500×g for 15 min. The supernatant was decanted and the cell pellets were resuspended in 0.5–2.5 mL of saccomanno solution, according to weight. Lastly, two drops of the cell suspension were placed onto each of six slides and smeared with capillary forces using a separate glass slide. Out of the six air-dried slides, two were stained according to the methods of Papanicolaou 17 for conventional cytology, two were stained according to the modified Feulgen reaction for DNA image cytometry 18 and the remaining two were stored for future reference.

Conventional sputum cytology

Sputum cytology examination and DNA image cytometry were independently performed without the knowledge of clinical results. The cytology slides were sent to the Institute of Cytology (Hannover, Germany) to be screened by cytotechnologists and cytopathologists from the institute. The results, presented according to a modified classification of Papanicolaou 19, were sent directly to the study administrator.

DNA image cytometry

In the present study, image cytometry was automatically carried out using the Cyto-Savant system (Cyto-Savant®; Oncometrics, Vancouver, Canada) adopted for sputum specimens 9. The feulgen–thionin-stained slides were used to quantify nuclear DNA content. After DNA staining, the nuclear-integrated optical density is the cytometric equivalent of its DNA content. The technology was adapted to the recommendations of the European Society for Analytical Cellular Pathology (ESACP) Task Force on standardisation of diagnostic DNA image cytometry 20. For each slide, almost 2,000 exfoliated cells were collected, which included: 1,000 epithelial cells; up to 200 suspicious nuclei with a DNA-index >1,125 and <2.5; up to 100 highly suspicious nuclei with a DNA index >2.5 times the normal DNA content; ∼200 each of lymphocytes, polymorphonuclear and eosinophil granulocytes; and 100 alveolar macrophages. Before measurement, the cell galleries were reviewed to eliminate the nuclei that were blurred or overlapping. The normal 2c reference value was established by measuring the normal well-preserved lymphocytes in each slide as internal reference cells, whose co-efficient of variation was <12.5% (sd/mean). Hence, the DNA content of normal epithelial nuclei, termed as euploidy, was 2±0.25c; a cell was defined as aneuploid if its DNA content deviated from 2±0.25c or the integer-valued exponents of 2c (e.g. 4±0.5c, 8±1c, 16±2c).

Two parameters were applied to describe the DNA content of each cell: the 5c exceeding rate (5cER) and the 2c deviation index (2cDI). 5cER was defined as the percentage of aneuploid cells having a DNA content >5c. 2cDI represented the mean squared deviation from the diploid value or the variance around the 2c peak. It was calculated as the sum of squares of the differences between the DNA values of single cells (ci) and the 2c value, divided by the number of measured cells (n), as follows 21: A third parameter, the malignancy grade (MG) was calculated from the 2cDI and 5cER using a modified formula as follows: This was used to determine the cytometric diagnosis as follows. 1) Grade I: benign, normal nuclear structures, euploidy, MG <0.10; 2) Grade II: suspicious samples with MG ≥0.10. 3) Grade 0: nonrepresentative because of inadequate diagnostic cells.

Bronchoscopy and pathology examination

All patients with severe dysplasia or higher grade changes seen with conventional sputum cytology, or grade II at image cytometry were called for bronchoscopy. Conventional white-light bronchoscopy (WLB) fluorescence was performed on subjects, followed by light-induced fluorescence endoscopy (LIFE). The findings were reported as normal or suspicious for tumour.

During bronchoscopy, biopsy specimens were taken from all areas suspicious of dysplasia or worse on WLB or LIFE examination, and at least from one side of the normal mucosa. Two to six biopsies were taken from every subject. The histopathological type and grade of diagnosis were determined according to the World Health Organisation (WHO) classification. From the resulting series of biopsies in each case, the most advanced degree of histological change with respect to lung malignancy was used as the final diagnosis in the analysis.

Diagnostic procedure

At the beginning of the study, all participants underwent spirometry, sputum induction and chest radiographical examination. For participants with a suspicious radiograph, further examinations, such as CT of the chest and bronchoscopy, were performed to obtain a final diagnosis. All patients with severe dysplasia or higher grade changes on cytological examination or grade II with DNA image cytometry were also called for bronchoscopy (conventional and autofluorescence).

Data analysis

The distribution of continuous variables was examined with respect to skewness and kurtosis. The unpaired t-test, nonparametric test (Kruskal-Wallis or the Mann-Whitney U-test) and Pearson's Chi-squared test were employed to determine the difference of results from various variables between different groups. All statistical tests were two-sided, and statistical significance was determined by a p-value <0.05. The receiver operating characteristic (ROC) curves for MG and 5cER were plotted to determine the suitable threshold and to estimate the efficiency of image cytometry relating to the “gold standard” clinically relevant lung cancers, which were defined as histopathologically confirmed severe dysplasia and/or higher grade lesions. The comparison between the areas under the ROC curves for MG and 5cER were made using the method of Delong et al. 22. The mean values for 2cDI, 5cER and MG were calculated along with the 95% confidence intervals (CI) for grade I and II image cytometry results, as well as for histologically classified normal, pre-invasive and invasive biopsies.

The study population characteristics in terms of DNA image cytometry diagnosis

The demographic and clinical characteristics of the study population are summarised in table 1⇓. In total, 2,480 smokers (71% males) were included, (mean±sd) age 59.6±5.9 yrs and 58.6±25 pack-yrs, who started smoking at age 17.7±3.4 yrs. The majority of the study population had no history of respiratory disease. Nearly half of the subjects had a family history of cancer and occupational exposure to harmful agents.

Subjects with suspicious cytometric results smoked more heavily than those with normal cytometric results. More subjects in the group with suspicious results had moderate or strong cough and harmful agents exposure history. It is worth mentioning that >80% of all subjects with moderate cough had >1-yr cough history. For the remaining features listed in the table, there were no statistically significant differences between the groups.

Initial automated sputum cytometry, conventional sputum cytology and chest radiograph screening findings

Out of 2,480 participants, 2,461 provided adequate sputum for automated sputum cytology and 2,449 for conventional sputum cytology. All 2,480 subjects underwent chest radiographical examination. To date, a total of 27 cancers (representing a prevalence of 1.1%) have been found, 20 cancer patients had grade II automated sputum cytology results and one had a positive conventiona sputum cytology result. Low-grade lesions (metaplasias, and mild and moderate dysplasias) were found in two out of the 24 cancer patients receiving cytological examination (table 2⇓).

In total, 176 subjects had grade II results by image cytometry, of which 20 were diagnosed as cancers, representing a sensitivity of 80%. One tumour was diagnosed in patients with positive cytology, representing a sensitivity of 4.2%. Out of 2,480 participants, 138 (5.6%) were suspect for lung cancer by chest radiographical examination, of which 17 lung cancers were diagnosed. Out of the 138 subjects with suspect results on chest films, 127 (92%) underwent a CT examination, 76 were considered to be suspect for malignant lesions by CT, of which 11 were finally diagnosed as cancers, the other six tumours were diagnosed by other diagnostic methods such as a sonography operation.

In total, cancer was detected in 27 patients by screening, 10 patients had both chest radiograph and sputum positive results, 10 had only cytometry suspicious results, one had both cytological and cytometry positive results, and six had only chest radiography suspect results.

Findings of bronchoscopy

The bronchoscopic findings for tumours were classified as normal or suspect. In total, 95 subjects who had grade II results detected by image cytometry underwent bronchoscopy (out of those six also had positive sputum cytology results). Seven were confirmed as clinically relevant lung cancer by pathological analysis. Out of six participants who had positive results on sputum cytology, one was confirmed by pathological examination. A total of 201 biopsies were taken, of which 49 had suspect results (table 3⇓). No severe dysplasias were diagnosed by pathology in biopsies with normal results. Using the authors' gold standard, there were no false–negative results using the bronchoscopy procedure, taking endoscopically abnormal mucosa and high-grade alterations in histopathology as the cut-off points. From biopsies which were found to be suspect for tumour under bronchoscopy, seven out of 40 patients were diagnosed with clinically relevant lung cancer and one was diagnosed with a benign tumour. The seven patients included three with severe dysplasia, two with carcinoma in situ, one with stage Ib and one with stage III lung cancer (table 3⇓). Out of the participants, 41 (43.2%) had low-grade lesions (27 metaplasias, nine mild dysplasias and five moderate dysplasias).

DNA image cytometric results and histopathological diagnosis

According to the WHO/International System for Staging Lung Cancer Histological Classification of Lung and Pleural Tumours 3, all histological diagnoses were classified into three main categories: 1) normal (benign); 2) pre-invasive lesions, including squamous dysplasia (mild, moderate and severe) and carcinoma in situ; and 3) malignant or higher grade lesions than carcinoma in situ.

Among the 176 cases with suspicious (grade II) cytometric results, 99 had histological results from biopsy by either bronchoscopy or surgery. The results of 5cER and 2cDI, as well as MG, of 99 participants with pathological results were plotted against the diagnoses obtained from the histopathological results in figure 1⇓. There was an association between the different phases in the development of malignancy and the value of 5cER, the median value of which increased from 0.034 to 0.051 and 0.070% with the progression from histological normal through pre-invasive to invasive tumour changes (Kruskal-Wallis test p = 0.019; Mann-Whitney U-test p = 0.021 for the group with histological normal and pre-invasive results; and p = 0.023 for the group with normal and invasive results). The median of MG significantly increased from 0.121 to 0.153 with the progression from histological normal to pre-invasive tumour changes (Mann-Whitney U-test p = 0.004). Nevertheless, a significant association between the progression of histopathological alterations and 2cDI was not found (Kruskal-Wallis Test p = 0.21).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Results of 5c exceeding rate (□), 2c deviation index () and malignancy grade (▪) versus histopathological diagnosis in normal (n = 66), pre-invasive (n = 17) and invasive (n = 16) participants. The lower and upper box boundaries represent the 25th and 75th percentiles. The horizontal lines under and above the box represent the 10th and 90th percentiles, respectively. The median value is shown as the horizontal line within the box. •: outliers.

ROC curves for MG and 5cER

To estimate the diagnostic accuracy of automated image cytometry using MG, or 5cER as decision parameters in predicting lung cancer, the optimum cut-off points were first derived for the two parameters. As the decision cut-off point of the test varied, the sensitivity and specificity of the test also changed. These values are plotted and joined to produce the ROC curve, in which the Y axis represents sensitivity (the proportion of true–positive results), the X axis represents 1–specificity, and the area under the curve indicates the accuracy of the test with different cut-off points. The optimum cut-off levels of MG of 0.10 and 5cER of 0.03 were determined by selecting the point on the ROC curve which maximised both sensitivity and 1–specificity. The accuracy of automated sputum cytology in predicting clinically relevant cancers was 87% (95% CI: 77.6–96.5) using MG as the decision parameter and 81.2% (95% CI: 69.9–92.5) using 5cER as the decision parameter, which is significantly lower than using MG (p<0.05; fig. 2⇓).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Receiver operator characteristics curve for malignancy grade (MG; ––) and 5c exceeding rate (5cER; – – –). Area under the curve for MG = 0.87 (0.776–0.965). The value of MG represented at the point on the curve closest to the upper left-hand corner was 0.10. At this level its sensitivity and specificity were 80 and 93.6%, respectively. Area under the curve for 5cER = 0.812 (0.699–0.925).

Histological type and stage of lesions found in the study versus DNA image cytometric results

Out of the 25 patients with cancers detected, one had no cytological results because of inadequate material. The main histological types of the cancers found in this study were adenocarcinoma and squamous cell cancer. All nine patients with squamous cell lung cancer and six with later stage adenocarcinoma had suspicious or grade II cytometric results. Only one out of four with a stage I lesion had suspicious cytometric results. Out of all cancer patients, only two had abnormal results by conventional sputum cytology (table 4⇓). There were four tumours with other histological types which are listed in table 4⇓, including one undifferentiated carcinoma, one metastasis of a nonlung cancer and two without pathological type characterisation. Among these cases, only one without a pathological result had a negative result on image cytometry.