Long-term Lung Cancer Risk Associated with Sputum Atypia: A 27-Year Follow-up Study of an Occupational Lung Screening Cohort in Yunnan, China

Lung cancer is currently the leading cause of cancer-related death both in China and around the world (1, 2). Results of the National Lung Cancer Screening Trial (NLST) demonstrated a 20% reduced mortality due to lung cancer with low-dose computed tomography (LDCT) screening among heavy smokers compared with chest radiography (3). Subsequently, several randomized controlled trials have reported a mortality reduction for lung cancer with LDCT screening (4–6).

However, LDCT screening still raises some questions. Precisely identifying a population in high-risk of developing lung cancer could improve the ratio between the benefits and harms of lung cancer screening (7, 8), and several studies have reported that an individual's screening history might be helpful for risk stratification (9–11). Furthermore, data from NLST also suggested that the effectiveness of LDCT screening might vary according to histology, and the LDCT benefit was not observed for squamous carcinoma and small cell lung cancer (SCLC), which are mainly located in central airways that were insensitive for LDCT screening (12, 13).

Sputum cytology is a noninvasive test for lung cancer, especially central-airway tumors. Early trials did not observe a mortality reduction of lung cancer from screening with chest X-ray and sputum cytology (14, 15). Accordingly, sputum cytology is currently not recommended for lung cancer screening. However, a combined mortality analysis from these trials showed that a modest benefit might have been present (16). Besides, including our previous study, several cohort studies found that sputum atypia detected 4 or more years before diagnosis was associated with increased lung cancer risk (17–19), and it was also used for identifying a candidate for lung cancer screening or chemoprevention trials (20, 21). Usually, sputum cytologic screening was conducted periodically, and previous study found that the association between sputum atypia and lung cancer incidence was greatest for those sputum samples collected 5 months or less before the diagnosis of lung cancer (17). However, few studies report the magnitude and temporal change of lung cancer risk according to different sputum cytologic results in previous screening rounds, especially in occupational population. Besides, there was still controversy about this risk in relation to histology (17, 19). In this study, we focus on the long-term lung cancer risk of sputum atypia based on the extended follow-up of an occupational screening cohort in Yunnan, China.

In 1992, to establish a biological specimen bank for the identification and validation of markers of early lung cancer, a dynamic prospective cohort study among tin miners was initiated in the Yunnan Tin Corporation (YTC). A total of 9,295 tin miners ≥ 40 years old who had 10 or more years of underground radon exposure and/or arsenic exposure were enrolled from 1992 to 1998. Detailed information on inclusion criteria was described previously (19, 22). Written informed consent was obtained for each participant. This study was conducted in accordance with Declaration of Helsinki, and was approved by the institutional review board of Cancer Hospital/Institute of Chinese Academy of Medical Sciences (201812190401002).

From 1992 to 1999, 8 rounds of annual lung cancer screening were conducted with chest radiograph and sputum cytology in YTC (no participant was enrolled in 1999). Of 9,295 participants, 9,094 received at least 1 sputum screening. Detailed information on chest radiograph and sputum cytology can be seen elsewhere (22). For sputum screening, induced or 3-day pooled spontaneous sputum specimens were collected. Sputum specimens were stored at room temperature in Saccomanno solution and were smeared on glass slides and stained with Pap for cytologic examination. Slides were reviewed independently by 2 well-trained cytopathologists. Sputum cytologic screening result was classified into degrees as follows: (i) 1: negative, (ii) 2–1: slight atypical metaplasia, which was considered to result from inflammatory or irritative process, but not from cancer, (iii) 2–2: moderate atypical metaplasia, which indicated a cytologic abnormality sufficient to repeat examination and surveillance, (iv) 2–3: grave atypical metaplasia, which implied that changes were suspicious for cancer, but not diagnostic, either because the abnormal cells were too few or the abnormalities were not sufficiently well defined, (v) 3: suspicious for cancer, (vi) 4: highly suspicious for cancer, (vii) 5–1: squamous carcinoma, (viii) 5–2: adenocarcinoma, (ix) 5–3: undifferentiated carcinoma (includes both large cell and small cell), and (x) other cancers (19, 23). In this study, we used moderate or worse atypia as a primary measure for long-term lung cancer risk.

At the baseline interview, data were collected with a standardized questionnaire, including demographic characteristics, work history, tobacco consumption, and previous medical history. Individuals who had smoked cigarettes and/or pipes (water or long-stem pipes) regularly for 6 months or longer were defined as smokers, while those who had a smoking duration of less than 6 months were considered nonsmokers (24). Pack-years were calculated as the cumulative exposure to cigarettes, pipes (1 g pipes = 1 cigarette), and total tobacco exposure. The cumulative radon and arsenic exposure for each participant were calculated by summing across the estimated working-level months (WLM) and index of arsenic-exposure months (IAEM) for each job held at the YTC before the date of enrollment, respectively (22). Moreover, exposure information was annually updated from 1992 to 1996. Thus, the cumulative radon and arsenic exposure were calculated to the end of 1996 for participants who entered into study before 1996. In this study, occupational radon and arsenic exposure were grouped into four quartiles (Q1 to Q4) based on each individual's cumulative radon or arsenic levels, respectively. For tobacco smoking, we presumed that the smoking status since 1996 was unchanged, since the rate of smoking cessation were as low as to 5.8%.

From 1992 to 1999, annual follow-up was conducted combined with screening. During the postscreening period after 1999, the first follow up was performed in 2005 and 2006. In 2019, an extended follow up was conducted, and the end date of this follow up was December 31, 2018. By the end of this extended follow up, 204 participants (2.2%) were lost to follow up, with a follow-up rate of 97.8%.

Lung cancer cases were confirmed via the following ways: (i) screen-detected cases with chest radiograph and/or sputum cytologic examination, (ii) interval case with negative screening results identified by hospital due to symptoms during the time from 1992 to 1999, and (iii) cases confirmed by the YTC cancer registry system, which was established in 1973 and received information of all YTC cancers from medical record system and the local hospital.

Two sets of statistical analyses were performed (Fig. 1). First, the analysis was based on the baseline sputum cytologic results, and person-years of follow up were calculated from the date of enrollment to the date of lung cancer diagnosis or censoring as of December 31, 2018 (whichever came first). Lung cancer incidence and incidence-rate ratios according to personal characteristics and sputum cytologic results were also calculated. The joint effects of sputum atypia and other factors including age, smoking, radon, and arsenic exposure were also analyzed.

Second, since a large proportion of sputum atypias were detected in annual screening rounds after baseline screening, we restricted the analysis in participants who received baseline (T0) and 3 subsequent annual (T1-T3) sputum screenings with the aim to further explore the magnitude and temporal change of lung cancer risk according to different sputum cytologic results in previous screening rounds, and person-years of follow up were calculated from the date of T3 to the date of lung cancer diagnosis or censoring as of December 31, 2018. The lung cancer risks associated with moderate or worse cytologic result for at least once, only in the first 2 rounds (T0 or T1), only in the last 2 rounds (T2 or T3), or in both T0-T1 and T2-T3 were compared with those with all negative results in T0-T4 rounds. The association between sputum atypia and lung cancer-risk was analyzed with time-varying covariate Cox regression model since the proportional hazards assumption was violated based on the results of the Schoenfeld residuals test. In the time-varying covariate Cox regression model, a sputum cytologic result*log of time, i.e., ln (t), was added. To control the confounding effect from the changes in various kinds of exposure during the long-term follow up, age, cumulative exposure of radon, arsenic and smoking (for current smokers), years since last exposure of radon, arsenic and smoking (for former smokers) were adjusted as time-varying covariate.

The effect of sputum cytologic result on lung cancer risk was also analyzed according to the different intervals of the follow-up period. In consideration of the increased risk of death from a cause other than lung cancer accompanied by aging, competing-risks regression analysis was conducted as a sensitivity analysis (25, 26). Finally, E value was calculated to assess the magnitude of the potential residual confounding (27). E value was defined as the minimum strength of association on the risk-ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome, conditional on the measured covariates, to fully explain away a specific exposure–outcome association. Statistical analysis was performed using Stata 14.0 software.

The personal and occupational characteristic–specific lung cancer incidence of 9,084 participants who received at least 1 sputum cytological screening is shown in Table 1. Most participants were male (93.3%), and over 50% were 50 years old or older. Most of participants were smokers (84.9%), and had radon and arsenic exposure. After an extended median follow-up of 18.3 years, 1,326 lung cancer cases were confirmed. Of these lung cancers, 1,294 had a definite diagnosis date, with an overall lung cancer incidence of 780.28/10⁵. Significant higher lung cancer incidence-rate ratios were observed in male smokers, and those who had lower education levels, higher occupational radon/arsenic exposure, and previous lung disease. The incidence-rate ratio of participants who had moderate or worse atypia compared with those with negative results was 4.52 [95% confidence interval (CI): 3.42–5.88].

Table 2 showed the association of baseline sputum cytologic results and lung cancer risk. After adjusting for other potential risk factors listed in Table 1, moderate or worse baseline cytologic result was associated with a significantly increased lung cancer risk, with adjusted hazard ratio HR of 6.62 (95% CI: 4.66–9.42), and this relative hazard significantly decreased with time. The association between baseline moderate or worse atypia and risk of lung cancer by tumor histology was also analyzed. As shown in Table 2, moderate or worse atypia was associated with a moderately and continuously increased risk of SCLC (adjusted HR = 3.04, 95% CI: 1.74–5.33), but not with adenocarcinoma. The association was substantially stronger for squamous cell lung cancer (adjusted HR = 13.68, 95% CI, 8.86–21.13) and was decreased significantly with time.

The joint effects of baseline sputum atypia and other risk factors on lung cancer risk are presented in Table 3. Compared with participants ≤ 60 years old who had negative cytologic results, the adjusted HRs for lung cancer associated with moderate or worse atypia were 13.68 (95% CI: 7.04–26.58) and 18.62 (95% CI: 10.78–32.15) for participants ≤ 60 or > 60 years old, respectively. Both of them demonstrated a significantly decreasing trend with time. Similarly, the lung cancer risks associated with moderate or worse sputum atypia in participants with higher level smoking, radon, and arsenic exposure were a little higher than in those participants with lower exposures.

A total of 4,269 participants received the first 4 consecutive rounds of sputum screening (T0-T3). Compared with negative screening results, the adjusted HR for at least one moderate or worse cytologic atypia was 7.27 (95% CI: 4.67–11.32). Besides, the adjusted HRs for moderate or worse cytologic atypia only in T0-T1 and only in T2-T3 were 3.23 (95% CI: 1.55–6.73) and 9.48 (95% CI: 5.68–15.82) respectively, while the adjusted HR for moderate or worse cytologic atypia in both T0-T1 and T2-T3 was as high as 37.44 (95% CI: 15.00–93.47). These results suggested that more recent or consistent cytologic results were associated with higher lung cancer risk, and also showed a significant decreasing trend over time.

To directly demonstrate the temporal pattern of lung cancer risk related to sputum atypia, the adjusted HRs in different time intervals during the follow up were estimated (Table 5). Significantly increased lung cancer risks concerning at least one moderate or worse cytologic atypia were observed in the first 10 years since follow up, with the adjusted HRs of 3.11 (95% CI: 2.37–4.07) and 3.25 (95% CI: 2.33–4.54) for baseline and the first 4 consecutive screening rounds respectively. Besides, the association was stronger among those who had recent or consistent moderate or worse atypia in the first 4 consecutive rounds of sputum screening with adjusted HRs of 4.14 (95% CI: 2.70–6.35) and 17.55 (95% CI: 8.32–37.03) respectively.

We also used competing-risks regression model to evaluate the association between sputum atypia and lung cancer risk, and the results are shown in Supplementary Table S1. No significant differences were observed between the results from the time-varying covariate Cox model (Tables 2 and 4,Table 5) and those from the competing-risks regression model.

In this study, the E values were 5.67 (CI: 4.17) and 5.95 (CI: 4.09) for the 10-year lung cancer risk following moderate or worse sputum atypia at baseline screening or at least one moderate or worse sputum atypia in the first 4 consecutive rounds respectively. These results suggested that no significant residual confounding exist in this study.

In this prospective study, although there was a decreasing trend, an up to 10 years increase in the risk of lung cancer associated with moderate or worse sputum atypia was observed. This association was stronger for recent and persistent atypia, and in terms of histology, for squamous carcinoma and SCLC.

Similar to our previous study and other studies, this extended follow-up study continues to confirm the association between increased lung cancer risk and sputum atypia (17–19 28, 29). Three hypotheses might explain this result. First, long-term, widespread exposure to smoking, occupational radon and arsenic, or other risk factors for lung cancer might lead to field cancerization (30). However, significantly increased lung cancer in relation to risk of baseline moderate or worse sputum atypia was also observed in participants who had relatively young age, lower exposure of smoking, radon and arsenic, which implied that field cancerization could not be fully responsible for the association between sputum atypia and increased lung cancer risk. Second, the high lung cancer risk among older individuals, those who had high levels of smoking, radon, and arsenic exposure implied a positive interaction between sputum atypia and other carcinogens. Finally, in addition to sputum atypia as an independent risk factor, exfoliated abnormal cells of lung cancer might also contribute to the stronger association between sputum atypia detected in more recent screening rounds and lung cancer risk (17).

Stepwise progression to invasive cancer was reported for both adenocarcinoma and squamous carcinoma. However, we did not find significantly increased risk of adenocarcinoma following sputum atypia. The main reason might be that abnormal cells that are exfoliated into the sputum are located more commonly in the central airways than in the periphery of the lung. Accordingly, preinvasive lesions of squamous carcinoma might be more easily to be detected by sputum cytologic examination, which resulted in a stronger association between sputum atypia and risk of squamous carcinoma compared with that for adenocarcinoma. Several studies have indicated that squamous cell carcinoma is developed in a step-wise pattern where the epithelium changes from normal to hyperplasia, metaplasia, mild, moderate, and severe dysplasia and then carcinoma in situ (31, 32). Besides, it is generally accepted that high-grade lesions are more likely to progress to invasive cancer than low-grade lesions (33). Similarly, in this study, 70.5% (31/44) of baseline sputum atypia worse than moderate progressed to lung cancer with a significantly higher frequency than that of moderate atypia. High-grade lesions might also progress more rapidly to invasive cancer compared with low-grade lesions. Compared with moderate atypia, more high-grade atypias might progress to squamous carcinoma in a relative shorter period, meanwhile, fewer high-grade atypias would regress to normal or low-grade level. Thus, more lung cancers would appear during the first few years of follow up. On the other hand, relatively fewer lung cancers would be found over time due to the regression of some moderate atypia during the later years of follow up. In an earlier long-term follow up study, of 14 participants with severe atypia, 6 (42.9%) were found to have lung cancer during the 9-year follow up, and most were found during the first year. In contrast, of 169 participants with moderate atypia, only 18 (10.6%) developed lung cancer during the follow up, and most of these occurred during the 6 to 9 years of follow up (31). High, rapid progression rate of high-grade atypia and low, slow progression rate of moderate atypia might be the reason for the decreasing trend of the association between sputum atypia and increased lung cancer risk during the long-term follow up. However, other studies found no significant difference in progression rate between individuals with or without severe dysplasia, and progression of metaplasia to invasive cancer within a relatively short time has also been reported, which challenged the concept of step-wise progression of preinvasive bronchial lesions to invasive cancer (34, 35).

Unlike, adenocarcinoma and squamous carcinoma, SCLC is believed to arise from severely molecularly-damaged epithelium without going through recognizable preneoplastic changes (36). However, in line with our previous study (19), we still observed a persistently increased risk of SCLC concerning sputum atypia. In contrast, the results of the Colorado study demonstrated a nonsignificant reduced risk of SCLC associated with sputum atypia (17). Molecular studies might help to elucidate the true relationship between sputum atypia and the risk of SCLC.

Sputum cytology was not recommended for lung cancer screening in most counties. However, controversy still exists for its effectiveness in lung cancer screening. In Japan, sputum cytology was still recommended for lung cancer screening (37). In clinical practice, sputum cytology is also the routine examination for other lung diseases, such as chronic cough, asthma in China (38). Molecular analysis of histopathologic grading might be helpful for a better understanding of the natural history of preinvasive cancers (39). For example, progression of bronchial dysplasia was reported to be associated with specific immune alterations (40). Characterizing the immune microenvironment of bronchial dysplasia will allow for a better understanding of progressive lesions of the central airway and advance the field of precision chemoprevention and lung cancer risk stratification.

To our knowledge, this study is the longest long-term evaluation of lung cancer risk in relation to sputum atypia. It is the first to demonstrate a decreasing trend of lung cancer risk associated with sputum atypia over time. However, significantly increased lung cancer risk could still be observed for up to 10 years since follow-up. The dynamic changes of lung cancer risk following sputum atypia might contribute to a refinement of selection criteria for lung cancer screening or chemoprevention trials (21). Besides, based on the results of NLST, noncalcified nodules identified on LDCT screening were predictive of lung cancer risk up to 10 or more years following the screen (41). In theory, LDCT and sputum are complementary, since the former is more efficient to detect peripheral lesion, while the latter is relatively more efficient for central lesions. Based on the results of these two studies, it is natural that noncalcified nodule and sputum atypia might also be complementary in high-risk identification and surveillance, especially given the time-varying effect of these two kinds of precursors of lung cancer risk. Additionally, biomarkers in the sputum would allow further refinement of screening selection criteria and discrimination of indeterminate pulmonary nodules (42).

There were some strengths to this study. Firstly, for the prospective cohort design, information about other risk factors was obtained before lung cancer occurrence. Secondly, through an extended follow up with a low rate of loss to follow up, the number of lung cancer cases doubled compared with our previous study, increasing the power of analysis. Meanwhile, this extended follow up allowed us to assess not only the magnitude but also the temporal pattern of the lung cancer risk in relation to sputum atypia. Finally, the robustness of the association between sputum atypia and lung cancer risk was confirmed by the sensitivity analysis including competing-risks regression model, joint effect analysis and the assessment of impact of residual confounding. However, limitations of this study should also be considered. The first one was that most participants were exposed to smoking, occupational radon, and arsenic exposure. Therefore, residual confounding might not have been fully eliminated. The second limitation was that the histology information is lacking for nearly half of lung cancer cases, which would decrease the statistical power when the analysis was conducted according to histology. In addition, the number of lung cancer cases in some subgroups according to other exposures was small, which also led to a reduced statistical power.

In conclusion, this study confirmed the long-term increase of lung cancer risk in individuals with moderate or worse sputum atypia. Sputum atypia might play a complementary role in identifying high-risk individuals for lung cancer screening, surveillance, or lung cancer chemoprevention. Besides, the risk decreased over time, especially for squamous cell lung cancer. Molecular sputum analysis is warranted to gain insight into the natural history of bronchial preinvasive lesions and to further quantify lung cancer risk.

The sponsor had no role in the design of the study, the collection and analysis of the 301 data, or the preparation of the manuscript.

No disclosures were reported.

Y. Fan: Conceptualization, formal analysis, funding acquisition, writing–original draft. Z. Su: Funding acquisition, investigation, methodology, writing–review and editing. M. Wei: Investigation, writing–review and editing. H. Liang: Funding acquisition, writing–review and editing. Y. Jiang: Project administration, writing–review and editing. X. Li: Funding acquisition, writing–review and editing. Z. Meng: Funding acquisition, writing–review and editing. Y. Wang: Writing–review and editing. H. Pan: Writing–review and editing. J. Song: Funding acquisition, writing–review and editing. Y. Qiao: Conceptualization, supervision, project administration, writing–review and editing. Q. Zhou: Conceptualization, supervision, writing–review and editing.

This work was supported by Cancer Foundation of China (grant number: CFC2020KYXM001, CFC2020KYXM002 to Y. Fan). This study was also partly funded from Tianjin Natural Science Foundation (grant number: 17JCYBJC25400 to Y. Fan, 18JCYBJC92100 to X. Li), NIH (grant number K01 1K01TW011190–01A1 to J. Song), Key R & D projects of Science and Technology Department of Sichuan (grant number 2020YFS0212 to H. Liang); National Natural Science Foundation of China (grant number 81971650 to Z. Meng). We gratefully acknowledge all participants who have participated in this study. We thank the staff of Office of Gejiu Municipal Leading Group for Cancer Prevention and Control, Gejiu City, Yunnan, China for their assistance in the collection of the follow-up data of the YTC screening cohort. We thank Dr. Nasra Mohamoud Ali (a native English speaker) for providing assistance in editing this manuscript.

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.