The effect of smoking on health-related quality of life in lung cancer patients: a systematic review
- 1Department of Psychology, University of Sheffield, Sheffield, UK
- 2Academic Unit of Clinical Oncology, Weston Park Hospital, Sheffield, UK
- Correspondence to Christine Rowland, Department of Psychology, University of Sheffield, Sheffield S10 2TN, UK;
Introduction Given poor survival rates for lung cancer, health-related quality of life (HRQoL) is very important. Smoking is prevalent among those diagnosed with lung cancer, and continued smoking is associated with compromised HRQoL in other patient groups.
Aims A systematic review was conducted to determine: (i) differences in HRQoL between lung cancer patients who smoke compared with those who quit or never smoked and (ii) changes in HRQoL in patients who continue to smoke after diagnosis compared with those who quit or never smoked.
Method Scopus, Medline, PubMed, PsychINFO and Web of Knowledge from January 1995 to June 2010 were searched. The included studies were assessed and given a score for quality.
Results Eight studies met the inclusion criteria. Four studies showed that lung cancer patients who smoked report impaired HRQoL compared with those who never smoked or had quit. Smokers reported significantly lower HRQoL than former smokers, who in turn reported lower HRQoL than never smokers. This finding remained consistent over time.
Conclusions When taking account of methodological quality, smoking is associated with poorer HRQoL in lung cancer patients. These results suggest that programmes are needed to address the specific support needs of this group and promote HRQoL during their final months. Longitudinal research is necessary to further understand the association between smoking and HRQoL.
Links between smoking and subsequent onset of lung cancer are well established.1 The prevalence of smoking has declined considerably in many countries (from 31% in 1998 to 21% in 2008 (UK)2 and 24% in 19983 to 21% in 2009 (USA)4) but delays between the onset of smoking and diagnosis of lung cancer, and increasing ageing among populations5 mean that lung cancer rates are expected to increase.6
Although 5-year survival rates for many cancers are encouraging (eg, prostate 99% USA,7 71% UK;8 breast 89% USA,7 81% UK8), the figures for lung cancer are much lower. The 5-year survival rate in the USA is 16%7 and just 7% in the UK.8 Many lung cancer patients are diagnosed at an advanced stage7 so that treatment is necessarily palliative. Invasive therapy may result in side effects including alopecia, nausea and vomiting, dysphagia, anorexia, breathlessness, and pain.9 Against this, modest benefits to survival must be balanced against the impact of treatments on symptom palliation and health-related quality of life (HRQoL). Elderly cancer patients and those nearing the end of life often regard HRQoL as more important than striving for increased survival time compared with younger patients or those in the earlier stages of their illness.10–12
HRQoL is often described as being multidimensional, incorporating psychological, social, physical, functional and spiritual domains.13–15 HRQoL assessment offers insight into well-being from the patient's perspective and provides a more comprehensive assessment than reliance on measures of physical function alone. Clinicians can overestimate patients’ HRQoL leading to recommendations that patient views are routinely assessed.16 ,17
A basic distinction has been made between generic and disease-specific measures of HRQoL. Generic measures (eg, the Medical Outcomes Study 36-item Short Form (SF-36)18 ,19) can be used to assess HRQoL in different patient groups and in ‘healthy’ populations. Alongside global HRQoL scores, generic measures include summary scores indicating physical or mental well-being. In contrast, disease-specific measures focus on issues relating to a particular condition and provide a greater degree of sensitivity (eg, European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire-C3020). The EORTC has also published site-specific cancer modules (eg, LC13 for lung21). A modular approach such as this offers a broad assessment of HRQoL alongside disease-specific coverage.
Symptom-specific scales focus directly on the symptoms associated with a particular illness (eg, Lung Cancer Symptom Scale (LCSS)22 ,23). Reliance on symptom scales alone can result in restricted assessment of HRQoL and does not allow comparison with other patient groups.
Two reviews have considered HRQoL in relation to lung cacner;24 ,25 Montazeri et al24 reported a general review which included literature published from 1970 to 1995 and concluded there was a need to regularly assess HRQoL using validated tools. Furthermore, it was suggested that family/support networks may have a substantial impact on HRQoL.24 Parsons et al25 reported a systematic review and meta-analysis addressing the effects of smoking cessation after diagnosis of lung cancer. Regardless of stage or tumour histology, smoking cessation improved prognostic outcomes in terms of mortality (HR 2.94; 95% CI 1.15 to 7.54), disease recurrence (HR 1.86; CI 1.01 to 3.41) and development of a second primary tumour (HR 4.31; CI 1.09 to 16.98). Neither review addressed whether smoking was associated with HRQoL.
Negative relationships between smoking and HRQoL have been found for patients with obstructive airway disease,26 asthma,27 head and neck cancer,28 ,29 and HIV.30 Considering the importance of smoking in the development of lung cancer, and of HRQoL to patients, little is known about the implications of continued smoking for HRQoL in these patients. Our pilot review confirmed the findings of Montazeri et al24 that no literature before 1995 addressed the relationship between smoking and HRQoL in lung cancer patients. We therefore conducted a systematic review of the literature from January 1995 to June 2010 with the aims to determine:
Differences in HRQoL between lung cancer patients who smoke compared with those who quit or never smoked.
Longitudinal effects on HRQoL in patients who continue to smoke after lung cancer diagnosis compared with those who quit smoking or never smoked.
We searched Scopus, Medline, PubMed, PsychINFO, and Web of Knowledge and Theses.com. The following search terms were combined with Boolean operators: ‘lung cancer’, ‘quality of life’, ‘smok*[ing][ers][er][e]’ and ‘tobacco’ (eg, ‘lung cancer’ AND ‘quality of life’ AND ‘smok*’). Searches were conducted in Web of Knowledge for the following specific HRQoL measures: ‘FACT-L’, ‘EORTC-QLQ-C30’, ‘EORTC-QLQ-LC13’ and ‘LCSS’. The search was limited to English language papers. The ‘in press’ web pages of each of the following journals were also searched: Lung Cancer, Psycho-Oncology and Chest. Two reviewers independently assessed titles and abstracts and retrieved articles suitable for inclusion. A final list of papers to be included was discussed and agreed without requiring a third party. Where a conference abstract was identified, the primary author was contacted for further details. Additional studies were identified by checking references and citation records of papers included in the review.
Inclusion and exclusion criteria
Studies were included which involved adults with a lung cancer diagnosis, reported primary research, commented directly on the link between smoking and HRQoL, and used validated HRQoL tools.
Exclusion criteria were studies which used only performance status measures, reviews and case studies.
We used the National Health Service Public Health Resource Unit Critical Appraisal Skills Programme (CASP) appraisal tool for cohort studies31 as a guide for appraising study quality. This outlines 12 questions that examine the focus of the study, appropriateness of methodology, and validity and reliability of results. A quality score was derived, identifying those studies which merited greater influence when drawing conclusions. Each question on the CASP guide was worth one point if clearly addressed or half a point if there was ambiguity (maximum score of 12).
Data extraction and synthesis
Each study that met the inclusion criteria was assessed independently by two reviewers (CR and SO). Information extracted from each study included: demographic characteristics, details of measurement tools, classification of smoking status, time since diagnosis, statistical results and overall findings. The reviewers achieved high inter-rater reliability (κ=0.88) and disagreement was resolved through discussion.
Study selection process
A total of 1838 records were identified, which included 754 citations from electronic databases and 1084 using other previously described methods. After removing duplicates and assessing relevance from the title, the abstracts of 67 articles were assessed for suitability. Full text articles were obtained for 20 papers, and further information was obtained from the principal author of one conference abstract (figure 1). Eight studies met the inclusion criteria32–39 and are summarised in tables 1⇓–3. They included seven journal articles, and one conference abstract with supplementary information from the author.
Characteristics of included studies
The majority of studies were conducted in the USA,32 ,34 ,35 ,37–39 plus one each from Sweden33 and India.36 Sample size ranged from 5137 to 101934 participants. Gender ratios were approximately even (ranging from 45.8%32 to 57%33 male subjects), although one sample comprised women only,35 and another included 87% male subjects.36 Five samples were comprised entirely of non-small cell lung cancer patients,32 ,33 ,35 ,38 ,39 three of which included disease stages I–III32 ,33 ,38 exclusively. Small cell lung cancer patients were included in three studies and comprised between 8%34 and 21.6%37 of the sample. The association between smoking and HRQoL was a primary objective in two studies34 ,39 and a secondary objective in six (table 1).
Time from diagnosis to assessment ranged from two weeks34 to over five years.39 Five studies were cross-sectional in design32–34 ,36 ,37 and three were longitudinal cohort studies (table 3).35 ,38 ,39
A variety of HRQoL measures were used. Four studies used generic measures, SF-3632 ,33 ,35 and WHO-QOL (BREF),36 and four used disease-specific measures or symptom scales, Functional Assessment of Cancer Therapy-Lung instrument (FACT-L),37 LCSS34 ,37 ,39 and EORTC-QLQ-C30+LC13.38 One study32 used both a generic (SF-36) and disease-specific questionnaire for cancer survivors (QOL-Survivors40–42). Only three studies reported a measure of HRQoL at diagnosis (table 3).36–38
Using the CASP appraisal tool,31 all included studies were considered to be methodologically sound. Quality scores ranged from 738 to 10 (table 3).39 Potential barriers to quality were small sample sizes so that a number of studies were possibly underpowered.32 ,33 ,35 ,38 The four papers that received the highest quality scores33 ,36 ,37 ,39 were given more weight in our conclusions.
The majority of studies organised smoking behaviour into at least three categories: ‘current’, ‘former’ and ‘never’ smokers. These categorisations were not usually further defined but one study identified never smokers as less than 100 cigarettes in their lifetime.34
Garces et al34 included assessment of pack years (PY) and packs-per-day (PPD). Mohan et al36 also classified smoking history by PY. Browning and colleagues37 included data from smokers only (table 2). Two studies32 ,35 employed biochemical verification of smoking status to combat the response bias that can occur with self-reporting.
Effect of smoking on HRQoL
Differences depending on smoking status
No significant difference in overall HRQoL dependent on smoker status was found in four studies.32 ,35 ,36 ,38 A trend was observed for those who had smoked the longest (assessed in PY) to have a lower HRQoL than those who had smoked for a shorter period, or who had never smoked.36 Garces et al,34 who found significant differences in HRQoL by smoking history, also reported a non-significant trend of decreased HRQoL for increased PPDs among a small group of persistent smokers.
Significant differences in HRQoL among smokers, former smokers and never smokers were reported in two studies33 ,34 with never smokers having the highest and smokers having the lowest HRQoL. Those who quit smoking prior to HRQoL assessment and did not resume smoking had HRQoL scores which were more similar to those of never smokers than current smokers.34 The sample in the seventh study comprised only smokers37 and the authors compared their findings with those of Garces et al.34 They found a substantially lower HRQoL between smokers in their sample and the HRQoL of all participants (smokers and non-smokers) reported by Garces et al, although statistical comparisons were not made.
There was evidence that some dimensions of HRQoL assessed by subscales and summary scores differed by smoking status. Myrdal and colleagues found that those who continued to smoke had significantly lower scores on the mental components summary of the SF-36 than former or never smokers.33 Sarna et al38 who reported no significant differences in global HRQoL by smoking status found that smokers reported significantly higher scores on the pain subscale than former or never smokers.
Longitudinal effects on HRQoL by smoking status
Only one study addressed this aim.39 Current smokers reported significantly worse HRQoL than former smokers who in turn reported significantly worse HRQoL than never smokers at two time points (<3 and >5 years since diagnosis). HRQoL declined significantly over time regardless of smoking status but the rate of decline was not different between groups (table 3).39
Although no studies were excluded on the basis of our quality assessment, we identified four33 ,36 ,37 ,39 that were rated of higher quality and were therefore given greater emphasis when drawing conclusions. There is evidence suggesting that HRQOL may be worse for those who smoked33 ,37 and continue to smoke after diagnosis39 compared with former or never smokers.
Of all studies reviewed, three provided evidence that lung cancer patients who smoke report impaired HRQoL compared with those who never smoked or had quit smoking.33 ,34 ,37 This was also apparent in longitudinal work.39
Four studies32 ,35 ,36 ,38 found no significant difference in HRQoL between smoker groups. Of these four papers, only the work of Mohan et al36 was identified as of higher quality and given greater weight. Evidence of a negative association between HRQoL and continued smoking has been shown to exist with other patient groups.26–30 While this association may have been assumed to exist for lung cancer patients, this review demonstrates that while the balance of evidence supports this assumption, the evidence base could be stronger.
Strengths and weaknesses
As with any systematic review the strengths of this work include the transparent search criteria, coverage of a variety of electronic databases, and citation and reference searching.
Although we excluded texts that were not in English language, most of the other language records identified did not meet our inclusion criteria.
Limitations with this review include the minimal volume and heterogeneity of the research identified which compromise our ability to reach definitive conclusions.
It is important to acknowledge the difficulties of conducting research with this patient group (advanced disease stage, older age, compromised HRQoL due to treatments and symptoms), and particularly the limitations for longitudinal work imposed by survival time. These difficulties may account to some extent for the limited volume of work identified. Based on the quality appraisal, we concluded that the work was generally methodologically sound. However, we noted the literature would benefit from more longitudinal research, precise measurement of smoking status, larger samples to provide greater statistical power, more representative samples and the selection of validated HRQoL tools.
The majority of work reviewed was cross-sectional. It is essential to identify the longer-term consequences of smoking on HRQoL and factors that influence changes in smoking behaviour given the implications for planning patient care.
Measurement of smoking status tended to be simplistic. Just two studies considered smoker behaviour beyond categorisations and quantified cigarette use (PPD/PY).34 ,36 Simple categorisations of smoker behaviour cannot readily accommodate changes in smoking behaviour nor account for differences between recent former smokers and those who quit before diagnosis. Studies may be subject to bias where patients describe their smoking behaviour inaccurately. Only two of the eight studies32 ,35 employed biochemical verification of patients’ reported smoking.
Critically, the majority of studies included only a small number of people who continued to smoke, making potential differences between smoker groups difficult to detect statistically.
Samples may not be representative of lung cancer patients as those with advanced disease appear under-represented. Two studies32 ,33 did not include any patients with stage IIIb or IV non-small cell lung cancer, and a number included only a small proportion with this staging.34 ,35 ,38 ,39 Not all samples included small cell lung cancer patients, highlighting the need for research that is more reflective of the general population.
Finally, many different HRQoL measures were used, adding to the difficulty in drawing conclusions across studies. Three generic tools (QOL-Survivor,40–42 WHO QOL,43 SF-3618 ,19), two disease-specific tools with additional lung cancer modules (FACT-L,44 EORTC-QLQ-C30+LC1320 ,21) and one symptom scale (LCSS22 ,23) were used. A modular approach to HRQoL assessment which allows for coverage of a variety of HRQoL domains with the incorporation of a disease-specific symptom scale was used in two studies.37 ,38 Sarna et al38 provided no data on the impact of smoking across HRQoL domains or for disease-specific items, but a narrative summary was given for the one significant finding reported. Browning et al37 did provide scores for all dimensions of the FACT-L but their sample was limited to smokers only and so comparisons between smoker groups were not possible (they additionally used the LCSS from which scores were compared with those from Garces et al34). Studies using a symptom scale34 ,37 reported scores for each symptom as well as global HRQoL. The additional dimensional/symptom-specific detail should be reported as it is of potential interest to clinicians.
In quitting smoking after diagnosis, lung cancer patients may gain a psychological boost; feeling a sense of control and perhaps that they are contributing to their battle against cancer.45 These factors may explain why smoking cessation might be related positively to HRQoL. However, further studies are necessary to elucidate the relative contributions of smoking in the context of other factors, such as socio-economic status or pre-existing conditions such as depression for HRQoL.
Healthy smokers with a previous history of depression who quit smoking are significantly more likely to experience a new occurrence of major depression.46 In a recent review of smoking and depression, it was shown that depression increases the risk of smoking, and smoking increases the risk of depression.47 Clinicians and smoking cessation counsellors should be vigilant for any evidence of mental health comorbidity in lung cancer patients who smoke in order to prevent a cycle of mental health problems and continued smoking.
Additionally, lung cancer patients may feel stigmatised or guilty about their contribution to the onset of disease, and believe that others blame them for their illness.48 ,49 Family members sometimes attribute blame to the patient for causing their disease. Zhang and Siminoff50 found family carers reported frustration and anger about patients’ persistent smoking. Family members might resent care roles and some professionals may offer less sympathy and support than they otherwise might.
It is important to establish whether the link between continued smoking and compromised HRQoL is evident in samples more representative of the lung cancer population. Critical too are the implications for medical treatment, and for any guilt and stigmatisation experienced by the patient. If continued smoking does impact negatively on HRQoL there may be implications for the role of family in the smoking choices of patients. Importantly, it may be possible to use this information at the time of diagnosis within the context of a ‘teachable moment’51 in order to promote smoking cessation among family members.
The studies reported in this review suggest a negative relationship between smoking and HRQoL in lung cancer patients. Despite the potential importance of this area we only identified eight studies that addressed smoking and HRQoL. With many lung cancers diagnosed at an advanced stage HRQoL is highly valued by patients and smoking cessation may be one way in which HRQoL could be improved. There may be clinical implications of continued smoking after lung cancer diagnosis with links to mental health problems such as depression, and implications for the role of family in smoking choices.
Research which clarifies the relationship between smoking and HRQoL is necessary in order to inform clinical care of these patients and provide recommendations for smoking cessation that are appropriate and acceptable to lung cancer patients.
We are grateful to Sheffield Hospitals Charitable Trust who funded this work. We also acknowledge Sam Otero for her contribution to data extraction and quality assessment.
Contributors CR: developed the review protocol, conducted data search and extraction, conducted quality review, interpreted data, drafted and edited article and approved the final submitted version; CE, RR, SD: contributed to the development of the review protocol, interpretation of data, drafting and editing article. They supervised the review process and approved the final submitted version.
Funding This work was funded by Sheffield Hospitals Charitable Trust (grant number: 080904). The funder remained separate to the research process and the development of this submission.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data.
Disclaimer The views expressed in this publication are those of the authors and have not been influenced by the funding body.
- Received 21 December 2011.
- Accepted 14 August 2012.
- Published Online First 11 October 2012
- Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions