Risk of primary malignant cancer after non–melanoma skin cancer in England: a national record linkage study
Running Title: Primary Malignant Cancer after Non-melanoma Skin Cancer
Eugene Ong1 MB ChB, Raph Goldacre1 BA, Uy Hoang1 FFPH, Rodney Sinclair2 MD, Michael Goldacre1 FFPH
1 Unit of Health–Care Epidemiology, Department of Public Health, University of Oxford, Rosemary Rue Building, Oxford, OX3 7LF
2 Department of Dermatology, University of Melbourne, Melbourne, Australia, VIC 3010
Corresponding author:Eugene Ong, University of Oxford Old Road Campus, Oxford OX3 7LF, United Kingdom.
Tel: +44 (0)1865 289377
Fax: +44 (0)1865 289379
Email:
Word Count: 3346 (excluding abstract/references); 27 references; 3 tables; supplementary
Page 1
ABSTRACT
Background
Conflicting evidence exists about whether non–melanoma skin cancer (NMSC) is associated with other malignancies.
Methods
Using hospital admissionand death data in England, we calculated rates of primary cancers before and after NMSC compared with a control cohort.
Results
The rate ratio for all cancers combined after NMSC was 1.30 (95% confidence interval 1.29-1.31). Significantly high RRswere found in 22 cancer sites. Cancer susceptibility was higher in younger people. Analysing NMSC after other cancers yielded similar results.
Conclusions
NMSC is associated with a broad spectrum of primary cancers, particularly at certain sites and in younger age–groups.
[word count abstract 100]
Key Words: NMSC, Association, Non-melanoma skin cancer, Primary Cancer
INTRODUCTION
The evidence about whether other cancers follow NMSC more commonly than by chance alone is inconclusive. Some studies show increased risks of a range of non-UV associated cancers [Wheless et al, 2010] in individuals with NMSC whilst others find risk reductions in certain cancers, which authors suggest to be due to protective effects of increased vitamin D from increased sunlight exposure [Grant et al, 2007].
We aimed to settle the controversy, systematically across a range of cancer sites and ages, using a record–linked dataset covering the whole of England. The dataset was large enough for us meaningfully to stratifyrisks by for cancer site, by age and time periods, with sizes of strata unachievable in previous published studies.
METHODS
Setting and dataset
We used linked health records in England from 1999 to 2011. The dataset contained Hospital Episode Statistics (administrative data routinely collected for each hospital admission that includes all day–case and inpatient admissions in all NHS hospitals) and data on mortality obtained from death registrations. The data were linked such that episodes of care, and any subsequent readmissions or death records, were brought together into a single composite record for each individual.
Subjects
A cohort of people with a record of day-case care or hospital inpatient admission for NMSC was constructed (‘exposure cohort’) for those with a principal diagnosis of NMSC as the reason for care, by identifying individuals’ first episode of day–case care, or hospital admission, for the condition in an NHS hospital during the study period. NMSC was defined using code C44 in the tenth revision of the International Classification of Diseases (ICD). ICD codes used for other cancers are shown in the Tables.
A ‘reference cohort’ was constructed (as a control group) by identifying the first admission for each individual with various other, mainly minor, medical and surgical conditions (see footnote to Table 1). Following standard epidemiological practice, we selected a wide range of different conditions.
Outcomes
We searched the database for any subsequent NHS hospital care for, or death from, cancer in these cohorts to obtain outcomes of subsequent primary cancers (specified in Tables).
Analysis
We calculated cancer rates for each individual cancer based on person–days, taking “date of entry” into each cohort as the date of first admission for NMSC, or condition in the reference cohort, and “date of exit” as the date of first cancer record, death, or the end of the data file, whichever was the earliest.
In comparing the rate of cancers in the NMSC cohort with the reference cohort, both cohorts were stratified by age (in five–year groups), sex, calendar year of first recorded admission, region of residence, and quintile of patients’ Index of Deprivation score (a measure of socio–economic status). The indirect method of standardisation was used, with the NMSC and reference cohortstaken together as the standard population. Rate ratios were calculated by taking the standardised rate of occurrence of cancer in the NMSC cohort relative to the reference cohort using (ONMSC/ENMSC)/(Oref/Eref), where ONMSC and ENMSCare the observed and expected number of cancer cases in the NMSC cohort, and Oref and Erefare the observed and expected number of cases in the reference cohort. The rate ratio confidence interval and 2 statistics for its significance were calculated as described elsewhere [Breslow et al. 1987].
In the above analyses, we excluded anyone with a record of NMSC prior to, or at the same time as, the non–NMSC cancer event (or reference cohort condition). We then reversed procedures by constructing cohorts of people with each non–NMSC cancer, each being an‘exposure cohort’, and following each to identify any subsequent NMSC.
RESULTS
NMSC followed by non–NMSC cancers
There were a total of 502,503 NMSC patients. Table S1 shows the number, age and sex distribution of people at entry to the NMSC cohort.
The rate ratio (RR) of malignant cancers in the NMSC cohort, compared with the reference cohort, across all ages, was 1.30 (95% CI 1.29–1.31). Results for all individual cancers are shown in Table 1. There were increased risks (at p<0.05) in 22 out of 30 cancers. Any differences between males and females were small (Table S2).
Non–NMSC cancers as exposures
There was a similar pattern of risk when analysing risks of NMSC after primary malignant cancers (Table 1).
Results by age
Tables 2 and 3 show associations stratified by age. The associations were generally strongest in younger age groups and declined with increasing age.
Short (<1 year), medium (1–4 years) and long–term (5+ years) associations
For most cancers, the excess risks of developing subsequent cancers decreased a little bit with time after admission but still remained significant at 5 years or more after NMSC admission (Table S3).
Absolute risk
Figure S1 shows the approximate absolute risk of acquiring subsequent primary malignant cancer in the NMSC and reference cohorts. Figures S2–4 compare the two cohorts’absolute risk of acquiring breast and colorectal cancers and rectal cancer specifically. The level of absolute risk at which screening for breast and colorectal cancer is started in the general populationis reached at a younger age in people with NMSC.
DISCUSSION
Strengths, weaknesses and potential biases
With 502,503 NMSC cases, this is the largest study on this topic so far, with high statistical power and precision in analyses subdivided by cancer site and age. To our knowledge, this is the first study to undertake a bi–directional analysis for a full range of primary malignant cancers. The main shortcoming is that we were unable to distinguish SCCs from BCCs because these cancers are coded together in the ICD and we are unable to link our records to histological samples. One type of NMSC might be more strongly associated with increased risks of subsequent primaries, but only subtle differences have been noted in studies that do differentiate SCCs and BCCs [Wheless et al. 2010].
We were unable to adjust for potential confounders like BMI, smoking and UV exposure. However, smoking–related cancers, such as lung, had a rate ratio of less than the average elevation of cancer site risk, and RRs for cancer overall were still significantly high after excluding melanoma from the analyses. This suggests that factors other than smoking and sun damage are at play. The effects of acquired risks, such as smoking and other behavioural factors, are cumulative, and one would expect an increasing relative risk with increasing age if they were major factors behind our associations. We found the opposite (higher RRs at younger ages), suggesting a genetic rather than acquired cause of association.
Our study does not include data from primary care or outpatient departments, only from inpatient admissions and daycase care. It is difficult to estimate accurately what proportion of the total number of recorded NMSC cases during the time period were captured by our study; however, most GPs in England do not operate on suspected NMSCs (NYCRIS, 2001) and most non-inpatient hospital procedures are likely to be recorded as day-case procedures. Surveillance bias might account for an initial increased rate of melanoma diagnosis, but is unlikely to account for the increased rates of other non–visible internal cancers, particularly where the subsequent cancer diagnosis is a long time after diagnosis of NMSC; many cancers showedan increasing RR with time.
Comparison with other studies
A meta–analysis combining three cohort studies [Wheless et al, 2010],accounting for individual level risk factors, like smoking, showed an overall relative risk of cancer after NMSC of 1.49 (1.12–1.98), similar to our RR of 1.30 (1.29–1.31). Our results show a specificity of association for certain cancers not easily explained by conventional risk factors.
Risks of certain cancers have been reported as low in people with NMSC [Grant et al, 2007]. It has been suggested that increased sunlight exposure and vitamin D levels play a protective role in their development. We did not find significantly low risks for these cancers.
Principal findings and Implications
NMSC was associated with an increased risk of a broad spectrum of primary malignant cancers. We highlight the high risks in young people in whom low cancer incidence rates can only be studied in large datasets like ours. The benefits of more precise characterization of those with NMSC who are at risk of specific malignancies would be considerable. Current guidelines do not cover any specific surveillance in NMSC patients other than for skin cancers; our results suggest specific surveillance might be warranted.
The fact that the associations were generally similar in both directions – NMSC before and after other cancers – is also striking. These findings support the hypothesis that shared early life factors cause a bi–directional association between NMSC and other cancers. Further work to elucidate why people with NMSC, particularly the young, are at increased risk of other malignancies could be an important step to a more fundamental understanding of carcinogenesis.
SUPPLEMENTARY INFORMATION
Further detail on methods
The hospital data were supplied by the English national Information Centre for Health and Social Care and mortality data by the Office for National Statistics.We considered that cancer rates in the reference cohort would approximate the general population of the region while allowing for migration in and out of it (migration data for individuals were unavailable).
In standardising, we applied the stratum–specific rates that were found in the standard population to the number of people in each stratum in the NMSC cohort and then, separately, to those in the reference cohort, to obtain the expected number of people with cancer in each stratum of the NMSC and reference cohort. Observed and expected numbers were then summed across all strata to give totals for all strata combined.
In this analysis, people were included in either non–NMSC cancer or reference cohort if they did not have an admission for NMSC before or at the same time as the admission for the exposure or reference cohort condition. In this way no individual was double–counted in the two analyses (i.e. NMSC before other cancer, and other cancer before NMSC); and people with an admission for NMSC and cancer, or reference condition and cancer, at the same time were excluded from the study altogether. All people eligible to be in the reference cohort were included in each cohort analysis (NMSC before and after each of the other cancers), meaning that there were numerically different ratios of reference cohort conditions to exposure cases in different age strata. Our use of all eligible numbers in the reference cohort maximises statistical power (nothing is gained by shedding controls to provide equal numbers in each stratum or cohort).The reference cohort is known from previous cancer association studies outcomes not to include conditions with atypical cancer rates [Fois et al. 2010; Goldacre et al. 2008].
Further Detail on Results
Calculations of absolute risk can only be approximate because we have no migration data and only limited years of follow up. There were 502,503 people in the NMSC cohort; and a primary cancer was known to have developed in 58,498 (11.6%) over a mean period of follow–up of about 5 years. This is an elderly population and therefore one in which large numbers of cancers, over a period of years, can be expected. There were 1,585 people aged under 25 with NMSC of whom 75 (4.7%) developed another primary malignant cancer in the “follow-up” period.
Absolute risk calculation: the interested reader can undertake similar approximations of ‘absolute risk’ for each site and age group using the data in Tables 1, 3 and 4.
Given that only 0.3% of the NMSC cohort were under 25, and that cancer incidence rates are generally very low in this age group, the absolute number of people with subsequent cancers was relatively small. 11 sites were, however, strongly associated with NMSC (p<0.002).
Further discussion
We have not adjusted for multiple comparisons of cancer sites, but, as the majority of rate ratios go in the same direction of excess risk, it is very unlikely indeed that these reflect the play of chance and multiple testing. The ubiquity and height of RRs, by site, of cancer risk associated with NMSC is very unusual and unlikely to be attributable to arterfacts of data collection or study design.
Tranplant
Transplant patients are known to have high risks of NMSC and cancer [Jensen et al, 1999]. Only a small proportion of our NMSC cohort is likely to have a transplant. The risk and pattern of cancers in people who have undergone transplantation, in whom the excess cancer risk particularly affects the kidney, liver and non-Hodgkin’s lymphoma [Engels et al, 2011], cannot fully account for our results.
T–cell lymphomas are usually coded separately, and might occasionally be miscoded as NMSC, but given their rarity compared to SCCs and BCCs this is unlikely to account for any important change in our results.
Reduced RR results
There were significantly reduced risks (p<0.05) of developing subsequent NMSC in five cancers. Most are cancers in which survival is poor, such as lung, pancreas and oesophagus. It is probable that if these patients subsequently developed relatively minor conditions like NMSC that they might not be referred for treatment, and any NMSC might remain unrecorded, resulting in falsely low rate ratios.
Additional References
Engels EA, Pfeiffer RM, Fraumeni JF Jr, Kasiske BL, Israni AK, Snyder JJ, Wolfe RA, Goodrich NP, Bayakly AR, Clarke CA, Copeland G, Finch JL, Fleissner ML, Goodman MT, Kahn A, Koch L, Lynch CF, Madeleine MM, Pawlish K, Rao C, Williams MA, Castenson D, Curry M, Parsons R, Fant G, Lin M (2011) Spectrum of cancer risk among US solid organ transplant recipients. JAMA 306(17):1891-901.
Fois AF, Wotton CJ, Yeates D, Turner MR, Goldacre MJ. Cancer in patients with motor neuron disease, multiple sclerosis and Parkinson's disease: record linkage studies. J Neurol Neurosurg Psychiatry 2010;81:215–21.
Goldacre MJ, Wotton CJ, Yeates D, Seagroatt V, Jewell D. Cancer in patients with ulcerative colitis, Crohn's disease and coeliac disease: record linkage study. Eur J Gastroenterol Hepatol 2008;20:297–304.
Jensen P, Hansen S, Møller B, Leivestad T, Pfeffer P, Geiran O, Fauchald P, Simonsen S. (1999) Skin cancer in kidney and heart transplant recipients and different long-term immunosuppressive therapy regimens. J Am Acad Dermatol 40:177-86.
Acknowledgements
David Yeates wrote the software package used for the analysis. Over many years, the linked data–files were built by Leicester Gill and Matt Davidson, Unit of Health–Care Epidemiology, University of Oxford.
Competing interests
None
Funding
The Unit of Health–Care Epidemiology is funded by the English National Institute for Health Research to analyse the linked data (grant reference number RNC/035/002). The views expressed in this paper do not necessarily reflect those of the funding body.
Ethical approval
Ethical approval for analysis of the record linkage study data was obtained from the Central and South Bristol Multi–Centre Research Ethics Committee (04/Q2006/176).
Author Contributions
EO, MG, UH and RS conceived the idea of the manuscript. RG undertook the data analysis. EO wrote the first draft. All authors contributed to the interpretation of data, reviewed successive drafts and approved the final version of the report.
References
Breslow N, Day N (1987) Statistical methods in cancer research. Volume II—The design and analysis of cohort studies. IARC Sci Publ82: 91 — 97
Grant WB(2007) A meta–analysis of second cancers after a diagnosis of nonmelanoma skin cancer: additional evidence that solar ultraviolet–B irradiance reduces the risk of internal cancers. J Steroid Biochem Mol Biol103:668–74.
NYCRIS (2001) Registration of skin cancer in Yorkshire. Northern & Yorkshire Cancer Registry & Information Service (accessed 20 Feb 2013)
Wheless J, Black J, Alberg AJ (2010) Nonmelanoma skin cancer and the risk of second primary cancers: a systematic review. Cancer Epidemiol Biomarkers Prev19:1686–95.
Table S1 Table 1: Age–distribution of people with NMSC in the study, the percentage who were women, the number of people in the reference cohort in each strata*, and the ratio of people in the reference cohort to each person in the NMSC cohort (reference:NMSC ratio)†
England (1999–Feb 2011)Age at admission to NMSC cohort / Number in NMSC cohort (% of total) / % women / Number in reference cohort / reference:NMSC ratio
0–24 / 1585 (0.3) / 58.0 / 2581158 / 1628.5
25–44 / 22398 (4.5) / 54.8 / 1982738 / 88.5
45–59 / 71655 (14.3) / 48.1 / 1434736 / 20.0
60+ / 406865 (81.0) / 45.5 / 2748818 / 6.8
Total / 502503 (100.0) / 46.3 / 8747450 / 17.4
*Conditions used in reference cohort, with Office of Population, Censuses and Surveys (OPCS) code edition 4 for operations and ICD10 code for diagnosis (with equivalent codes used for other coding editions): adenoidectomy (OPCS4 E20), tonsillectomy (F34+F36), appendectomy (H01–H03), total hip replacement (W37–W39), total knee replacement (W40–W42), cataract (ICD 10 H25), squint (H49–H51), otitis externa/media (H60–H67), varicose veins (I83), haemorrhoids (I84), upper respiratory tract infections (J00–J06), deflected septum, nasal polyp (J33+J34.2), impacted tooth and other disorders of teeth (K00–K03), inguinal hernia (K40), in–growing toenail and other diseases of nail (L60), sebaceous cyst (L72.1), bunion (M20.1), internal derangement of knee (M23), superficial injury and contusion (S00, S10, S20, S30, S40, S50, S60, S70, S80, S90), dislocations, sprains and strains (S03, S13, S23, S33, S43, S53, S63, S73, S83, S93), head injury (S06), selected limb fractures (S42, S52, S62, S82, S92).