Skip to main content
Log in

Combined effect of tobacco smoking and alcohol drinking in the risk of head and neck cancers: a re-analysis of case–control studies using bi-dimensional spline models

  • Published:
European Journal of Epidemiology Aims and scope Submit manuscript


The synergistic effect of tobacco smoking and alcohol consumption on the risk of head and neck cancers has been mainly investigated as a cross-product of categorical exposure, thus leading to loss of information. We propose a bi-dimensional logistic spline model to investigate the interacting dose–response relationship of two continuous exposures (i.e., ethanol intake and tobacco smoking) on the risk of head and neck cancers, representing results through three-dimensional graphs. This model was applied to a pool of hospital-based case–control studies on head and neck cancers conducted in Italy and in the Vaud Swiss Canton between 1982 and 2000, including 1569 cases and 3147 controls. Among never drinkers and for all levels of ethanol intake, the risk of head and neck cancers steeply increased with increasing smoking intensity, starting from 1 cigarette/day. The risk associated to ethanol intake increased with incrementing exposure among smokers, and a threshold effect at approximately 50 g/day emerged among never smokers. Compared to abstainers from both tobacco and alcohol consumption, the combined exposure to ethanol and/or cigarettes led to a steep increase of cancer risk up to a 35-fold higher risk (95 % confidence interval 27.30–43.61) among people consuming 84 g/day of ethanol and 10 cigarettes/day. The highest risk was observed at the highest levels of alcohol and tobacco consumption. Our findings confirmed a combined effect of tobacco smoking and alcohol drinking on head and neck cancers risk, providing evidence that bi-dimensional spline models could be a feasible and flexible method to explore the pattern of risks associated to two interacting continuous-exposure variables.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save

Springer+ Basic
EUR 32.99 /Month
  • Get 10 units per month
  • Download Article/Chapter or Ebook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2

Similar content being viewed by others


  1. IARC. IARC Monographs on the evaluation of carcinogenic risk of chemicals to humans. Alcohol drinking. IARC Sci Publ No. 44. Lyon: IARC. 1988.

  2. IARC. IARC Monographs on the evaluation of carcinogenic risks to Humans. Vol. 83: Tobacco smoke and involuntary smoking. Lyon: IARC Press. 2004.

  3. Hashibe M, Brennan P, Chuang SC, et al. Interaction between tobacco and alcohol use and the risk of head and neck cancer: pooled analysis in the International Head and Neck Cancer Epidemiology Consortium. Cancer Epidemiol Biomarker Prev. 2007;18:541–50.

    Article  Google Scholar 

  4. Zhao LP, Kolonel LN. Efficiency loss from categorizing quantitative exposures into qualitative exposures in case-control studies. Am J Epidemiol. 1992;136:464–74.

    CAS  PubMed  Google Scholar 

  5. Weinberg CR. How bad is categorization? Epidemiology. 1995;6:345–7.

    Article  CAS  PubMed  Google Scholar 

  6. Greenland S. Dose–response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology. 1995;6:356–65.

    Article  CAS  PubMed  Google Scholar 

  7. Desquilbet L, Mariotti F. Dose–response analyses using restricted cubic spline functions in public health research. Stat Med. 2010; doi:10.1002/sim.3841.

  8. Polesel J, Dal Maso L, Bagnardi V, et al. Estimating dose–response relationship between ethanol and risk of cancer using regression spline models. Int J Cancer. 2005;114:836–41.

    Article  CAS  PubMed  Google Scholar 

  9. Polesel J, Talamini R, La Vecchia C, et al. Tobacco smoking and the risk of upper aero-digestive tract cancers: a reanalysis of case-control study using spline models. Int J Cancer. 2008;122:2398–402.

    Article  CAS  PubMed  Google Scholar 

  10. Franceschi S, Talamini R, Barra S, et al. Smoking and drinking in relation to cancers of the oral cavity, pharynx, larynx, and esophagus in northern Italy. Cancer Res. 1990;50:6502–7.

    CAS  PubMed  Google Scholar 

  11. Levi F, Pasche C, La Vecchia C, et al. Food groups and risk of oral and pharyngeal cancer. Int J Cancer. 1998;77:705–9.

    Article  CAS  PubMed  Google Scholar 

  12. Franceschi S, Levi F, Dal Maso L, et al. Cessation of alcohol drinking and risk of cancer of the oral cavity and pharynx. Int J Cancer. 2000;85:787–90.

    Article  CAS  PubMed  Google Scholar 

  13. Talamini R, Vaccarella S, Barbone F, et al. Oral hygiene, dentition, sexual habits and risk of oral cancer. Br J Cancer. 2000;83:1238–42.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Talamini R, Bosetti C, La Vecchia C, et al. Combined effect of tobacco and alcohol on laryngeal cancer risk: a case–control study. Cancer Causes Control. 2002;13:957–64.

    Article  CAS  PubMed  Google Scholar 

  15. D’Avanzo B, La Vecchia C, Katsouyanni K, et al. Reliability of information on cigarette smoking and beverage consumption provided by hospital controls. Epidemiology. 1996;7:312–5.

    Article  PubMed  Google Scholar 

  16. Ferraroni M, Decarli A, Franceschi S, et al. Validity and reproducibility of alcohol consumption in Italy. Int J Epidemiol. 1996;25:775–82.

    Article  CAS  PubMed  Google Scholar 

  17. Ruppert D, Wand MP, Carroll RJ. Semiparametric regression. Cambridge: Cambridge University Press; 2003.

    Book  Google Scholar 

  18. Akaike H. Information theory and an extension of the maximum likelihood principle. In: Second international symposium on information theory. Budapest. 1973:267–281.

  19. Barón AE, Franceschi S, Barra S, et al. A comparison of the join effects of alcohol and smoking on the risk of cancers across sites in the upper aerodigestive tract. Cancer Epidemiol Biomarker Prev. 1993;2:519–23.

    Google Scholar 

  20. Rothman KJ, Greenland S, Lash TL. Modern epidemiology. Philadelphia: Lippincott Williams & Wilkins; 2008.

    Google Scholar 

  21. Seitz HK, Stickel F, Homann N. Pathogenetic mechanisms of upper aerodigestive tract cancer in alcoholics. Int J Cancer. 2004;108:483–7.

    Article  CAS  PubMed  Google Scholar 

  22. Chiolero A, Wietlisbach V, Ruffieux C, et al. Clustering of risk behaviors with cigarette consumption: a population-based survey. Prev Med. 2006;42:348–53.

    Article  PubMed  Google Scholar 

Download references


This work was supported by grant from the Italian Association for Cancer Research and the Swiss League and Swiss Research against Cancer [Grants Number KFS-700 and OCS-1633]. Dr. Di Maso’s work was partially supported by a grant from Fondazione Umberto Veronesi. The authors wish to thank Drs. Renato Talamini, Silvia Franceschi, Luigi Barzan, Eva Negri, Paola Zambon, and Prof. Fabio Barbone for their help in study coordination, and Mrs Luigina Mei for editorial assistance.

Conflict of interest

The authors declare that they have no conflict of interests.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Jerry Polesel.

Appendix: univariate logistic spline model

Appendix: univariate logistic spline model

Let y = [y 1,…, y n ]′ be a n × 1 vector of realizations of a random variable Y with binomial distribution so that π = Pr{Y = 1} and x = [x 1,…, x n ]′ be a n × 1 vector of related realizations of a continuous predictor X of π. The additive logistic regression model:

$$\log {\text{it}}\left( \pi \right) = \log \left[ {\frac{\pi }{1 - \pi }} \right] = \alpha + f\left( X \right)$$

where f is an arbitrary spline function, i.e. a piecewise polynomial truncated function defined as:

$$f\left( X \right) = \beta_{0} + \mathop \sum \limits_{d = 1}^{D} \beta_{1,d} X^{d} + \mathop \sum \limits_{k = 1}^{K} \beta_{2,k} \left( {X - t_{k} } \right)_{ + }^{D}$$

where D is the degree of polynomial, t k is the \(k\text{th}\) knot location, K is the number of knots, and

$$\left( {X - t_{k} } \right)_{ + }^{D} = \left\{ {\begin{array}{ll} \left( {X - t_{k } } \right)^{D}, &{\text{if}} \quad X \ge t_{k} \\ 0, &{\text{otherwise}} \\ \end{array} } \right.$$

indicates a value of zero for negative values of the argument. The intercept α is absorbed into f and the continuity on knots is guaranteed by contrasts (·)+. The log-likelihood function of the additive logistic regression model as:

figure a

where \(\beta = \left[ {\beta_{0} ,\beta_{1,1} , \ldots ,\beta_{1,D} ,\beta_{2,1} , \ldots ,\beta_{2,K} } \right]^{\prime}\) is the (D + K + 1) × 1 vector of parameters and \(\overset{\lower0.2em\hbox{$\smash{\scriptscriptstyle\smile}$}}{x}_{i} = \left[ {1, x_{i} ,x_{i}^{2} , \cdots ,x_{i}^{D} ,\left( {x_{i} - t_{1} } \right)_{ + }^{D} , \cdots ,\left( {x_{i} - t_{K} } \right)_{ + }^{D} } \right]^{{^{\prime} }}\) is the (D + K + 1) × 1 vector of covariates for the \(i\text{th}\) observation.

The maximum likelihood estimator of parameters \(\widehat{\beta }\) can be found solving:

figure b

The pointwise (1 − α) confidence band is \(\widehat{\beta }X \pm z_{{\left( {1 - {\raise0.7ex\hbox{$\alpha $} \!\mathord{\left/ {\vphantom {\alpha 2}}\right.\kern-0pt} \!\lower0.7ex\hbox{$2$}}} \right)}} \left( {XVX^{\prime} } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-0pt} \!\lower0.7ex\hbox{$2$}}}}\)where

figure c

is a (D + K + 1) × (D + K + 1) covariance matrix.

Bi-dimensional logistic spline model

Let x = [x 1,…, x n ]′ and z = [z 1,…, z n ]′ be two n × 1 vector of related realizations of two continuous predictors X and Z. The corresponding spline function:

$$\begin{aligned} f(X,Z)& = \beta _{0} + \sum\limits_{{d_{{x = 1}} }}^{{D_{x} }} {\beta _{{1,d_{x} }} } X^{{d_{x} }} + \sum\limits_{{k_{x} = 1}}^{{K_{x} }} {\beta _{{2,k_{x} }} } (X - t_{{k_{x} }} )_{ + }^{{D_{x} }} + \sum\limits_{{d_{z} = 1}}^{{D_{z} }} {\beta _{{3,d_{z} }} Z^{{d_{z} }} } \hfill \\ &\quad + \sum\limits_{{k_{z} = 1}}^{{K_{z} }} {\beta _{{4,k_{z} }} } (Z - t_{{k_{z} }} )_{ + }^{{D_{Z} }} + \sum\limits_{{d_{{x = 1}} }}^{{D_{x} }} {} \sum\limits_{{d_{{z = 1}} }}^{{D_{z} }} {\beta _{{5,d_{x} ,d_{z} }} X^{{d_{x} }} Z^{{d_{z} }} } \hfill \\ &\quad + \sum\limits_{{d_{{z = 1}} }}^{{D_{z} }} {} \sum\limits_{{k_{{x = 1}} }}^{{K_{x} }} {\beta _{{6,d_{z} ,k_{x} }} (X - t_{{k_{x} }} } )_{ + }^{{D_{x} }} Z^{{d_{z} }} \hfill \\ &\quad + \sum\limits_{{d_{{x = 1}} }}^{{D_{x} }} {} \sum\limits_{{k_{{z = 1}} }}^{{K_{z} }} {\beta _{{7,d_{x} ,k_{z} }} X^{{d_{x} }} (z - t_{{k_{z} }} } )_{ + }^{{D_{Z} }} \hfill \\ &\quad + \sum\limits_{{k_{{x = 1}} }}^{{K_{x} }} {} \sum\limits_{{k_{{z = 1}} }}^{{K_{z} }} {\beta _{{8,k_{x} ,k_{z} }} (X - t_{{k_{x} }} } )_{ + }^{{D_{x} }}(Z - t_{{k_{z} }} )_{ + }^{{D_{z} }} \hfill \\ \end{aligned}$$

where, D x and D z are the degrees of the polynomials, \(t_{{k_{x} }}\) and \(t_{{k_{z} }}\) are the \(k_{x}\text{th}\) and \(k_{z}\text {th}\) knot location, K x and K z are the number of knots in the distribution, respectively, of X and Z continuous predictors,

$$\left( {X - t_{{k_{x} }} } \right)_{ + }^{{D_{x} }} = \left\{ {\begin{array}{ll} {(X - t_{{k_{x} }} )^{{D_{x} }} ,} &\quad {{\text {if}}\quad X \ge t_{{k_{x} }} } \\ 0 &\quad {\text {otherwise}} \\ \end{array} } \right.$$


$$\left( {Z - t_{{k_{z} }} } \right)_{ + }^{{D_{z} }} = \left\{ {\begin{array}{ll} {(Z - t_{{k_{z} }} )^{{D_{z} }} ,} &\quad {{\text{if}}\quad Z \ge t_{{k_{z} }} } \\ 0 &\quad {\text{otherwise}} \\ \end{array} } \right.$$

indicate the zero value for negative values of the argument for X and Z respectively.

The parameters and the corresponding confidence intervals will be estimated through the methodology described above.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Dal Maso, L., Torelli, N., Biancotto, E. et al. Combined effect of tobacco smoking and alcohol drinking in the risk of head and neck cancers: a re-analysis of case–control studies using bi-dimensional spline models. Eur J Epidemiol 31, 385–393 (2016).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: