From 573a8dab35e91f67c9cb9ca04ae9e4e8db192468 Mon Sep 17 00:00:00 2001 From: Duncan Gillespie Date: Tue, 6 Jun 2023 10:02:46 +0100 Subject: [PATCH] review and amend methods for tobacco-alcohol risk interaction --- R/RRFunc.R | 10 ++++---- R/TobAlcInt.R | 6 ++--- vignettes/Methods_for_STAPM_disease_risks.Rmd | 24 +++++++++++++++---- ...moking_and_the_risks_of_adult_diseases.Rmd | 9 ------- vignettes/refs/disease-risks.bib | 19 +++++++++++++++ 5 files changed, 47 insertions(+), 21 deletions(-) diff --git a/R/RRFunc.R b/R/RRFunc.R index 8997202..c030f35 100644 --- a/R/RRFunc.R +++ b/R/RRFunc.R @@ -770,9 +770,9 @@ RRFunc <- function( )] # Combine the tobacco and alcohol relative risks - # accounting for the synergy index - #data[ , (d) := (1 + ((get(d_alc) - 1) + (get(d_tob) - 1)) * get(d_si))] - data[ , (d) := get(d_alc) * get(d_tob) * get(d_si)] # formula changed to align with corrected formula for the synergy index from Prabhu Am J Gastroenterol 2014; 109:822–827; doi: 10.1038/ajg.2014.71 + # accounting for the Synergy Index - interaction of the additive scale + data[ , (d) := 1 + (((get(d_alc) - 1) + (get(d_tob) - 1)) * get(d_si))] + #data[ , (d) := get(d_alc) * get(d_tob) * get(d_si)] # Synergy Factor - multiplicative scale - NOT used - Prabhu Am J Gastroenterol 2014; 109:822–827; doi: 10.1038/ajg.2014.71 data[ , (d_si) := NULL] @@ -780,8 +780,8 @@ RRFunc <- function( # Combine the tobacco and alcohol relative risks # *without* accounting for the synergy index - #data[ , (d) := 1 + ((get(d_alc) - 1) + (get(d_tob) - 1))] - data[ , (d) := get(d_alc) * get(d_tob)] + data[ , (d) := 1 + ((get(d_alc) - 1) + (get(d_tob) - 1))] # additive scale + #data[ , (d) := get(d_alc) * get(d_tob)] # multiplicative scale } diff --git a/R/TobAlcInt.R b/R/TobAlcInt.R index 4a987dd..d75e92a 100644 --- a/R/TobAlcInt.R +++ b/R/TobAlcInt.R @@ -66,15 +66,15 @@ TobAlcInt <- function( "Oesophageal_SCC" ) - # Calculate the synergy index. + # Calculate the Synergy Index. if(disease %fin% str1) { alc1_tob0 <- rr[ , alc1_tob0] alc0_tob1 <- rr[ , alc0_tob1] alc1_tob1 <- rr[ , alc1_tob1] - #si <- (alc1_tob1 - 1) / ((alc1_tob0 - 1) + (alc0_tob1 - 1)) - si <- alc1_tob1 / (alc1_tob0 * alc0_tob1) # corrected to match Prabhu Am J Gastroenterol 2014; 109:822–827; doi: 10.1038/ajg.2014.71; + si <- (alc1_tob1 - 1) / ((alc1_tob0 - 1) + (alc0_tob1 - 1)) + #si <- alc1_tob1 / (alc1_tob0 * alc0_tob1) # Synergy Factor - multiplicative scale - NOT USED - Prabhu Am J Gastroenterol 2014; 109:822–827; doi: 10.1038/ajg.2014.71; } else { diff --git a/vignettes/Methods_for_STAPM_disease_risks.Rmd b/vignettes/Methods_for_STAPM_disease_risks.Rmd index 13e5433..0549bc7 100644 --- a/vignettes/Methods_for_STAPM_disease_risks.Rmd +++ b/vignettes/Methods_for_STAPM_disease_risks.Rmd @@ -237,15 +237,31 @@ RR_{former}=1+[RR_{current}-1][1-p(f)], in which the relative risk ($RR$) associated with current smoking is scaled according to our estimates of the proportional reduction in the risk ($p$) associated with their number of years ($f$) as a former smoker. After someone has been quit for 40 years, we assume their risk reverts back to be the same as a never smoker. # Tobacco–Alcohol interactions in disease risk -See the smoking and disease risks report [@Webster2018] for a short section on tobacco-alcohol risk interactions. +When we talk about tobacco-alcohol interactions in disease risk, what we mean is that we want to account for any additional increase in risk of disease that occurs for people who both smoke and drink alcohol. See VanderWeele and Knol [-@vanderweele2014tutorial] for an explanation of measures of interaction. The rationale for considering this risk interaction is that for some diseases there can be physiological interactions between substances in alcoholic beverages and in tobacco smoke that mean that when both alcohol and tobacco are consumed, then the disease-causing effect of either one is greater. -There is evidence for the risk of disease in someone who consumes both tobacco and alcohol ($RR_{ta}$) being higher than would be expected from combining the independent risks from tobacco ($RR_t$) and alcohol ($RR_a$). This additional risk due to tobacco-alcohol interaction (over and above the combination of the independent contributions to risk from tobacco and alcohol) can be expressed as a "synergy factor" (1) [@Prabhu2014;@Hashibe2009]. +In 2015, we conducted a scoping review across all diseases for which tobacco and alcohol are causal factors to ascertain the extent of evidence on interaction effects. We included only interactions with a meta-analysis of effect size. The only diseases for which we found suitable meta-analyses to inform the tobacco–alcohol interaction in risk was cancers of the oral cavity, pharynx, larynx (for which we used estimated of head and neck cancer overall: alcohol alone 1.06, tobacco alone 2.37, tobacco and alcohol 5.73) [@Hashibe2009], and Squamous Cell Carcinoma (SCC) of the oesophagus (alcohol alone 1.21, tobacco alone 1.36, alcohol and tobacco 3.28) [@Prabhu2014]. Interactions estimated from odds ratios were assumed to be equivalent to interactions estimated from relative risks. + +Prabhu et al. [-@Prabhu2014] reported interaction effects using a metric of multiplicative interaction called the "Synergy Factor". The Synergy Factor gives the factor by which the relative risk of disease in someone who both smokes and drinks ($RR_{ta}$) exceeds the product of the relative risks in someone who only smokes ($RR_{t}$) and someone who only drinks ($RR_{a}$): + +\begin{equation} +\text{Synergy Factor} = \frac{RR_{ta}}{RR_t \times RR_a}. +\end{equation} + +In the STAPM modelling, interaction estimates from the literature are applied on the additive rather than multiplicative scale by estimating a metric called the "Synergy Index" [@rothman1986]. The Synergy Index measures the extent to which the relative risk for someone who both smokes and drinks exceeds 1, and whether this is greater than the sum of the extent to which the relative risk of people who only smoke or only drink exceeds 1: \begin{equation} -SF = \frac{RR_{ta}}{(RR_t - 1) + (RR_a - 1)} +\text{Synergy Index} = \frac{RR_{ta}-1}{(RR_t-1)+(RR_a-1)}. \end{equation} -Data on the synergistic effects of tobacco and alcohol risks are stored in `tobalcepi::tob_alc_risk_int` and applied in modelling by the function `tobalcepi::TobAlcInt()`. +For head and neck cancer overall, the Synergy Index is estimated to be 3.09, and for oesophageal SCC the Synergy Index is 1.99. + +Data on the synergistic effects of tobacco and alcohol risks are stored in the package repository in the spreadsheet `tobalcepi/vignettes/inst/tob_alc_interactions_180119.csv` and as package data in `tobalctobalcepi::tob_alc_risk_int`. + +The interaction effects are applied in the STAPM modelling using the functions `tobalcepi::RRFunc()` and `tobalcepi::TobAlcInt()`. The function `tobalcepi::TobAlcInt()` contains code that calculates the Synergy Index according to the above formula. The function `tobalcepi::RRFunc()` then combines the relative risks for diseases that are related to both tobacco and alcohol using the additive scale formula: + +\begin{equation} +RR_{ta} = 1 + \left[\left((RR_a - 1) + (RR_t - 1)\right) \times \text{Synergy Index}\right]. +\end{equation} # References diff --git a/vignettes/Smoking_and_the_risks_of_adult_diseases.Rmd b/vignettes/Smoking_and_the_risks_of_adult_diseases.Rmd index 4368f0e..a339d5b 100644 --- a/vignettes/Smoking_and_the_risks_of_adult_diseases.Rmd +++ b/vignettes/Smoking_and_the_risks_of_adult_diseases.Rmd @@ -124,15 +124,6 @@ This document presents our list of **52** adult diseases related to smoking and 1. Since publication of the RCP report, Cancer Research UK (CRUK) produced their own disease list and risk sources for cancers attributable to modifiable risk factors, including tobacco and alcohol [@Brown2018]. Discussions with CRUK shaped the disease definitions in our updated Sheffield disease list for alcohol. Where there are differences in the risk sources used in the RCP report and CRUK's work, we take the estimate that matches most closely to our disease definitions, or the more recent estimate. -# Tobacco–Alcohol interactions in disease risk -There is limited evidence for the risk of disease in someone who consumes both tobacco and alcohol ($RR_{ta}$) being higher than would be expected from combining the independent risks from tobacco ($RR_t$) and alcohol ($RRa$). This additional risk due to tobacco-alcohol interaction (over and above the combination of the independent contributions to risk from tobacco and alcohol) can be expressed as a "synergy factor" (1) [@Prabhu2014;@Hashibe2009]. - -\begin{equation} -SF = \frac{RR_{ta}}{(RR_t - 1) + (RR_a - 1)} -\end{equation} - -We conducted a scoping review across all diseases for which tobacco and alcohol are causal factors to ascertain the extent of evidence on interaction effects. We include only interactions with a meta-analysis of effect size. The only diseases for which we have found suitable meta-analyses to inform the tobacco–alcohol interaction in risk are cancers of the oral cavity, pharynx, larynx and oesophagus. To evaluate the interaction between tobacco and alcohol and the risk of cancer of the oral cavity, pharynx and larynx, Hashibe et al. [-@Hashibe2009] conducted a pooled analysis within the INHANCE consortium. For the oral cavity and pharynx, the SF was estimated to be 3.09 (CI 1.82–5.23) i.e. smoking and drinking together causes around a 3-fold increase in the risk of head and neck cancers over and about the independent contributions of tobacco and alcohol. For cancer of the larynx, Hashibe et al. [-@Hashibe2009] estimate the SF to be 1.62 (CI 0.85–3.09). For cancer of the oesophagus, Prabhu et al. [@Prabhu2014] estimated the increase in risk of Squamous Cell Carcinoma in people who both smoke and drink to be characterised by a synergy factor of 1.85 (CI 1.45–2.38). - ```{r smkrelrisks, eval = T, warning = F, out.extra='', fig.pos = "H", echo=F, fig.width = 8, fig.height = 10, fig.cap = "Relative risks and 95 percent confidence intervals in current smokers for 52 conditions attributable to smoking.", cache = F, fig.align = "center"} diff --git a/vignettes/refs/disease-risks.bib b/vignettes/refs/disease-risks.bib index 22c6a31..2ae5033 100644 --- a/vignettes/refs/disease-risks.bib +++ b/vignettes/refs/disease-risks.bib @@ -1,6 +1,25 @@ % Encoding: UTF-8 +@article{vanderweele2014tutorial, + title={A tutorial on interaction}, + author={VanderWeele, Tyler J and Knol, Mirjam J}, + journal={Epidemiologic methods}, + volume={3}, + number={1}, + pages={33--72}, + year={2014}, + publisher={De Gruyter}, + Doi = {10.1515/em-2013-0005} +} + +@book{rothman1986, + author = {Rothman, Kenneth J.}, + title = {Modern epidemiology /}, + publisher = {Little Brown,}, + year = {1986.}, + address = {Boston :} +} @techreport{purshouse2009nice,