This update follows the review and analysis protocol that were published previously, with a few minor amendments. Firstly, the systematic review was conducted according to the preferred reporting items for systematic review and meta-analysis (PRISMA) 2020 checklist (rather than the 2009 version) ¹⁹ . Secondly, the original methodology covered 13 categories of nicotine products (chewing tobacco, combustible cigarettes, cigarillos, cigars, cut tobacco, dipping tobacco, electronic cigarettes, heat-not-burn devices, NRT, non-tobacco pouches, snus, water pipe tobacco and western pipe tobacco). This review adds two categories for bidis and smokeless products from the rest of the world, thereby expanding the selection criteria compared with the 2020 iteration, as well as focusing the dipping and chewing tobacco categories on U.S. varieties only. The full Population-Intervention-Comparison-Outcomes-Study type (PICOS) question addressed in this review is outlined in Table 1.

Systematic searches were conducted using specific search terms pertaining to the health risks of nicotine products on May 9 ^th 2022 in the MEDLINE ( Pubmed) and NIH clinical trials ( ClinicalTrials.gov) databases (see extended data). Due to the broad scope of the searches, the most relevant literature was targeted by searching at the title and abstract levels. The publication lists returned by the searches were exported and screened at the title, abstract and full-text levels according to pre-defined inclusion and exclusion criteria ( Table 2). The screening steps were completed by one researcher and confirmed by a second. During each screening step, the reason for exclusion of each individual publication was recorded and the level of evidence assessed for the final shortlisted publications using the Oxford evidence-based medicine level of evidence scale ²⁰ .

The shortlisted studies were analyzed in detail to extract health risk data and relevant meta-data, including the nicotine product, brand, disease/symptom, methodology used, measurement and unit, error in measurement, significance of measurement, geographic location, sample size and conflict of interest.

For the toxin emissions/content studies, only data points for Group 1 International Agency for Research on Cancer (IARC) carcinogens were extracted. Data were sought for NNN and NNK, formaldehyde, 2-amino-naphthalene, 4-aminobiphenyl, benzo(a)pyrene, 1,3-butadiene, benzene, vinyl chloride, ethylene oxide, arsenic, chromium-IV and cadmium for inhalable products, and NNN and NNK, benzo(a)pyrene, arsenic, chromium-IV and cadmium for ingestible products. In addition, studies that reported sample collection conducted before 1 ^st January 1990 were excluded due to lack of relevance to the current market.

For the epidemiological studies, only adjusted odds ratios pertaining to tobacco-related diseases were extracted and only data points that adjusted for smoking were included. Data were sought for myocardial infarction, stroke, cardiovascular disease, coronary heart disease, atrial fibrillation, heart disease, asthma, asthma attack, bronchitis, wheeze, COPD, CHD mortality, cardiovascular disease mortality, chronic lower respiratory disease, cerebrovascular disease mortality and chronic obstructive pulmonary disease mortality for the non-cancer outcomes. Data were sought for oral cancer, oropharyngeal cancer, mouth cancer, lip cancer, tongue cancer, cancers of the upper aero-digestive tract, head and neck cancer, larynx cancer, esophageal cancer, hypopharyngeal cancer, lung cancer, pancreatic cancer, stomach cancer, bladder cancer, rectal cancer, gastrointestinal cancer, colorectal cancer, anal cancer, liver cancer, cardia cancer, kidney cancer, cervical cancer, non-Hodgkin’s lymphoma, acute lymphoblastic lymphoma, acute myeloid lymphoma, chronic myeloid lymphoma multiple myeloma and cancer mortality for the cancer outcomes. Where stratification was available, data points for current users were prioritized over former/ever users, data points for both sexes were prioritized over single sex datasets (within the same study), and non- or never users of any other nicotine product were prioritized over dual users. If stratification was provided by consumption level, moderate consumption matching as closely as possible to the assumptions used in the LCR analysis were selected. In applying these additional selection criteria to the data, we aim to adjust for confounders, reduce the risk of bias and focus on risk associated with the same consumption levels across all analyses. No automation tools were used for data extraction and all data points were verified by a second reviewer. The information extracted included the nicotine product, brand, disease/symptom, methodology used, measurement and unit, error in measurement, significance of measurement, geographic location, sample size and conflict of interest.

The extracted and harmonized dataset was analyzed in two main segments: lifetime cancer risk and epidemiological data.

Lifetime cancer risk . The lifetime cancer risk (LCR) of each nicotine product was calculated separately for inhalable and smokeless products. The inhalable products LCR calculation was based on the methodology outlined by Stephens ⁹ , whereas the LCR of the smokeless products was determined using the methodology outlined by the FDA ¹⁰ .

The LCR of each inhalable nicotine product was calculated from the toxin emissions data by adjusting the OEHHA unit risk values ¹² . A total of 12 toxins are included in the inhalable nicotine product analysis. The toxins selected are all International Agency for Research on Cancer (IARC) Group 1 carcinogens (“known carcinogens”) for which OEHHA unit risk values and emissions data were available ¹³ . The protocol used in this study deviates slightly from that of Stephens. Firstly, whereas in Stephen’s protocol Group 1 and Group 2B carcinogens are included, only Group 1 carcinogens were included here ¹⁸ . The Group 1 carcinogens were selected because this grouping applies to toxins for which “sufficient evidence” of cancer in humans has been observed, whereas in the Group 2B category only “limited” or “inadequate” evidence of cancer in humans is available, meaning that the classification in this group is primarily based on evidence from experimental animals or mechanistic studies ²¹ . In this review, we exclude evidence from animal or in vitro studies, therefore Group 2B carcinogens (as well as Group 2A and 3) were also excluded from this toxin emissions analysis. Secondly, we convert the unit risk values to risk per µg or ng per breath in order to match the units of the converted toxin concentrations, which are reported per puff. This is contrary to the methodology outlined by Stephens, which takes unit risk values in their original form of risk per µg per m ³ and adjusts the smoke/vapour toxicants to match these units.

For each carcinogen, the OEHHA unit risk values were sourced and converted from risk per μg per m ³ to risk per μg per breath by assuming the average breath volume of a healthy human is 500 mL ¹⁴ . The toxin emissions of the nicotine products were reported in varying units. For instance, combustible cigarette studies reported toxin emissions as μg or ng per stick, whereas electronic cigarette studies reported toxin emissions as μg or ng per 150 puffs. Therefore, the toxin emissions data for each product were converted to per puff values. In order to make this conversion, the average number of puffs per product per session was extracted from puff topography studies in the scientific literature (see extended data ¹¹ ). The cancer potency of each nicotine product was calculated by adjusting the unit risk values with the observed masses of toxins in the emissions from each inhaled nicotine product using Equation 1:

P i = ∑ j = 1 m C i , j U j

Equation 1: Cancer potency of the nicotine product

Where P _i is the cancer potency of the ith nicotine product, C _i,j is the mass of the jth toxin in the ith nicotine product and U _j is the unit risk for the jth toxin. The cancer potency (P _i) represents the excess cancer risk associated with continuous lifetime use of each nicotine product. To put the cancer potency values into real-world context, the lifetime cancer risk was calculated by adjusting the cancer potency values for average consumption patterns of each product, using Equation 2:

L C R i = P i D i B

Equation 2: Lifetime cancer risk of inhalable nicotine products

Where LCR _i is the lifetime cancer risk of the ith nicotine product, P _i is the cancer potency of the ith nicotine product, D _i is the average daily number of puffs taken by users of each nicotine product and B is the average number of breath taken in one day (40,000 breaths, equivalent of 20 m ³ breathed per day). The lifetime cancer risk (LCR _i) represents the excess cancer risk associated with average daily use of each nicotine product over the course of a person’s lifetime.

For the non-inhaled (smokeless) products, the estimated lifetime cancer risk (ELCR) equation as defined by the FDA was used ¹⁰ . This equation calculates the lifetime cancer risk based on adjustment of the cancer slope factor for each carcinogen with the observed amounts of toxins measured in smokeless tobacco products and average consumption of the products:

E L C R i = ∑ j = 1 m C i , j I R j A B j E F j E D j B W A T C S F j

Equation 3: Estimated lifetime cancer risk of smokeless nicotine products

Where ELCR _i is the estimated lifetime cancer risk of the ith product, C _i,j is the concentration of the jth toxin in the ith product, IR _i,j is the intake rate of the jth toxin in the ith product, AB _i,j is the absorption rate of the jth toxin in the ith product, EF _i,j is the exposure frequency of the jth toxin in the ith product, ED _i,j is the exposure duration of the jth toxin in the ith product, BW is the body weight of the average user, AT is the averaging time of use and CSF _j is the cancer slope factor of the jth toxin.

Epidemiological data analysis . Risk ratios, odds ratios and hazard ratios were extracted from the epidemiological studies and a set of meta-analyses were performed to determine the relative risk of cardiovascular disease, respiratory disease, cancer and mortality in users of each nicotine product compared to non-users of any nicotine products. The epidemiological data extracted from the systematic literature searches was screened to include only relative risk values that compared current users of a single nicotine product to non-users of any nicotine product. Relative risk values were excluded if they were unadjusted for tobacco smoking. Where more than one study was available for a specific disease, the best available evidence was selected according to the Oxford Center for Evidence-based Medicine Level of Evidence Scale ²⁰ . For instance, if a meta-analysis of prospective cohort studies and a single case-control study were available, the former would be included and the latter would be excluded.

The remaining data after screening were grouped by disease type into cardiovascular disease, respiratory disease, cancer and mortality. The cancer category was further broken out into oral, other head and neck, lung, gastrointestinal and other. Meta-analyses of the relative risk data for each disease category were conducted using a random-effects model in the Comprehensive Meta-analysis (CMA) software Version 3.3, with a statistical significance threshold of α = 0.05. The meta-analysis could also be conducted in the open-access alternative, the metaphor package of R. A final meta-analysis was conducted to obtain overall relative risk ratios of cancer and non-cancer diseases for each nicotine product.

In the LCR analysis, several of the combustible products lacked data points for multiple toxins. As the LCR was calculated for each toxin and then summed for each product, missing data points significantly skewed the result towards lower risk. In order to compensate for this, the missing data points for combustibles were filled with values from combustible cigarettes, for which all data points were available. The assumption being that the mechanism of combustion is likely to produce a similar profile of toxins. This assumption was applied to 92% of the data points for bidis, 58% for cut tobacco, 50% of the data points for cigars, 50% for cigarillos and 33% for water pipe tobacco. The room for error due to this assumption should be noted, particularly for bidis where almost all of the data points are assumed from combustible cigarettes.

In the epidemiological data analysis, no data were available for the cut tobacco category and very few studies differentiated between factory-made and roll-your-own cigarette smokers. Therefore, it was assumed that the studies investigating cigarette smokers most likely included a mix of factory-made and roll-your-own, with both identifying as combustible cigarette smokers. Based on this assumption, as well as the assumption that the two products would carry comparable risk, we used the same data for the combustible cigarettes and cut tobacco categories. No other assumptions were made in the epidemiological analysis and missing data was simply recorded using the data completeness score.

For each product, a data completeness score was derived which shows to what extent data points were available for each product and category of data. It serves to give an indication of the risk of error due to missing data for each product. The data completeness is calculated as a percentage, where all the possible data points for each product represent 100% and any missing data is calculated as a fraction of this total as follows.

Equation 1: Data completeness for each nicotine product

Where DC _i is data completeness for the ith nicotine product, n _a,i is the number of data points available for the ith nicotine product and n _t,i is the total number of possible data points for the ith nicotine product. The higher the data completeness percentage, the lower the risk of bias due to missing data.

The RRH combines the results of the lifetime cancer risk and epidemiological analysis, with a weighting system that accounts for the completeness of the dataset. In order to integrate these two analyses into a combined risk score for each nicotine product, an arbitrary scale from 0 to 100 was defined, with 0 representing non-users of nicotine products and 100 representing users of combustible cigarettes. Combustible cigarettes were selected as the top of the scale because they were the highest risk product in both of the analyses. Non-user groups were the control or baseline in both analyses, therefore no correction was required to the lower end of the scale. After converting both analyses onto a 100-point scale, a weighting of 68 was applied to the lifetime cancer risk analysis and 54 to the epidemiological analysis, according to the data completeness scores for each component.

The sensitivity of the RRH to each analysis was determined by simulating several weightings of the lifetime cancer risk and epidemiological data and assessing their outcomes on the risk hierarchy (see extended data). The analyses were weighted 68:54, 1:1 and 54:68, producing three simulations of the RRH.

Data was extracted from the scientific literature into a Microsoft Excel Version 16.41 spreadsheet. Microsoft Excel was used to conduct the lifetime cancer risk calculations. Comprehensive Meta-analysis (CMA) software Version 3.3 was used to conduct statistical meta-analyses in the epidemiological data-analysis. Calculation of the combined risk scores that were incorporated into the RRH was completed in Microsoft Excel.

A PRISMA flow diagram of the updated literature searches is shown in Figure 1. In this update, a total of 23,781 studies were identified in the literature searches. Of these, 5,981 were excluded as duplicates. At the title screening step, 14,876 studies were excluded, primarily due to lack of relevance to the PICOS question. This left 2,907 publications that were assessed in greater detail at the abstract level, where 2,027 studies were excluded. This left 880 studies to be examined at the full-text review, where 810 studies were excluded and 70 were included from the latest round of searches. This made a total of 123 unique studies included in the analysis.

The 70 new studies consist of 40 new toxin emissions studies and 30 new epidemiological studies. The new epidemiological studies consist of 12 meta-analyses, 10 single cohort studies, five case-control studies, two cross-sectional studies and one randomized controlled trial ( Table 3). The sample sizes for cohort and cross-sectional studies are over 1,000, over 100 for case-control studies (combined case and controls) and they range from 33 to 75 for different groups in the randomized controlled trial. In total, 14 studies include only male participants, the rest are mixed with only one study limited to an exclusively female population. The confounder adjustment varies across the studies mostly because the risk for different diseases, populations and products are affected by different confounders. All included odds ratios are adjusted for smoking and the majority are also adjusted for other common risk factors, such as age, sex and socio-economic factors. Only two studies on NRT and snus have a declared conflict of interest. For the former, one of the three study authors is an employee of a consulting company that has provided services to GlaxoSmithKline consumer healthcare (a marketer of NRT products) and the other study is jointly funded by Philip Morris Products, Swedish Match and the European Smokeless Tobacco Council.

There are 40 new toxin emissions studies included in this update ( Table 4). The new data covers combustible cigarettes, cigars, U.S. dipping tobacco, e-cigarettes, heated tobacco products, snus, cut tobacco, smokeless (rest of world), non-tobacco nicotine pouches and NRT. The protocols used to determine toxin emissions are mostly well described and the majority comply with a published standard (e.g., international standards organization or Health Canada Intense). Most studies use the HCI condition with a puff volume of 55 mL, puff duration of 2 or 3 seconds and puff interval of 30 seconds. For cigars and cut tobacco, a puff volume of 35 mL is used. The EN ISO 17294-2:2016/EN 13805:2014 was used for non-tobacco pouches, NRT and one of the snus studies. The remaining studies used in-house methods or protocols referenced elsewhere in the scientific literature. Of the 41 studies, 10 were funded by and/or conducted by researchers at tobacco companies and five studies had another declared conflict of interest.

Toxin emissions/content data were available for all the product categories excluding western pipe tobacco, which was therefore excluded from the LCR analysis. The calculated cancer potency, assumed consumption, LCR, and number of excess cancer cases per 100,000 are presented in Table 5, as well as the data completeness score. The data completeness, including the combustible product assumption, is 100% for all products except the nicotine inhalator, which only has 25% of the data points. For bidis, over 90% of the data is assumed from combustible cigarettes meaning that the actual LCR may be prone to shift if more data becomes available. This is also true to a lesser extent for cut tobacco, cigarillos, cigars and water pipe tobacco.

The LCR data for each product is also plotted as a value relative to NRT in Figure 2. Combustible cigarettes have a relative LCR of 655.3, bidis of 652.2 cut tobacco of 645.5, cigarillos of 547.3, cigars of 330.1 and water pipe tobacco of 325.5, compared to NRT. There was not a significant amount of new data in the combustible categories so the associated LCRs have not shifted significantly compared with the 2020 iteration of this study. Bidis enter the scale just below combustible cigarettes, although it is important to note that the data available for bidis was very limited so this position may be subject to change. The smokeless (rest of world) category enters the scale at 223.3, with very large error bars representing the wide range of toxicity of products in this category. For example, the study by Zakiullah et al investigates the toxin content of 30 brands of Pakistani naswar and reports that the cadmium concentration ranges from below the detection limit (<0.01) to 9.2 mg/kg, whereas for other product categories the ranges were generally smaller, such as NRT with a cadmium concentration of 0.01 – 0.029 µg/g.

The heat-not-burn and electronic cigarette products have relative LCRs of 29.7 and 22.3, respectively. U.S. dipping tobacco, U.S. chewing tobacco and snus have values of 26.4, 14.9 and 6.4, respectively. The non-tobacco pouches have displaced NRT at the bottom of the scale with a relative score of 0.5. This is due to new data uncovered in this iteration reporting higher values of chromium and cadmium for nicotine gum, compared with non-tobacco nicotine pouches. However, it should be noted that most of the lower data points for both products (non-tobacco nicotine pouches and NRT) are below the limit of detection of the instruments. In these cases, we use the limit of detection as the lower limit, but this may be an overestimation and it is not possible to accurately quantitate these products relative to each other on the scale.

In total, 151 risk ratios across 12 categories of nicotine products were included in the epidemiological analysis, compared with 101 risk ratios across eight categories of nicotine products in the 2020 iteration. Heat-not-burn, non-tobacco pouches and cigarillos are not represented in the epidemiological analysis. In this update, new data has been added for NRT, electronic cigarettes, smokeless (rest of world), bidis, snus, U.S. chewing tobacco, U.S. dipping tobacco, water pipe tobacco and western pipe tobacco. Meta-analyses of cancer and non-cancer outcomes for each nicotine product were conducted and are presented in Table 6.

The completeness of the epidemiological data ranges from 33% for smokeless (rest of world), water pipe tobacco and electronic cigarettes to 100% for combustible cigarettes, cut tobacco, cigars and snus. More complete datasets are representative of the relative risk for a wider range of diseases, whereas less complete datasets generally represent a narrower range. While there is significant overlap in the diseases represented for most products, there remain significant gaps in the dataset for certain products.

For cancer outcomes, the products with the highest risk ratios are smokeless (rest of world) (RR 3.675, CI 95% 1.166 – 11.585), combustible cigarettes (RR 2.964, CI 95% 1.579 – 5.565) and cut tobacco (RR 2.964, CI 95% 1.579 – 5.565), followed by bidis (RR 2.884, CI 95% 1.75 – 4.752), water pipe tobacco (RR 2.639, CI 95% 1.635 – 4.258), western pipe tobacco (RR 1.924, CI 95% 1.345 – 2.751) and cigars (RR 1.672, CI 95% 1.218 – 2.295) ( Figure 3). In all cases, users of these products have a statistically significant risk of cancer compared to non-users of nicotine products. The risk ratios associated with dipping tobacco (RR 1.31, CI 95% 0.685 – 2.503) and chewing tobacco (RR 1.205, CI 95% 0.906 – 1.601) are above one, but the lower limit of the 95% confidence interval is below one, meaning that these products are not associated with a significantly higher risk than non-use of nicotine products. The same is true for NRT (RR 1.017, CI 95% 0.977 – 1.058) and snus (RR 0.998, CI 95% 0.902 – 1.104). All four of these reduced risk products also have p-values > 0.05.

For non-cancer risk, the order is slightly different compared with cancer risk. Combustible cigarettes and cut tobacco (RR 1.941, CI 95% 1.676 – 2.248) occupy the highest position, followed by water pipe tobacco (RR 1.83, CI 95% 1.095 – 3.057), bidis (RR 1.708, CI 95% 1.542 – 1.891) and western pipe tobacco (RR 1.707, CI 95% 1.363 – 2.139). Further down the chart are snus (RR 1.242, CI 95% 1.007 – 1.533), U.S. chewing tobacco (RR 1.207, CI 95% 1.119 – 1.301), cigars (RR 1.202, CI 95% 1.077 – 1.341) and the smokeless (rest of world) category (RR 1.166, CI 95% 1.08 – 1.258). Finally, the lowest risk ratios are for U.S. dipping tobacco (RR 1.063, CI 95% 0.881 – 1.282) and electronic cigarettes (RR 0.908, CI 95% 0.666 – 1.238). The non-cancer risk for all products is statistically significant, except for the U.S. dipping tobacco and electronic cigarettes category.

The combined risk scores were derived by integrating the LCR and epidemiological analyses and plotting them on a scale from 0 to 100, with 0 representing non-users of any nicotine products and 100 representing users of combustible cigarettes ( Figure 4).

Combustible products are associated with the highest risk scores ranging from 40 to 100, with combustible cigarettes at 100, cut tobacco at 99, bidis at 93, cigarillos at 84, water pipe tobacco at 66, western pipe tobacco at 61 and cigars at 40. This high toxicity part of the relative risk hierarchy also includes the smokeless (rest of world) category with a score of 51. All of the products in this higher part of the hierarchy, marked with red bars in Figure 4, have error bars that overlap with other high risk products, and carry a significantly higher risk compared with products in the lower part of the hierarchy (marked by green bars in Figure 4). With relative risk scores of less than 10, the lower part of the hierarchy includes U.S. chewing tobacco at 9, U.S. dipping tobacco at 8, snus at 6, heat-not-burn devices at 5, electronic cigarettes at 3, NRT at 0.4 and non-tobacco nicotine pouches at 0.1.

The data completeness has increased compared to the 2020 iteration, with combustible cigarettes, smokeless (rest of world), cigars, U.S. dipping tobacco, U.S. chewing tobacco and snus now above 75% complete. Bidis, water pipe tobacco and western pipe tobacco are 40–50% complete, with many of the missing data points being filled by the assumption of combustible tobacco values for toxin emissions. Heat-not-burn devices, electronic cigarettes, NRT and non-tobacco pouches are around 50% complete, with the epidemiological analysis accounting for most of the missing data points. As such, their position on the relative risk hierarchy is determined mostly by the LCR analysis. The lowest data completeness scores are cigarillos and cut tobacco, which have only 20% of the data points complete. For cut tobacco, the dataset has mostly been assumed from combustible cigarettes in both the LCR and epidemiological analysis and taking account of this assumption the data completeness score is 100%. For cigarillos, most of the dataset for the LCR (toxin emissions) analysis was assumed from combustible cigarettes and there is no data for this product in the epidemiological analysis. Overall, data completeness has increased from 2020, but there are still significant gaps to fill and therefore, there remains a level of uncertainty due to missing data in the scores of most products.

The overall order of products in the relative risk hierarchy is consistent with the original analysis completed in 2020, although some product categories have shifted one place up or down the scale. Notably, western pipe and water pipe tobacco have inversed, as well as U.S. dipping and U.S. chewing tobacco, heat-not-burn and snus, and non-tobacco pouches and NRT. These changes are driven by new data and adjustments to the scope of the categories, for example the limitation of the chewing and dipping tobacco category to U.S. varieties only and the creation of the smokeless (rest of world) and bidis category.

Overall, the data completeness for the epidemiological analysis has increased from 33% in 2020 to 54% in this update and the LCR analysis from 57% to 68% across all the categories of nicotine products. While this is a reasonable increase in the completeness of both analyses, there are still gaps that could affect the position of products in the spectrum moving forward.

In the LCR analysis, the seven highest scoring products have remained in the same position as in the 2020 iteration, and none have changed by more than 1%. Bidis enter the spectrum between combustible cigarettes and cut tobacco, although only one data point was available for the toxin emissions on bidis and the rest of the data was filled based on the assumption that bidis would emit at least the same toxins as a combustible cigarette. The heat-not-burn and electronic cigarette products have shifted higher on the scale in the LCR analysis. This shift can be attributed to incorporation of data from new studies published since 2020 and completion of the data sets, which are now both 100% complete compared to only 91% and 50% previously. The ingestible categories have all shifted slightly higher on the scale, in part due to the change in methodology where ingestible NRT data now serves as the referent. The exception to this is the non-tobacco pouches which have shifted lower due to new data on their toxin content. The rest of world smokeless category enters the scale with a lower LCR than any of the combustible products, but higher than the reduced-risk categories. The range, represented by the error bars, for the smokeless (rest of world) category is very broad, due to the variety of Indian and south Asian smokeless tobacco products that comprise this category and their wide range of measured toxin content. Some products have measured toxin content on par with combustibles, whereas some have toxin contents on par with the lower risk products.

The overall order of the nicotine products in the epidemiological analysis in this update is identical to the 2020 iteration, with the sole exception of cigars which are now above U.S. dipping and chewing tobacco. This change can be explained primarily by the limitation of this category to U.S. varieties of dipping and chewing tobacco, resulting in the removal of data for other types of chewing and dipping tobacco associated with higher risk ratios. This update brings the epidemiological analysis more in line with the order observed for the toxin emissions data, where U.S. dipping and U.S. chewing tobacco are both below cigars and sit much closer to the newer products.

This update of the relative risk hierarchy is broadly in agreement with previous work that estimated the relative harms of nicotine products. Compared to the spectrum presented by Nutt and colleagues, the overall order of the products is the same except for pipes, cigars and water pipe tobacco, which in our spectrum appear as water pipe, western pipe and cigars (from highest to lowest) ¹⁷ . Another point of divergence is that our spectrum suggests that there is a larger difference in risk between combustibles and other nicotine products. While the risk spectrum of Nutt and colleagues shows a significant drop in risk from cigarettes and small cigars to pipes and other combustibles, followed by an incremental decrease in risk down to electronic cigarettes, followed by a significant drop to NRT, our spectrum shows a high risk across much of the combustible and smokeless (rest of world) categories, with incremental decreases from cigarillos down to cigars, followed by a significant gap between cigars and the other product categories (U.S. chewing and dipping tobacco, snus, heat-not-burn, electronic cigarettes, NRT and non-tobacco pouches). This is more consistent with the division of products proposed by Abrams and colleagues, defining “extreme toxicity” as the combustible categories, “much less harm” as the smokeless tobacco (U.S./European), e-cigarettes and NRT grouping, and then “no harm” applying to non-users of any products ¹⁰ . In agreement with this categorization, we define two categories of products; those associated with high risk and those associated with reduced risk of tobacco-related diseases.

The high risk nicotine products category includes combustible cigarettes, cut tobacco, bidis, cigarillos, water pipe tobacco, western pipe tobacco, smokeless (rest of world) and cigars. All products in the high risk category have a combined risk score of between 40 and 100, and are associated with significant harm to health supported by the results of both analyses. The higher risk associated with these products is driven by high levels of toxin exposure, which significantly harm their users health.

Cigars occupy the lowest position in the high risk products group at 40. Despite being a combustible product, cigars are often consumed without inhalation into the lungs, which may be a factor determining the lower risk observed in epidemiological analyses. While the epidemiological analysis is based on real-world consumption, the toxin emissions analysis models only the toxin content and number of puffs per day, without accounting for how the toxins are absorbed into the body. On the other hand, the smokeless (rest of world) category shows much higher cancer risk than any of the other products in the epidemiological analysis, compared with the toxin content analysis. This may be driven by the difference in the varieties represented. In the toxin emissions analysis, studies include varieties of Indian smokeless tobacco, betel quid, zarda and gul, whereas in the epidemiological analysis the products represented are nass, mishri, paan/betel quid, pan-tobacco, khaini, zarda, mawa and gutka.

The reduced risk products are U.S. chewing and dipping tobacco, snus, heat-not-burn, electronic cigarettes, NRT and non-tobacco nicotine pouches. All of these products have a combined risk score of less than 10. The reduced risk category equates to the “much less harm” category defined by Abrams and colleagues, and all products therein carry significantly reduced risk compared with the combustible products. In the epidemiological analysis, the reduced risk products were found to carry a non-statistically significant risk of cancer compared with non-use of any nicotine product, based on the data available. The lifetime cancer risk also shows a significant drop between products in the high risk category and those in the reduced risk category.

This study has a number of limitations, the first of which is the availability of the input data. While there are fewer missing data in this iteration than in the 2020 version, there are still gaps which could give rise to error. For this reason, it is important to interpret the results in the context of the data completeness scores highlighted throughout the presentation of the data. Secondly, there is also a risk of error arising from a lack of accountability of products that have not been studied in the scientific literature. The sample selection bias of this study determines that we are only able to represent products for which data was available, which may not correlate with what is currently on the market. Thirdly, the study methodology is limited to average consumption patterns and usage of a single nicotine product. Therefore, it does not represent the potential risk associated with higher consumption levels of the products or dual/poly-use of the products. Fourthly, the epidemiological meta-analysis combines odds ratios for several diseases, which effectively generalizes a multitude of diseases to “cancer” and “non-cancer” categories. While generalization of multiple datasets is not unprecedented in meta-analyses, where calculation of odds ratios for all-cause mortality often involves a similar process, it is important to note that these values must be interpreted carefully. For example, if the generalized cancer risk ratio is 2, the lung cancer ratio may still be much higher or the risk for stomach cancer much lower, as is the case for combustible cigarettes. The risk ratios derived here are meaningless at the level of individual diseases and should not be used as an indicator for specific disease risk. Rather they give an indication of the average risk across a wide variety of diseases, when several confounders are accounted for.

This update of the relative risk hierarchy defines two categories of products, high risk and reduced risk nicotine products. None of the nicotine products assessed in this study were found to have zero risk, and not using any nicotine product would be the only way to eliminate all risks associated with them. However, the results of this analysis are supportive of the potential to reduce harms caused by high-risk nicotine products by switching to reduced risk products. While it is very important to remain vigilant to new data about novel nicotine products, and to be ready to develop these products with a view to reducing health risks further, reduced risk products can be an effective compromise for reducing the harms of nicotine consumption via combustible tobacco, while giving users the freedom to switch from high risk products on their own terms.

There is a growing body of data demonstrating that nicotine replacement can be a powerful tool for smoking cessation while carrying minimal risks to the user. Indeed, NRT is a widely accepted treatment for smoking cessation. There is also accumulating evidence for other low risk products, such as electronic cigarettes, in their potential to help people who would not otherwise succeed with NRT ⁹³ . Beyond the scientific literature, the potential of reduced risk products is further borne out in real-world data. In Sweden, the use of snus rose dramatically between the 1940s and 1960s in the male population and smoking rates decreased sharply ⁹⁴ . This smoking trend continues to the present day and the smoking rate in Sweden is only 5% of the population, the lowest in Europe, while around 20% still use snus ⁹⁴ . The result for public health is that in 2020 Swedish men also had the lowest incidence of tobacco-related cancers across Europe ⁹⁵ .

While this study does not cover the efficacy of reduced risk products for driving smoking cessation, the results are consistent with the real-world data from Sweden suggesting that switching from high risk to reduced risk nicotine products could lead to a decrease in tobacco-related disease in the long-term.