The number of quit attempts it takes a smoker to quit successfully is a commonly reported figure among smoking cessation programmes, but previous estimates have been based on lifetime recall in cross-sectional samples of successful quitters only. The purpose of this study is to improve the estimate of number of quit attempts prior to quitting successfully.

We used data from 1277 participants who had made an attempt to quit smoking in the Ontario Tobacco Survey, a longitudinal survey of smokers followed every 6 months for up to 3 years beginning in 2005. We calculated the number of quit attempts prior to quitting successfully under four different sets of assumptions. Our expected best set of assumptions incorporated a life table approach accounting for the declining success rates for subsequent observed quit attempts in the cohort.

The estimated average number of quit attempts expected before quitting successfully ranged from 6.1 under the assumptions consistent with prior research, 19.6 using a constant rate approach, 29.6 using the method with the expected lowest bias, to 142 using an approach including previous recall history.

Previous estimates of number of quit attempts required to quit may be underestimating the average number of attempts as these estimates excluded smokers who have greater difficulty quitting and relied on lifetime recall of number of attempts. Understanding that for many smokers it may take 30 or more quit attempts before being successful may assist with clinical expectations.

Uses multiple methodologies to improve upon estimates of the number of attempts it takes a smoker to quit.

Novel use of life table methodology to overcome issues of censoring.

Population estimates of cessation derived from a large, longitudinal, population representative study designed to assess smoking cessation every 6 months for up to 3 years.

Findings reflect a population average and are not individually predictive.

The definition used of a successful quit was cessation to 1 year, but significant likelihood of relapse for some after that point.

Tobacco use is the number one cause of preventable mortality. Five million deaths each year are attributable to smoking, with an estimated rise to as much as 10 million deaths per year by the 2030s.

Mark Twain is reported to have said, ‘Quitting smoking is easy: I've done it thousands of times’. Smoking cessation is a difficult and complex process, and smokers use many methods and approaches to achieve cessation. Knowing how many quit attempts it takes an average smoker to quit is important as it can frame different narratives about the quitting process. This information can be used to inform smoking cessation advice and important messages to stay engaged in the process of quitting. Alternatively, knowing the average number of quit attempts may be a deterrent to continue trying.

For this reason, the number of quit attempts it takes before smokers are successful at quitting is a statistic that figures prominently in much of the literature on smoking cessation programmes. For instance, the American Cancer Society suggests that it takes 8–10 quit attempts,

We know that the likelihood of quitting smoking on any given quit attempt is low;

One study suggests that these estimates are too low. Borland

If we assume that each quit attempt is an independent event, this allows us to use basic probability to come up with the geometric mean, which is calculated by dividing 1 by the probability of that event. If it does take five to seven attempts on average as suggested, then one must assume that the success rate for each attempt is somewhere between 14% and 20%—a figure that does not consistently align with the literature for long-term success.

However, the real number of attempts needed to succeed should be even higher, as the assumption of independence of quit attempts is not realistic. We should expect that the average chance of success will decrease over repeated quit attempts. Those who will have an easier time quitting will succeed on early attempts, whereas those who struggle may repeatedly try and relapse, making the average success rate for early quit attempts higher than subsequent ones. Partos and colleagues found that smokers who had one or more recent failed quit attempts were indeed less likely to quit successfully.

It may be possible to improve on the estimate of the number of quit attempts prior to quitting successfully. A standard life table translates cross-sectional age-specific mortality rates into a theoretical cohort lasting from birth to death.

The purpose of this paper is to improve on the literature estimating the number of attempts it takes to quit smoking successfully by calculating the number under four different sets of assumptions.

We primarily used data from the Ontario Tobacco Survey (OTS), a population-based longitudinal panel survey of smokers (18 years of age and over) in Ontario, Canada. Smokers were followed for up to 3 years at ∼6-month intervals (baseline survey plus 6 follow-ups) with recruitment beginning in 2005. Recruitment was staggered so that the survey interviews were conducted between 2005 and 2011. The smoker response rate was 61%. Participants were selected through random digit dialling stratified by region of Ontario, and characteristics of the sample were consistent with census data from Ontario.

Characteristics of current smokers at baseline in the Ontario Tobacco Survey who had made at least one quit attempt (n=1277)

Characteristic | n | Per cent |
---|---|---|

Sex | ||

Female | 732 | 57.3 |

Male | 545 | 42.7 |

Age | ||

18–24 | 137 | 10.7 |

25–34 | 218 | 17.1 |

35–44 | 291 | 22.8 |

45–54 | 340 | 26.6 |

55–64 | 185 | 14.5 |

65+ | 106 | 8.3 |

Marital status | ||

Married | 721 | 56.5 |

Single | 556 | 43.5 |

Education | ||

Less than high school | 192 | 15.0 |

High school or more | 1085 | 85.0 |

Children at home | ||

No | 746 | 58.4 |

Yes | 531 | 41.6 |

Ever used pharmacotherapy | ||

No | 530 | 41.5 |

Yes | 747 | 58.5 |

Ever used behavioural therapy | ||

No | 846 | 66.2 |

Yes | 431 | 33.8 |

Number of previous quit attempts at baseline | ||

0 | 97 | 7.6 |

1 | 250 | 19.6 |

2 | 244 | 19.1 |

3 | 211 | 16.5 |

4 | 123 | 9.6 |

5 or more | 352 | 27.6 |

Frequency of smoking at baseline | ||

Daily smoker (every day) | 1092 | 85.5 |

Occasional smoker (less than daily) | 185 | 14.5 |

Heaviness of smoking index | ||

Low | 728 | 57.0 |

Medium/high | 549 | 43.0 |

Heaviness of smoking index calculated from daily cigarette consumption and the time to first cigarette after waking (scores of 0–2 were classified as low, 3–6 as high).

Among eligible participants at baseline, only 400 (10.1%) did not provide follow-up data. Those without follow-up data were younger (37.6 vs 44.2 years) and more likely to be male than were those who were included in the analysis; however, there was no difference in smoking behaviours (self-perceived addiction, intentions to quit, daily smoking status and number of cigarettes smoked per day) for those with follow-up data compared with those without follow-up data. Of those participants who made an eligible quit attempt, 91 were lost to follow-up before a successful or unsuccessful quit attempt could be ascertained (see

Number of eligible participants lost to follow-up prior to ascertainment of outcome

Follow-up number | Number lost prior to outcome ascertainment |
---|---|

1 | 0* |

2 | 1 |

3 | 12 |

4 | 30 |

5 | 7 |

6 | 41 |

*Participants were required to complete at least one follow-up to be eligible for inclusion.

This study takes four different approaches to estimating the number of quit attempts each with their own sets of limitations. These analyses are descriptive of the data. The first method is a count of the number of attempts; the second is the calculation of the mean based on a simple probability calculation; Methods 3 and 4 rely on a novel use of a life table analysis. Probability of quitting successful for each of the methods, both overall and quit attempt number specific, is based on observational data from the OTS. We calculated the estimate for each method across the overall sample, as well as within subpopulations of smoking frequency at baseline (daily, less than daily).

Formula for Method 1:

This method corresponds to the assumptions that have been used previously to calculate the average number of quit attempts before quitting successfully using cross-sectional surveys. We expect theoretically that this method will underestimate the number of quit attempts given the known issues with ability to recall quit attempts, which will be much more severe over a lifetime compared with a 6-month period.

The median and mean attempt numbers are calculated from the geometric distribution—the probability distribution of the number of Bernoulli trials prior to achieving one success, using the formula:

The probability of long-term success was based on the percentage of quit attempts that were successful among all quit attempts among attempts observed in the OTS survey (170 successful quits out of 3362 attempts, 5.1%) and for daily and occasional smokers separately (4.9% (n=2930) and 6.0% (n=432), respectively).

Theoretically, we expect this calculation to underestimate the number of quit attempts. Quit attempts are unlikely to be independent events as probability of success decreases with multiple quit attempts (see

Estimated probability of successful quitting (quit attempt of longer than 1 year) by quit attempt number in the Ontario Tobacco Survey. The top figure includes only observed quit attempts, while the bottom figure also includes reported quit attempts prior to study entry. Dotted lines represent upper and lower CIs. A life table analysis was used to calculate probabilities (n=1277).

To do this, we first calculated the probability of a successful quit per quit attempt observed. We estimated the probability of success for each quit attempt using a standard life table approach.

We observed quit attempt ‘age’, that is, the cumulative attempt number, and coded whether each attempt was successful or unsuccessful. Then, we calculated the following quantities:

n_{i}, the number of people who have not quit successfully (>1 year) at quit attempt i (normally the number of survivors for that age interval);

We adjusted for the influence of right censoring (quit attempts that are not observed as they occur after the period of study observation) using the standard actuarial method by calculating n_{i}: subtracting half the number in the period that was censored after a quit attempt was observed. This calculation provides an estimate of the probability of success by quit attempt number, and cumulative probability of success. The expected average number of quits prior to quitting successfully was then obtained by calculating the area under the curve (equivalent to the restricted mean).

We expect that this method will modestly underestimate the number of quit attempts. Ignoring previous quitting history will attribute a greater probability of success to earlier rather than later attempts (many of the first and second attempts observed are likely to be people who are making their third or higher lifetime quit attempt), which will produce an overestimate. We also do not know what the effect on success rates would be of quit attempts made in relatively short succession after a failed attempt, compared to attempts made further apart. Given that earlier probabilities will have a bigger effect on the area under the curve that is used to produce the mean, we expect that the overall bias in this method will be a relatively modest underestimate, and provide the best estimate of the number of quit attempts.

We expect that this method will overestimate the effect. Unlike with Method 3, Method 4 censors the data from successful smokers, estimating, in essence, the average number of quit attempts among those with difficulty in quitting. The method has a similar but opposite issue to that of Method 1, recalled life attempts among successful smokers. Where Method 1 excludes information from people who have not yet quit, underestimating the result, Method 4 does not account for people who have already quit, overestimating the effect. However, the use of lifetime recall quit attempt will underestimate the total number of quit attempts.

Estimate of the number of quit attempts made prior to successfully quitting for 1 year or longer among 1277 smokers in the Ontario Tobacco Survey for four different methods of calculation

Expected number of quit attempts prior to quitting successfully | |||||
---|---|---|---|---|---|

Method | Key assumption | Expected direction of bias | Overall (n=1277) | Daily (n=1092) | Occasional (n=185) |

Method 1 | |||||

Recalled quit attempts, among successful quitters | Recall of quit attempts over a lifetime is valid. Only successful quitters are included | Underestimate: only includes successful quitters, recall over a lifetime is poor | 6.3 | 6.8 | 4.7 |

Method 2 | |||||

Constant rate assumption | Every quit attempt has the same chance of success, no matter how many previous quit attempts there have been | Underestimate: chance of success decreases by number of previous quit attempts | 19.6 | 20.4 | 16.7 |

Method 3 | |||||

Life table, observed quit success rates | The success rate of quit attempts varies by ‘quit attempt age’, as observed during the period of the study | Offsetting: overestimates chance of success on quit attempt, may underestimate on subsequent attempts | 29.6 | 29.5 | 11.4 |

Method 4 | |||||

Life table, recalled lifetime quit numbers | The success rate of quit attempts varies by ‘quit attempt age’, as observed and recalled | Probable overestimate: recall over a lifetime is underestimated, but calculation does not include successful quitters who had quit prior to the study | 142.0 | 142.3 | 21.0 |

Successful quit attempts were measured as abstinence of 1 year or longer.

Absolute survival probability (per cent who have not quit successfully) of reported quit attempts of longer than 1 year during the first 18 months of observation in the Ontario Tobacco Survey. The top figure includes only observed quit attempts, while the bottom figure also includes reported quit attempts prior to study entry. Dotted lines represent upper and lower CIs. A life table analysis was used to calculate survival probabilities (n=1277).

The average number of quit attempts is likely significantly higher than what is generally communicated to smokers. Under what we consider to be the best set of assumptions (Method 3: Life Table, Observed Quit Rates), this study suggests that a current smoker tries to quit on average 30 times or more before successfully quitting for 1 year or longer. The life table approach within this population-representative longitudinal cohort overcomes some of the challenges of estimating the number of quit attempts it takes a smoker to quit for good. Previous estimates of how many quit attempts it takes to quit used populations of former smokers exclusively; these estimates are significantly biased by difficulty in recalling past quit attempts, and by not accounting for the large number of people who try to quit but who are never successful.

We selected Method 3: Life Table, Observed Quit Rates as the best set of assumptions as it accounts for the biases (lifetime recall, independence assumption, exclusion of unsuccessful quitters) that affect Method 1 (Recalled Attempts Among Successful Quitters) and Method 2 (Constant Rate Assumption), and consequently provides a better estimate of the average number of attempts prior to quitting successfully than those two methods. Arguably, Method 4: Life Table, Recalled Lifetime Quit numbers improves upon Method 3 as it uses a distribution of success per quit attempt that may be more realistic. That is, Method 4 demonstrates that, as expected, the highest success rates are not on the first lifetime quit attempt, but rather peak after a number of quit attempts, before declining. However, the downside of Method 4 is that it is likely underestimating the quit success rates, particularly for the early attempts as we were unable to measure the quit success of people who quit prior to study entry and who conceivably were people who found quitting easier. The implication is that the 30-attempt estimate is likely still underestimating the true average number of attempts, but the true estimate is likely lower than that obtained for Method 4.

The estimate of 30 attempts is consistent with the estimate of quitting behaviour from Borland

It is important to note that this is a descriptive estimate, and not a normative one. This has two implications. The first is that it is important to consider the results as an average of a population, rather than any one individual experience. Despite a high number of quit attempts needed to quit was found on average in this study, many individual smokers are able to quit successfully on relatively few attempts.

The second implication of this is, as a descriptive statistic, the estimates here in this study describe the experience of this population, given the population distribution of these factors at the time. To compare the estimate of average number of quit attempts from one jurisdiction or time period to another may theoretically require standardisation of these factors to the Ontario population at this time. While there are many factors that affect the ability to quit smoking, these estimates can be considered accurate for this population. The correspondence between the estimate found in this study (6.1) to those prior estimates that were made under the similar sets of assumptions (ie, five to seven attempts)

This study defined quit attempt as a ‘serious quit attempt’ as self-perceived by the smoker. Other definitions of quit attempts that require smokers to be abstinent for a certain amount of time (such as 24 hours or 1 week) add significant bias by excluding those smokers who are least likely to quit and have the most difficulty maintaining abstinence.

Quit success was assessed at 1 year. While many smokers can maintain abstinence after 1 year of successful quit, over 37.1% of those may subsequently relapse.

The average number of quit attempts prior to a successful quit is a number that is inherently difficult to capture with full certainty; nevertheless, this study improves upon the estimation of the average number of quit attempts. Most accepted previous estimates have relied on recalled data from cross-sectional studies, and have not addressed the issues of censored data from smokers who have not yet or will not ever quit. The life table methodology with the use of longitudinal survey data addresses some of these previous limitations to provide a more realistic estimate of the experience of smokers.

Knowing that many smokers try and fail on multiple attempts leads to three potential implications. The first suggests that we need to recognise that quit attempts occur frequently, and that further increasing the frequency of these attempts and the rate of success on these attempts may be required to reduce smoking prevalence in a population. For any given quit attempt to be successful, many more resources may be needed, particularly evidence-based resources such as pharmaceutical aids and counselling.

More research is needed to understand how best to communicate to smokers about the number of quit attempts prior to successful quitting. Our study does demonstrate that it takes many more quit attempts to succeed than previously estimated. It is possible that being realistic about the chances of success and the length of time it may take to become a successful quitter may help facilitate a positive, long-term relationship with healthcare providers. Smoking is a chronic condition, and an appropriate estimation may help avoid treating quit attempts as discrete, acute events. On the other hand, communicating the specific number to patients may not be helpful. It may be that some smokers may be discouraged by hearing how difficult it can be to quit smoking. Further research to help customise risk messaging and supportive interventions is required for those who have a number of previously failed attempts to quit and for new smokers just beginning their journey to quitting.

MC conceived the study. MC, AP and LD performed the analysis. JEC, SJB, PS and RS designed and implemented the survey. All authors contributed to writing and interpretation.

The Ontario Tobacco Research Unit and the Ontario Tobacco Survey are supported by the Ontario Ministry of Health and Long-Term Care. This work is supported by Canadian Cancer Society grant #702160 (MC). The funders had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript, or the decision to submit the paper for publication.

None declared.

University of Toronto and University of Waterloo Research Ethics Boards.

Not commissioned; externally peer reviewed.

Research teams may apply for access to OTS data through one of the following university-based data libraries: Propel Centre for Population Health Impact—Population Health Data Repository at the University of Waterloo (