When the age limit of alcohol and tobacco use was raised from 16 to 18 years in the Netherlands, the Health Ministry launched a nationwide campaign to raise public awareness. Naturally, it would be of interest to quantify the effect of introducing such a campaign. For example, by looking at the change in the number of 17-year olds being admitted to hospital due to excessive drinking. In the lecture on frequency measures you learned how to count, and now we can discuss how to compare. In this lecture, we will discuss measures of effect for the exposure we are interested in, under the assumption that the groups are otherwise similar. At the end of this lecture, you'll be able to construct a two-by-two table and calculate both absolute and relative measures of effect. You'll be able to calculate risk and rate differences, risk and rate ratios, attributable risks, and the number needed to treat. In epidemiology, we use effect measures to compare disease frequencies between different groups, typically defined on their exposure status. Finding a difference may mean that an exposure causes disease or if an exposure is treatment or policy that this has had an effect on the disease risk. The most basic presentation of the frequency of a binary outcome across two exposure groups is the two-by-two table. This table is simply a square divided into four spaces and sometimes described as a contingency table or cross-table, and during this course we'll use the convention of placing the outcome on top and the exposure on the left-hand side. We can now fill in how many individuals fall within each category. For now we start by simply labeling the spaces as A, B, C, and D. As discussed in the lecture on frequency measures, incidence can be a cumulative incidence or an incidence rate. These represent the absolute risk or rate of an outcome occurrence. To obtain an absolute measure of effect, we can calculate the difference in incidence between the two exposure groups, thereby quantifying the excess risk or rate of disease in the exposed group compared to the unexposed group. The unexposed group is traditionally used as a comparator or reference, but this may also be another exposure or treatment depending on your study. Just remember that knowing which group was used as the comparator is key to interpreting effect measures. Let's go to an example. Say we have data on a trial comparing open versus laparoscopic surgery, also known as minimally invasive surgery, for gallbladder removal. If you compare often wound infections occur within 30 days after surgery, we might observe that the cumulative incidence was 5 over 100 in the open surgery group, and 1 over 100 in the laparoscopic surgery group. The resulting risk difference would be the difference between these two, or 4 over 100. Our interpretation of this risk difference would be that there were 4 additional cases of wound infection per 100 people in the open surgery group. In other words, the risk difference is calculated by subtracting the cumulative incidence in the unexposed group from the cumulative incidence in the group with the exposure. To calculate the rate difference, we modify the two-by-two table to include person-time, often expressed as person-years, and subtract the incidence rate in the unexposed group from that in the exposed group. While similar to the interpretation of the risk difference, don't forget that person-time is included. In other words, the rate difference is the amount of additional cases per a certain number of person-years in the exposed group when compared with the unexposed group. Now, let's go back to the example on underage drinking given at the start of this lecture. Imagine this campaign was implemented in one city, city A, and not implemented in another, city B. In city A we observed that six 17-year olds are hospitalized due to excessive alcohol consumption, over a summed person-time of 60,000 person-years. While ten 17-year olds are hospitalized in city B over 50,000 person-years. The incidence rate difference would again be the difference or 1 per 10,000 person-years. We would interpret this as that for every 10,000 person-years, city A had one less case compared to city B, where the campaign was not implemented. When given the choice, researchers often prefer to report absolute measures of effect as they help to assess how much impact eliminating exposure might have on public health. Relative measures of effect, compare the risk or rate in exposed and unexposed groups on a relative scale. This shows how strong the association is between an exposure and a disease on this relative scale. Both for risks of disease shown on the top and for rates of disease shown on the bottom, the same tables are used as before. Taking the ratio of the cumulative incidences and incidence rates, produces the relative risk and incidence rate ratio respectively. Their interpretation is then: how many times higher or lower the risk or rate is in the exposed group compared to the unexposed group. Let's now fill in these tables again with our two examples on wound infection and hospitalization due to excessive drinking. If we calculate the relative risk, we observe that patients in the open surgery group have five times the risk of wound infection of the laparoscopic surgery group. For the incidence rate ratio, we observe to in city A the rate of hospitalization due to excessive drinking in 17 year olds was half the rate observed in city B. Another risk measure is the attributable risk which quantifies how much of the disease occurrence of the exposed individuals is due to the exposure itself. Importantly, it is also directly related to the relative risk. For example, smoking has a relative risk of 10 for lung cancer. If indeed causal, the attributable risk is 90 percent. That means that 90 percent of lung cancer in smokers is indeed due to smoking. Another form of the attributable risk is the population attributable risk, which is not concerned with the excess risk of disease in the exposed, but rather in the entire population. Therefore, the population attributable risk represents the proportion of disease incidence in the population which would be eliminated if the exposure was removed. Of course again only, when there is a causal relation. A risk difference of three per 1,000 is difficult to get a feel for. Another measure that may be helpful is therefore the number needed to treat, the NNT. This measure provides an easily interpretable estimate of how many individuals one should treat to avoid one additional adverse outcome, and is simply the inverse of the risk difference. For example, imagine you observe that individuals exposed to treatment X have a cumulative incidence of heart attack of 30 percent across a certain time period, while those who were not exposed to treatment X have a cumulative incidence of 50 percent. The number needed to treat here would be one divided by 0.5 minus 0.3, so five. In other words, five individuals would need to be treated with treatment X to avoid one heart attack. I hope it has become clear how to make a two-by-two table and that you are now able to calculate both absolute and relative measures of effect. All of these are straightforward for incidence rates, but require some additional assumptions for risks, where the time window of observation is crucial.