Contraceptives fail a lot more often than you may have realized. Those failure rate statistics that you see are actually a bit misleading — they are for failure rates in a given year. Over time, the risk of failure is compounded.
It's one of those things that makes sense if you stop and think about it, but a series of interactive charts by Gregor Aisch and Bill Marsh in last weekend's New York Times really helps put it in perspective. Here are figures for three of the fifteen contraceptive methods presented in their visualization:
These charts, while not entirely accurate (see the Considerations section, below), illustrate a little-understood fact about contraceptive failure: The longer you use a given contraceptive – whether it's the pill, condoms, spermicides, or something else – the more chances it has to fail. The same is true for any game of chance carrying an element of risk. Imagine playing Russian roulette with a hundred-chamber revolver. You load one bullet into the cylinder, give it a spin, snap it in place and pull the trigger. Your odds of firing off a round are just 1 in 100. In fact, they're just 1 in 100 every time you repeat the process. But repeat that process enough times – load, spin, snap, pull; load, spin, snap, pull; load, spin, snap, pull – and the overall risk of the gun going off is compounded. The same logic applies to condoms, or the pill, neither of which is 100% effective. On a long enough timeline, probability says they (and most other contraceptive methods) will probably fail.
Photo Credit: Selbe and Lily | CC BY-NC-ND 2.0
Or, in the case of contraceptives, more likely what will happen is you'll drop the ball. The prescription will run out on your birth control and you'll miss a couple of days, lowering your body's guard against unwanted pregnancy. Or you'll accidentally tear the condom putting it on. In other words: The longer you use a given contraceptive, the more opportunities there are for you to use it improperly or inconsistently.
The fact that none of us is perfect highlights the distinction between what James Trussell, a professor of economics and public affairs at Princeton's Office of Population Research, refers to in a 2011 review of contraception failure in the United States as "perfect use" and "typical use." Aisch and Marsh, who used the statistics from Trussell's paper to create their charts, characterize these two standards as follows:
Typical use: This is the norm, reflecting the effectiveness of each method for the average couple who do not always use it correctly or consistently.
Perfect use: A measure of the technical effectiveness of each method, but only when used exactly as specified and consistently followed. Few couples, if any, achieve flawless contraceptive use, especially over long time periods.
The difference between theory and practice is how you get charts like this one:
Assuming typical male condom use by a heterosexual couple, the number of women projected to experience an unplanned pregnancy over a period of ten years is about 86 out of 100. Eighty-six percent! Compare that to the 92% of women projected to have an unplanned pregnancy over the same time period relying only on the withdrawal – or "pull out" – method. (Condoms, of course, provide protection against STDs that the withdrawal method does not, so I hesitate to draw the comparison, but that's a difference of just 6% over ten years. Needless to say, I was surprised by how close together those numbers were.)
For male condoms used on a ten-year timeline, the difference between pregnancy rates during perfect use (18%) and typical use (86%) is almost 70%. That's huge. With the exception of the ovulation method (perfect use 26%, typical use 94%), that's a bigger difference than any other contraceptive method. According to Trussell, this gap tells us something important about condoms.
"The difference between pregnancy rates during imperfect use and pregnancy rates during perfect use reveals how forgiving of imperfect use a method is," he explains. "The difference between pregnancy rates during typical use and pregnancy rates during perfect use reveals the consequences of imperfect use."
The upshot – and this is arguably the most important takeaway – is that the condom is forgiving of neither imperfect use nor inconsistent application. The risk associated with user error (or neglect), in other words, is high. This seems to suggest that if you want birth control to be effective, you need to minimize user interaction. The data support this. Here's Aisch and Marsh's chart for the pill, Evra patch and NuvaRing:
Used correctly and consistently, these contraceptive methods are projected to result in an unplanned pregnancy in just 3% of women over a ten year period. Over the same time period, that's lower than the projected rate of unplanned pregnancy in women who have been sterilized.
But humans aren't perfect. We forget, misremember, and misuse, and so we see a projected rate of pregnancy (again, over 10 years) of not 3%, but 61%. This is one reason sterilization, hormonal implants, and IUDs are so effective – you can't really use them improperly. It's set it and forget it. On the chart for the Levornorgestrel IUD, the line for typical use and the line for perfect use are one and the same:
There are some important things to keep in mind when looking at Aisch and Marsh's graphs. You'll be relieved to learn, for example, that the true likelihood of unplanned pregnancy is actually lower than the "typical use" figures they present.
Number 1 is that nearly all of the numbers given for "typical use" in Aisch and Marsh's graphs are overestimates. Here's what Trussell had to say on that point when I e-mailed him:
[Aisch and Marsh] used my estimates of first-year failure rates during typical and perfect use to produce their 10-year graphs. Doing so is fine for perfect use. But it results on overestimates for typical use. The reason is that the least adherent users become pregnant, leaving behind a group that over time is less and less likely to fail. I told [Bill Marsh] this. There are no studies looking at ten years of use except for female sterilization. I encouraged him to instead use the attached chart that would have made the same point just as effectively (and correctly).
Here is the chart Trussell sent me:
Source: Trussell J, Guthrie KA. Choosing a contraceptive: efficacy, safety, and personal considerations. In: Hatcher RA, Trussell J, Nelson AL, Cates W, Kowal D, Policar M (eds). Contraceptive Technology: Twentieth Revised Edition. New York NY: Ardent Media; 2011. p. 45-74
Number 2 is that Trussell's original figures are estimates, made based on data collected from large populations. Probabilities of pregnancy for most of the reviewed contraceptive methods were estimated based on data collected in the 1995 and 2002 National Surveys of Family Growth (NSFG), a nationally representative sample of users. Unintended pregnancy rates of other methods were based on surveys and clinical investigations. (Typical-use estimates for male condoms, for instance, relied on responses to the 1995 and 2002 NSFG, while unintended pregnancy estimates for perfect male condom use were based on "the only three studies of the male condom meeting modern standards of design, execution, and analysis.")
The upshot? These numbers can't tell you with any certainty what your likelihood of an unwanted pregnancy is, because the only person who can say how consistent you are with your chosen method of birth control is you. What they can tell you is how forgiving your method of birth control is, or – equally important – how woefully ineffective it is, even when used flawlessly (looking at you, spermicides).
Number 3 is that these estimates have been corrected for underreporting of abortion and over-reporting of contraceptive use, both of which occur, and neither of which is very well understood. As Trussell notes:
Thus, biases in opposite directions affect these estimates. Pregnancy rates based on the NSFG alone would tend to be too low because induced abortions (and contraceptive failures leading to induced abortions) are underreported but would tend to be too high because contraceptive failures leading to live births are overreported. We reason that the former bias is the more important one.
Number 4 is that "typical use" as Trussell defines it and "typical use" as you define it don't have to mean the same thing – and if you're trying not to get pregnant, they really shouldn't. Here's an expanded definition, from Trussell's review:
Typical use does not imply that a contraceptive method was always used. In the NSFG and in most clinical trials, a woman is 'using' a contraceptive method if she considers herself to be using that method. So, typical use of the condom could include actually using a condom only occasionally, and a woman could report that she is 'using' the pill even though her supplies ran out several months ago. In short, 'use'—which is identical to 'typical use'—is a very elastic concept that depends entirely on an individual woman's perception.
If your individual perception of "typical use" hews more toward I use a condom every time I have sex than it does toward I have condoms in my nightstand that expired before Netflix was a thing that I use when I remember and/or feel like it, your odds of getting pregnant are going to be a lot lower.
Check out Trussell's excellent, very readable review of U.S. contraceptive failure here.
Check out the rest of Aisch and Marsh's contraceptive visualizations over at The New York Times.