How likely is it that birth control could let you down?
November 26, 2014 10:28 PM   Subscribe

The NYT calculates the probability of pregnancy using 15 common birth control methods, for up to 10 years of both "typical" and "perfect" use. Protip: the graphs do slidey comparison things on mouseover!
posted by Ragini (55 comments total) 15 users marked this as a favorite


 
Man, I love my IUD.
posted by Windigo at 10:32 PM on November 26, 2014 [8 favorites]


The math is wrong in this article, it was debunked two months ago when published.
posted by Nelson at 10:34 PM on November 26, 2014 [2 favorites]


Holy crap the pregnancy rate for male condoms is 86% over 10 years with typical use? That can't be right. What the tits are y'all even doing with those things?

Just got my IUD. Love it to death so far. Such a weight off, and no more nauseating pills to take. Haven't named her yet-- any suggestions?
posted by WidgetAlley at 10:35 PM on November 26, 2014


The math is wrong in this article, it was debunked two months ago when published.

I was just coming in here to complain that they messed up the binomial probability function. It shouldn't just be p^x like they used, it should be (n ways to get x)p^x(1-p^(n-x), right?
posted by dialetheia at 10:39 PM on November 26, 2014 [1 favorite]


Here's a good takedown of the article from (of all places) The American Conservative. And another from iO9.

Shame on us for not having better public health data about long-term contraceptive use and efficacy, and having to extrapolate from single-year data slices, though.
posted by WidgetAlley at 10:40 PM on November 26, 2014 [5 favorites]


I don't think that saying it was "debunked" is accurate. Some people have taken issue with the statistical methodology.

E.g. this article in American Conservative discusses some perceived shortcomings of the 10-year extrapolation, although I'd be lying if I said that my spidey sense wasn't tingling at the source.

This analysis by io9 has some interesting alternative infographics that may be less problematic.

Basically: the NYT numbers are probably worst-case estimates. The rate of 1st-year contraceptive failure may not be the same as 2nd-year, 3rd-year, 10th-year, or Nth-year failure rates, so simple arithmetic extrapolation isn't necessarily a good model. However, there aren't a lot of long-term longitudinal studies on contraception effectiveness except for sterilization, because people don't tend to use a single method consistently for that long. So I don't think that it's completely without merit, either.
posted by Kadin2048 at 10:43 PM on November 26, 2014 [5 favorites]


Ooooor what WidgetAlley said in fewer words. That too.
posted by Kadin2048 at 10:45 PM on November 26, 2014


It always weirds me out a little that they give numbers for "typical use" where "typical use" includes not using that method of birth control consistently. It's like if I were trying to find out how many calories are in the salad I made and the information includes a bacon cheeseburger in the average because a lot of people cheat on their diets and eat bacon cheeseburgers instead of salad.

I understand why they use the "typical use" numbers. I really do. I just wish it were obvious to people that if you use a condom sometimes and don't use a condom other times it isn't going to work very well.
posted by Justinian at 10:47 PM on November 26, 2014 [30 favorites]


I just wish it were obvious to people that if you use a condom sometimes and don't use a condom other times it isn't going to work very well.

I think that's obvious to people -- no one expects a condom to work if it is sitting on the bedside table. But how else do you capture that a very large percentage of people who "use condoms" actually dislike them enough to in fact not use them an awful lot of the time? It's like the recent FPP about research into methods for swallowing pills, where it turns out that people having trouble swallowing pills is a significant component of medication non-compliance. Finding the reasons for why the spread is sometimes so large between perfect use and real-world use is where real gains are going to be had.

Or, more or less what the io9 article said: 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.

I don't know enough about statistics to criticize the numbers given, but intuitively they look way too high unless a lot of people are being counted who aren't doing anything at all to prevent pregnancies.
posted by Dip Flash at 11:01 PM on November 26, 2014


I think it's interesting that withdrawal shows comparable success rates to condoms - for both "perfect" and "imperfect" use. So it appears that withdrawal is therefore be a perfectly legitimate option for people not worried about disease and who aren't interested in or can't handle more invasive or hormonal methods.
posted by cilantro at 11:25 PM on November 26, 2014 [1 favorite]


Holy crap the pregnancy rate for male condoms is 86% over 10 years with typical use? That can't be right. What the tits are y'all even doing with those things?

Trying to prevent STDs? Taking a small step in responsibility for birth control?

TMI time! TMI time! Back in the day, my boyfriend and I used the following:

He: condom

She: diaphragm and spermicide

No babbys was made in the relationship.
posted by potsmokinghippieoverlord at 11:43 PM on November 26, 2014 [4 favorites]


I totally recognize how important this information--including its nuances--is to the majority of the population.

But, as always, I am saddened that it focuses solely on pregnancy and not on STI prevention, particularly amongst non-cis-het populations. We`re here too!
posted by feckless fecal fear mongering at 12:26 AM on November 27, 2014 [2 favorites]


Haven't named her yet-- any suggestions?

AlleyWidget?
posted by pompomtom at 12:52 AM on November 27, 2014 [15 favorites]


I'm suprised to see condom [female] is more effective than condom [male]. It does say "These charts show probabilities of pregnancy while using each method, alone..."

Huh.
posted by vapidave at 1:10 AM on November 27, 2014


A female condom is a different device than a male condom, if that was unclear.
posted by mbrubeck at 1:27 AM on November 27, 2014


It always frustrates me that nuvaring gets lumped in with the pill. Remembering to take a pill at the same time every day is super different from remembering to replace a nuvaring once a month.
posted by prefpara at 3:19 AM on November 27, 2014 [4 favorites]


(Unsurprisingly) it turns out this sort of data is already available from a number of reputable public health sources!

A quick search of Planned Parenthood didn't turn up a handy comparison chart like is up on the wall of every single doctor I've ever gone to for annual check-ups, but effectiveness data for each individual birth control method is among the details that they provide when you click on each link, plus copious information to help you determine which birth control methods you can choose in order to get closer to the perfect use stats, based on your own individual needs and your and your partners' behaviors.

Our Bodies Ourselves embeds a chart in their discussion of perfect versus typical use rates, they explain why they talk about pregnancy rate instead of failure rate, and have links to some primary sources. They note that the original surveys covered a longer time period, and that the one-year pregnancy rate was calculated in a particular way from the survey data, which is likely relevant if one wants to reverse this calculation to look at multi-year pregnancy rates.
posted by eviemath at 4:17 AM on November 27, 2014 [5 favorites]


Depo is an injection....does the 'typical use' line reflect that many women don't stick closely enough to the dosing schedule? As 46% over 10 years seems insane when the 'perfect use' line is more or less flat.
posted by snuffleupagus at 5:35 AM on November 27, 2014 [1 favorite]



Just got my IUD. Love it to death so far. Such a weight off, and no more nauseating pills to take. Haven't named her yet-- any suggestions?

The widget?, Widget's widget?
posted by notreally at 5:47 AM on November 27, 2014 [1 favorite]


I would love to see some figures for using condoms and also using plan B as a contingency if there's some kind of failure or screwup.

I think it's interesting that withdrawal shows comparable success rates to condoms - for both "perfect" and "imperfect" use. So it appears that withdrawal is therefore be a perfectly legitimate option for people not worried about disease and who aren't interested in or can't handle more invasive or hormonal methods.

Yeah, withdrawal is more effective than most sex ed talks would have you believe. I think it's not often recommended because a. It's probably harder than condoms to achieve "perfect use" with and b. The potential for transmitting disease.
posted by Gymnopedist at 5:47 AM on November 27, 2014 [1 favorite]


What the tits are y'all even doing with those things?

Well, if that's where you're trying to put the condom...
posted by dances with hamsters at 5:48 AM on November 27, 2014


Regarding The American Conservative: I cannot vouch either way for the accuracy of that particular article, but the magazine itself regularly contains excellent articles. It is not your typical rag, conservative or otherwise.

...

Also, who the hell is using withdrawal for ten straight years?
posted by Sticherbeast at 6:14 AM on November 27, 2014


(I actually do find that TAC takedown to be very persuasive, but I'm busy and too untrained/dumb/distracted to really weigh in myself.)
posted by Sticherbeast at 6:24 AM on November 27, 2014


I wish the Mortal Kombat announcer were involved in some of these situations. Perfect withdrawal.
posted by tapesonthefloor at 6:38 AM on November 27, 2014 [5 favorites]


Typical use rates are useful for population statistics. If you want to know how your citizens got so fat, the fact that they sometimes eat a bacon cheeseburger instead of a salad is important.

But those rates are worse than useless when making decisions about personal use, where you can factor in how consistent and responsible you're likely to be. My wife took the Pill for over 20 years and she was fanatically devoted to making sure she took it each day, so we never really worried about failure, because it's well documented that nearly all failures of the Pill are due to not taking it consistently.

On the other hand someone who knew she was less consistent might make the decision to go for an IUD on practical grounds. That would be a legitimate use of the numbers to make a rational decision.

But the NYT article seems to be aimed at scare mongering, tarring the very idea of birth control with worst-case assumptions that don't apply to a lot of individual users.
posted by localroger at 6:47 AM on November 27, 2014


Also, who the hell is using withdrawal for ten straight years?

If pornography is to be believed, every act of intercourse should conclude with the man pulling out, and then spraying a geyser of sperm onto the grimacing face of his partner.
I wonder if this is behind the recent reduction in teenage pregnancies?
posted by Flashman at 7:08 AM on November 27, 2014 [2 favorites]


When long acting reversible contraception (IUDs, Implanon, etc) are actively promoted over the pill, condoms, patches, and Depo shots among teenage girls seeking contraception, and insertion of such methods is available that same day, pregnancy is 22 times less likely to occur. That's the multiple of 22.

This is probably the greatest story in medicine of the last year and should lead to a revolution in unplanned teen pregnancy. It's almost malpractice to prescribe short acting hormonal contraception to a teenager, or anyone else who thinks they might not use it correctly.
posted by Slarty Bartfast at 7:34 AM on November 27, 2014 [16 favorites]


They left out abstinence!

/hamburger
posted by oddman at 7:45 AM on November 27, 2014


Thanks for posting the debunking articles, WidgetAlley. The American Conservative article is the one I was thinking of, explaining some of the statistical fallacies in this article. (Summary: "typical use" confuses things, and you can't extrapolate from 1 year rates like they do.) This visualization is beautiful, but in this case I think the statistics are so bad that the article is harmful and misleading.

I wonder if this is behind the recent reduction in teenage pregnancies?

Nah, that's the result of the normalization of anal sex ;-)
posted by Nelson at 7:51 AM on November 27, 2014


They left out abstinence!

/hamburger


I would love to see the 'typical use' stats for that.
posted by Catseye at 7:53 AM on November 27, 2014 [7 favorites]


I knew the implant was really effective, but I didn't know it was actually more effective than sterilisation. Is that due to how often sterilisation fails, which I presume is rare but pretty difficult to pick up if it occurs after you've been given the all-clear? As opposed to the implant where you can check it's still there pretty easily, provided it was put in properly in the first place. Given the details above it could be statistical weirdness/lack of proper studies instead, I guess the key message is that they are both really effective and very difficult for the user to fuck up.
posted by penguinliz at 8:12 AM on November 27, 2014


Trying to prevent STDs? Taking a small step in responsibility for birth control?

Oh, absolutely, I just meant that for the pregnancy rate to be 86% over ten years, at least occasionally people had to be wearing them on their heads by accident or something or not using them at all, as Justinian points out.

In public health we distinguish between "efficacy" and "effectiveness" and "compliance": efficacy is how well something works when used in a controlled setting like a randomized clinical trial (pretty much the 'perfect use' category), effectiveness is how well something works in a normal treatment setting (so the subset of folks who use condoms every time but may be using them incorrectly or they're breaking or whatever), and compliance (who uses the device or medication as prescribed and at the times prescribed, and who doesn't.) It seems like the charts do a really good job of showing efficacy, but obscure effectiveness and compliance by using an intention-to-treat type of analysis, where you count noncompliance in the same null category as ineffective. It's a great statistical method for clinical trials because it doesn't obscure possible null results in the dropout rates of the study, but it doesn't make sense here; ineffectiveness and noncompliance are fundamentally different things and different issues (both important and totally worthy of study, as Dip Flash mentions, but lumping them together obscures the important context around both of them.)

AlleyWidget?

Close the thread and go home, that's perfect and we're done here.
posted by WidgetAlley at 9:15 AM on November 27, 2014 [5 favorites]


I would love to see the 'typical use' stats for that.

One of my professors worked on the wonderful ADD Health survey, one of the best surveys of adolescent health and behavior in the United States, and one thing they asked was, "Have you ever had sexual intercourse with an opposite-sex partner? (Penis in vagina)?" something like that (the questions are great-- very frank and straightforward.) Turns out abstinence does have a failure rate.
posted by WidgetAlley at 9:20 AM on November 27, 2014 [2 favorites]


I think it's interesting that withdrawal shows comparable success rates to condoms - for both "perfect" and "imperfect" use.

It really, really doesn't: the failure rates for perfect use of withdrawal are double that of the male condom. 1-year: 4% vs 2%. 10-year: 34% vs 18%.

What it does show is that imperfect use of condoms and of withdrawal are both fairly comparably bad -- as noted above, because "imperfect" includes "sometimes didn't use it."
posted by We had a deal, Kyle at 9:30 AM on November 27, 2014 [2 favorites]


It shouldn't just be p^x like they used, it should be (n ways to get x)p^x(1-p^(n-x), right?

Nah, it should just be 1-pr(no pregnancy that year)^n.

But those rates are worse than useless when making decisions about personal use, where you can factor in how consistent and responsible you're likely to be.

Surely, ex ante, the smartest bet is that you're going to use them in ways similar to other users. Or, differently, if we looked at the probability of becoming pregnant among couples who intended to be consistent and careful or who "knew" they were going to be consistent and careful, we'd almost certainly find that probability is much closer to the typical-use probability than the perfect-use probability.

Betting that you're going to be special and different from most people is, for most people, not a very smart bet.
posted by ROU_Xenophobe at 9:32 AM on November 27, 2014 [2 favorites]


One thing nobody has ever adequately explained to me : when talking about male condom failure, are they just talking about breakage? Or are there other, less-noticeable ways for them to fail? Like, if you take the condom off and you see it isn't broken, are you in the clear? Or do condoms sometimes have "invisible" defects, like microscopic holes that allow sperm to get through?

I've asked this question a million times over the years, and nobody's ever given me a satisfactory answer.
posted by evil otto at 9:33 AM on November 27, 2014


when talking about male condom failure, are they just talking about breakage? Or are there other, less-noticeable ways for them to fail?

"Perfect use" failure is breakage and leakage. "Typical use" failure includes those rare breaks and leaks, plus the much more common "I was losing my erection so I took the damn thing off" and "It was making me dry out so I told him to not bother with it this time."
posted by Dip Flash at 9:52 AM on November 27, 2014 [2 favorites]


Like, if you take the condom off and you see it isn't broken, are you in the clear?

Correct/perfect use is never to reuse a condom, and that's reuse; there could be sperm in there (if only a little bit) and you could transfer it to the outside by handling it. If it comes off for any reason, correct use is to get a new one.

Which is to say, perfect or correct use is hard.
posted by ROU_Xenophobe at 9:57 AM on November 27, 2014


Correct/perfect use is never to reuse a condom, and that's reuse

Good grief. Evil otto didn't say anything about putting the condom back on after use.
posted by Shmuel510 at 10:12 AM on November 27, 2014 [5 favorites]


Tryptophan is a hell of a drug?
posted by ROU_Xenophobe at 10:13 AM on November 27, 2014 [4 favorites]


I've asked this question a million times over the years
But never at the right place, apparently. It gets asked a lot on Scarleteen. The answer is always the same:
Condoms fail when they are not used for all genital contact from start to finish, used improperly or if they break, tear or slip off: they most often fail because of not being used for all genital contact. If a condom fails, it will be fairly obvious, and emergency contraception can be used. Condoms can also fail if a partner withdraws without holding the base of the condom while he does, as the condom may slip off.
posted by Too-Ticky at 10:15 AM on November 27, 2014 [2 favorites]


It should be noted that, as the iO9 piece points out, these "10 year" rates are simply extrapolations of one-year rates.
posted by aaronetc at 12:02 PM on November 27, 2014


Betting that you're going to be special and different from most people is, for most people, not a very smart bet.

Yet it has worked for us in multiple venues. Either we are that special and different, or the metrics are whacked.
posted by localroger at 6:38 PM on November 27, 2014


It really, really doesn't: the failure rates for perfect use of withdrawal are double that of the male condom. 1-year: 4% vs 2%. 10-year: 34% vs 18%.

Depends on how you look at it but the 10-year extrapolation does show how minor differences become more significant in the long run. I have some difficulty believing anybody actually makes "perfect" use of withdrawal for ten years, anyway - I suspect it's not too many who can say that for one - which of course is the real problem. I'd also be very interested in what happens to that number in the era of plan B.

I was kind of surprised that the diaphragm performed worse than withdrawal in the ideal case. Though I can kind of see how it might not be that reliable a setup. How many people even use a diaphragm anymore?
posted by atoxyl at 12:09 AM on November 28, 2014


My philosophy I guess is that being on the same page about what happens if someone does get pregnant is far more important than exactly what you do for birth control. In a long-term relationship I mean - outside obvs you gotta just wear the damn condom.
posted by atoxyl at 12:16 AM on November 28, 2014


I don't mean like "oh just get an abortion it's easy" more that if an unplanned pregnancy is going to be a major conflict situation you're kind of playing with fire just having sex on the regular!
posted by atoxyl at 12:26 AM on November 28, 2014


Betting that you're going to be special and different from most people is, for most people, not a very smart bet.

Many people are actually quite different from other people. When it comes to things like curating data and extrapolating probabilities, this is actually quite a big deal.

One may define "typical usage" of condoms to include taking the condom off, occasionally neglecting to use one, etc., but that will not magically turn every single condom user into a person who takes their condom off, occasionally neglects to use one, etc. Put another way, a fictive generality is not a real world specific: it may be true that X% of condom users make certain mistakes, but this does not govern how any given individual acts. If you aren't making those mistakes, then it makes no sense to pretend that you will, merely because some sloppy calculation assumed a probability that people in general would.

That's not even getting into all the other problems with these numbers.
posted by Sticherbeast at 4:54 AM on November 28, 2014


And of course this goes the other way as well - "well, I didn't misuse condoms that way" doesn't disprove that people in general have a certain failure rate for condom use. (Not that anyone said such a thing.) But, that still goes to the heart of the problem of tacitly assuming that typical condom users are a homogeneous group.
posted by Sticherbeast at 5:02 AM on November 28, 2014


With the advent of big data, I'm always surprised at how crappy birth control data is. I mean, I get that the data is difficult to collect and verify etc, but it's such important data. My dream data set would have:

A 1-5 scale with different compliance rates. For example, a 5 would be perfect use, and a 1 could be didn't even use the product 50% of the time, with 2-4 somewhere in between. The graph would have a sweet slider that would should the failure rate for each amount of compliance, as well as the fraction of the population that is however much compliant. It would be really interesting to see the differences there.

A comparison to fertility rates (and the confidence interval) for people not using any birth control

A stat comparing frequency of reproductive sex and pregnancy rate for different birth controls. I'd be interested in knowing if people that use diaphragms just have less of the type of sex that can get you pregnant.

Possibly longer time frames, but that's actually less interesting to me than the other data points. A year is fine, and relates better to how people actually use and switch birth control.
posted by fermezporte at 6:14 AM on November 28, 2014 [1 favorite]


the 10-year extrapolation does show how minor differences become more significant in the long run

The 10-year extrapolation is nonsense. It's killing me to read people on MeFi discussing these numbers like they mean anything. If you care about this information, please read this critique.
posted by Nelson at 8:12 AM on November 28, 2014


I can understand why The American Conservative doesn't like the numbers, since a world where contraception failure is relatively commonplace implies a world where abortion remains a critical backstop to most couples' contraception strategies. And the numbers are derived from a little toy model and don't claim to be otherwise. But I don't quite get why people are so insistent that birth control in particular should only ever be talked about with fully-specified models that take all relevant variables int o account. Hell, the whole implicit binary DV model is suspect; a better approach would be to run a count model on number of pregnancies.

The numbers do indeed tell you something. If you grabbed a random sexually-active couple and made them use condoms or the pill or whatever for ten years, in the way that people actually tend to use them, you'd expect many or most of them to get pregnant at least once unless they used sterilization or implants. This corresponds reasonably well to the data (from 1994, so the rate is presumably lower now) that about half of American women 15-44 had experienced at least one unintended pregnancy, so I don't see why it's so obviously unbelievable as a casual ballpark measure.
posted by ROU_Xenophobe at 9:09 AM on November 28, 2014


There's no "typical use" line for vasectomy because there's no wrong way to use a vasectomy.
posted by jepler at 9:12 AM on November 28, 2014


I don't see why it's so obviously unbelievable as a casual ballpark measure.

I have no problem with casual ballpark measures. My complaint is these precise line graphs that suggest this is some sort of statistical certainty. It's not, it's gross extrapolation. And it's dangerous, you get discussions like we have here where people draw incorrect conclusions like "condoms aren't much better than withdrawal".

I don't like the American Conservative's political stance either, but that critique is ad hominem. The article makes a couple of very simple arguments about why you can't take 1 year pregnancy rates and extrapolate those to 10 year rates. Fertility changes over 10 years. People's actaul use of contraception changes over 10 years. Frequency of sexual activity changes over 10 years. In no way can you pretend a 1 year trial is repeatable and independent so you can just extrapolate pn and call it a day.

Accidental pregnancies happen all the time, with birth control. We don't really have any meaningful numbers about what that rate is for people over long periods of time, like 10 years. These pretty graphs are misleading. I'm a huge fan of the NYT Graphics department and this kind of data journalism. This example is just really bad.
posted by Nelson at 9:26 AM on November 28, 2014 [1 favorite]


I can understand why The American Conservative doesn't like the numbers, since a world where contraception failure is relatively commonplace implies a world where abortion remains a critical backstop to most couples' contraception strategies.

Not only is this merely an attempt at an ad hominem argument, but the ad hominem argument is not even responding to any kind of point of view expressed in the article. Contrariwise, the author bemoans the prospect of people becoming less assiduous about using contraception. One might as well say that you oppose The American Conservative because you are a devout follower of Enver Hoxha. It's just nonsense.
posted by Sticherbeast at 11:33 AM on November 28, 2014


"It always weirds me out a little that they give numbers for "typical use" where "typical use" includes not using that method of birth control consistently."

Well it's not much use telling folks how well something works if it's done perfectly for 10 years. I don't do anything perfectly, let alone when handling a torture device or method for my penis. The deal with humans is you really have to work with and around them instead of trying to corral them into behaving in some specific way deemed best. If you know how people actually do things you can learn more than if you pretend they are ideal agents.
posted by GoblinHoney at 11:23 AM on November 29, 2014


« Older Munchkin the Teddy Bear gets her exercise   |   DIY Diagnosis: How an Extreme Athlete Uncovered... Newer »


This thread has been archived and is closed to new comments