Natural Family Planning Apples to Artificial Contraception Oranges
Longtime Vox Nova readers will be familiar with my habit of writing about Natural Family Planning from a supportive yet self-critical point of view. I seem to have found a kindred spirit. Over at Roma Locuta Est, Jake Tawney has recently published a post called The Effectiveness of Natural Family Planning. In it he dismantles the idea that NFP is 99% effective.
Now, before you get your dander up you need to know two things. First of all, Jake is supportive of Church teaching and is, in fact, a certified NFP teacher. Second, you need to have a look at his math. As far as I can tell, he is spot on when he argues that, in practice, NFP is something closer to 67% effective over the full course of a couple’s fertile time together.
This may look somewhat alarming at first glance, but it is important to note that Jake’s observations apply almost as well to artificial methods and that none of them approach the advertised 99% effectiveness either. (I have, in previous posts, suggested NFP is as effective as most other methods, but without resorting to numbers.) In fact, when I give talks to young people I often ask them to put up their hands if they know anyone who got pregnant while using a condom or the pill. Even among younger audiences, a significant number – I’d guess a large minority – of hands go up.
The fact is that both those who would sell you artificial contraceptives and those who would “sell” you NFP are happy with statistical models that give them effectiveness ratings well above what your own empirical experience is likely to indicate. And both can blame poorer than expected performance on “user failure.”
I don’t want to go into this too much more here. Jake does a nice job of it himself and I recommend you go check out his post as well as the comments following it, in which he gives more detail on his mathematical model in response to queries from his readers. His basic conclusion is that we should not use its supposed 99% effectiveness as a selling point, or at the very least not as the primary selling point, of NFP. Such a tactic is bound to backfire when people actually live in the 67% world. Rather, Jake says, we should do it because it is what we believe to be right.
I agree with Jake, but what I do want to do here is take the opportunity to highlight one other reason why I feel that comparing NFP to artificial contraception statistically is not particularly productive. I have noted before that NFP differs from artificial contraception because it consists in, basically, not doing something. Now, from a statistical point of view, it is tricky to compare the effects of not doing something with the effects of doing something.
Of course, NFP can be used to 100% effectiveness. One can totally abstain and one will not get pregnant. But the fact that it is not 100% effective means that the people using it actually do engage in “pregnancy-causing behavior” on occasion. This behavior is, for NFP users, different than for artificial contraception users. Let me try to explain.
AC users are either using AC or they are not. If you didn’t take your pill or didn’t use your condom, then you weren’t using your AC. User failure is very easy to identify. There is no equivalent here for NFP users. There is nothing to point to in order to say, “Oh, we didn’t use our NFP that night.” All that means is that we didn’t abstain (something which the AC users also didn’t do).
But if not abstaining is tantamount to user failure then NFP becomes, by definition, 100% effective! Every single pregnancy is the result of user failure.
No, in order for NFP to be usefully compared to AC in terms of effectiveness rates there must be an equivalent kind of method failure. If NFP is not to be 100% effective by definition there must be cases where people who are absolutely sure of their infertility make love and get pregnant. But is this actually possible? Can you be as certain of your infertility as of the fact that you took your pill consistently or that you are wearing a condom? I think not.
The thing about using NFP is that you are conscious of the fact that, even with the best information you can gather, you are never 100% sure about the state of your fertility. You can be fairly certain, and less certain, but you can’t be 100% sure. You yourself have to decide just how certain you need to be before deciding whether or not to make love. A couple that is in a drastic situation (e.g., if pregnancy threatens the life of the woman) may feel that they are almost never certain enough. But if individual couples have to decide how certain is certain enough for them, how can this be reflected in studies comparing effectiveness rates? I, for one, don’t think it can be.
This means at least two things. First of all, it means that much of the time NFP couples are in a kind of “yellow-light” phase in which they are uncertain about the status of their fertility, a phase that has no equivalent for AC couples. And without such an equivalent, statistical comparisons mean less than they seem to on the surface. Secondly, it means that typical NFP users, unlike typical AC users, are conscious that pregnancy-causing behavior always has the chance of causing pregnancy. Armed with such knowledge, NFP users are much better equipped to deal with method or user failure, no matter how common or rare, than those who have been tricked into believing that sex doesn’t make babies.
Brett Salkeld is a doctoral student in theology at Regis College in Toronto. He is a father of two (so far) and husband of one. He is the co-author of How Far Can We Go? A Catholic Guide to Sex and Dating.
Comments are closed.
Ah, the yellow light time! I would argue that if you are willing to have a baby early on, you will get much better at figuring out your “uncertain” times and know to avoid them. I gave birth twice in the first 30 months of my marriage but have not become pregnant again in the following 26 months. We have learned that, for me, we can only be (reasonably)certain of avoiding pregnancy during the final 7-8 days of my cycle, so we limit sex to that time. It’s difficult, but after having two in a row so quickly, I really wanted some time to have the older in preschool or kindergarten and the younger potty-trained before taking on another baby.
For the record, we have a timetable set for ceasing our “avoidance,” as we believe we are reaching the limit of justifying NFP usage for the spacing of children, and don’t have any other serious factor that would warrant avoiding pregnancy. If and when a third child is born, we’ll have to think about that again.
The effectiveness rate, however, is usually quoted in reference to how long a couple can go after marriage before conceiving their first child. Any studies on its effectiveness later in the marriage after the first child or two?
Julie,
Are you my wife incognito? Our situations are eerily similar!
Sorry, Brett, unless my husband has taken to using a different online alias than usual, I’m not your wife. But it’s always nice to know that we’re not alone out there!
Brett, this is kind of sad in a sense. My biggest fear is that if this stat goes heavily viral, left-wing groups or people who HATE the Church’s support of NFP and ban on contraception and abortion will use it as a weapon to strike at NFP and further convince people the mainstream (yet immoral and drastically consequential) standard methods of fertility “regulation” are the way to go.
They already believe NFP doesn’t work. Hopefully they learn that their own preferred methods don’t fare any better.
I skimmed Jake’s argument and I take it with a grain of salt. I am not a bio-statistician, but I am a mathematician and humble enough to both understand a lot of the math involved and to recognize the many methodological problems (that are beyond me) involved in defining and determining “effectiveness rates.” Jake’s argument is facile, but I suspect that it would break down under careful analysis by a bio-statistician who works in the field of fertility analysis.
Of course, if Jake is a bio-stats guy, I will need to go back and review his argument again more carefully.
David,
I certainly appreciate the time you took to skim my article. Quite frankly, I was a bit surprised at the rate in which the piece took off. I also appreciate your call for a bio-statistician. Certainly there are nuances that occur throughout one’s entire fertility life that will impact the effectiveness rates of NFP. I even took some pains to point out a few of these nuances, some in the post itself and others in the comments. For this very reason, I would never tie myself to the 67% lifetime figure. However, when the one-year rate is extrapolated to a smaller time period, say the first five years of married life, the multiplicative property can be assumed to be more accurate than a longer-term extrapolation.
I think that perhaps you missed the forrest for the trees. The point was never that 67% (or some of the other numbers in the article) were perfectly accurate. The point was that the 99% that is quoted is a one-year rate, and a one-year rate can be highly misleading for most folks. When they hear “99% effective” they almost always think in terms of lifetime fertility (of 100 couples who practice NFP, only one will experience an unintended pregnancy … ever). Feel free to take issues with the actual numbers – you are correct, I am no bio-statitician – and I readily admit that the lifetime model would be far more complicated. This does not, however, negate either of the main points:
1. Method effectiveness is different than user effectiveness, the latter being always lower than the former (this is from the research literature, not from my own calculations), and when people report 99% they typically fail to mention the distinction.
2. “99%” (or lower if user effectiveness) is a one-year rate. Over more than a year, the rate is surely something different, more than likely worse.
If you have some specific objections to these two points, I am happy to hear them. At the very least, thank you for your comments and interest in the piece.
God bless,
Jake Tawney
Jake,
Thanks so much for joining us in the discussion of your post. I hope you’ll stick around.
Brett
Jake,
thank you for a patient reply to a very quick, off the cuff response.
With regards to your two main points: yes, I completely agree that user effectiveness and method effectiveness are very different.
With regards to your second point, I think that lifetime versus one year effectiveness will be different, but I am not convinced they will diverge rapidly. In your model you assumed that each year is an independent event (so as to use the binomial distribution model) but this strikes me as a very iffy assumption. With any method it is reasonable to assume practice makes perfect, so certainly lifetime user effectiveness may be asymptotic to a value only a bit lower than one year effectiveness. (Something like this may be going on in the normalizations you mentioned seeing in the literature.) I don’t know which model is better: hence my call for a bio-statistician or at least good references in the literature.
Yes, it is true that years are not completely independent, hence the binomial model is an estimate. It is also true that fertility itself declines as the year go on, so we would expect later years in a marriage to have a higher “efficiency rate”. However, the literature tries to use a cross section of couples, not just new couples, so a 99% quoted one-year rate already has an “average” of sorts built into it, and as such, the years approach independence. I know the situation is more complicated than this, but to simplify it a bit, suppose that five consecutive years have the following efficiencies: 91%, 92%, 93%, 94%, and 95%. If we treat them as is, the combined efficiency is 68%. If we use the average (93%) and compound over five years, the result is 69% … pretty close.
In fact, this actually serves to strengthen my conclusion. The product of the actual probabilities per year is always less than (or equal) to the average of those probabilities to the nth power by the AM-GM inequality. Thus, while using the average is an estimate, the actual number should be lower than the one’s arrived at by assuming an average, not higher as you surmised.
I hope this helps.
David Cruz-Uribe,
I’m unclear what sort of magic you think would be going on which would turn a “X% of the study group became pregnant while using the method effectively during one year” into something else, nor how taking the studies to say what they say is “facile”.
Certainly, there are aspects of planning or conducting such a study which might, depending on what you were trying to do, require a bio-statistician, but ascertaining what a one year effectiveness rate statistic means over a period of years (until you stretch that out enough that you need to calculate how a given person’s fertility would vary over time) doesn’t even require vary advanced math abilities, much less specialized bio-stats knowledge.
If you think there’s a specific problem with the argument, describing what the problem is would probably be much more persuasive than vaguely invoking the need for a knowledge of bio-statistics in order to interpret what is, on the face of it, a very simple statistic — indeed one that is routinely given to laymen with no knowledge of the field.
DC, I addressed some of this in my exchange with Jake. For the rest: I know enough to know that I am ignorant, and that there are many deep methodological issues involved in creating effective and accurate measurements of fertility, etc. The math may be “simple” but its application is subtle. Thus my invocation of professionals.
We are speaking of two different statistics. In one case, we are speaking of the odds of getting a blackjack for any given hand. In the other case, we are speaking of the odds of getting a blackjack playing consistently over a finite period. Other terms that are thrown out are theoretical versus effective error.
Each statistical measure has its defects. A 70% lifetime effective rate means there is:
30% chance of having one unanticipated child
9% chance of having two unanticipated children
3% chance of having three unanticipated children
The second and third numbers will be lower in practice due to a year of fertility being consumed each time.
In contrast, not actively trying to prevent pregnancy results in a 15-20% monthly pregnancy rate; hence one definition of infertility being not able to achieve pregnancy within one year.
This is basic statistics.
If the chance of not getting pregnant on one occasion .99/1.00 = .99
If they have sex two times the chance of not getting pregnant is
(.99/1.00) times (.99/1.00) = .9801 (.99 squared)
If they have sex three times the chance of not getting pregnant is
(.99/1.00) times (.99/1.00) times (.99/1.00) = .9702 (.99 to the third power)
If they have sex one hundred times the chance of not getting pregnant is .99 to the 100th power = .3660
Or about one chance in three that they will not get pregnant over this period.
But on each occasion the probability is .99.
Numbers rounded to 4 decimal places, this assmes a lot of constants that are most likely not constant.
Use the scientific calculator under accessories in Microsoft Windows
http://faculty.uncfsu.edu/dwalla
Hank, this calculation of probabilities assumes that the events are independent. For a simple model this is fine, but for more realistic modeling, this assumption is called into question. And once you lose the assumption of independence, the calculation of contingent probabilities is much more subtle.
David is right here. The events do have to be independent in order to apply to multiplicative property. However, the literature uses an average probability across the various years, so the exponentiation of this average can serve as an estimate (probably an overestimate). See my previous explanation for more details.
Sorry. I meant to add this above … the .99 is not in reference to one act of sex. That is not how the Pearl index is calculated. In the NFP literature, when they say 99%, they mean that of 100 women (of various levels of NFP experience) who used a method for one year, 1 experienced an unintended pregnancy during that year. So the multiplication would have to occur as a function of years not as a function of how many times the couple had sex. (I will not repeat here the distinction between user and method effectiveness, but that was the other part of my argument to dismiss the 99% figure.)
Just out of curiousity, did my explanation of why NFP is difficult to compare statistically to AC make much sense to the reader? It is something I feel like I have trouble articulating.