CHA Family Medicine Residency

Tufts family medicine residents have "the best of both worlds"

Leave a comment

UberDiagnosis – Chapter 5: the NNT and NNH

You’ve heard about the “Number Needed to Treat” and the “Number Needed to Harm.” This statistic guesstimates how many people we need to screen in the clinic/hospital before we find a benefit or harm from a treatment.

Has your family or friends ever heard of the NNT? Have your patients ever asked: “what’s the NNH for this?


Well, this is an exciting year for biostatistics because NNT and NNH have hit pop culture and the mainstream media in a big way!


First, it started in October 2014 in the Wired Magazine. Then, PBS spread the word in December 2014.
Next, it hit the NY Times two months ago:How to Measure a Medical Treatment’s Potential for Harm


Then the AFP decided to get in on the action in this month’s AAFP: Introducing Medicine by the Numbers: A Collaboration of The NNT Group and AFP

This new series will appear in the online-only edition of AFP. Each month, medical editors from AFP and will select a topic to feature. We will use’s color-coding to quickly convey the relative merits of an intervention, and present the numbers for benefit and harm in a summary box. A discussion outlines the background evidence and the rationale for the rating, accompanied by key supporting references.



If you’ve worked with me, you might have heard me quoting things like “1 out of 8 to 9 people who are treated for AOM or sinusitis with antibiotics with GI upset and diarrhea.”

Where did I get these facts? How do I know them off the top of my head?

From the great website! In addition to facts on treatment, they also have launched a new section on diagnosis (and most relevantly here, on harms associated with overdiagnosis.)


Here are some of the relevant links to check out below. How often have you asked yourself a question about these clinical issues in the past month?

Primary Care: Diagnosis


Primary Care: Treatment



1 Comment

The Natural History of Benign Thyroid Nodules

Friday, we discussed the UBER sexy topic of thyroid nodules:

Durante et al, “The Natural History of Benign Thyroid Nodules,” JAMA. 2015; 313(9):926-935

Why do we care?

We all have “that patient” who has to get repeat thyroid ultrasounds and aspirations for either a nodule or multinodular disease. More than 90% of detected nodules are clinically insignificant benign lesions, but we are picking up more thyroid nodules everyday with our non-evidence-based annual physicals and incidentalomas on CT/MRI.

Currently, established guidelines for initial biopsy include:

– nodule size: > 1 cm

– sonographic characteristics: hypoechogenicity, irregular margins, taller-than-wide shape, intranodular vascular spots, microcalcifications

Furthermore, current guidelines recommend serial ultrasound exams for benign thyroid nodules and repeat needle biopsy if a nodule grows by 20%. The problem is that not much is known about the correlation between nodule growth and actual cancer.

This was a prospective study that followed some middle-aged Italian folks over five years with benign nodules, confirmed via laboratory testing and needle aspiration at the baseline evaluation. They received serial ultrasounds and a needle aspiration if they experienced nodule growth of 20% or if the ultrasound met the above-mentioned criteria for biopsy.

P: About a thousand Italian people, with a mean age of 52 years, 82% of whom were female. After loss to follow up and people who no longer met criteria (developed thyroid dysfunction), they were left with 992 patients.

I: This study wasn’t really about an “intervention,” rather it followed growth of nodules over time, so the groups were split into those who experienced growth vs…

C: Those who experienced no growth.

O: In 69% of patients, the thyroid nodules remained stable. 18% shrank spontaneously. In 2%, the nodules *disappeared.* 15% grew, most of which were in patients who had multiple nodules at baseline. 9% developed new nodules. “Over the course of the 5 years, 37% met re-aspiration criteria (growth or concerning features on ultrasound).” (growth or concerning features on ultrasound). Of these, 98.9% confirmed the baseline benign diagnosis.  Thyroid cancer was detected in 0.3% of the 1567 original nodules. 


– 1k patients in the grand scheme of things isn’t a whole lot, but it helps that this study underlines what we already suspected about thyroid nodules

– This doesn’t help us with younger people, in whom thyroid cancer is more prevalent

– 5 years of follow up is a drop in the bucket compared to the usual length of time our patients get follow up (eons) — additional studies should have a longer follow up time

– This was the most boring article EVER. Seriously. OMG.

Bottom Line: Don’t freak out about thyroid nodules! And the guidelines should probably space out ultrasounds because most of them won’t grow, anyway.

Leave a comment

Early Imaging for Back Pain

Today we discussed early imaging in back pain.

Jarvik JG et al. “Association of Early Imaging for Back Pain with Clinical Outcomes in Older Adults.” JAMA 2015; 313 (11): 1143-1153.

Population: 5k patients who are, according to the #’s, mostly white, over 65, with a high school or higher education, but with a smattering of other races and educational levels thrown in. Drawn from Harvard Vanguard, Kaiser Permanente of NoCA, and Henry Ford health systems.
Excluded: people with cancer or back surgery

Intervention: early back imaging < 6 weeks from “index visit”

Control: propensity-matched controls that did NOT undergo imaging

Outcome: Patient self-reported pain/function/depression at different intervals, mainly 3-, 6-, and 12-month intervals. Basically, 1300-1500 dollars more usage per patient in those who undergo early imaging. The short version is: early imaging largely results in same outcomes with more health care costs. Proportion of cancer diagnoses were not statistically different between the two groups!

Limitations of study: patient-reported data and a morass of CPT, ICD9, and RVU data pulled from medical records. How to account for osteoporotic patients?

Short version: you can safely avoid getting early (< 6 weeks) imaging in older folks who present with your basic back pain visit without worrying about “missing something.”

Why did this study matter? Most back pain literature is based on younger < 65 year old peeps.

Questions? Concerns? Comments?

Leave a comment

GI diet diary challenge!

We played a game on Tuesday as a part of the R2 conference on Dyspepsia and Changes in Bowel Habits.

  • A 36 year old Italian-American woman presents to you with a history of GERD for follow up. She brings in a diet diary with the following foods.
    • Which foods aggravate her symptoms?
    • Which foods alleviates her symptoms?

Here’s the answers we assembled.









Now, the harder challenge:

  • She returns with a new diagnosis of IBS that she self-diagnosed. She brings in (another!) diet diary with the following foods.
    • Which foods aggravate her symptoms?
    • Which foods alleviates her symptoms?



Can you think of any tasty recipes you would recommend to this woman with GERD and IBS?

Leave a comment

UberDx! 2

By the way, we appreciate all comments. It’s hard to know whether this stuff is just getting deleted, or scoffed at, yawned at, laughed at, marveled at or what.

Hopefully you read the wiki at
If not, please go for it. It’s short. Try to get an idea of Overdiagnosis Bias.
We had planned a second informative message for this week, but then this YouTube showed up. You gotta see this:



Lung-Cancer Screening with Low-Dose Computed Tomography

Bonjou, zanmi!  Today, in Journal Club, we discussed:

Lung-Cancer Screening with Low-Dose Computed Tomography

Gould, MK. NEJM 6 Nov 2014 pp1813-1820

The issue:

Lung CA is the leading cause of death from cancer in both men and women in the US. Despite the advances of modern medicine, only 18% of patients with lung CA are still alive 5 years after diagnosis. Early trials of chest x-rays and sputum screening don’t decrease lung cancer mortality. This engendered studies worldwide that demonstrated the sensitivity of CT scanning, particularly the low dose varieties that reduce radiation by 75-80% (but are still 10-15 times higher than plain films). The pitfalls of these studies are that they were vulnerable to lead-time bias (read: same mortality, earlier diagnosis) and overdiagnosis.

The data:

The National Lung Screening Trial enrolled 50,000 people at 33 US centers who were:

  • 55-74 years of age
  • At least 30 pack year hx
  • If a former smoker, had to have quit within previous 15 years

Randomly assigned to three rounds of annual screening with low-dose CT vs chest radiography. 20% reduction in lung cancer mortality was shown (247 vs 309 deaths per 100,000 patient-years) which translated into 3 fewer deaths from lung CA per 1000 high risk patients who underwent low-dose CT screening. This magnitude of benefit is at least as great as that reported for breast cancer mortality with annual (not biennial) mammo screening among women 50-59 years of age. Which is to say, more lead-time bias.

Of course: 39% of the participants had at least one positive result, and more than 95% of these findings were falsely positive. 10% of participants underwent tissue sampling. This adds to the data that show the reduction in lung-cancer mortality observed in the NLST is not as clear as the author is purporting.

Funky stuff:

  • USPSTF jumped on the bandwagon in Dec 2013 and released a grade B recommendation in favor of annual low-dose CT screening for high risk patients (55-80 w/ 30py smoking hx who are either currently smoking or quit within the last 15 years).
  • Conflicting committee opinions on how to apply the NLST data have resulted in how Medicare will cover this.
  • The greatest benefits of screening are found in those at highest risk for death from lung cancer, who, really, have the least to gain from screening than those at low risk!

Conclusions we drew from the conclusions

  • Annual lung cancer screening of high risk and former smokers with low dose CT mirrors annual breast cancer screening in 50-59 year old women with mammography. i.e. increases lead time bias without making a real impact on our patients’ qualities of life.
  • Though the article’s recs say that screening with low-dose CT prevents one in five deaths from lung cancer, we can’t fiddle with the numbers to make this work. Fellows: could you help us with this?
  • Lung cancer screening is NOT a single test, it is ANNUAL testing and the article makes no judgment about cost-effectiveness.
  • False positives Each examination is approximately 20 times as likely to yield a false positive as it is to reveal lung cancer.
    • Of those false positives, 5% will require invasive evaluation.
  • No surprise here: screening for lung CA with low dose CT is not a substitute for smoking cessation.

EBM points:

Always look for the NNT on these things. Per The NNT (

1 in 217 are helped (prevented death)
1 in 4 were harmed by false positive
1 in 30 were harmed by unnecessary surgery
1 in 161 were harmed by surgical complication

Per the NNT: “…Despite these caveats the significant and surprisingly large reduction in mortality using CT screening in this trial is promising. Because this is the first high quality randomized trial of CT screening it will take multiple further trials to confirm the benefit, and it will be critical to apply these data only to people at very high risk unless future trials expand to include others.”.

I dunno, y’all, our discussion made me more negative about it than The NNT seems to think.

Danielle (PGY-3)