Is the United States Preventive Services Task Force still a voice … · 2017. 5. 12. · States Preventive Services Task Force (USPSTF) has earned a reputation for its independent,

Is the United States Preventive Services Task Forcestill a voice of caution?Recent recommendations issued by the USPSTF have surprised some experts who say they maylead to overtesting and overtreatment. Jeanne Lenzerinvestigates

Jeanne Lenzer associate editor

The BMJ

Independent analyses of medical research are prized by doctorsconcerned about industry bias. Created in 1984, the UnitedStates Preventive Services Task Force (USPSTF) has earned areputation for its independent, objective guidance on clinicalpreventive services, and many have long viewed itsrecommendations as a bulwark against the increasing excessesof these services.However in 2009, the task force was the subject of a backlashafter issuing an unenthusiastic “grade C” recommendationregarding routine mammography screening of women aged 40to 49. In 2012, USPSTF found itself again in the hot seatfollowing its “grade D” recommendation against routine prostatecancer screening. Grade D recommendations are issued whenthe task force finds “moderate or high certainty that the servicehas no net benefit or that the harms outweigh the benefits.”Following these conservative recommendations, industry fundedlobbying efforts resulted in Congress seeking to terminate thetask force by cutting off funding to the Agency for HealthcareResearch and Quality (AHRQ), the government agency thatconvenes the task force. Several bills attempted to “zero out”AHRQ’s funding.1 2 In December 2015, Congress sent a warningshot to the agency and cut 8% of its budget.1

It is unclear what effects these threats have had. But in the pastfew years, the task force has issued several recommendationsthat are far more liberal in promoting interventions, which someexperts say will lead to overtesting and overtreatment.3-8

In 2013, USPSTF recommended screening high risk smokerswith low dose computed tomography screening.9 In January andFebruary 2016 it recommended routine depression screeningfor all adults and all teens,10 11 and in November 2016 itrecommended treatment with statins for primary prevention ofcardiovascular disease, replacing its 2008 recommendation thataddressed screening only and made no recommendationsregarding statins.12

The stakes are high. Task force recommendations are often usedto develop clinical practice guidelines and serve as a basis forinstitutional quality assurance measures. Grade A and B

recommendations also have financial implications, as USinsurers are mandated by law to pay for these services (althoughnot downstream expenses).13

Rigorous science?Some task force recommendations rely on questionable researchmethodologies, including the use of “indirect” evidence; failureto include unpublished data; failure to check reported outcomesversus pre-specified outcomes; and, in at least one case, to havepromulgated broad screening recommendations in the absenceof a single randomised controlled screening trial (RCT).Consider the task force’s 2016 recommendation to screen allchildren aged 12 years and older and all adults for depression,including pregnant and postpartum women.10 11 USPSTF assignedthese recommendations a B grade, meaning that “there is highcertainty that the net benefit is moderate or there is moderatecertainty that the net benefit is moderate to substantial.” Thetask force’s recommendations are in contrast with Britain andCanada which recommend against routine screening.14 15

However USPSTF’s recommendations occurred in the absenceof any RCTs assessing the health outcomes of screening versusnot screening in children, and just one such trial in adults whichreported that screening “leads to a modest increase in recognitionrates, but does not have consistently positive effects on patientoutcomes.”16

Brett Thombs, professor of psychiatry at McGill University,who has published research critical of the task forcerecommendations,6 told The BMJ: “In the absence of any trialevidence that screening would benefit patients, there is realconcern that these recommendations may lead to more harmthan good.”Because there are no RCTs of depression screening versus noscreening that show a health benefit, the task force relied onevidence from studies of the accuracy of screening tests and,separately, of depression treatment approaches.6

John Ioannidis, one of Thombs’ co-authors and professor ofmedicine at Stanford University, said that indirect evidence

[email protected]

For personal use only: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2017;356:j743 doi: 10.1136/bmj.j743 (Published 2017 February 21) Page 1 of 4

Feature

FEATURE

should not even be considered screening evidence. He said itignores the harms of screening and fails to take into account thesometimes important differences between the benefits of treatingan individual with a clinically apparent condition comparedwith those who are detected only through screening. Ioannidissaid that the harm-to-benefit ratios can vary greatly in these twopopulations.Historically, indirect evidence has led to failed and harmfulscreening recommendations, according to Barnett Kramer,director of the division of cancer prevention at the NationalCancer Institute, who told The BMJ that proponents of prostatecancer screening with the prostate specific antigen test andovarian cancer screening with Ca-125 relied on “insufficientand weak indirect evidence.” He said that subsequent directevidence—using RCTs comparing screened and unscreenedindividuals—demonstrated that “the weight of evidence is thatthe harms are not likely to be exceeded by benefits.”17 18 (In 2012USPSTF recommended against screening for these cancers.Both reviews are currently being updated.17 19)Kirsten Bibbins-Domingo, current USPSTF chair and professorat University of California, San Francisco, said: “We evaluatethe available evidence around preventive services by assessinga variety of valid trial designs and rigorously examining studiesfor potential bias. For adult depression, RCTs that randomizedwhether physicians received screening results showed positiveoutcomes. For children, the recommendation was made basedon the strength of the evidence chain.”For non-elderly adults the USPSTF contends that four trials, inaddition to the one that found no clinical outcome benefit,16 arescreening studies that support its recommendation.20-23 Accordingto the task force, these trials “screened all patients fordepression, enrolled only those screening positive, and returnedresults of screening to clinicians in the intervention grouponly.”10 The USPSTF told The BMJ that, if anything, this wouldhave led to more conservative estimates of screening benefit.24

Critics point out, however, that none of the four studies includedan unscreened control group, and three of the studies comparedoptimized with usual care, meaning that any benefit in healthoutcomes could be because of optimized care and not screening(though not all found a benefit).20-23 The fourth study providedusual care to both arms but found that the intervention arm, inwhich providers were told of screening results, “did not lead toimproved patient outcomes.”22

Wanda Filer, president of the American Academy of FamilyPhysicians, told The BMJ that the academy supported the taskforce recommendations on depression screening. Respondingto criticisms about their research methodology, she said, “It’snot the job of the task force to create a research agenda; it’s tolook at what’s out there and to make an interpretation. Is itperfect? No. And maybe more research is necessary, but theyare dealing with the hand they were dealt.”

Reliance on published, industry fundedstudiesA second concern stems from the fact that USPSTFrecommendations have been based on evidence reviews thathave not always included unpublished data.Researchers who led the evidence reviews underlying thedepression screening recommendations for children andnon-pregnant adults told The BMJ that they did not requestunpublished data from the US Food and Drug Administration(FDA) or the manufacturers—contrary to the agency’s own

guidance which states that “requests should be made to industryfor additional sources of unpublished data.”25

Many systematic reviews, even outside USPSTF, do not includeunpublished data from regulators and manufacturers.26 Howeverresearch by Erick Turner, a psychiatrist and former FDAreviewer, has highlighted the dangers of omitting unpublisheddata. In 2008 Turner showed that whereas FDA reviewersdescribed 38 antidepressant studies as positive and 36 asnegative or questionable, the published literature representingthese same studies showed 48 studies as positive and a merethree as negative or questionable.27 He told The BMJ that it wasimportant to take steps to mitigate industry and publication biaswhen conducting reviews, saying, “Relying solely on selectivelypublished, sponsor generated journal articles is a bit like relyingon TV ads to tell you which kind of car is the best.”Albert Siu, immediate past chair of the task force, defended itsreliance on published data, telling The BMJ that although“publication bias is a very important question . . . it takes us 24to 30 months from start to finish to release a report, and gettingnon-published information would drag this out for years.” Helater added by email, “The task force also relies on publisheddata because it has been vetted by the peer review process,which—while imperfect—can address many sources of potentialbias and methodological limitations that would be difficult tosort through independently.”But others rejected this logic. Jeffrey S Flier, former dean ofHarvard Medical School, told The BMJ that although peerreview is intended to improve the accuracy and truthfulness ofpublished data, “a surprising amount of what we publish afterpeer review is still not true.” Flier added that skilled reviewerswho analyse unpublished data might be able to increase thelikelihood that a conclusion is actually correct.Carl Heneghan, professor of evidence based medicine at theUniversity of Oxford added: “The assessment of evidence fortreatment effects should not be undermined by reporting biasthat occurs when evidence synthesis is based on journalpublications alone. The major issue is the under-reporting ofharms which has the effect to overestimate the net benefits.”The AHRQ, however, stands by Siu’s position, stating that theagency “places the greatest weight on the published data.”The evidence report underlying USPSTF’s statinrecommendations has similar shortcomings. Roger Chou,professor of general internal medicine and director of the PacificNorth west Evidence Based Practice Center, which conducteda systematic review of statins for the USPSTF, told The BMJthat they did not request data from the Cholesterol TreatmentTrialists’ (CTT) Collaboration because “CTT has not shareddata in the past, so we did not expect that they would be willingto share data this time.”The evidence review led by Chou relied on 19 studies of primaryprevention with statins compared with placebo or no statins.12

The reviewers reported an absolute risk difference in all causemortality of −0.40% (95% CI, −0.64% to −0.17%); howeverthey were forced to rely almost entirely on published, industryfunded studies (a single non-industry funded study weighted ascontributing 0.2% of the mortality calculation was included).8 28

Chou discounted the importance of obtaining the CTT data onstatins for primary prevention, however, stating, “I don’t thinkit is likely that having the individual patient data would changethe findings—the results are very stable across studies and inmultiple sensitivity and subgroup analyses.”Other limitations—each known to potentially inflatebenefits—raise questions about the reliability of the task forcerecommendation: four of the 19 studies did not report

For personal use only: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2017;356:j743 doi: 10.1136/bmj.j743 (Published 2017 February 21) Page 2 of 4

FEATURE

information on all cause mortality and therefore could not beincluded in the mortality calculation; one major trial was stoppedearly for benefit29 30; and several of the 19 trials failed tospecifically ask about common adverse effects of statins.8 12

Despite these limitations, the task force issued a grade Brecommendation for the use of statins in people aged 40 to 75years without a history of cardiovascular disease and with oneor more risk factors and a calculated 10 year risk score of 10%or greater.12

Outsourcing evidence reviewsThe USPSTF generally does not conduct its own systematicreviews of the evidence, but often outsources the work toevidence based practice centers (EPCs),31 after which it issuesa recommendation.The apparent liberalization of recent screening recommendationsraises questions of whether financial conflicts could affect taskforce recommendations. However, The BMJ found that bothUSPSTF members and the individual EPC researchers selectedto work on reviews were almost entirely free of financialconflicts.Kenneth Lin, associate professor of family medicine atGeorgetown University and former task force staffer, led thetask force’s in-house evidence review that led to the grade Drecommendation against prostate cancer screening. An intensebacklash followed.Lin told The BMJ that negative reactions against earlierconservative recommendations may, however, affect how taskforce members interpret evidence—along with a desire not to“deprive” people of an option for screening that could be madefree with a grade A or B recommendation.AHRQ requires task force members and EPC researchers todisclose financial interests in excess of $1000 (£802; €941).The agency can manage such conflicts by requiring divestment,barring the individual from voting or participation, or bygranting an exemption if the agency believes their expertise isnecessary.32 33

While individual task force and EPC members are mostly freeof financial conflicts, several EPCs receive industry funding.This creates potential institutional conflicts of interest— researchgroups may be incentivized to accept study designs favouredby the sponsor in order to maintain funding streams. Over time,sponsors may shift contracts away from research groups thatare too demanding about research design and award contractsto more pliant research groups.34 35 Given the history ofCongressional punishment of the AHRQ for publishing resultsbelieved to interfere with industry interests, conflicts at theinstitutional level—rather than individual researcher level—areof concern.1 36

A spokesperson for RTI International—part of the RTIInternational-University of North Carolina EPC that providedan evidence review for the USPSTF’s childhood depressionscreening recommendations—told The BMJ that RTI receivedaround $161m of its $807m funding in 2014 from “commercialand non-governmental” sources. Kaiser’s EPC—whichconducted the adult depression review—gets 10% of its fundingfrom industry, and another EPC, Duke Clinical ResearchInstitute, receives 63.3% of its $331m annual revenue fromindustry.37

Too much medicineSome experts contacted by The BMJ said that the USPSTF’sadvice for depression screening will lead to inappropriatetreatment.The task force acknowledged that the “evidence on treatmentbenefit for pregnant women is primarily for nonpharmacologicinterventions (cognitive behavioral therapy, for example),” andthat there is “evidence of potential serious fetal harms frompharmacologic treatment.”These cautions echo the findings of an EPC report commissionedby the AHRQ from the Pacific North west EPC based in OregonHealth and Science University. This EPC receives no industryfunding. That review, led by Marian McDonagh, reported that“the majority of the evidence was indirect” and was “related tothe use of selective serotonin reuptake inhibitors (SSRIs) takenduring pregnancy [with]…little evidence” regarding the use ofnon-pharmacological interventions.38 McDonagh and colleaguesreported that data were mostly derived from observationalstudies of pregnant women who did and did not takeantidepressants for any reason, and that “the proportions ofwomen with depression in either group are rarely reported oranalyzed.”The researchers said that they could not draw any conclusionsabout the effects of SSRIs on maternal depression or on itseffects on the fetus or infant during pregnancy and breastfeeding, as the evidence was “conflicting” and “insufficient.”Nonetheless, the USPSTF concluded that the “likelihood ofthese serious harms is low,” and recommended screening forall pregnant and postnatal women. The task force added that“clinicians are encouraged to consider cognitive behavioraltherapy or other evidence based counseling interventions whenmanaging depression in pregnant or breastfeeding women,” andthat screening “should be implemented with adequate systemsin place to ensure accurate diagnosis, effective treatment, andappropriate follow-up.”Allen Frances, chair of the task force that wrote the Diagnosticand Statistical Manual (DSM-IV), and a well known critic ofoverdiagnosis, told The BMJ that current services for severelymentally ill people were already strained to bursting point. “Wedon’t need to create an army of mislabeled healthy people. Weshould first take adequate care of people who are already sickand in urgent need.”7

Alan Roth, chair of the primary care council of the Right CareAlliance, a non-profit organization based in Boston that worksto ensure patients are neither undertreated nor overtreated, raisedsimilar concerns. He told The BMJ that he often saw poor andmiddle class patients who were unable to access mental healthservices.39 40 He said that depression screening tests—some ofwhich, like Pfizer’s PHQ2 and PHQ9 tests, are made bycompanies that manufacture antidepressants—are “certainlyvery sensitive but we know they’re not very specific.” Withoutadequate services in place to detect false positive test resultsand to provide non-drug treatment, Roth said that screeningtests were likely to drive overtreatment with antidepressants.41 42

A better way forwardSome experts say that before making population based screeningand prevention recommendations independent researchers shouldanalyze more forms of raw data such as clinical study reportsand patient level data, as was performed for the Cochrane reviewof oseltamivir and the paroxetine Study 329 re-analysis.43 44

For personal use only: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2017;356:j743 doi: 10.1136/bmj.j743 (Published 2017 February 21) Page 3 of 4

FEATURE

When such rigoros analyses are not possible, it is important toacknowledge the resulting uncertainty, says Heneghan, whoco-authored the Cochrane review. Heneghan told The BMJ thatclinical practice recommendations “should ideally aim to reduceuncertainty around practice based interventions and increasethe quality of care; but use of low quality evidence often hasthe opposite effect.”He added, “In the face of no evidence, or very low qualityevidence, guideline writers should refrain from makingrecommendations. Indeed many guideline bodies would betterserve clinical practice by making fewer recommendations:reflecting more of the uncertainty around treatment decisions.What we need are fewer recommendations and more high qualityevidence to base decisions on. Currently we seem to be seeingthe exact opposite.”In light of the AHRQ’s critical role in making screening andprevention recommendations, better funding of the agency couldenhance its ability to conduct in-depth and in-house analysesthat would be less subject to the demands of industry and moreresponsive to the public interest.

Special thanks to Peter Doshi.Competing interests: JL has coauthored articles with John Ioannidisand Kenneth Lin.Provenance and peer review: Commissioned; externally peer reviewed.

1 Leventhal R. AHRQ sees budget cuts, but agency remains in place. 2015. www.healthcare-informatics.com/news-item/ahrq-sees-budget-cuts-agency-remains-place.

2 Frieden J. House committee votes to defund AHRQ: attempt to restore funding fails onvoice vote. 2015. www.medpagetoday.com/publichealthpolicy/washington-watch/52294.

3 Bach PB, Mirkin JN, Oliver TK, et al. Benefits and harms of CT screening for lung cancer:a systematic review. JAMA 2012;356:2418-29. doi:10.1001/jama.2012.5521 pmid:22610500.

4 Borgmeyer C. Evidence lacking to support or oppose low dose CT screening for lungcancer, says AAFP. 2014. www.aafp.org/news/health-of-the-public/20140113aafplungcarec.html.

5 Prasad V, Lenzer J, Newman DH. Why cancer screening has never been shown to “savelives”--and what we can do about it. BMJ 2016;356:h6080. doi:10.1136/bmj.h6080 pmid:26740343.

6 Thombs BD, Ziegelstein RC, Roseman M, Kloda LA, Ioannidis JP. There are norandomized controlled trials that support the United States Preventive Services TaskForce Guideline on screening for depression in primary care: a systematic review. BMCMed 2014;356:13. doi:10.1186/1741-7015-12-13 pmid:24472580.

7 Frances A. Screen everyone for depression? Good intention, very bad idea. 2016. www.newscientist.com/article/2075249-screen-everyone-for-depression-good-intention-very-bad-idea.

8 Redberg RF, Katz MH. Statins for primary prevention: the debate is intense, but the dataare weak. JAMA 2016;356:1979-81. doi:10.1001/jama.2016.15085 pmid:27838702.

9 Moyer VA. U.S. Preventive Services Task Force. Screening for lung cancer: U.S.Preventive Services Task Force recommendation statement. Ann Intern Med2014;356:330-8.pmid:24378917.

10 Siu AL, Bibbins-Domingo K, Grossman DC, et al. US Preventive Services Task Force(USPSTF). Screening for depression in adults: US Preventive Services Task Forcerecommendation statement. JAMA 2016;356:380-7. doi:10.1001/jama.2015.18392 pmid:26813211.

11 Siu AL. U. Preventive Services Task Force. Screening for depression in children andadolescents: US Preventive Services Task Force recommendation statement. Ann InternMed 2016;356:360-6. doi:10.7326/M15-2957 pmid:26858097.

12 Bibbins-Domingo K, Grossman DC, Curry SJ, et al. US Preventive Services Task Force.Statin use for the primary prevention of cardiovascular disease in adults: US PreventiveServices Task Force recommendation statement. JAMA 2016;356:1997-2007. doi:10.1001/jama.2016.15450 pmid:27838723.

13 Health Resources and Services Administration. What are federally qualified health centers(FQHCs)? www.hrsa.gov/healthit/toolbox/RuralHealthITtoolbox/Introduction/qualified.html.

14 Joffres M, Jaramillo A, Dickinson J, et al. Canadian Task Force on Preventive HealthCare. Recommendations on screening for depression in adults. CMAJ 2013;356:775-82.doi:10.1503/cmaj.130403 pmid:23670157.

15 UK National Screening Committee. Screening for depression. 2015. http://legacy.screening.nhs.uk/policydb_download.php?doc=522.

16 Williams JW Jr, , Mulrow CD, Kroenke K, et al. Case-finding for depression in primarycare: a randomized trial. Am J Med 1999;356:36-43. doi:10.1016/S0002-9343(98)00371-4 pmid:10320115.

17 Moyer VA. U.S. Preventive Services Task Force. Screening for prostate cancer: U.S.Preventive Services Task Force recommendation statement. Ann Intern Med2012;356:120-34. doi:10.7326/0003-4819-157-2-201207170-00459 pmid:22801674.

18 Fedorkow D. ACP Journal Club. Annual screening with CA-125 and transvaginalultrasonography did not reduce ovarian cancer incidence or mortality. Ann Intern Med2011;356:JC3-8. doi:10.7326/0003-4819-155-6-201109200-02008 pmid:21930845.

19 Moyer VA. U.S. Preventive Services Task Force. Screening for ovarian cancer: U.S.Preventive Services Task Force reaffirmation recommendation statement. Ann InternMed 2012;356:900-4. doi:10.7326/0003-4819-157-11-201212040-00539 pmid:22964825.

20 Wells KB, Sherbourne C, Schoenbaum M, et al. Impact of disseminating qualityimprovement programs for depression in managed primary care: a randomized controlledtrial. JAMA 2000;356:212-20. doi:10.1001/jama.283.2.212 pmid:10634337.

21 Rost K, Nutting P, Smith J, Werner J, Duan N. Improving depression outcomes incommunity primary care practice: a randomized trial of the quEST intervention. QualityEnhancement by Strategic Teaming. J Gen Intern Med 2001;356:143-9. doi:10.1111/j.1525-1497.2001.00537.x pmid:11318908.

22 Bergus GR, Hartz AJ, Noyes R Jr, et al. The limited effect of screening for depressivesymptoms with the PHQ-9 in rural family practices. J Rural Health 2005;356:303-9. doi:10.1111/j.1748-0361.2005.tb00099.x pmid:16294652.

23 Jarjoura D, Polen A, Baum E, Kropp D, Hetrick S, Rutecki G. Effectiveness of screeningand treatment for depression in ambulatory indigent patients. J Gen Intern Med2004;356:78-84. doi:10.1111/j.1525-1497.2004.21249.x pmid:14748864.

24 O’Connor E, Rossom RC, Henninger M, et al. Screening for depression in adults: anupdated systematic evidence review for the US Preventive Services Task Force. EvidenceSyntheses No. 128. www.ncbi.nlm.nih.gov/books/NBK349027.

25 Agency for Healthcare Research and Quality. Methodology. www.ahrq.gov/research/findings/evidence-based-reports/technical/methodology/index.html.

26 Schroll JB, Bero L, Gøtzsche PC. Searching for unpublished data for Cochrane reviews:cross sectional study. BMJ 2013;356:f2231. doi:10.1136/bmj.f2231 pmid:23613540.

27 Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R. Selective publication ofantidepressant trials and its influence on apparent efficacy. N Engl J Med 2008;356:252-60.doi:10.1056/NEJMsa065779 pmid:18199864.

28 Vaccarino V, Bremner JD, Kelley ME. JUPITER: a few words of caution. Circ CardiovascQual Outcomes 2009;356:286-8. doi:10.1161/CIRCOUTCOMES.109.850404 pmid:20031850.

29 Ridker PM, Macfadyen JG, Nordestgaard BG, et al. Rosuvastatin for primary preventionamong individuals with elevated high-sensitivity c-reactive protein and 5% to 10% and10% to 20% 10-year risk. Implications of the Justification for Use of Statins in Prevention:an Intervention Trial Evaluating Rosuvastatin (JUPITER) trial for “intermediate risk”. CircCardiovasc Qual Outcomes 2010;356:447-52. doi:10.1161/CIRCOUTCOMES.110.938118 pmid:20736443.

30 Bassler D, Briel M, Montori VM, et al. STOPIT-2 Study Group. Stopping randomized trialsearly for benefit and estimation of treatment effects: systematic review and meta-regressionanalysis. JAMA 2010;356:1180-7. doi:10.1001/jama.2010.310 pmid:20332404.

31 Agency for Healthcare Research and Quality. Evidence based practice centers programoverview. 2015. www.ahrq.gov/research/findings/evidence-based-reports/overview/index.html.

32 Effective Healthcare. Agency for Healthcare Research and Quality. AHRQ evidence basedpractice center policy on financial and nonfinancial interests. www.effectivehealthcare.ahrq.gov/ehc/assets/File/epcdisclosureform.pdf.

33 US Preventive Services Task Force. Conflict of interest disclosures. www.uspreventiveservicestaskforce.org/Page/Name/conflict-of-interest-disclosures.

34 Lenzer J. Truly independent research?BMJ 2008;356:a1332. doi:10.1136/bmj.a1332 pmid:18719008.

35 Brownlee S, Lenzer J. Painkiller trial raises questions for FDA, Pfizer center for publicintegrity. 2009. www.publicintegrity.org/articles/entry/1203.

36 Deyo RA, Psaty BM, Simon G, Wagner EH, Omenn GS. The messenger under attack --intimidation of researchers by special-interest groups. N Engl J Med 1997;356:1176-80.doi:10.1056/NEJM199704173361611 pmid:9099665.

37 Duke Clinical Research Institute. Annual report: answering today’s questions for tomorrow’sworld. 2015. https://dcri.org/about/annual-report.

38 McDonagh M, Matthews A, Phillipi C, et al. Antidepressant treatment of depression duringpregnancy and the postpartum period. 2014. www.ncbi.nlm.nih.gov/books/NBK233904.

39 Bishop TF, Press MJ, Keyhani S, Pincus HA. Acceptance of insurance by psychiatristsand the implications for access to mental health care. JAMA Psychiatry 2014;356:176-81.doi:10.1001/jamapsychiatry.2013.2862 pmid:24337499.

40 Bishop TF, Ramsay PP, Casalino LP, Bao Y, Pincus HA, Shortell SM. Care managementprocesses used less often for depression than for other chronic conditions in US primarycare practices. Health Aff (Millwood) 2016;356:394-400. doi:10.1377/hlthaff.2015.1068 pmid:26953291.

41 Dowrick C, Frances A. Medicalising unhappiness: new classification of depression risksmore patients being put on drug treatment from which they will not benefit. BMJ2013;356:f7140. doi:10.1136/bmj.f7140 pmid:24322400.

42 Mojtabai R. Clinician-identified depression in community settings: concordance withstructured-interview diagnoses. Psychother Psychosom 2013;356:161-9. doi:10.1159/000345968 pmid:23548817.

43 Jefferson T, Jones M, Doshi P, Spencer EA, Onakpoya I, Heneghan CJ. Oseltamivir forinfluenza in adults and children: systematic review of clinical study reports and summaryof regulatory comments. BMJ 2014;356:g2545. doi:10.1136/bmj.g2545 pmid:24811411.

44 Le Noury J, Nardo JM, Healy D, et al. Restoring Study 329: efficacy and harms ofparoxetine and imipramine in treatment of major depression in adolescence. BMJ2015;356:h4320. doi:10.1136/bmj.h4320 pmid:26376805.

Published by the BMJ Publishing Group Limited. For permission to use (where not alreadygranted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions

For personal use only: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2017;356:j743 doi: 10.1136/bmj.j743 (Published 2017 February 21) Page 4 of 4

FEATURE