Optimization of Clinical Breast Examination

Article Outline

• Abstract
• Materials and Methods
• Results
• Discussion
• Conclusions
• References
• Copyright

Abstract 

Background

Breast examination is necessary for evaluation of the 8% to 17% of cancers missed by mammograms, but it is being done less often and less effectively. To improve the use of breast examination, we tested whether a technique to focus attention could improve the call rate (the percent of examinations leading to further evaluation), a measure of quality, without retraining in examination technique.

Methods

Clinicians were randomized to complete 1 of 2 dedicated, de-identified forms after routine breast examination: a long form intended to focus attention by requesting general breast descriptors along with clinical information and breast examination findings (10 clinicians recorded 964 examinations) or a short form recording only clinical information and examination findings (11 clinicians recorded 558 examinations). There was no technique retraining. Study call rates were compared with historical controls (298 breast examinations by 16 clinicians).

Results

The call rates by the study groups of clinicians were similar, but the call rate using either form (8.3%) was significantly higher than the call rate in the preceding year when no dedicated form was used (4.7%; P=.031), suggesting a Hawthorne effect in which altering conditions of data collection (using the dedicated forms) functioned as an independent variable. Surveillance, Epidemiology, and End Results data predicted 3.4 cancers in all 1822 patients; 4 cancers were found.

Conclusion

Breast examination call rate doubled when attention was focused on examination results using a dedicated form, and we found the anticipated cancers. Breast examination quality can be improved by focusing clinician attention without retraining in technique.

Keywords: Clinical breast examination, Reminders, Retraining, Screening

Time is short in a typical primary care practice,¹ and there is a tendency to substitute the technology of mammography for the manual skill of clinical breast examination.² Clinical breast examination, however, is necessary for detection and evaluation of the 8% to 17% of breast cancers not seen by mammograms.³, ⁴ Yet, performance of breast examination has decreased,², ⁵, ⁶ and poor application of breast examination skills is the leading cause of delayed diagnosis of breast cancer.⁷, ⁸

Unlike ordering screening mammograms, skill at performance of breast examination is not a common quality indicator, probably because it is difficult to assess clinical application of breast examination skills.

Assessing clinical examination quality is daunting. Routine breast examination is a low-yield procedure, and estimating quality by proctoring (as opposed to in vitro detection of simulated masses in simulated breasts, a high-yield procedure, eg, 5 simulated masses in 1 simulated breast⁹ or simulated masses in 5 of 6 breast models¹⁰) would require numerous concurrent, duplicate examinations by an expert to know whether clinically relevant areas were being detected or missed by the clinician being assessed. There is no combination of visual or palpable characteristics of any focal area in the breast that is sufficiently specific or sensitive to determine whether a specific finding in a woman's breast is so clearly clinically relevant (or not relevant) to her future health that a clinician would be deficient if he/she missed those areas, that is, when experienced clinicians examine the same patients, they often disagree on what has been found¹¹ and what is clinically relevant.¹²

The metric used to circumvent these challenges is the call rate: the percent of breast examinations with findings for which the examiner refers the woman for further evaluation. Call rate is based on the fact that normal breasts are often asymmetric or have areas that are irregular.¹³, ¹⁴ It is not possible to know by palpation whether such areas are malignant or not, but if a clinician does not observe a low, but expected (ie, nonzero), rate of such abnormalities, it is assumed that the clinician is unlikely to be detecting all the palpable abnormalities that are clinically relevant. Thus, a low call rate indicates either inattention to detail or clinical judgment is being used to ignore some findings, even though the latter is the most common cause of delayed diagnosis of breast cancer.⁷ Because any breast mass is cancer until proved otherwise, identification of all breast masses as reflected in the call rate, not identification of cancers specifically, is the appropriate quality measure for breast examination. Call rate can be defined for any set of breast examinations; thus, calculation of call rate eliminates the need to proctor many examinations. Stated another way, call rate consists of true positives plus false positives. Because it is impossible to distinguish false positives from true positives while performing the examination, it is likely that both increase as the call rate increases. Thus, unless one postulates no false-negative breast examination findings (an unlikely condition¹²), a low call rate implies lower identification of both true and false positives and thus a higher false-negative rate. Call rate (alternatively tabulated as referral to specialists) has been reported for clinical follow-up of examination technique retraining with silicone breast models¹⁰ and for large screening trials.¹⁵, ¹⁶

Because of the importance of breast examination (even though it is a low-yield procedure), we asked whether an attention-focusing device—initially focusing attention by asking clinicians to describe the breasts in general—would increase the call rate as an indicator of improved examination quality. To comply with requirements of the Health Insurance Portability and Accountability Act, we used dedicated, de-identified forms to collect data. We anticipated that the effort required to observe and record a description of the breasts would focus attention and increase the call rate. We found that using either prospective form was sufficient to improve call rate.

Back to Article Outline

Materials and Methods 

With institutional review board approval, between November 2004 and June 2005, clinicians at a community health center recorded breast examination findings by marking a breast graphic. For Health Insurance Portability and Accountability Act compliance, we gathered data and tabulated results from the no-carbon-required copy page (that obscured identifiers except age) of a dedicated breast examination form.

All clinicians attended a 30-minute presentation about the clinical utility of breast examination that included a video emphasizing complete coverage of the breast area without emphasizing a strip pattern of examination. Clinicians were randomized to use either a short form that asked the clinician to record abnormal findings or a long form that asked the clinician to describe the breasts in general and to record abnormal findings. On the long form, 3 responses were used to describe the breasts: outline on the graphic the area where it was not possible to palpate ribs through the breast (a property called durity⁸); describe breast nodularity as “average” or “more” or “less than” average relative to the clinician's own experience; and 3 indicate whether the breasts were symmetric.

Neither group was retrained in technique. Clinicians using the long form had a brief session with a volunteer subject to demonstrate the meaning of the descriptive terms. All forms had a place to record mammography results (before or after the examination, with after being more typical), whether the patient had noticed a mass, and follow-up of clinical or mammographic findings when appropriate. Examination was considered positive if the clinician marked any area of interest on the breast graphic. A nurse investigator (JP) completed forms with subsequent mammography results and diagnoses when applicable.

To assess whether the use of a form by itself influenced clinical practice, we created a retrospective, historical comparison group for which the same information was abstracted from a convenience sample of 300 women (based on chart availability) drawn from all women who received preventive gynecologic care at the center in the 12 months preceding November 2004 (Figure 1).

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  Flowchart for study patients and historical controls.

Health center clinicians were board-certified Family Physicians, General Internists, Nurse Practitioners, and Certified Physician Assistants with an average postgraduate experience greater than 10 years.

The expected number of cancers was estimated by weighting Surveillance, Epidemiology, and End Results (SEER) data according to the number of women in the study in each age group. Tabulated data were analyzed in Data Desk (Data Description Inc, Ithaca, NY) and Excel (Microsoft Corp, Redmond, Wash).

Back to Article Outline

Results 

We recorded 1522 breast examinations prospectively, 558 short forms and 964 long forms (Figure 1). The 2 patient groups were of similar age and had similar rates of self-identification of a mass before the clinical breast examination. There were more abnormal mammogram findings in the group using the short form (Table 1). Approximately 3% of all patients were seen twice for a screening examination during the study. The clinicians using the long form described 90% and 9% of women as having symmetric and asymmetric breasts, respectively (1% did not answer). The same clinicians described 38% of breasts as less nodular than average, 40% of breasts as average nodularity, 19% of breasts as more nodular than average, and 1% of breasts as very coarse nodularity; 2% did not answer. When the patient thought she had a mass, 36% of the time the clinician thought the examination findings were normal. The majority of women did not have screening mammograms in the 12 months before their breast examination: A total of 631 study patients (42%) were aged less than 40 years; of these, 40 (6.3%) had a screening mammogram before their examination. Of the 891 women aged more than 40 years, 420 (47%) had a screening mammogram before their examination.

Table 1. Call Rate of Clinical Breast Examination in 1522 Study Patients According to Use of Long or Short Form to Record Clinical Breast Examination

	Clinician Used Long Form	(%)	Clinician Used Short Form	(%)	P
n =	964		558
Median age	42 y		43 y
Mean age±SD	41±14 y		42±14 y
Patient felt a mass	73	7.60%	43	7.70%	.92
Abnormal mammogram	19	2.00%	22	3.90%	.022
Clinician found a mass	84	8.71%	43	7.70%	.494
Clinician found a mass that patient had not noticed	35	3.63%	18	3.22%	.71
Clinician found a mass patient had not noticed (no positive mammogram)	33	3.42%	17	3.04%	.69
Biopsy done	15	1.56%	15	2.69%	.126
Breast cancera	2	0.31%	0	0	.40

SD = standard deviation.

Clinicians using the long and short forms had a similar overall call rate, call rate when the patient had not identified a mass, and call rate when the patient had not found a mass and there was not a positive mammogram (Table 1). Because results in the 2 groups were essentially the same, the 2 groups were combined for comparison with historical controls from the preceding year.

A total of 107 of 300 historical controls were aged less than 40 years (36%). Of these controls, 6 (5.6%) had a screening mammogram, whereas 139 of 193 women aged 40 years and older (72.4%) had screening mammograms. Eleven controls (3.6%) did not have a recorded breast examination. Historical controls had more positive mammograms, but women in the study group were more likely to have found a mass themselves before coming to the clinic (Table 2). Therefore, conditions at the time of examination may have biased results toward finding more abnormal examination results for the control patients because there were more abnormal mammogram findings or for the study patients because more women came to the clinic believing they had a mass. When control women thought they had felt a mass, the clinician disagreed and thought a mass did not exist for 45% of the women.

Table 2. Call Rate of Clinical Breast Examination for Study Patients versus Historical Controls

	Study Patients	(%)	Historical Controls	(%)	P
n =	1522		300
Mean age±SD	42±14 y		44±13 y
Median age	42 y		44 y
Patient felt mass	116	7.62%	11	3.67%	.014
Abnormal mammogram	41	2.69%	24	8.03%	<.0001
No. with possible CBE	1522		298a
Clinician found mass	127	8.34%	14	4.70%	.031
Clinician found a mass that patient had not noticed	53	3.48%	7	2.34%	.32
Clinician found a mass patient had not noticed (no positive mammogram)	50	3.28%	6	2%	.24
Biopsy done	30	1.97%	18	6%	<.0001
Breast cancer	2	0.13%	2	0.66%	.069

CBE = clinical breast examination; SD = standard deviation.

During the study, clinicians were significantly more likely to find an abnormality on breast examination than during the historical control period (8.3% vs 4.7%, P=.031). The same trend was seen when women had not found a mass on self-examination and when the patient had not found a mass and there was not a positive mammogram (Table 2). Fewer biopsies were performed during the study period than during the control period (1.9% vs 6%; P<.001).

SEER data predicted 2.2 invasive breast cancers in the study group, and 2 invasive breast cancers were found, 1 initially by a clinician and 1 initially by the patient (Table 3). SEER predicted 0.6 noninvasive cancers. None were found in the study group, but only 30% of women had screening mammograms. SEER data predicted 0.5 invasive cancers in the historical group; 1 was found, initially by the patient. SEER predicted 0.1 noninvasive cancers; 1 intraductal cancer was found by mammography. Thus, 4 breast cancers were observed in all groups combined (Table 3) compared with the 3.4 cancers expected.

Table 3. Cancers Found in Study and Control Patients

CBE = clinical breast examination; IDC = invasive ductal carcinoma; DCIS = ductal in situ carcinoma.

One inflammatory cancer (described as “indurated subcutaneous tissue”) was initially treated as an infection. There were 6 other presumed infections, and each of them had an obvious source of infection, such as a skin lesion.

For 1822 patients combined, there were 595 screening mammograms, 152 diagnostic mammograms, 135 diagnostic ultrasound examinations, and 48 biopsies. In addition to 4 breast cancers (Table 3), there were 12 fibroadenomas, 20 biopsies of benign fibrocystic changes, 11 cysts confirmed by ultrasound, and B-cell lymphoma in an axillary node (a study patient with the long form [Table 1]; the lymphoma is not included in the study analysis because it is not a breast cancer). For the 141 women with an abnormal breast examination result, the abnormality resolved for 24 (17%) at the time of a follow-up examination. Twenty of these 24 women were aged less than 50 years, so short-term changes related to their menstrual cycle would not be unexpected. If the 12 women whose clinical finding resolved within 6 months are excluded as possible false positives, the call rate is 7.6% for study cases (116/1522) and 4.4% (13/298; P=.045) for controls; but 11 of these 12 women were aged less than 50 years. Ten women (0.56%) had cosmetic implants.

For the group of 1822 women, the sensitivity of breast examination was 0.75 and the specificity was 0.92, giving a likelihood ratio of 9.9 for a positive examination. Three of 141 abnormal breast examination results were cancer for a positive predictive value of 2.1%.

Back to Article Outline

Discussion 

Clinical breast examination is a dynamic process. For 17% of women with an abnormality, the abnormality resolved before their planned follow-up examination, and when the patient thought she had a mass, the clinician thought it was normal tissue in 36% to 45% of cases. Nevertheless, in this prospective study of breast examination, we found 2 invasive cancers against 2.2 predicted by age-weighted SEER data, and in the combined prospective and retrospective studies, we found 4 cancers against the 3.4 predicted by SEER. Given that the number of cancers detected approximates the number expected, it is unlikely that we missed cancers. Thus, we can comment on focusing attention as a single step to improve breast examination.

Our study depends on examination results more than most other settings because the proportion of study women who received both an examination and a mammography was only 30%. This reflects 2 influences: mammography is not indicated or only reimbursed every other year for much of this age group, whereas women often presented annually for breast examination in conjunction with a gynecologic examination, and many women could not afford mammography in this relatively indigent population.

When clinicians' attention was focused by being asked to complete a dedicated form to record any abnormality on breast examination, the “call rate” was 7.7% to 8.7% in all circumstances. When the patient had not found a mass, the call rates were 3.2% and 3.6%. When the patient had not found a mass, and there was no positive mammogram, the call rate was 3.0% to 3.4%. These latter call rates, uninfluenced by mammography or patient observation, are similar to the uninfluenced call rates for breast examination in large breast cancer screening trials (4.9% for an initial examination with 3.8% for follow-up examinations in the Ontario Breast Cancer Screening Trial;¹⁵ 3.5% for initial and 2.9% for follow-up in the Canadian Breast Cancer Screening Database¹⁶).

Because the call rate using both forms approximated the call rate in demonstration studies, we compared the call rate using the forms with the call rate before the study to see if we might have induced a so-called Hawthorne effect in which changing the conditions of a test itself changes the results, that is, the change in condition (in this case, using either of the 2 forms) becomes an independent variable.¹⁷ Comparison of study results with historical controls showed that asking clinicians to complete a dedicated form for the study (either the short or the long form) influenced practice and increased the call rate, and although the call rate was increased, the biopsy rate was lower in the study groups. Improved call rate using a dedicated form is similar to improved compliance with guidelines with clinical reminders.¹⁸ We add the observation that a form can improve the results of a clinical task, not just the rate at which the task is performed.

We have not compared focusing attention to retraining in the vertical strip 3-pressure method of breast examination with silicone breast models,⁹, ¹⁰, ¹⁹ but we achieved better clinical compliance with a dedicated form (100%) than the one study to report clinical follow-up of practices of vertical strip 3-pressure method trainees (24%),¹⁰ and we had a call rate similar to that in the Canadian Breast Cancer Screening Database, in which 3 of 4 centers used the vertical strip examination pattern.¹⁶ Thus, our results show that increasing attention to the examination (via a dedicated form) improves the call rate of experienced clinicians, and although the Hawthorne effect may “fatigue” with time, use of a form is presumably less inhibiting because it does not require 5 minutes per breast as does the vertical strip 3-pressure method.⁹

We attempted to focus attention because breast examination is a low-yield procedure that can become perfunctory. In our study, only 7% to 8% of patients had a clinical finding, but the positive predictive value of breast examination was only 2.1%. The likelihood ratio of a positive examination was 9.9, which is similar to the likelihood ratio of 10.6 that can be calculated from the sensitivity and specificity published by the Group Health Cooperative in Seattle.³ This likelihood ratio is larger than for common clinical tests, such as fecal occult blood testing (likelihood ratio 1.68²⁰) and listening for an inguinal bruit (likelihood ratio 4.8²¹), but the prevalence of breast cancer is so low (4/1822) that this likelihood ratio still has limited predictive value.

Back to Article Outline

Conclusions 

The limited predictive value of breast examination makes its role in screening a conundrum because the alternative (ie, mammograms) does not image 8% to 17% of breast cancers.³, ⁴ In larger series, clinicians are the first to identify 1 in 10 breast cancers, approximately one third of which have negative mammograms.⁷ It would be difficult to abandon these opportunities to detect breast cancer, and thus difficult to abandon breast examination. We believe a more effective approach is to retain clinical breast examination but to develop tools to focus clinician attention rather than to demand more clinician time for a more elaborate technique.

Back to Article Outline

References 

1. Yarnall KS, Pollak KI, Ostbye T, et al. Primary care: is there enough time for prevention?. Am J Public Health. 2003;93:635–641
   • View In Article
   • MEDLINE
   • CrossRef
2. Burns RB, Freund KM, Ash AS, et al. As mammography use increases, are some providers omitting clinical breast examination?. Arch Intern Med. 1996;156:741–744
   • View In Article
   • MEDLINE
3. Green BB, Taplin SH. Breast cancer screening controversies. J Am Board Fam Pract. 2003;16:233–241
   • View In Article
   • MEDLINE
4. Goodson WH, Grissom NA, Moore DH, Dribas FM. Streamlining clinical breast examination. J Natl Cancer Inst. 2005;97:1476–1477
   • View In Article
   • CrossRef
5. Coleman EA, Feuer EJ NCI Breast Cancer Screening Consortium. Breast cancer screening among women 65 to 74 years of age in 1987-88 and 1991. Ann Intern Med. 1992;117:961–966
   • View In Article
   • MEDLINE
6. Chagpar AB, McMasters KM. Trends in mammography and clinical breast examination: a population-based study. J Surg Res. 2007;140:214–219
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (88 KB)
   • CrossRef
7. Goodson WH, Moore DH. Causes of physician delay in the diagnosis of breast cancer. Arch Intern Med. 2002;162:1343–1348
   • View In Article
   • MEDLINE
   • CrossRef
8. Goodson WH, Moore DH. Overall clinical breast examination as a factor in delayed diagnosis of breast cancer. Arch Surg. 2002;137:1152–1156
   • View In Article
   • MEDLINE
   • CrossRef
9. Vetto JT, Petty JK, Dunn N, et al. Structured clinical breast exam (CBE) training results in objective improvement in CBE skills. J Cancer Educ. 2002;17:124–127
   • View In Article
   • MEDLINE
10. Campbell HS, Fletcher SW, Pilgrim CA, Morgan TM, Lin S. Improving physicians' and nurses' clinical breast examination: a randomized controlled trial. Am J Prev Med. 1991;7:1–8
    • View In Article
    • MEDLINE
11. Smiddy FG The Yorkshire Breast Cancer Group. Observer variation in recording clinical data from women presenting with breast lesions. Br Med J. 1977;2:1196–1199
    • View In Article
    • MEDLINE
12. Boyd NF, Sutherland HJ, Fish EB, et al. Prospective evaluation of physical examination of the breast. Am J Surg. 1981;142:331–334
    • View In Article
    • MEDLINE
    • CrossRef
13. Loughry CW, Price TE, Morek WM, et al. Right and left breast volume distribution comparisons in normal and tumor containing breasts. Cancer Detect Prev. 1987;10:215–221
    • View In Article
    • MEDLINE
14. Losken A, Fishman I, Denson DD, et al. An objective evaluation of breast symmetry and shape differences using 3-dimensional images. Ann Plast Surg. 2005;55:571–575
    • View In Article
    • MEDLINE
15. Chiarelli AM, Halapy E, Nadalin V, et al. Performance measures from 10 years of breast screening in the Ontario Breast Screening Program, 1990/91 to 2000. Eur J Cancer Prev. 2006;34–42
    • View In Article
16. Bancej C, Decker K, Chiarelli A, et al. Contribution of clinical breast examination to mammography screening in the early detection of breast cancer. J Med Screen. 2003;10:16–21
    • View In Article
    • MEDLINE
    • CrossRef
17. Parsons HM. What happened at Hawthorne?: new evidence suggests the Hawthorne effect resulted from operant reinforcement contingencies. Science. 1974;183:922–932
    • View In Article
18. Dexter PR, Perkins S, Overhage JM, et al. A computerized reminder system to increase the use of preventive care for hospitalized patients. N Engl J Med. 2001;345:965–970
    • View In Article
    • MEDLINE
    • CrossRef
19. Steiner E, Austin DF, Prouser NC. Detection and description of small breast masses by residents using a standardized clinical breast exam curriculum. J Gen Intern Med. 2007;23:129–134
    • View In Article
    • CrossRef
20. Collins JF, Lieberman DA, Durbin TE, et al. Accuracy of screening for fecal occult blood on a single stool sample obtained by digital rectal examination: a comparison with recommended sampling practice. Ann Intern Med. 2005;142:81–85
    • View In Article
21. Khan NA, Rahim SA, Anand SS, et al. Does the clinical examination predict lower extremity peripheral arterial disease?. JAMA. 2006;295:536–546
    • View In Article
    • CrossRef