Home > Results - Noninvasive Technologies for...

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Dolor RJ, Patel MR, Melloni C, et al. Noninvasive Technologies for the Diagnosis of Coronary Artery Disease in Women [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Jun. (Comparative Effectiveness Reviews, No. 58.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Results

The flow of articles through the literature search and screening process is depicted in Figure 2. Of the 8,231 citations identified by our searches, 634 were duplicates. A manual search identified an additional 445 citations for a total of 8,042 citations. After applying inclusion/exclusion criteria at the title-and-abstract level, 1,772 full-text articles were retrieved and screened. Of these, 1,662 articles were excluded at the full-text screening stage. We excluded 1376 (83 percent) for not reporting data on women and 615 (37 percent) for looking only at a population with known CAD. (Note that an article may have been excluded for more than one reason.) The final set comprised 110 articles representing 104 studies.

Figure 2 depicts the flow of articles through the literature search and screening process. Of the 8231 citations identified by our searches, 634 were duplicates. A manual search identified an additional 445 citations for a total of 8042 citations. After applying inclusion/exclusion criteria at the title-and-abstract level, 1772 full-text articles were retrieved and screened. Of these, 1662 articles were excluded at the full-text screening stage, with 110 articles representing 104 studies remaining for data abstraction. Of the 104 studies, 1 was an RCT, 79 were prospective observational, and 24 were retrospective observational with study cohorts comprised of individuals who presented for NIT testing and received diagnostic coronary angiography.

Figure 2

Literature flow diagram. Abbreviations: CAD = coronary artery disease; KQ = Key Question; NIT = noninvasive technology; SR = systematic review

Of the 104 studies, 1 was an RCT, 79 were prospective observational, and 24 were retrospective observational with study cohorts comprising individuals who presented for NIT testing and received diagnostic coronary angiography (100 studies) or another NIT modality only (4 studies). The four studies without coronary angiography compared ECHO with ECG22,23 or ECG with SPECT.24,25 Three of these studies were applicable to Key Question (KQ) 3,2224 and one was applicable to KQ 2.25 Of the 94 studies included in the KQ 1 results, 5 reported NIT versus NIT comparisons in addition to coronary angiography.2630

Appendix E provides a complete list of articles excluded at the full-text screening stage, with reasons for exclusion.

A summary graph of the QUADAS ratings for all 104 studies is shown in Figure 3. (Refer to Table D-2 in Appendix D for a summary table of the QUADAS quality scores for diagnostic accuracy of the 104 studies included in this review.) A majority of studies uniformly applied a reference test (i.e., coronary angiography) that was independently performed, used blinded interpretation of the reference and index test, and had sufficient detail about the index test to allow for replicability. Many studies had a high risk of spectrum bias (i.e., patient representation of those who would receive the test in practice), a poor description of study withdrawals, unclear descriptions of clinical data available during test interpretation, and lacked descriptions of uninterpretable or intermediate test results. There is a possibility that sensitivity and specificity values may be biased because of subjects included in the studies that did not represent the spectrum of the population of interest; we explored the impact of the underlying prevalence of CAD in the population on our findings.

Figure 3 shows a summary graph of the Quality Assessment for Diagnostic Accuracy Studies (QUADAS) ratings for all 104 studies. A majority of studies uniformly applied a reference test (i.e., coronary angiography) that was independently performed, used blinded interpretation of the reference and index test, and had sufficient detail about the index test to allow for replicability. Many studies had a high risk of spectrum bias (i.e., patient representation of those who would receive the test in practice), a poor description of study withdrawals, unclear descriptions of clinical data available during test interpretation, and lacked descriptions of uninterpretable or intermediate test results. There is a possibility that sensitivity and specificity values may be biased because of a nonrandom selection of subjects from the population of interest; however, available data did not provide sufficient information to consider possible adjustment for such bias.

Figure 3

QUADAS elements used to rate diagnostic accuracy.

Key Question 1. Diagnostic Accuracy of NITs

KQ 1. What is the accuracy of one NIT in diagnosing obstructive and nonobstructive CAD when compared with another NIT or with coronary angiography in women with symptoms suspicious for CAD?

  • Exercise ECG stress test, including resting ECG technology (e.g., multifunctional cardiogram)
  • Exercise/stress ECHO with or without a contrast agent
  • Exercise/stress radionuclide myocardial perfusion imaging, including SPECT and PET
  • CMR imaging
  • Coronary CTA

Key Points

Individual study performance characteristics were evaluated for each testing modality, and summary receiver operating characteristic (SROC) curves were calculated. These analyses demonstrated:

  • Overall, within a given testing modality, the summary sensitivities and specificities were similar for both types of populations (known and no known CAD) and for all studies when compared with good-quality studies.
  • When accounting for only the good-quality studies, it appeared that the diagnostic accuracy of detecting CAD in women was better (in descending order) for coronary CTA, SPECT, ECHO, CMR, and ECG, although the strength of evidence varied markedly for different modalities.
  • For the newer technologies (i.e., CMR and coronary CTA), more studies in women would be needed to support the point estimates given the wide confidence intervals (CIs) on the test performance.
  • For women without previously known CAD, there were statistically significant differences between the performance of the available modalities (p < 0.001). The sensitivity of ECHO and SPECT was significantly greater than that of ECG. Specificity of ECG was less than that of CMR (borderline) and of ECHO.
  • In the subset of studies that were good-quality and where there was no known CAD in the included population, there were statistically significant differences between performance of tests (p = 0.006), with the specificity of ECG being less than that of CMR and ECHO. Our ability to quantify the difference between test performance of the modalities between men and women was inhibited by the limited number of studies that reported both sexes separately in their analysis.
  • In exploratory analysis of the difference between test performance in men and women, the ECG and coronary CTA modalities were both less sensitive and less specific in women than in men. The ECHO and SPECT modalities, although less sensitive, appeared to be more specific in women. The lower specificity of the ECG modality in women, however, is the only estimate that was determined to be a statistically significant difference.

Detailed Synthesis

In KQ 1 we sought to determine the accuracy of each NIT modality in diagnosing obstructive and nonobstructive CAD when compared with coronary angiography in women with symptoms suspicious for CAD. For this analysis, we included 94 studies describing comparative diagnostic accuracy of NITs. Of these 94 studies, 78 studies included sufficient data to estimate the sensitivity and specificity of the NIT compared with coronary angiography. This included 41 studies examining exercise/stress ECG (13 good quality, 22 fair, 6 poor); 22 examining exercise/stress ECHO (8 good quality, 13 fair, 1 poor); 30 examining exercise/stress radionuclide myocardial perfusion imaging (e.g., SPECT or PET) (10 good quality, 15 fair, 5 poor); 6 examining CMR (5 good, 1 fair); and 8 examining coronary CTA (4 good quality, 4 fair).

For each testing modality, we used the individual performance characteristics to calculate an SROC curve and to estimate the summary sensitivity and specificity and CIs of the modality compared with coronary angiography. We present forest plots of the individual study estimates of sensitivity and specificity of each NIT for diagnosing CAD in women. Error bars in these plots represent 95 % CIs; the dashed vertical line represents the summary sensitivity and specificity for the included studies. The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95-percent confidence region around it. In our primary analyses, we evaluated these performance characteristics in the population of women who had no previously known CAD. In secondary analyses we explored a broader patient population including those studies that had women from a mixed population of known and no known CAD. We also assessed the impact on our findings if, in each population, we restricted our analyses to those studies that were assessed to be good quality. We then compared the performance characteristics of the NIT modalities with each other. In a final exploratory analysis, we evaluated the test performance of the modalities in women compared with men. All secondary analyses involved the use of separate generalized linear mixed models with covariates for disease state (no known versus mixed), NIT modality (ECG, ECHO, SPECT, CMR, CTA), or sex (women versus men).

ECG

We identified 41 studies evaluating the accuracy of exercise/stress ECG compared with coronary angiography (Table 2).2729,3168 This table lists those studies that focused purely on women with no known CAD and then follows with the additional studies that included a mixed population of known and no known CAD. Within these populations, good-quality studies are listed first, followed by those of fair and poor quality. Twenty-nine of the ECG studies reported accuracy data in women with no known CAD, and these are the studies used in our primary analysis.

Table 2. Summary of accuracy data evaluating ECG for diagnosing CAD.

Table 2

Summary of accuracy data evaluating ECG for diagnosing CAD.

In our secondary analysis, we evaluated the accuracy of exercise/stress ECG in diagnosing CAD in mixed populations of known and no known CAD. This analysis included 1 study that reported additional data for a mixed population61 and an additional 12 studies that reported findings with mixed populations of known and no known CAD. Two of these studies evaluated the use of resting ECG,33,49 and two studies35,50 evaluated pharmacological stressed ECG. All other studies evaluated exercise/stress ECG.

Primary Analysis: Population of Women With No Known CAD

The 29 studies represent findings on ECG use in 3,391 women (sample size ranging from 10 to 580 women). Of these studies, 10 were good quality, 15 were fair quality, and 4 were poor quality. Sensitivity varied from 32 to 91 percent, and specificity varied from 40 to 100 percent; the median sensitivity was 61 percent, and the median specificity was 68 percent. Figure 4 presents forest plots of the individual study estimates of sensitivity and specificity of ECG for diagnosing CAD in women with no known CAD. Error bars represent 95 percent CIs; the dashed vertical line represents the summary sensitivity and specificity for the included studies.

Figure 4 presents forest plots of the individual study estimates of sensitivity and specificity of ECG for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 29 included studies. The 29 studies represent findings on ECG use in 3391 women (sample size ranging from 10 to 580 women). In these studies, sensitivity varied from 32 to 91 percent, and specificity varied from 40 to 100 percent; the summary sensitivity was 62 percent, and the summary specificity was 68 percent.

Figure 4

Accuracy of ECG in women with no known CAD.

Figure 5 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 62 percent (95% CI, 55 to 68 percent) and specificity of 68 percent (95% CI, 63 to 73 percent). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95-percent confidence region around it.

Figure 5 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 62 percent (95% CI, 55 to 68%) and specificity of 68 percent (95% CI, 63 to 73%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 5

SROC curve for ECG in women with no known CAD.

The prevalence of CAD on coronary angiogram in these 29 studies ranged from 18 to 67 percent with a mean prevalence of 41 percent. In the individual studies, the positive predictive value (PPV) ranged from 29 to 100 percent, and the negative predictive value (NPV) ranged from 40 to 100 percent. The positive likelihood ratio (LR+) ranged from 0.98 to 3.00, and the negative likelihood ratio (LR−) ranged from 0.18 to 1.03. Using the summary sensitivity and specificity of 62 and 68 percent, respectively, we calculated an overall PPV of 57 percent and NPV of 72 percent. Similarly, we calculated summary LR+ of 1.94 and LR− of 0.56.

Accuracy of ECG in 10 Good-Quality Studies

Next, we evaluated the accuracy of ECG compared with coronary angiography in the 10 good-quality studies. In these studies, sensitivity varied from 32 to 91 percent, and specificity varied from 46 to 81 percent; the median sensitivity was 71 percent, and the median specificity was 58 percent. Figure 6 presents forest plots of the individual study estimates of sensitivity and specificity of ECG in 10 good-quality studies for diagnosing CAD in women with no known CAD.

Figure 6 presents forest plots of the individual study estimates of sensitivity and specificity of ECG in 10 good-quality studies for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 10 included studies. In these studies, sensitivity varied from 32 to 91 percent, and specificity varied from 46 to 81 percent; the summary sensitivity was 70 percent, and the summary specificity was 62 percent.

Figure 6

Accuracy of ECG in 10 good-quality studies in women with no known CAD.

Figure 7 presents an SROC curve with an average sensitivity of 70 percent (95% CI, 58 to 79 percent) and specificity of 62 percent (95% CI, 53 to 69 percent).

Figure 7 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 70 percent (95% CI, 58 to 79%) and specificity of 62 percent (95% CI, 53 to 69%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 7

SROC curve for ECG in 10 good-quality studies in women with no known CAD.

The prevalence of CAD in these 10 good-quality studies ranged from 18 to 67 percent with a mean prevalence of 38 percent. In the individual studies, PPV ranged from 29 to 77 percent, and NPV ranged from 50 to 96 percent. LR+ ranged from 1.30 to 2.71 and LR− from 0.18 to 0.84. Using the summary sensitivity and specificity of 70 and 62 percent, respectively, we calculated an overall PPV of 53 percent and NPV of 77 percent. Similarly, we calculated summary LR+ of 1.84 and LR− of 0.48.

Secondary Analysis: Mixed Population of Women With Known and No Known CAD

We performed a secondary analysis where we expanded our inclusion criteria to include studies whose patient population included a mix of women with known CAD and women with no known CAD. This expanded inclusion criteria allowed an additional 12 studies to be included in the analysis and an additional 83 patients from one study (totaling 41 studies). The 41 studies represent findings on ECG use in 4946 women (sample size ranging from 10 to 613 women). Of these 41 studies, 13 were good quality, 22 were fair quality, and 6 were poor quality (Table 2).

In these 41 studies, sensitivity varied from 26 to 96 percent, and specificity varied from 1 to 100 percent; the median sensitivity was 61 percent, and the median specificity was 65 percent. Figure 8 presents forest plots of the individual study estimates of sensitivity and specificity of ECG for diagnosing CAD in women from mixed populations.

Figure 8 presents forest plots of the individual study estimates of sensitivity and specificity of ECG for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 41 included studies. In these 41 studies, sensitivity varied from 26 to 96 percent, and specificity varied from 1 to 100 percent; the summary sensitivity was 61 percent, and the summary specificity was 65 percent.

Figure 8

Accuracy of ECG in women from mixed populations.

Figure 9 presents an SROC curve demonstrating an average sensitivity of 61 percent (95% CI, 54 to 67 percent) and specificity of 65 percent (95% CI, 58 to 72 percent).

Figure 9 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 61 percent (95% CI, 54 to 67%) and specificity of 65 percent (95% CI, 58 to 72%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 9

SROC curve for ECG in women from mixed populations.

The prevalence of CAD in these 41 studies ranged from 11 to 67 percent with a mean prevalence of 42 percent. In the individual studies, PPV ranged from 4 to 100 percent, and NPV ranged from 1 to 100 percent. LR+ ranged from 0.30 to 5.42 and LR− from 0.05 to 55.3. Using the summary sensitivity and specificity of 61 and 65 percent, respectively, we calculated an overall PPV of 56 percent and NPV of 70 percent. Similarly, we calculated summary LR+ of 1.74 and LR− of 0.60.

Accuracy of ECG in 13 Good-Quality Studies

Next, we evaluated the accuracy of ECG compared with coronary angiography in the 13 good-quality studies. In these studies, sensitivity varied from 26 to 91 percent, and specificity varied from 33 to 81 percent; the median sensitivity was 71 percent, and the median specificity was 58 percent. Figure 10 presents forest plots of the individual study estimates of sensitivity and specificity of ECG in 13 good-quality studies for diagnosing CAD in women from mixed populations.

Figure 10 presents forest plots of the individual study estimates of sensitivity and specificity of ECG in 13 good-quality studies for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 13 included studies. In these studies, sensitivity varied from 26 to 91 percent, and specificity varied from 33 to 81 percent; the summary sensitivity was 65 percent, and the summary specificity was 60 percent.

Figure 10

Accuracy of ECG in 13 good-quality studies in women from mixed populations.

Figure 11 presents an SROC curve demonstrating an average sensitivity of 65 percent (95% CI, 52 to 76 percent) and specificity of 60 percent (95% CI, 52 to 68 percent).

Figure 11 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 65 percent (95% CI, 52 to 76%) and specificity of 60 percent (95% CI, 52 to 68%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 11

SROC curve for ECG in 13 good-quality studies in women from mixed populations.

The prevalence of CAD in these 13 good-quality studies ranged from 18 to 67 percent with a mean prevalence of 37 percent. In the individual studies, PPV ranged from 28 to 77 percent, and NPV ranged from 31 to 96 percent. LR+ ranged from 0.39 to 2.71 and LR− from 0.18 to 2.21. Using the summary sensitivity and specificity of 65 and 60 percent, respectively, we calculated an overall PPV of 49 percent and NPV of 75 percent. Similarly, we calculated summary LR+ of 1.62 and LR− of 0.58.

ECHO

We identified 22 studies evaluating the accuracy of exercise/stress ECHO compared with coronary angiography (Table 3).27,28,35,39,41,46,5456,61,67,6979 Fourteen of these studies reported accuracy data in women with no known CAD, and these are the studies used in our primary analyses. In our secondary analyses, we evaluated the accuracy of the ECHO in diagnosing CAD including an additional 8 studies that reported findings with mixed populations of known and no known CAD as well as additional patients from one study61 that had data for those with no known CAD. None of the identified ECHO studies used contrast.

Table 3. Summary of accuracy data evaluating ECHO for diagnosing CAD.

Table 3

Summary of accuracy data evaluating ECHO for diagnosing CAD.

Primary Analysis: Population of Women With No Known CAD

The 14 studies represent findings on ECHO use in 1289 women (sample size ranging from 14 to 192 women). Of these studies, five were good-quality, eight were fair-quality, and one was poor-quality. Sensitivity varied from 57 to 90 percent, specificity varied from 37 to 96 percent; the median sensitivity was 79 percent, and the median specificity was 82 percent. Figure 12 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women with no known CAD.

Figure 12 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 14 included studies. In these studies, sensitivity varied from 57 to 90 percent, specificity varied from 37 to 96 percent; the summary sensitivity was 79 percent, and the summary specificity was 83 percent.

Figure 12

Accuracy of ECHO in women with no known CAD.

Figure 13 presents an SROC curve with an average sensitivity of 79 percent (95% CI, 74 to 83 percent) and specificity of 83 (95% CI, 74 to 89 percent).

Figure 13 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 79 percent (95% CI, 74 to 83%) and specificity of 83 (95% CI, 74 to 89%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 13

SROC curve for ECHO in women with no known CAD.

The prevalence of CAD in these 14 studies ranged from 29 to 66 percent with a mean prevalence of 44 percent. In the individual studies, PPV ranged from 60 to 90 percent, and NPV ranged from 37 to 96 percent. LR+ ranged from 1.25 to 18.44 and LR− from 0.13 to 0.60. Using the summary sensitivity and specificity of 79 and 83 percent, respectively, we calculated an overall PPV of 78 percent and NPV of 83 percent. Similarly, we calculated summary LR+ of 4.65 and LR− of 0.25.

Accuracy of ECHO in Five Good-Quality Studies

Next, we evaluated the accuracy of ECHO compared with coronary angiography in the five good-quality studies. In these studies, sensitivity varied from 68 to 87 percent, specificity varied from 71 to 96 percent; the median sensitivity was 80 percent, and the median specificity was 82 percent. Figure 14 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women with no known CAD.

Figure 14 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the five included studies. In these studies, sensitivity varied from 68 to 87 percent, specificity varied from 71 to 96 percent; the summary sensitivity was 79 percent, and the summary specificity was 85 percent.

Figure 14

Accuracy of ECHO in five good-quality studies in women with no known CAD.

Figure 15 presents an SROC curve with an average sensitivity of 79 percent (95% CI, 69 to 87 percent) and specificity of 85 percent (95% CI, 68 to 94 percent).

Figure 15 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 79 percent (95% CI, 69 to 87%) and specificity of 85 percent (95% CI, 68 to 94%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 15

SROC curve for ECHO in five good-quality studies in women with no known CAD.

The prevalence of CAD in these five good-quality studies ranged from 34 to 49 percent with a mean prevalence of 40 percent. In the individual studies, PPV ranged from 60 to 90 percent, and NPV ranged from 86 to 90 percent. LR+ ranged from 2.73 to 18.44 and LR− from 0.16 to 0.33. Using the summary sensitivity and specificity of 79 and 85 percent, respectively, we calculated an overall PPV of 78 percent and NPV of 86 percent. Similarly, we calculated summary LR+ of 5.27 and LR− of 0.25.

Secondary Analysis: Mixed Population of Women With Known and No Known CAD

We performed a secondary analysis where we expanded our inclusion criteria to include studies whose patient population included a mix of women with known CAD and women with no known CAD. This expanded inclusion criteria allowed an additional eight studies to be included in the analysis and an additional group of patients from one study (totaling 22 studies). The 22 studies represent findings on ECHO use in 1944 women (sample size ranging from 7 to 192 women). Of these 22 studies, 8 were good quality, 13 were fair quality, and 1 was poor quality (Table 3).

In these 22 studies, sensitivity varied from 40 to 93 percent, and specificity varied from 37 to 100 percent; the median sensitivity was 79 percent, and the median specificity was 84 percent. Figure 16 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women from mixed populations.

Figure 16 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 22 included studies. In these 22 studies, sensitivity varied from 40 to 93 percent, and specificity varied from 37 to 100 percent; the summary sensitivity was 78 percent, and the summary specificity was 86 percent.

Figure 16

Accuracy of ECHO in women from mixed populations.

Figure 17 presents an SROC curve demonstrating an average sensitivity of 78 percent (95% CI, 73 to 83 percent) and specificity of 86 percent (95% CI, 79 to 91 percent).

Figure 17 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 78 percent (95% CI, 73 to 83%) and specificity of 86 percent (95% CI, 79 to 91%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 17

SROC curve for ECHO in women from mixed populations.

The prevalence of CAD in these 22 studies ranged from 26 to 86 percent with a mean prevalence of 46 percent. In the individual studies, PPV ranged from 43 to 100 percent, and NPV ranged from 37 to 100 percent. LR+ ranged from 1.25 to 18.44 and LR− from 0.08 to 0.74. Using the summary sensitivity and specificity of 78 and 86 percent, respectively, we calculated an overall PPV of 82 percent and NPV of 82 percent. Similarly, we calculated summary LR+ of 5.57 and LR− of 0.26.

Accuracy of ECHO in Eight Good-Quality Studies

Next, we evaluated the accuracy of ECHO compared with coronary angiography in the eight good-quality studies. In these studies, sensitivity varied from 40 to 87 percent, and specificity varied from 71 to 100 percent; the median sensitivity was 80 percent, and the median specificity was 84 percent. Figure 18 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women from mixed populations.

Figure 18 presents forest plots of the individual study estimates of sensitivity and specificity of ECHO for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the eight included studies. In these studies, sensitivity varied from 40 to 87 percent, and specificity varied from 71 to 100 percent; the summary sensitivity was 77 percent, and the summary specificity was 89 percent.

Figure 18

Accuracy of ECHO in eight good-quality studies in women from mixed populations.

Figure 19 presents an SROC curve demonstrating an average sensitivity of 77 percent (95% CI, 65 to 85 percent) and specificity of 89 percent (95% CI, 76 to 95 percent).

Figure 19 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 77 percent (95% CI, 65 to 85%) and specificity of 89 percent (95% CI, 76 to 95%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 19

SROC curve for ECHO in eight good-quality studies in women from mixed populations.

The prevalence of CAD in these 8 good-quality studies ranged from 26 to 64 percent with a mean prevalence of 40 percent. In the individual studies, PPV ranged from 43 to 100 percent, and NPV ranged from 70 to 94 percent. LR+ ranged from 2.06 to 18.44 and LR− from 0.16 to 0.74. Using the summary sensitivity and specificity of 77 and 89 percent, respectively, we calculated an overall PPV of 82 percent and NPV of 85 percent. Similarly, we calculated summary LR+ of 7.0 and LR− of 0.26.

SPECT

We identified 30 studies evaluating the accuracy of SPECT compared with coronary angiography (Table 4).26,28,37,39,40,43,50,52,54,62,67,68,71,77,78,8094 Fourteen of these studies were conducted exclusively in women with no known CAD, and these are the studies used in our primary analysis. In our secondary analysis, we evaluated the accuracy of SPECT in diagnosing CAD in mixed populations of known and no known CAD, including 16 additional studies.

Table 4. Summary of accuracy data evaluating SPECT for diagnosing CAD.

Table 4

Summary of accuracy data evaluating SPECT for diagnosing CAD.

Primary Analysis: Population of Women With No Known CAD

The 14 studies represent findings on SPECT use in 1000 women (sample size ranging from 19 to 184 women). Of these studies, four were good quality, nine were fair quality, and 1 was poor quality. Sensitivity varied from 62 to 93 percent, and specificity varied from 50 to 91 percent; the median sensitivity was 82 percent, and the median specificity was 81 percent. Figure 20 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with no known CAD.

Figure 20 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 14 included studies. In these studies, sensitivity varied from 62 to 93 percent, and specificity varied from 50 to 91 percent; the summary sensitivity was 81 percent, and the summary specificity was 78 percent.

Figure 20

Accuracy of SPECT in women with no known CAD.

Figure 21 presents an SROC curve with an average sensitivity of 81 percent (95% CI, 76 to 86 percent) and specificity of 78 percent (95% CI, 69 to 84 percent).

Figure 21 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 81 percent (95% CI, 76 to 86%) and specificity of 78 percent (95% CI, 69 to 84%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 21

SROC curve for SPECT in women with no known CAD.

The prevalence of CAD in these 14 studies ranged from 14 to 67 percent with a mean prevalence of 45 percent. In the individual studies, PPV ranged from 36 to 89 percent, and NPV ranged from 50 to 91 percent. LR+ ranged from 1.86 to 9.14 and LR− from 0.12 to 0.47 respectively. Using the summary sensitivity and specificity of 81 and 78 percent, respectively, we calculated an overall PPV of 75 percent and a negative predictive value of 83 percent. Similarly, we calculated summary LR+ of 3.68 and LR− of 0.24.

Accuracy of SPECT in Four Good-Quality Studies

Next, we evaluated the accuracy of SPECT compared with coronary angiography in the four good-quality studies. In these studies, sensitivity varied from 62 to 93 percent, and specificity varied from 50 to 91 percent; the median sensitivity was 83 percent, and the median specificity was 68 percent. Figure 22 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with no known CAD.

Figure 22 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the included studies. In these studies, sensitivity varied from 62 to 93 percent, and specificity varied from 50 to 91 percent; the summary sensitivity was 83 percent, and the summary specificity was 72 percent.

Figure 22

Accuracy of SPECT in four good-quality studies in women with no known CAD.

Figure 23 presents an SROC curve with an average sensitivity of 83 percent (95% CI, 52 to 95 percent) and specificity of 72 percent (95% CI, 37 to 92 percent). It is important to note that given the small number of studies and wide confidence intervals that these summary statistics should be interpreted with caution.

Figure 23 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 83 percent (95% CI, 52 to 95%) and specificity of 72 percent (95% CI, 37 to 92%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 23

SROC curve for SPECT in four good-quality studies in women with no known CAD.

The prevalence of CAD in these 4 good-quality studies ranged from 14 to 67 percent with a mean prevalence of 38 percent. In the individual studies, PPV ranged from 36 to 79 percent, and NPV ranged from 78 to 93 percent. LR+ ranged from 1.86 to 8.0 and LR− from 0.14 to 0.47. Using the summary sensitivity and specificity of 83 and 72 percent, respectively, we calculated an overall PPV of 65 percent and NPV of 88 percent. Similarly, we calculated summary LR+ of 2.96 and LR− ratio of 0.24.

Secondary Analysis: Mixed Population of Women With Known and No Known CAD

We performed a secondary analysis where we expanded our inclusion criteria to include studies whose patient population included a mix of women with known CAD and women with no known CAD. This expanded inclusion criteria allowed an additional 16 studies to be included in our analysis (totaling 30 studies). The 30 studies represent findings on SPECT use in 2157 women (sample size ranging from 14 to 243 women). Of these 30 studies, 10 were good quality, 15 were fair quality, and 5 were poor quality (Table 4).

In these 30 studies, sensitivity varied from 15 to 100 percent, and specificity varied from 27 to 100 percent; the median sensitivity was 83 percent, and the median specificity was 81 percent. Figure 24 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women from mixed populations.

Figure 24 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 30 included studies. In these studies, sensitivity varied from 15 to 100 percent, and specificity varied from 27 to 100 percent; the summary sensitivity was 82 percent, and the summary specificity was 81 percent.

Figure 24

Accuracy of SPECT in women from mixed populations.

Figure 25 presents an SROC curve demonstrating an average sensitivity of 82 percent (95% CI, 77 to 87 percent) and specificity of 81 percent (95% CI, 74 to 86 percent).

Figure 25 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 82 percent (95% CI, 77 to 87%) and specificity of 81 percent (95% CI, 74 to 86%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 25

SROC curve for SPECT in women from mixed populations.

The prevalence of CAD in the 30 studies ranged from 14 to 96 percent with a mean prevalence of 54 percent. In the individual studies, PPV ranged from 36 to 100 percent, and NPV ranged from 27 to 100 percent. LR+ ranged from 1.89 to 24 and LR− from 0 to 0.89. Using the summary sensitivity and specificity of 82 and 81 percent, respectively, we calculated an overall PPV of 84 percent and NPV of 79 percent. Similarly, we calculated summary LR+ of 4.32 and LR− of 0.22.

Accuracy of SPECT in 10 Good-Quality Studies

Next, we evaluated the accuracy of SPECT compared with coronary angiography in the 10 good-quality studies. In these studies, sensitivity varied from 62 to 94 percent, and specificity varied from 50 to 100 percent; the median sensitivity was 81 percent, and the median specificity was 84 percent. Figure 26 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with from mixed populations.

Figure 26 presents forest plots of the individual study estimates of sensitivity and specificity of SPECT for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the 10 included studies. In these studies, sensitivity varied from 62 to 94 percent, and specificity varied from 50 to 100 percent; the summary sensitivity was 82 percent, and the summary specificity was 79 percent.

Figure 26

Accuracy of SPECT in 10 good-quality studies in women from mixed populations.

Figure 27 presents an SROC curve demonstrating an average sensitivity of 82 percent (95% CI, 72 to 88 percent) and specificity of 79 percent (95% CI, 66 to 87 percent).

Figure 27 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 82 percent (95% CI, 72 to 88%) and specificity of 79 percent (95% CI, 66 to 87%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 27

SROC curve for SPECT in 10 good-quality studies in women from mixed populations.

The prevalence of CAD in these 10 good-quality studies ranged from 14 to 96 percent with a mean prevalence of 56 percent. In the individual studies, PPV ranged from 36 to 100 percent, and NPV ranged from 20 to 94 percent. LR+ ranged from 1.86 to 9.72 and LR− from 0.07 to 0.47. Using the summary sensitivity and specificity of 82 and 79 percent, respectively, we calculated an overall PPV of 84 percent and NPV of 78 percent. Similarly, we calculated summary LR+ of 3.90 and LR− of 0.23.

CMR

We identified six studies evaluating the accuracy of CMR compared with coronary angiography (Table 5).26,9599 Five of these studies reported accuracy data in women with no known CAD, and these are the studies used in our primary analysis. In our secondary analysis, we evaluated the accuracy of CMR in diagnosing CAD in mixed populations of known and no known CAD, including additional data from two studies95,96 and data from one additional study.99

Table 5. Summary of accuracy data evaluating CMR for diagnosing CAD.

Table 5

Summary of accuracy data evaluating CMR for diagnosing CAD.

Primary Analysis: Population of Women With No Known CAD

The five studies represent findings on CMR use in 501 women (sample size ranging from 30 to 184 women). All five of these studies were rated good quality. In these studies, sensitivity varied from 58 to 83 percent, and specificity varied from 59 to 96 percent; the median sensitivity was 75 percent, and the median specificity was 88 percent. Figure 28 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women with no known CAD.

Figure 28 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the five included studies. In these studies, sensitivity varied from 58 to 83 percent, and specificity varied from 59 to 96 percent; the summary sensitivity was 72 percent, and the summary specificity was 84 percent.

Figure 28

Accuracy of CMR in women with no known CAD.

Figure 29 presents an SROC curve with an average sensitivity of 72 percent (95% CI, 55 to 85 percent) and specificity of 84 percent (95% CI, 69 to 93 percent).

Figure 29 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 72 percent (95% CI, 55 to 85%) and specificity of 84 percent (95% CI, 69 to 93%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 29

SROC curve for CMR in women with no known CAD.

The prevalence of CAD in these 5 studies ranged from 14 to 43 percent with a mean prevalence of 27 percent. In the individual studies, PPV ranged from 30 to 84 percent, and NPV ranged from 59 to 96 percent. LR+ ranged from 1.83 to 21.67 and LR− from 0.17 to 0.54. Using the summary sensitivity and specificity of 72 and 84 percent, respectively, we calculated an overall PPV of 62 percent and NPV of 89 percent. Similarly, we calculated summary LR+ of 4.5 and LR− of 0.33.

Secondary Analysis: Mixed Population of Women With Known and No Known CAD

We performed a secondary analysis where we expanded our inclusion criteria to include studies whose patient population included a mix of women with known CAD and women with no known CAD. This expanded inclusion criteria allowed an additional 64 patients from one study95 and an additional 45 patients from another study96 to be included in our analysis—as well as 168 patients from a third study that was not included in our primary analysis99 (totaling 6 studies). The 6 studies represent findings on CMR use in 778 women (sample size ranging from 30 to 184 women). Five of these studies were good-quality, and one was fair quality (Table 5).

In these 6 studies, sensitivity varied from 58 to 92 percent, and specificity varied from 59 to 91 percent; the median sensitivity was 80 percent, and the median specificity was 83 percent. Figure 30 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women from mixed populations.

Figure 30 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the six included studies. In these studies, sensitivity varied from 58 to 92 percent, and specificity varied from 59 to 91 percent; the summary sensitivity was 78 percent, and the summary specificity was 84 percent.

Figure 30

Accuracy of CMR in women from mixed populations.

Figure 31 presents an SROC curve demonstrating an average sensitivity of 78 percent (95% CI, 61 to 89 percent) and specificity of 84 percent (95% CI, 74 to 90 percent).

Figure 31 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 78 percent (95% CI, 61 to 89%) and specificity of 84 percent (95% CI, 74 to 90%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 31

SROC curve for CMR in women from mixed populations.

The prevalence of CAD in the 6 studies ranged from 7 to 56 percent with a mean prevalence of 32 percent. In the individual studies, PPV ranged from 30 to 93 percent, and NPV ranged from 59 to 91 percent. LR+ ranged from 1.83 to 10.28 and LR− from 0.102 to 0.54. Using the summary sensitivity and specificity of 78 and 84 percent, respectively, we calculated an overall PPV of 69 percent and NPV of 89 percent. Similarly, we calculated summary LR+ of 4.88 and LR− of 0.26.

Accuracy of CMR in Five Good-quality Studies

Next, we evaluated the accuracy of CMR compared with coronary angiography in the five good-quality studies. In these studies, sensitivity varied from 58 to 91 percent, and specificity varied from 59 to 91 percent; the median sensitivity was 75 percent, and the median specificity was 86 percent. Figure 32 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women from mixed populations.

Figure 32 presents forest plots of the individual study estimates of sensitivity and specificity of CMR for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the five included studies. In these studies, sensitivity varied from 58 to 91 percent, and specificity varied from 59 to 91 percent; the summary sensitivity was 72 percent, and the summary specificity was 84 percent.

Figure 32

Accuracy of CMR in five good-quality studies in women from mixed populations.

Figure 33 presents an SROC curve demonstrating an average sensitivity of 72 percent (95% CI, 55 to 85 percent) and specificity of 84 percent (95% CI, 69 to 93 percent).

Figure 33 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 72 percent (95% CI, 55 to 85%) and specificity of 84 percent (95% CI, 69 to 93%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 33

SROC curve for CMR in five good-quality studies in women from mixed populations.

The prevalence of CAD in these five good-quality studies ranged from 14 to 56 percent with a mean prevalence of 37 percent. In the individual studies, PPV ranged from 30 to 93 percent, and NPV ranged from 59 to 91 percent. LR+ ranged from 1.83 to 10.28 and LR− from 0.102 to 0.54. Using the summary sensitivity and specificity of 72 and 84 percent, respectively, we calculated an overall PPV of 72 percent and NPV of 84 percent. Similarly, we calculated summary LR+ of 4.5 and LR− of 0.33.

Coronary CTA

We identified eight studies evaluating the accuracy of coronary CTA compared with coronary angiography (Table 6).66,98,100105 Five of these studies were exclusively in women with no known CAD, and these are the studies used in our primary analysis. In our secondary analysis, we evaluated the accuracy of coronary CTA in diagnosing CAD in mixed populations of known and no known CAD, including the three additional studies.

Table 6. Summary of accuracy data evaluating coronary CTA for diagnosing CAD.

Table 6

Summary of accuracy data evaluating coronary CTA for diagnosing CAD.

Primary Analysis: Population of Women With No Known CAD

The five studies represent findings on coronary CTA use in 474 women (sample size ranging from 30 to 280 women). Of these studies, three were good quality, two were fair quality, and none were poor quality. Sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 91 percent; the median sensitivity was 89 percent, and the median specificity was 78 percent. Figure 34 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women with no known CAD.

Figure 34 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the five included studies. In these studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 91 percent; the summary sensitivity was 93 percent, and the summary specificity was 77 percent.

Figure 34

Accuracy of coronary CTA in women with no known CAD.

Figure 35 presents an SROC curve with an average sensitivity of 93 percent (95% CI, 69 to 99 percent) and specificity of 77 percent (95% CI, 54 to 91 percent).

Figure 35 presents a summary receiver operating characteristic (SROC) curve with an average sensitivity of 93 percent (95% CI, 69 to 99%) and specificity of 77 percent (95% CI, 54 to 91%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 35

SROC curve for coronary CTA in women with no known CAD.

The prevalence of CAD in the 5 studies ranged from 16 to 60 percent with a mean prevalence of 30 percent. In the individual studies, PPV ranged from 23 to 87 percent, and NPV ranged from 46 to 91 percent. LR+ ranged from 1.58 to 11 and LR− from 0 to 0.41. Using the summary sensitivity and specificity of 93 and 77 percent, respectively, we calculated an overall PPV of 63 percent and NPV of 96 percent. Similarly, we calculated summary LR+ of 4.04 and LR− of 0.09.

Accuracy of Coronary CTA in Three Good-Quality Studies

Next, we evaluated the accuracy of coronary CTA compared with coronary angiography in the three good-quality studies. In these studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 91 percent; the median sensitivity was 86 percent, and the median specificity was 73 percent. Given the small number of studies and the specific point estimate and CIs of these studies, our meta-analytic modeling was not able to reach convergence and provide a summary sensitivity and specificity for this set of studies.

The prevalence of CAD in these 3 good-quality studies ranged from 16 to 27 percent with a mean prevalence of 21 percent. In the individual studies, PPV ranged from 23 to 80 percent, and NPV ranged from 91 to 100 percent. LR+ ranged from 1.58 to 11 and LR− from 0 to 0.41.

Secondary Analysis: Mixed Population of Women With Known and No Known CAD

We performed a secondary analysis where we expanded our inclusion criteria to include studies whose patient population included a mix of women with known CAD and women with no known CAD. This expanded inclusion criteria allowed three additional studies to be included in our analysis (totaling eight studies). The 8 studies represent findings on coronary CTA use in 690 women (sample size ranging from 30 to 280 women). Of these eight studies, four were good quality, four were fair quality, and none were poor quality (Table 6).

In these 8 studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 100 percent; the median sensitivity was 92 percent, and the median specificity was 88 percent. Figure 36 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women from mixed populations.

Figure 36 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women from mixed populations. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the eight included studies. In these studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 100 percent; the summary sensitivity was 94 percent, and the summary specificity was 87 percent.

Figure 36

Accuracy of coronary CTA in women from mixed populations.

Figure 37 presents an SROC curve demonstrating an average sensitivity of 94 percent (95% CI, 81 to 98 percent) and specificity of 87 percent (95% CI, 68 to 96 percent).

Figure 37 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 94 percent (95% CI, 81 to 98%) and specificity of 87 percent (95% CI, 68 to 96%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 37

SROC curve for coronary CTA in women from mixed populations.

The prevalence of CAD in these 8 studies ranged from 11 to 60 percent with a mean prevalence of 31 percent. In the individual studies, PPV ranged from 23 to 100 percent, and NPV ranged from 46 to 100 percent. LR+ ranged from 2.54 to 13.36 and LR− 0 to 0.41. Using the summary sensitivity and specificity of 94 and 87 percent, respectively, we calculated an overall PPV of 76 percent and NPV of 97 percent. Similarly, we calculated summary LR+ of 7.23 and LR− of 0.069.

Accuracy of Coronary CTA in Four Good-Quality Studies

Next, we evaluated the accuracy of coronary CTA compared with coronary angiography in the four good-quality studies. In these studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 91 percent; the median sensitivity was 85 percent, and the median specificity was 80 percent. Figure 38 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women with from mixed populations.

Figure 38 presents forest plots of the individual study estimates of sensitivity and specificity of coronary CTA for diagnosing CAD in women with no known CAD. Error bars represent 95% CIs; the dashed vertical line represents the summary sensitivity and specificity for the included studies. In these studies, sensitivity varied from 70 to 100 percent, and specificity varied from 46 to 91 percent; the summary sensitivity was 83 percent, and the summary specificity was 77 percent.

Figure 38

Accuracy of coronary CTA in four good-quality studies in women from mixed populations.

Figure 39 presents an SROC curve demonstrating an average sensitivity of 83 percent (95% CI, 58 to 94 percent) and specificity of 77 percent (95% CI, 40 to 94 percent). It is important to note that given the small number of studies and wide confidence intervals that these summary statistics should be interpreted with caution.

Figure 39 presents a summary receiver operating characteristic (SROC) curve demonstrating an average sensitivity of 83 percent (95% CI, 58 to 94%) and specificity of 77 percent (95% CI, 40 to 94%). The ROC curve illustrates the tradeoff between sensitivity and specificity since the threshold that defines a positive test result varies from the most stringent to the least stringent. Open circles represent individual study estimates of sensitivity and specificity. The black circle indicates the average sensitivity and specificity estimate of the study results, and the dashed circle represents the 95 percent confidence region around it.

Figure 39

SROC curve for coronary CTA in four good-quality studies in women from mixed populations.

The prevalence of CAD in these 4 good-quality studies ranged from 16 to 39 percent with a mean prevalence of 25 percent. In the individual studies, PPV ranged from 23 to 81 percent, and NPV ranged from 89 to 100 percent. LR+ ranged from 1.58 to 11 and LR− from 0 to 0.41. Using the summary sensitivity and specificity of 83 and 77 percent, respectively, we calculated an overall PPV of 55 percent and NPV of 93 percent. Similarly, we calculated summary LR+ of 3.61 and LR− of 0.22.

KQ 1 Summary

Table 7 and Figure 40 show the summary of the diagnostic accuracy of ECG, ECHO, SPECT, CMR, and coronary CTA modalities in women presented in Figures 439. The information is presented separately for no known CAD and mixed CAD populations, as well as for all studies separately from the good-quality studies. Overall, within a given modality, the summary sensitivities and specificities were similar for both types of populations (known and no known CAD) and for all studies when compared with good-quality studies. When accounting for only the good-quality studies, it appeared that the diagnostic accuracy of detecting CAD in women with unknown CAD was better (in descending order) for coronary CTA, SPECT, ECHO, CMR, and ECG. For the newer technologies (i.e., coronary CTA and CMR), more studies in women would be needed to support these findings since the 95% CIs were quite wide.

Table 7. Summary of accuracy of NITs compared with coronary angiography for diagnosing CAD in women.

Table 7

Summary of accuracy of NITs compared with coronary angiography for diagnosing CAD in women.

Figure 40 shows the summary of the diagnostic accuracy of ECG, ECHO, SPECT, CMR, and coronary CTA modalities in women with no known CAD. It appeared that the diagnostic accuracy of detecting CAD in women was better (in descending order) for coronary CTA, SPECT, ECHO, CMR, and ECG. For the newer technologies (i.e., coronary CTA and CMR), more studies in women would be needed to support these findings since the 95% CIs were quite wide.

Figure 40

Summary of accuracy of NITs compared with coronary angiography for diagnosing CAD in women with no known CAD (all studies).

To minimize the risk of spectrum bias, our primary analysis focused on women with no known CAD. We also explored mixed populations of women with known and no known CAD in sensitivity analyses. These analyses did not demonstrate a significant difference in terms of the sensitivities and specificities from our primary analysis. We also explored whether the accuracy of the modalities were correlated with the underlying prevalence of disease in the population of interest. The mean prevalences and 95% CIs for ECG, SPECT, ECHO, CMR, and coronary CTA with the population of women with no previously known CAD were 0.41 (0.36 to 0.46), 0.44 (0.34 to 0.55), 0.43 (0.37 to 0.50), 0.26 (0.14 to 0.44), and 0.29 (0.13 to 0.54), respectively. We evaluated whether these prevalences were different across modalities using a random-effects model and did not find a statistically significant difference (p = 0.17). Thus, these analyses did not indicate any specific trend or relationship between prevalence and the NIT’s sensitivity or specificity. There did appear to be an increase in the sensitivity of CMR over time, although the wide confidence intervals for this characteristic highlight the uncertainty in this trend.

We assessed the risk of verification bias by exploring the studies in our analysis that did not complete a coronary catheterization in all of the patients who underwent the NIT. In the population of women with no previously known CAD, this represented one study of SPECT,52 one study of ECHO,79 three studies of ECG,29,52,58 and no studies of CMR or coronary CTA. Given the small number of total studies with this potential bias, we felt confident that our primary results were minimized for verification bias.

We explored the potential for publication bias across the different modalities in our four populations of interest (studies of women with no known CAD, good-quality studies of women with no known CAD, studies of women from mixed populations, and good-quality studies of women from mixed populations). Our analyses did not provide evidence for publication bias, with our p values ranging from 0.093 to 0.95.

In a final analysis, we explored whether there was a statistically significant difference between the diagnostic accuracy of testing modalities in women using a generalized linear mixed model, with NIT modality and disease state (no known and mixed CAD) as covariates in the model. Our analyses determined that for women with no previously known CAD, there were differences between the performance of the available modalities (p < 0.001). The sensitivity of ECHO and SPECT was significantly higher than that of ECG. Specificity of ECG was less than that of CMR (borderline) and of ECHO. We similarly explored the differences among the modalities in the subset of studies that were good-quality and also where there was no known CAD in the included population. These analyses again demonstrated differences between performance of tests (p = 0.008) with the specificity of ECG being less than that of CMR and ECHO.

Comparative Accuracy of NIT Modalities in Men

Although it was not the primary goal of this systematic review, we also evaluated, when possible, the accuracy of the five NIT modalities in male patients and specifically how the accuracy of these modalities differed between men and women. Most of the studies included in our analysis, however, did not include data on both sexes. Specifically, of the 41 included studies evaluating ECG in a mixed population, only 20 included data on men as well as women. Similarly for the 22 ECHO, 30 SPECT, 6 CMR, and 8 CTA included studies, only 9, 11, 3, and 7 respectively included data on men.

Although limited, the available studies provided enough data for men to determine summary sensitivity and specificity estimates and to evaluate whether the accuracy of these modalities differed between men and women (Table 8). In Tables 26, we provide the accuracy data for the individual studies included in our analysis that had male representation. In Table 7 we list the summary sensitivities and specificities calculated with an SROC curve as described in our primary women analyses. Given the reduced number of available studies, we focused on studies with populations of either no known CAD or a mix of known and no known CAD. When comparing the accuracy of the modalities between men and women enrolled in the same studies, the ECG and coronary CTA modalities were both less sensitive and less specific in women. The ECHO, CMR, and SPECT modalities, although less sensitive, appeared to be more specific in women. The lower specificity of the ECG modality in women, however, is the only estimate that was determined to be a statistically significant difference.

Table 8. Summary of accuracy of NITs for diagnosing CAD in men compared with women from mixed populations.

Table 8

Summary of accuracy of NITs for diagnosing CAD in men compared with women from mixed populations.

Key Question 2. Predictors of Diagnostic Accuracy

KQ 2. What are the predictors of diagnostic accuracy (e.g., age, race/ethnicity, body size, heart size, menopausal status, functional status, stress modality) of different NITs in women?

Key Points

  • Significant variability existed around diagnostic accuracy among studies examining each NIT modality.
  • Only one study compared diagnostic accuracy for women with diagnostic accuracy for men.
  • The studies reviewed did not examine functional status, age, or body size—the main predictors examined included heart size, pretest probability of CAD, race/ethnicity, postmenopausal status, and beta blocker use.
  • There was heterogeneity in the types of predictors reported.
  • Studies that examined heart size varied on the effect on diagnostic accuracy by stress modality. These studies suggest that increased heart size reduces the specificity of stress ECG, ECHO, and SPECT.
  • One fair-quality study of 51 women reported that beta blocker use reduces the specificity of stress ECG and the sensitivity and specificity of SPECT. Withholding beta blockers prior to exercise stress testing is common to allow patients to achieve a target (or higher) heart rate in assessing for ischemia. One study showed that the PPV increases as the pretest probability of CAD increases for stress ECG and ECHO.
  • Insufficient evidence was available to draw definitive conclusions about predictors given the small number of studies for each predictor and for each modality, as well as the combination of predictor by modality.

Detailed Synthesis

Many factors are reported in the literature that affect the diagnostic accuracy of noninvasive testing in women, including (1) higher prevalence in women of nonobstructive CAD (microvascular abnormalities, mitral valve prolapse), (2) less predictive symptomatology, (3) limited exercise tolerance because of older age, obesity, and diabetes at initial diagnosis, (4) different response to exercise than men, (5) lower peak exercise values, (6) lower increase in the left ventricular ejection fraction, (7) an increase in cardiac output by enhancing end-diastolic volume, (8) inappropriate catecholamine release, (9) hormonal influences of estrogens mimicking a digitalis-like false-positive ECG response, (10) anatomic differences affecting stress test results, (11) breast attenuation artifacts, (12) smaller coronary artery size, (13) smaller left ventricular chamber size, (14) higher prevalence of single-vessel disease, and (15) poor left ventricular opacification on echocardiography.

For KQ 2, we examined studies for the following nine predictors of diagnostic accuracy of different NITs in women: age, race/ethnicity, body size, heart size, menopausal status, functional status, stress modality, cardiac risk factors, and pretest probability of CAD. We identified 11 studies25,28,31,38,50,55,67,69,83,95,97 that described diagnostic accuracy, but only 9 of these studies examined predictors of diagnostic accuracy. Four of the 11 studies were considered good quality, 5 were fair quality, and 2 were poor quality.

Findings of Diagnostic Accuracy by Predictor

Findings of diagnostic accuracy by predictor in the studies we reviewed are summarized in Tables 912. Of the nine predictors we originally searched for, we found studies that addressed four of the predictors: (1) age combined with (2) menopausal status, (3) race/ethnicity, and (4) heart size. We also found two other types of predictors reported in women: (5) pretest probability based on cardiac risk factors and (6) use of beta blockers:

Table 9. Age and menopausal status as a predictor.

Table 9

Age and menopausal status as a predictor.

Table 10. Race/ethnicity as a predictor.

Table 10

Race/ethnicity as a predictor.

Table 11. Heart size as a predictor.

Table 11

Heart size as a predictor.

Table 12. Other potential predictors.

Table 12

Other potential predictors.

  • The study by Cin,, et al.31 examined the diagnostic accuracy of stress ECG in postmenopausal women ages 55 to 64.
  • Two studies examined race/ethnicity: One study by Vashist,, et al.83 compared the diagnostic accuracy of SPECT across three race/ethnic categories—African American, Hispanic, and Asian—and another study by Yeih,, et al.50 examined the diagnostic accuracy of ECG testing and SPECT in Asian women in Taiwan.
  • Four studies examined heart size as a predictor of diagnostic accuracy: Lu,, et al.28 and Gebker,, et al.,95 compared left ventricular hypertrophy (LVH) with no LVH for ECG, ECHO, SPECT, and CMR. The study by Klem,, et al.,97 examined heart size in grams for CMR testing, and Siegler,, et al.,25 examined heart size for ECG alone.
  • Three studies, Yeih,, et al.,,50 Marwick,, et al.,55 and Ho,, et al.,67 examined pretest probability as a predictor of accuracy for ECG, dobutamine ECHO, and CMR.
  • The study by Yeih,, et al.,50 examined the use of beta blockers on diagnostic accuracy in ECG and SPECT.
  • We identified no studies that examined age alone, functional status, or body size as predictors of diagnostic accuracy in women.

The following sections focus on differences in predictors, organized by NIT modality. Summaries are in Tables 1316.

Table 13. Summary table for ECG.

Table 13

Summary table for ECG.

Table 16. Summary table for CMR.

Table 16

Summary table for CMR.

ECG

Six studies25,28,31,38,50,55 of stress ECG assessed predictors of diagnostic accuracy in women, including age/menopausal status, disease probability, use of beta blockers, and heart size. There was some variability in diagnostic accuracy using ECG testing (both exercise and with dobutamine) (Table 13).

Overall accuracy. For exercise ECG testing, overall sensitivity ranged from 67 to 86 percent, and specificity ranged from 44 to 78 percent. Yeih,, et al.,50 examined ECG testing with dobutamine, with a resulting sensitivity of 43 percent and specificity of 83 percent. Ho,, et al.,67 examined ECG with exercise treadmill test and found that the sensitivity was 71 percent and specificity was 44 percent.

Age/menopausal status. Cin,, et al.,31 examined age and menopausal status among women ages 55 to 64 (postmenopausal) and found that ECG testing performed moderately better than in other studies not targeting this age group (sensitivity of 86 percent, specificity of 60 percent). Unfortunately, that study did not have a younger or older age group to compare findings with and was rated poor quality.

Disease probability. Marwick,, et al.,55 compared ECG testing performance across disease probability categories (low, intermediate, and high as determined by a combination of type of chest pain and age) and found that ECG testing performed better in the low-probability group compared with the intermediate- or high-probability group. The quality of this study was fair.

Beta blockers. In the study by Yeih,, et al.,50 which compared low and high probability calculated according to age and symptoms, ECG testing performed better in the high-probability group, as expected. This study evaluated only Asian women in Taiwan. Women receiving beta blockers had fewer false positives compared with women not taking beta blockers at the time of the ECG testing.

Heart size. In the study by Lewis,, et al.,38 ECG testing performed better after excluding nondiagnostic cases (sensitivity 67 percent versus 43 percent, specificity 78 percent versus 66 percent). In the study by Lu,, et al.,28 ECG testing performed better in patients without LVH compared with those with LVH (specificity 69 percent versus 31 percent), and this finding may relate to heart size or baseline ECG strain associated with LVH.

Siegler,, et al.,25 examined 1011 patients to determine if heart size could lead to higher false-positives rates among patients undergoing stress ECG; 482 women were enrolled in the study. The prevalence of CAD was 28 percent among women enrolled. Overall sensitivity and specificity for women undergoing stress ECG was 60 and 80 percent respectively. There was a significant association between ECG outcome and heart size among women (p = 0.03), where smaller hearts were associated with a lower specificity compared with normal size hearts.

ECHO

Four studies28,55,67,69 examined the predictors of diagnostic accuracy for exercise ECHO in women, specifically disease probability and heart size (Table 14). There was significant variability in the diagnostic accuracy of stress ECHO imaging.

Table 14. Summary table for ECHO.

Table 14

Summary table for ECHO.

Overall accuracy. The overall sensitivity ranged from 61 to 93 percent. The specificity ranged from 71 to 91 percent. Shin,, et al.69 examined 464 patients to determine potential factors that could lead to a higher false-positive rate among patients undergoing exercise ECHO. There were 162 women enrolled in the study. The prevalence of CAD was 34 percent among the women enrolled. For women undergoing exercise ECHO, the overall sensitivity was 82 percent and specificity was 71 percent, which was less than for men in the study. PPV was 59 percent and NPV was 88 percent; LR+ was 2.82 and LR− was 0.26.

Marwick,, et al.,55 found an overall sensitivity of 80 percent (p = 0.050) and specificity of 81 percent (p < 0.004) for exercise ECHO compared with exercise ECG. PPV was 71 percent and NPV was 87 percent. Exercise ECHO was compared with exercise ECG, which had a sensitivity of 77 percent and specificity of 56 percent; LR+ was 4.28 and LR− was 0.25.

Lu., et al.,28 examined the diagnostic accuracy of dipyridamole and dobutamine stress ECHO modalities. The study enrolled 76 Asian women from Taiwan. The prevalence of CAD was 41 percent. For dipyridamole, sensitivity was 61 percent and specificity was 91 percent. PPV was 83 percent and NPV was 77 percent; LR+ was 6.9 and LR− was 0.42. A dobutamine stress ECHO was more sensitive (87 percent) and less specific (82 percent). (Note that the p-value for comparing the sensitivity was 0.02, and for specificity the exact p-value was not reported but was said to be not statistically significant.) PPV was 77 percent and NPV was 90 percent; LR+ was 4.9 and LR− was 0.16.

Disease probability. The study by Marwick, et al.,55 also compared the diagnostic accuracy of exercise ECHO in women with different pretest probabilities of CAD (i.e., high probability, intermediate probability, and low probability). The prevalence in each group was 69, 40, and 14 percent respectively. The sensitivity was highest (88 percent) among patients with low probabilities of CAD, second highest (82 percent) among those with high probability of CAD, and lowest (76 percent) among those with an intermediate probability of CAD. However, specificity was highest (86 percent) among patients with intermediate probability of CAD, the next highest (80 percent) was among patients with highest probability of CAD, and the lowest (78 percent) was among those with a low probability of CAD. The quality of this study was fair. In this study, the diagnostic accuracy changed depending on the prevalence of CAD. Ho, et al.,67 examined the diagnostic accuracy of dobutamine stress ECHO. The study enrolled 51 women from Taiwan. The prevalence of CAD was 27 percent. The sensitivity was 93 percent and specificity was 82 percent; PPV was 87 percent and NPV was 90 percent; LR+ was 5.12 and LR− was 0.08. This study also compared the diagnostic accuracy of dobutamine stress ECHO in patients with different coronary risk factors and different pretest probabilities of CAD.

In women with two or more CAD risk factors, Ho, et al., found that the sensitivity was similar to the overall sensitivity (92 percent), but the specificity decreased (67 percent). PPV was 85 percent and NPV was 80 percent; LR+ was 2.75 and LR− was 0.13. For women with zero or one CAD risk factor, sensitivity was 94 percent, specificity was 88 percent; PPV was 93 percent and NPV was 89 percent; LR+ was 7.53 and LR− was 0.07.

In addition, in women with at least a 50-percent pretest probability of CAD, the sensitivity did not change dramatically from the overall sensitivity (94 percent), but the specificity increased (88 percent). PPV was 90 percent and NPV was 86 percent; LR+ was 3.78 and LR− was 0.07. In women with a less than 50 percent pretest probability of CAD, the sensitivity was 91 percent and specificity was 86 percent. PPV was 83 percent and NPV was 92 percent; LR+ was 6.36 and LR− was 0.11. One limitation of this study is that neither beta blockers nor calcium channel blockers were withheld.

Heart size. Heart size was also examined in the study by Lu, et al.,28 specifically, the effect that LVH had on diagnostic accuracy. Studies have suggested that LVH is commonly associated with ECG repolarization abnormalities, such as ST elevations in the absence of wall motion abnormalities or other evidence of inducible ischemia, leading to a potentially higher false-positive rate.106 The presence of LVH led to a lower specificity in both dipyridamole and dobutamine (81 percent versus 69 percent) stress test modalities. The sensitivities were not reported. The overall quality of this study was good.

SPECT

Four studies28,50,67,83 of SPECT assessed predictors of diagnostic accuracy in women, including race/ethnicity, heart size and use of beta blockers. There was considerable variability in diagnostic accuracy using SPECT (Table 15).

Table 15. Summary table for SPECT.

Table 15

Summary table for SPECT.

Overall accuracy. Overall sensitivity ranged from 71 to 90 percent and specificity ranged from 27 to 88 percent. In the study by Ho, et al.,67 the prevalence of CAD was 27 percent. This study enrolled 51 women, but only 44 of them received SPECT. The overall sensitivity and specificity for SPECT was 79 percent and 75 percent respectively. PPV was 79 percent and NPV was 75 percent; LR+ was 3.17 and LR− was 0.28. It should be noted that, as in the Vashist, et al., study, SPECT was performed by either dipyridamole or dobutamine infusions. Determination of which agent to be used was based on patient preference.

Race/ethnicity. Vashist, et al.,83 examined race/ethnicity differences in performance of SPECT. This retrospective study was noted to have several limitations, including combining exercise SPECT with dipyridamole and dobutamine. This study found a difference in prevalence of CAD but did not find a difference in diagnostic accuracy between Hispanics and African Americans. In the study by Yeih, et al.,50 that focused on Asian women, the sensitivity was similar to other race/ethnic groups, but the specificity was significantly higher than that observed in Vashist, et al.,83 (87 versus 27 percent).

Beta blockers. In the study by Yeih, et al.,50 beta blockers reduced the diagnostic accuracy compared with no beta blockers (sensitivity 67 versus 77 percent, specificity 78 versus 93 percent). This fits with current clinical practice where beta blockers are withheld 48 hours prior to exercise stress to allow patients to achieve a higher (or target) heart rate for the assessment of ischemia.

Heart size. In the study by Lu, et al.,28 SPECT was found to have fewer false positives in the group without LVH versus the group with LVH (specificity 66 versus 31 percent).

CMR

Two studies95,97 of CMR examined the influence of heart size on diagnostic accuracy in women (Table 16). The overall sensitivities were similar at 84 percent versus 85 percent, and the specificities ranged from 86 to 88 percent.

Overall accuracy. Klem, et al.,97 studied 136 women at two academic medical centers who presented with chest pain. A multicomponent cardiac MRI (CMR test) that consisted of adenosine stress, rest perfusion, and delayed-enhancement CMR was used. The overall sensitivity and specificity of all three modes were found to be 70 and 81 percent respectively. PPV was 74 percent and NPV was 85 percent. These were reflected in a patient sample where the prevalence of significant CAD was 27 percent. LR+ was 5.96 and LR− was 0.35. Sensitivity and specificity decreased to 78 and 56 percent if perfusion CMR was used alone; LR+ also decreased to 1.77, and LR− increased to 0.39. Also, when coronary angiography results were examined for each patient, the sensitivity for multicomponent CMR was highest when patients with significant CAD had at least two-vessel disease with a sensitivity of 100 percent, as opposed to 71 percent when the patients had only single-vessel disease. The specificity was unaffected by the extent of CAD and remained 88 percent regardless if the patient had single-vessel or multiple-vessel disease. The overall quality of the study was good.

The study conducted by Gebker, et al.95 examined the effectiveness of dobutamine stress CMR in the detection of CAD in women compared with men. The study evaluated 183 women and 541 men for suspected CAD. The prevalence of CAD in women and men was 54 percent (99 women) and 74 percent (365 men). The overall sensitivity and specificity of dobutamine stress CMR was found to be 85 and 86 percent in women and 86 and 83 percent in men. PPV and NPV were 88 and 83 percent in women and 94 and 67 percent in men; LR+ was 5.94 and LR− was 0.18 in women compared with 5.12 and 0.17 in men. The sensitivity of detecting CAD was greater in the presence of multiple-vessel CAD than in single-vessel CAD in both men and women (91 percent compared with 81 percent in women and 91 percent compared with 82 percent in men).

Heart size. Klem, et al.,97 also analyzed study data to determine if the heart size of patients affected the ability to detect CAD. Investigators divided the patients into two groups—those with LV mass less than 97 grams (defined as “small” hearts) and those with LV mass greater than 97 grams (defined as “large” hearts). The prevalence of significant CAD was found to be similar and was 29 percent in those with small hearts and 26 percent in those with large hearts. The reported sensitivity was 69 percent in those with small hearts and 95 percent in those with large hearts. The authors suggested that this was caused by the limitations in spatial resolution with the 1.5 Tesla MR magnet and by the fact that a smaller heart leads to a smaller number of image pixels that are available to visualize the left ventricular wall.97

Gebker, et al.,95 also attempted to understand the relationship between heart size, in particular LVH, and the ability to detect the presence of CAD in women compared with men. Previous studies have shown that the presence of LVH causes a higher rate of false-negative studies.107 Women undergoing dobutamine CMR with evidence of LVH had a lower sensitivity but higher specificity than women without LVH (sensitivity 80 percent versus 87.5 percent, specificity 91 percent versus 84 percent). Men undergoing dobutamine CMR with evidence of LVH had a lower sensitivity and a higher specificity than men without LVH (sensitivity 79 percent versus 90 percent, specificity 95 percent versus 78 percent).

KQ 2 Summary

To summarize, there was insufficient available evidence to draw definitive conclusions about predictors given the small number of studies for each predictor and for each modality, as well as the combination of predictor by modality. The main predictors examined included heart size, pretest probability of CAD, race/ethnicity, postmenopausal status, and beta blocker use. No studies examined functional status, age alone, or body size. Significant variability around diagnostic accuracy existed among the studies examining each stress modality, and studies that examined heart size varied on the effect on diagnostic accuracy by stress modality.

Key Question 3. Use of NITs To Improve Risk Stratification, Decisionmaking, and Clinical Outcomes

KQ 3. Is there evidence that the use of NITs (when compared with other NITs or with coronary angiography) in women improves:

KQ 3a.

Risk stratification/prognostic information?

KQ 3b.

Decisionmaking regarding treatment options (e.g., revascularization, optimal medical therapy)?

KQ 3c.

Clinical outcomes (e.g., death, myocardial infarction, unstable angina, hospitalization, revascularization, angina relief, quality of life)?

Key Points

  • There were insufficient data to demonstrate that the use of specific NITs (compared with coronary angiography) routinely provided incremental risk stratification, prognostic information, or other meaningful information to improve decisionmaking.
  • Compared with other NITs, stress ECG testing had higher rates of indeterminate results in women, which limited the ability to compare risk stratification with low-risk and high-risk findings from other stress imaging studies.
  • There was insufficient evidence on the comparative effectiveness of different NITs to have an impact on clinical decisionmaking that leads to improved patient outcomes in women.

Detailed Synthesis

For KQ 3, we examined studies that reported prognostic, outcome, or decisionmaking data comparing one NIT with another NIT or with coronary angiography in women with symptoms suspicious for CAD. We identified 13 studies,2224,30,38,52,83,99,108112 of which 3 were considered good quality, 9 were fair quality, and 1 was poor quality. The majority of the comparative studies evaluated (n = 7) provided information on risk stratification and prognostic information. Five of these studies evaluated clinical outcomes (two studies with information on both risk stratification and clinical outcomes), and two studies were aimed at clinical decisionmaking.

In order to evaluate the ability of NITs to provide incremental information on risk stratification, prognosis, and decisionmaking, we evaluated studies that reported clinical outcomes for at least two NITs or coronary angiography. Although several studies in women described observational cohorts of women undergoing an NIT who were followed for findings related to clinical outcomes, these studies were generally excluded since they were limited by the population risk studied, and they did not provide information on evidence for comparative effectiveness. Table 17 summarizes the findings for KQ 3a, 3b, and 3c.

Table 17. Summary of findings for KQ 3.

Table 17

Summary of findings for KQ 3.

KQ 3a: Risk Stratification and Prognostic Information

Eight studies (two good quality, six fair)38,52,83,99,108111 provided evidence on risk stratification and prognosis. Of these studies, two108,109 from the Women’s Ischemia Study Evaluation (WISE) study evaluated the prognostic significance of CMR for women with suspected myocardial ischemia but without significant obstructive CAD (< 50 percent stenosis) on coronary angiography. In the study by Doyle, et al.,108 two imaging variables—global magnetic resonance-myocardial perfusion imaging ratio of average peak signal-to-normalized uptake slope at stress and ejection fraction—were predictive of events including death, MI, and hospitalization for worsening angina by univariable Cox regression modeling (hazard ratio 0.516; 95% CI, 0.314 to 0.848; p < 0.05 and hazard ratio of 0.949; 95% CI, 0.911 to 0.990; p < 0.05, respectively). Kaplan-Meier survival curves indicated a significant difference in time to adverse events between risk groups (log-rank 15.0, p < 0.0001). Annualized event rates were 12 percent in the high-risk group (2 deaths and 10 hospitalizations for worsening angina) and 4 percent in the not high-risk group (2 MIs and 9 hospitalizations for worsening angina).

Johnson, et al.,109 found that of the 74 women without CAD undergoing CMR, 19 percent had a cardiovascular event (1 heart failure, 12 unstable angina admission, 2 other vascular events). Women without CAD and a negative CMR had an 87-percent cumulative 3-year event-free survival rate. Women with CAD and positive (abnormal) CMR and women in the WISE study (control group, women with angiographically defined CAD) had lower event-free rates (57 percent; p = 0.009, and 52 percent; p < 0.0001 respectively). Among women without CAD, the rate of hospitalization for angina was lower in those with normal CMR compared with abnormal CMR (12 percent versus 36 percent; p < 0.05).

Two other studies52,83 described the low event rate and good prognosis in patients with low-risk findings on SPECT with diagnostic validation from coronary angiography.52,83 In a retrospective analysis, Vashist, et al.,83 specifically studied the prognostic value of myocardial perfusion imaging in minority women (African American, Hispanic, Asian) and found that low-risk perfusion scanning signified a favorable prognosis for mortality at 2 years. Five of the 54 patients (9.3 percent) had died at 2 years (3 with intermediate-risk scan and 2 with high-risk scan). In the prospective study by Mieres, et al.,52 evaluating the prognostic accuracy of SPECT in symptomatic postmenopausal women (n = 46) with an intermediate pretest likelihood for CAD, the cumulative 3- and 5-year event-free survival was 97 and 94 percent for normal myocardial perfusion scintigraphy (MPS) compared with 60 and 48 percent for those with abnormal MPS findings (p < 0.0001). Using ECG results for risk stratification, a negative exercise ECG was associated with 3- and 5-year event-free survival rates of 89 and 72 percent. When ECG and MPS results were included in a multivariable model, only MPS findings retained statistical significance (p = 0.017).

The prospective analysis by Coelho-Filho, et al.,99 evaluated the prognostic value of stress CMR in a consecutive group of women (n = 177). The myocardial extent of ischemia (ISCH-SCORE) was the strongest predictor of MACE events (cardiac death and acute MI) both with univariate and multivariate analysis: hazard ratio (95% CI) 1.36 (1.23 to 1.5) and 1.49 (1.31 to 1.69) respectively. Women with evidence of ischemia also had a higher annual rate of MACE and cardiac death compared with women without ischemia (15.1 percent and 8.2 percent versus 0.3 percent and 0 percent respectively).

Finally, three studies38,110,111 demonstrated the limitations and prognostic significance of ECG compared with coronary angiography and clinical outcomes. Morise, et al.,110 initially developed an exercise ECG score, including clinical and ECG variables to help risk stratify women with suspected CAD into groups of gradually increasing frequency of coronary disease and death. The score was then applied to women enrolled in the WISE study to assess the ability of the pretest and new exercise scores to stratify risk in women with low prevalence of angiographically defined CAD.111 Using the pretest score, a Kaplan-Meier curve analysis of the composite endpoint (death, MI, stroke, or late revascularization) revealed a clear separation between the low-risk group and the intermediate- and high-risk groups for as long as 4 years. The intermediate- and high-risk groups were separable for as long as 1.5 years but thereafter became less clearly separable. Using the exercise population, the number of events decreased, and the Kaplan-Meier curve displayed a less clear separation and a nonsignificant difference between the curves (p = 0.11). Using data from the WISE study, Lewis, et al.,38 found that the overall sensitivity of ECG was 31 percent for obstructive CAD on coronary angiography and that the inability to perform the ECG, rather than findings on ECG, predicted MI, death, or heart failure. These data emphasize some of the limitations of using ECG in women.

KQ 3b: Decisionmaking for Treatment Options

We found two studies23,112 with information on clinical decisionmaking that compared different NIT strategies. The prospective analysis by Wong, et al.,112 evaluated rates of referral for arteriography and revascularization according to sex. In this study, men were more likely to be referred for percutaneous transluminal coronary angioplasty or coronary artery bypass grafting than were women (59.4 percent versus 32.8 percent; odds ratio 3.0; 95% CI, 2.0 to 4.6). This difference in referral rate seemed to be linked to higher incidence of significant CAD in men (56.2 percent) compared with women (16.4 percent). When accounting for the difference in CAD incidence between men and women, there was no significant difference in revascularization rates, thus no difference in unnecessary referrals.

The study by Sanfilippo, et al.,23 was an RCT of three testing strategies in women with chest pain. This fair-quality study included 158 women randomized to ECG, exercise stress ECHO, or dobutamine stress ECHO. All tests resulted in a diagnosis of cardiac chest pain or noncardiac chest pain or were indeterminate. The study found that ECG had a higher likelihood of indeterminate results, and patients with noncardiac chest pain from all modalities had low event rates. Although these two studies were informative, there remains a significant need for studies that evaluate clinical decisionmaking to improve treatment options for patients.

KQ 3c: Clinical Outcomes

We found four studies that provided data on the comparative clinical outcomes for different NITs.22,24,30,38 As previously stated, the study by Lewis, et al.,38 identified limitations in diagnostic accuracy with ECG (overall sensitivity of 31 percent) and associated limitations in prognostic impact. Event-free survival was shorter in women who did not undergo ECG. The proportional hazards model that included ECG, CAD, and age showed estimated hazard ratios (95% CI) of 0.42 (0.18 to 0.97), 3.88 (1.72 to 8.73), and 1.01 (0.98 to 1.05) respectively (p = 0.0003). These results indicate a decreased risk of an event for women who underwent ECG, regardless of outcome, compared with those who did not undergo ECG.

The study by Dodi, et al.,22 evaluated 244 women who underwent ECG and exercise ECHO for symptoms suspicious for CAD. This study found that the prognostic information with stress ECHO (odds ratio 40) for death or MI was significantly above the effect of a positive ECG (odds ratio 3.5).

The study by Raman, et al.,30 enrolled 23 women who had a positive nuclear scan (SPECT) referred for coronary angiography and who underwent dobutamine stress CMR to assess for ischemia. Followup in this study lasted 20 ± 8 months, and there were no reported MI, hospitalizations, or death in their study sample.

The study by Shaw, et al.,24 evaluated women with an intermediate pretest likelihood of CAD. In that study, two diagnostic strategies, exercise ECG and SPECT, had a different effect on the 2-year posttest outcomes for MACE (cardiac death, nonfatal MI) or hospital admission for acute coronary syndrome or heart failure. MACE-free survival was found to be identical (98 percent) for women randomized to the ECG or SPECT arm (p = 0.59), and the observed 2-year MACE rate was 1.7 percent for ECG and 2.3 percent for SPECT (relative hazard, 95% CI for MACE in SPECT arm versus ECG arm was 1.3; 0.5 to 3.5; p = 0.59). It was noted that the study was underpowered for the MACE outcome (post hoc analysis power of 15 percent at 0.05 significance level).

No articles of coronary CTA reported clinical outcomes data.

KQ 3 Summary

We identified 13 comparative studies for KQ 3 that evaluated NITs for risk stratification, prognosis, and decisionmaking affecting clinical outcomes. Two studies reported that women with abnormal CMR and normal coronary angiography had lower event-free survival rates. One study found that an abnormal SPECT resulted in a lower event-free survival rate. One study found that a negative stress ECG and diagnosis of noncardiac chest pain translated into lower event rates. Another study found that a positive stress ECHO had higher prognosis of worse cardiovascular events than a positive stress ECG. However, the studies were small and underpowered, and therefore all these findings would require significant confirmation and replication in larger studies with women. Overall, the literature identified was insufficient in demonstrating that the use of a specific NIT provided incremental risk stratification, prognostic information, or other meaningful information to improve decisionmaking and improve patient outcomes. There were specific limitations for the populations studied, including baseline risk, comparative outcomes, and relationship to diagnostic accuracy.

Key Question 4. Safety Concerns and Risks

KQ 4. Are there significant safety concerns/risks (i.e., radiation exposure, access site complications, contrast agent-induced nephropathy, nephrogenic systemic fibrosis, anaphylaxis, arrhythmias) associated with the use of different NITs to diagnose CAD in women with symptoms suspicious for CAD?

Key Points

  • Data specific to women on access site complications, contrast agent-induced nephropathy, nephrogenic systemic fibrosis, or anaphylaxis associated with NITs were not reported in any of the studies included in this report.
  • One study showed a significantly lower rate of supraventricular tachycardia (SVT) in women undergoing dobutamine ECHO compared with men.
  • A study of adenosine thallium SPECT showed higher rates of ST depression in women compared with men.
  • The results of one study suggested that lifetime attributable risk of both cancer incidence and fatal cancer incidence associated with a single coronary CTA examination was approximately twice as high in women compared with men for 50-, 60-, and 70-year old patients. These estimates are derived from projected data rather than from direct observation. The mean effective dose associated with coronary CTA in four studies ranged from 13.7 to 16.0 mSv for women and 11.1 to 16.4 for men. Radiation safety issues were not discussed in the studies that reported on other NITs.
  • Other than higher mean effective radiation doses for coronary CTA studies for women compared with men, from three of the four studies reporting radiation exposure levels, there is insufficient evidence to conclude that safety concerns, risks, or radiation exposure associated with different NITs to diagnose CAD in patients with symptoms suspicious for CAD differ significantly between women and men.

Detailed Synthesis

Tests and procedures are associated with varying degrees of risk. In the case of NITs intended to diagnose CAD in women with symptoms suspicious for CAD, physiological stress, contrast agents, and exposure to radiation may cause immediate or long-term harm. Exercise or pharmacological stressors, for example, may cause arrhythmias, acute or worsening ischemia, hypotension, or cardiac arrest. The contrast agents used in conjunction with ECHO, radionuclide myocardial perfusion imaging, CMR, and coronary CTA can cause anaphylaxis, nephrotoxicity, arrhythmias, or thromboembolism. The Food and Drug Administration (FDA) has issued a boxed warning for ECHO contrast agents that contain microscopic gas-filled spheres as well as for gadolinium-based contrast agents because of apparent risk of causing nephrogenic systemic fibrosis. CMR introduces the unique risk from metallic objects in or on the body (e.g., an aneurysm clip) causing bodily harm if moved by the magnetic forces used in CMR. Furthermore, radiation exposure associated with radionuclide myocardial perfusion imaging and coronary CTA may be carcinogenic. These safety concerns and risks may differ for women compared with men, in part because of differences in radiosensitivity between female and male reproductive organs as well as differences between reproductive organs and other organs or tissues. In the next sections, we summarize the evidence pertaining to safety concerns and risks of NITs among women, as reported and discussed in the studies included in this review (Table 18).

Table 18. Adverse effects of different NITs for screening of CAD in women.

Table 18

Adverse effects of different NITs for screening of CAD in women.

For KQ 4, we examined studies that reported data pertinent to safety concerns or risks associated with the use of NITs to diagnose CAD in women. We identified 13 studies,27,28,35,39,67,72,89,95,96,103,105,113,114 of which 9 were good quality and 5 were fair quality.

ECG

Five studies reported sex-specific safety data among women who underwent exercise or stress ECG and at least one other diagnostic test for CAD.27,28,35,39,67 Of these studies, one was a good-quality RCT, two were good-quality prospective cohort studies, and two were fair-quality prospective cohort studies, representing a total of 413 women. Only one study28 reported side effects specifically for stress ECG. In the study by Lu, et al.,28 of 76 hypertensive women, rhythm disturbances were noted in 11 percent of subjects, as were frequent and severe premature ventricular contractions (PVCs). The reported adverse event rate associated with exercise ECG in this group of patients was 0 percent for symptomatic hypotension, dyspnea, nausea or vomiting, severe headache, flushing, left branch bound block (LBBB), and supraventricular tachycardia (SVT). The other four studies reported side effects associated with ECHO but not with exercise/stress ECG.

ECHO

Data pertaining to safety in women who underwent exercise/stress ECHO testing were reported in six studies.27,28,35,39,67,72 Of these, four were good-quality prospective cohort studies, and two were fair-quality prospective cohort studies, representing a total of 513 women. One study28 compared dobutamine ECHO with dipyridamole ECHO, and five studies27,35,39,67,72 used dobutamine as the pharmacological stressor.

In the study by Lu, et al.,28 that reported adverse events among 76 hypertensive women who underwent dobutamine ECHO, dipyridamole ECHO, and exercise ECG, the rates of adverse events associated with dobutamine ECHO were 4 percent for symptomatic hypotension, 0 percent for dyspnea, 1 percent for nausea or vomiting, 3 percent for severe headache, 0 percent for flushing, 16 percent for rhythm disturbances, 13 percent for frequent and severe PVCs, 1 percent for LBBB, and 1 percent for SVT. In contrast, the rates of adverse events associated with dipyridamole ECHO were 1 percent for symptomatic hypotension, 3 percent for dyspnea, 7 percent for nausea or vomiting, 12 percent for severe headache, 13 percent for flushing, 4 percent for rhythm disturbances, 4 percent for frequent and severe PVCs, and 0 percent each for LBBB and SVT.

In the study by Laurienzo, et al.,39 that evaluated transesophageal dobutamine stress ECHO, 2 out of 84 women (2.4 percent) developed supraventricular arrhythmias, and 3 (3.6 percent) had intolerance to the probe. The study by Elhendy, et al.,71 reported the symptoms and complications of dobutamine ECHO (with atropine administered as indicated) in 96 women and 210 men. Rates of reported events among the women were 5 percent for nausea, 0 percent for flushing, 2 percent for dizziness, 1 percent for anxiety, 4 percent for chills, 5 percent for headache, 1 percent for symptomatic hypotension, 38 percent for typical angina, 2 percent for SVT, 0 percent for atrial fibrillation, 1 percent for VT < 10 beats, and 0 percent for VT > 10 beats. Women experienced significantly lower rates (at the p < 0.05 level) of SVT and runs of VT < 10 beats compared with men who experienced these events at rates of 9 percent and 7 percent, respectively. A study by Lewis, et al.,72 that evaluated dobutamine ECHO in 92 women reported early termination of the stress test in 2 percent of patients because of VT or sustained SVT. Eight participants (9 percent) experienced dyspnea or extreme anxiety but did not require the study to be prematurely terminated, while 18 participants (20 percent) experienced mild symptoms of nausea and 8 participants (9 percent) had lightheadedness. A study of a cohort of 114 women by Lehmkuhl, et al.,27 reported incidence rates of 2.6 percent for arterial hypotension, 17 percent for PVCs, and 1.7 percent for nonsustained VT with a maximum of 7 beats associated with dobutamine.

Finally, a study by Ho, et al.,67 reported the following complications during dobutamine infusion for ECHO testing: frequent ventricular premature contractions (24 percent); chest pain (24 percent); palpitations (20 percent); frequent atrial premature contractions (18 percent); ST-segment change (16 percent); atrial fibrillation (2 percent); nonsustained ventricular tachycardia (2 percent); hypotension (2 percent); headache (2 percent); and yawning (2 percent).

SPECT

Data pertaining to safety in women who underwent exercise/stress SPECT were reported in four prospective cohort studies28,39,67,89 representing 294 women. Three studies were good-quality and one was fair-quality. The study by Lu, et al.,28 evaluated technetium-99 sestamibi SPECT, the study by Ho, et al., 67 compared dobutamine ECHO with SPECT, coronary angiography, and exercise ECG, and two studies, by Laurienzo, et al.,39 and Mohiuddin, et al.,89 evaluated thallium-201 myocardial perfusion imaging. Only one of the four studies89 reported sex-specific safety data associated with SPECT. In this study of adenosine thallium-201 myocardial perfusion imaging, the rates of adverse effects of adenosine were 41 percent for flushing, 25 percent for neck or jaw pain, 30 percent for dyspnea, 12 percent for lightheadedness, 10 percent for nausea, 8 percent for headache, 4 percent for second-degree atrioventricular (AV) block, 1 percent for third-degree AV block, 48 percent for hypotension, and 20 percent for miscellaneous. Compared with men in the same study, women experienced significantly higher rates of chest pain (21 percent in men) and ST segment depression (8 percent in men) but had no significant differences in rates of other side effects.

CMR

Two studies reported data pertaining to safety in women undergoing CMR. A study by Gebker, et al.,95 reported safety data in women undergoing dobutamine stress CMR. This good-quality, prospective cohort study included 204 consecutive women and 541 men with suspected and known CAD scheduled for clinically indicated coronary angiography. In general, severe side effects likely attributable to dobutamine occurred uncommonly but tended to occur less often in women than men, with incidences of severe dyspnea of 1 percent versus 0.7 percent; severe increase in blood pressure of 0.5 percent versus 0.6 percent; paroxysmal atrial fibrillation in 1.5 percent versus 2.4 percent; incidence of ventricular tachycardia in 0.5 percent versus 0.6 percent—all in women compared with men. None of these incidences of side effects in women were statistically significantly different. The study by Merkle, et al.,96 was a good-quality prospective cohort study that included 77 women who underwent both CMR and coronary angiography. This study reported no adverse events associated with adenosine infusion. Neither of the two studies reported adverse events potentially associated with CMR itself; the adverse events assessed in these two studies were limited to the pharmacological stress component of the testing procedure.

Coronary CTA

Four studies included sex-specific data on radiation dose associated with coronary CTA.103,105,113,114 All four were prospective cohort studies that compared coronary CTA with conventional coronary angiography. Of these, two were good quality and two were fair quality. Collectively, they included 486 women.

The estimated radiation exposure associated with a single 64-slice, contrast-enhanced coronary CTA was 14.4 mSv for women compared with 11.1 mSv for men in a study by Weustink, et al., (2007)113 evaluating the accuracy of a 32-slice dual-source CT. A study by Dewey, et al.,103 that used a 16-slice multislice CT scanner reported that the effective dose of a 16-slice coronary CTA examination was significantly higher by approximately 17 percent for women compared with men (13.7 ± 1.2 mSv versus 11.7 ± 0.9 mSv, p < 0.001). The largest contributor to dose among women were the lungs (average of 5.2 mSv, 37.8 percent of the effective dose), with breasts contributing 24.5 percent of the effective dose (3.35 mSv, on average). A study by Dharampal, et al.,105 included 280 women and 636 men. Single-source CT was used for the 385 patients enrolled between July 2004 and March 2006, and dual-source CT was used for the 531 patients enrolled between April 2006 and April 2009. Unlike the previous two studies, this study found the mean effective radiation dose for single-source CT to be slightly higher in men compared with women with levels of 16.4 mSv (SD = 1.1) and 16.0 mSv (SD = 1.3) respectively (p = 0.002). The mean effective radiation dose for dual-source CT was lower compared with single-source CT and was not significantly different between the sexes, with levels of 14.4 mSv (SD = 4.6) and 15.2 mSv (SD = 4.8) for women and men, respectively (p = 0.10).

The fourth study, by Weustink, et al.,,114 included sex-specific radiation data and involved 436 symptomatic patients (301 men, 135 women; mean age of 61.6 years) who underwent both conventional coronary angiography and coronary CTA. Standard and ECG pulsing were performed in 327 and 109 patients, respectively. The authors of this study applied the Biological Effects of Ionizing Radiation (BEIR) VII approach115 to estimate sex-dependent and age-dependent whole-body lifetime attributable risk of cancer incidence and mortality from a single coronary CTA examination. Risks were estimated for 50-, 60-, and 70- year-old men and women for each of three coronary CTA techniques: no ECG pulsing, standard ECG pulsing, and optimal ECG pulsing. The findings of this study suggest that lifetime attributable risk of both cancer incidence and fatal cancer incidence was approximately double in women, compared with men for 50-, 60-, and 70-year old patients. Attributable risk was highest for no ECG pulsing and lowest for optimal ECG pulsing across all three age groups and both sexes. Attributable risk was highest for 50-year old patients and lowest for 70-year patients across all three ECG pulsing approaches and both sexes. Lifetime attributable risk of cancer associated with a single coronary CTA examination with standard ECG pulsing was estimated at approximately 0.15 percent for 60-year-old women and 0.08 percent for 60-year-old men. Lifetime attributable risk of fatal cancer associated with a single coronary CTA examination with standard ECG pulsing was estimated at approximately 0.13 percent for 60-year-old women and 0.07 percent for 60-year-old men. These estimates are derived from projected data rather than from direct observation. Of note, reproductive organs are generally more sensitive to radiation than other tissues or organs; radiation exposure to reproductive organs may therefore result in higher projected cancer risk.

KQ 4 Summary

Thirteen studies reported data pertinent to safety concerns or risks associated with the use of NITs to diagnose CAD in women with suspected CAD. Nine of these studies were rated good quality and four fair quality. Data pertinent to safety concerns specifically for women for a given NIT were reported in six studies for ECHO, five for coronary CTA, two for CMR, and one each for exercise/stress ECG and SPECT.

Data specific to women on access site complications, contrast agent-induced nephropathy, nephrogenic systemic fibrosis, or anaphylaxis associated with NITs were not reported in any of the studies included in this review. There was insufficient information in the extant literature to draw conclusions about sex-specific concerns about arrhythmias associated with different NITs. Total-body radiation exposure from coronary CTA examinations appeared to be higher in women compared with men; lifetime attributable risk of both cancer incidence and fatal cancer incidence associated with a single coronary CTA examination was estimated in one study to be twice as high in women compared with men. However, recent advancements in technology have reduced the radiation exposure for coronary CTA, suggesting that these estimates may not be applicable to newer testing protocols. Radiation safety issues were not discussed for NITs other than coronary CTA.

Cover of Noninvasive Technologies for the Diagnosis of Coronary Artery Disease in Women
Noninvasive Technologies for the Diagnosis of Coronary Artery Disease in Women [Internet].
Comparative Effectiveness Reviews, No. 58.
Dolor RJ, Patel MR, Melloni C, et al.

AHRQ (US Agency for Healthcare Research and Quality)

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...