Gene expression profiling and expanded immunohistochemistry tests to guide the use of adjuvant chemotherapy in breast cancer management: a systematic review and cost-effectiveness analysis

BACKGROUND

Gene expression profiling (GEP) and expanded immunohistochemistry (IHC) tests aim to improve decision-making relating to adjuvant chemotherapy for women with early breast cancer.

OBJECTIVE

The aim of this report is to assess the clinical effectiveness and cost-effectiveness of nine GEP and expanded IHC tests compared with current prognostic tools in guiding the use of adjuvant chemotherapy in patients with early breast cancer in England and Wales. The nine tests are BluePrint, Breast Cancer Index (BCI), IHC4, MammaPrint, Mammostrat, NPI plus (NPI+), OncotypeDX, PAM50 and Randox Breast Cancer Array.

DATA SOURCES

Databases searched included MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, EMBASE and The Cochrane Library. Databases were searched from January 2009 to May 2011 for the OncotypeDX and MammaPrint tests and from January 2002 to May 2011 for the other tests.

REVIEW METHODS

A systematic review of the evidence on clinical effectiveness (analytical validity, clinical validity and clinical utility) and cost-effectiveness was conducted. An economic model was developed to evaluate the cost-effectiveness of adjuvant chemotherapy treatment guided by four of the nine test (OncotypeDX, IHC4, MammaPrint and Mammostrat) compared with current clinical practice in England and Wales, using clinicopathological parameters, in women with oestrogen receptor-positive (ER+), lymph node-negative (LN-), human epidermal growth factor receptor type 2-negative (HER2-) early breast cancer.

RESULTS

The literature searches for clinical effectiveness identified 5993 citations, of which 32 full-text papers or abstracts (30 studies) satisfied the criteria for the effectiveness review. A narrative synthesis was performed. Evidence for OncotypeDX supported the prognostic capability of the test. There was some evidence on the impact of the test on decision-making and to support the case that OncotypeDX predicts chemotherapy benefit; however, few studies were UK based and limitations in relation to study design were identified. Evidence for MammaPrint demonstrated that the test score was a strong independent prognostic factor, but the evidence is non-UK based and is based on small sample sizes. Evidence on the Mammostrat test showed that the test was an independent prognostic tool for women with ER+, tamoxifen-treated breast cancer. The three studies appeared to be of reasonable quality and provided data from a UK setting (one study). One large study reported on clinical validity of the IHC4 test, with IHC4 score a highly significant predictor of distant recurrence. This study included data from a UK setting and appeared to be of reasonable quality. Evidence for the remaining five tests (PAM50, NPI+, BCI, BluePrint and Randox) was limited. The economic analysis suggests that treatment guided using IHC4 has the greatest potential to be cost-effective at a £20,000 threshold, given the low cost of the test; however, further research is needed on the analytical validity and clinical utility of IHC4, and the exact cost of the test needs to be confirmed. Current limitations in the evidence base produce significant uncertainty in the results. OncotypeDX has a more robust evidence base, but further evidence on its impact on decision-making in the UK and the predictive ability of the test in an ER+, LN-, HER- population receiving current drug regimens is needed. For MammaPrint and Mammostrat there were significant gaps in the available evidence and the estimates of cost-effectiveness produced were not considered to be robust by the External Assessment Group.

LIMITATIONS

Methodological weaknesses in the clinical evidence base relate to heterogeneity of patient cohorts and issues arising from the retrospective nature of the evidence. Further evidence is required on the clinical utility of all of the tests and on UK-based populations. A key area of uncertainty relates to whether the tests provide prognostic or predictive ability.

CONCLUSIONS

The clinical evidence base for OncotypeDX is considered to be the most robust. The economic analysis suggested that treatment guided using IHC4 has the most potential to be cost-effective at a threshold of £20,000; however, the evidence base to support IHC4 needs significant further research.

STUDY REGISTRATION

PROSPERO 2011:CRD42011001361, available from www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42011001361.

Echocardiography in newly diagnosed atrial fibrillation patients: a systematic review and economic evaluation

OBJECTIVE

To investigate the clinical effectiveness and cost-effectiveness of transthoracic echocardiography (TTE) in all patients who are newly diagnosed with atrial fibrillation (AF).

DESIGN

Narrative synthesis reviews were conducted on the prognostic and diagnostic accuracy of TTE for, and prevalence of, pathologies in patients with AF. Databases were searched from inception. MEDLINE searches were conducted from March to August 2010, and reference lists of articles checked. There were 44 diagnostic accuracy studies, five prognostic studies, and 16 prevalence studies accepted into the review. Given the complexity of the many pathologies identified by TTE, the variety of potential changes to clinical management, and paucity of data, the model focused on changes to oral anticoagulation (OAC). The mathematical model assessed the cost-effectiveness of TTE for patients with AF who were not routinely given OAC, assuming, if left atrial abnormality was detected, that the higher risk of stroke warranted OAC; this meant that patients with a CHADS2 (cardiac failure, hypertension, age, diabetes, stroke doubled) score of 0 [dabigatran etexilate (Pradaxa, Boehringer Ingelheim)/rivaroxaban (Xarelto, Bayer Schering)] or 0/1 (warfarin) were included. A simplified approach evaluated the additional quality-adjusted life-years (QALYs) required in order for TTE to be perceived as cost-effective at a threshold of £20,000 per QALY.

SETTING

Transthoracic echocardiography is usually performed in cardiology clinics but may be used in primary or non-specialist secondary care.

PARTICIPANTS

Patients with newly diagnosed AF.

INTERVENTION

Transthoracic echocardiography.

MAIN OUTCOME MEASURES

Prognosis, diagnostic sensitivity or specificity of TTE, prevalence of pathologies in patients with AF, cost-effectiveness and QALYs.

RESULTS

Prognostic studies indicated that TTE-diagnosed left ventricular dysfunction, increased left atrial diameter and valvular abnormality were significantly associated with an increased risk of stroke, mortality or thromboembolism. There was a high prevalence (around 25-30%) of ischaemic heart disease, valvular heart disease and heart failure in patients with AF. Diagnostic accuracy of TTE was high, with most pathologies having specificity of ≥ 0.8 and sensitivity of ≥ 0.6. The mathematical model predicted that when the CHADS2 tool is used the addition of TTE in identifying patients with left atrial abnormality appears to be cost-effective for informing some OAC decisions. In the simplified approach a threshold of 0.0033 was required for a TTE to be cost-effective.

CONCLUSIONS

When CHADS2 was used, the addition of TTE in identifying patients with left atrial abnormality was cost-effective for informing some OAC decisions. A simple analysis indicates that the number of QALYs required for TTE to be cost-effective is small, and that if benefits beyond those associated with a reduction in stroke are believed probable then TTE is likely to be cost-effective in all scenarios. Our findings suggest that further research would be useful, following up newly diagnosed patients with AF who have undergone TTE, to study treatments given as a result of TTE diagnoses and subsequent cardiovascular events. This could identify additional benefits of routine testing, beyond stroke prevention. Studies assessing the proportion of people with a CHADS2 score of 0 or 1 that have left atrial abnormality would provide better estimates of the cost-effectiveness of TTE, and allow more accurate estimates of the sensitivity and specificity of TTE for identifying left atrial abnormality in AF to be obtained.

STUDY REGISTRATION

PROSPERO CRD42011001354.

FUNDING

The National Institute for Health Research Health Technology Assessment programme.

Increased cortisol levels in hair of recent Ecstasy/MDMA users

Previous research has revealed an acute 8-fold increase in salivary cortisol following self-administrated Ecstasy/MDMA in dance clubbers. It is currently not known to what extent repeated usage impacts upon activity of the hypothalamic-pituitary-adrenal axis over a more prolonged period of time. This study investigated the integrated cortisol levels in 3-month hair samples from recent Ecstasy/MDMA users and non-user controls. One hundred and one unpaid participants (53 males, 48 females; mean age 21.75 years) completed the University of East London recreational drug use questionnaire, modified to cover the past 3-months of usage. They comprised 32 light recent Ecstasy/MDMA users (1-4 times in last 3 months), 23 recent heavy MDMA users (+5 times in last 3 months), and 54 non-user controls. Volunteers provided 3 cm hair samples for cortisol analysis. Hair cortisol levels were observed to be significantly higher in recent heavy MDMA users (mean = 55.0 ± 80.1 pg/mg), compared to recent light MDMA users (19.4 ± 16.0 pg/mg; p=0.015), and to non-users (13.8 ± 6.1 pg/mg; p<0.001). Hence the regular use of Ecstasy/MDMA was associated with almost 4-fold raised hair cortisol levels, in comparison with non-user controls. The present results are consistent with the bio-energetic stress model for Ecstasy/MDMA, which predicts that repeated stimulant drug use may increase cortisol production acutely, and result in greater deposits of the hormone in hair. These data may also help explain the neurocognitive, psychiatric, and other psychobiological problems of some abstinent users. Future study design and directions for research concerning the psychoneuroendocrinological impact of MDMA are also discussed.

Use of lean principals to improve flow of patients with fractured neck of femur--the HOPE study

We describe the implementation of a care pathway for patients with fractured neck of femur (NOF) using Lean and Six Sigma principles. After introduction of the Lean pathway, 32 patients out a total of 86 (37%) with fractured NOF were admitted to the Trauma Ward within 4 hours of presentation to the hospital; prior to implementation this was 16 patients out of a total of 59 (27%). Post-Lean an earlier mean theatre start time of 8.40am was achieved, resulting in a 38 minute increase in daily theatre time. An additional 52 patients (12%) received surgery within 24 hours of admission, resulting in 1 night length of stay reduction. Lean methodology proved an effective method to guide change resulting in an improved journey for the patient and significant workflow gains.

Determinants of survival in patients with chronic myeloid leukaemia treated in the new era of oral therapy: findings from a UK population-based patient cohort

OBJECTIVES

To examine contemporary survival patterns in the general population of patients diagnosed with chronic myeloid leukaemia (CML), and to identify patient groups with less than optimal outcomes.

DESIGN

Prospective population-based cohort.

SETTING

The UK's Haematological Malignancy Research Network (catchment population 3.6 million, with >2000 new haematological malignancies diagnosed annually).

PARTICIPANTS

All patients newly diagnosed with CML, from September 2004 to August 2011 and followed up to 31 March 2013.

MAIN OUTCOME MEASURE

Incidence and survival.

RESULTS

With a median diagnostic age of 59 years, the CML age standardised (European) incidence was 0.9/100 000 (95% CIs 0.8 to 0.9), 5-year overall survival was 78.9% (72.3 to 84.0) and 5-year relative survival 88.6% (81.0 to 93.3). The efficacy of treatment across all ages was clearly demonstrated; the relative survival curves for those under 60 and over 60 years being closely aligned. Survival findings were similar for men and women, but varied with deprivation; the age and sex adjusted HR being 3.43 (1.89 to 6.22) for deprivation categories 4-5 (less affluent) versus 1-3 (more affluent). None of these differences were attributable to the biological features of the disease.

CONCLUSIONS

When therapy is freely provided, population-based survival for CML is similar to that reported in clinical trials, and age loses its prognostic significance. However, although most of the patients with CML now experience close to normal lifespans, those living in more deprived areas tend to have poorer outcomes, despite receiving the same clinical care. A significant improvement in overall population outcomes could be achieved if these socioeconomic differences, which may reflect the treatment compliance, could be eliminated.

Identification of PLX4032-resistance mechanisms and implications for novel RAF inhibitors

BRAF inhibitors improve melanoma patient survival, but resistance invariably develops. Here we report the discovery of a novel BRAF mutation that confers resistance to PLX4032 employing whole-exome sequencing of drug-resistant BRAF^V600K melanoma cells. We further describe a new screening approach, a genome-wide piggyBac mutagenesis screen that revealed clinically relevant aberrations (N-terminal BRAF truncations and CRAF overexpression). The novel BRAF mutation, a Leu505 to His substitution (BRAF^L505H), is the first resistance-conferring second-site mutation identified in BRAF mutant cells. The mutation replaces a small nonpolar amino acid at the BRAF-PLX4032 interface with a larger polar residue. Moreover, we show that BRAF^L505H, found in human prostate cancer, is itself a MAPK-activating, PLX4032-resistant oncogenic mutation. Lastly, we demonstrate that the PLX4032-resistant melanoma cells are sensitive to novel, next-generation BRAF inhibitors, especially the ‘paradox-blocker’ PLX8394, supporting its use in clinical trials for treatment of melanoma patients with BRAF-mutations.

Computational analysis in cancer exome sequencing

Exome sequencing in cancer is a powerful tool for identifying mutational events across the coding region of human genes. Here, we describe computational methods that use exome sequencing reads from cancer samples to identify somatic single nucleotide variants (SNVs), copy number alterations, and short insertions and deletions (InDels). We further describe analytical methods to generate lists of driver genes with more mutational events than expected by chance.

New data, strategies, and insights for Listeria monocytogenes dose-response models: summary of an interagency workshop, 2011

Listeria monocytogenes is a leading cause of hospitalization, fetal loss, and death due to foodborne illnesses in the United States. A quantitative assessment of the relative risk of listeriosis associated with the consumption of 23 selected categories of ready-to-eat foods, published by the U.S. Department of Health and Human Services and the U.S. Department of Agriculture in 2003, has been instrumental in identifying the food products and practices that pose the greatest listeriosis risk and has guided the evaluation of potential intervention strategies. Dose-response models, which quantify the relationship between an exposure dose and the probability of adverse health outcomes, were essential components of the risk assessment. However, because of data gaps and limitations in the available data and modeling approaches, considerable uncertainty existed. Since publication of the risk assessment, new data have become available for modeling L. monocytogenes dose-response. At the same time, recent advances in the understanding of L. monocytogenes pathophysiology and strain diversity have warranted a critical reevaluation of the published dose-response models. To discuss strategies for modeling L. monocytogenes dose-response, the Interagency Risk Assessment Consortium (IRAC) and the Joint Institute for Food Safety and Applied Nutrition (JIFSAN) held a scientific workshop in 2011 (details available at http://foodrisk.org/irac/events/). The main findings of the workshop and the most current and relevant data identified during the workshop are summarized and presented in the context of L. monocytogenes dose-response. This article also discusses new insights on dose-response modeling for L. monocytogenes and research opportunities to meet future needs.

Thermal inactivation of a single strain each of serotype O26:H11, O45:H2, O103:H2, O104:H4, O111:H⁻, O121:H19, O145:NM, and O157:H7 cells of Shiga toxin-producing Escherichia coli in wafers of ground beef

For each of two trials, freshly ground beef of variable fat content (higher: 70:30 %lean:%fat; lower: 93:7 %lean:%fat) was separately inoculated with ca. 7.0 log CFU/g of a single strain of Escherichia coli serotypes O26:H11, O45:H2, O103:H2, O104:H4, O111:H⁻, O121:H19, O145:NM, and O157:H7. Next, ca. 3-g samples of inoculated beef were transferred into sterile filter bags and then flattened (ca. 1.0 mm thick) and vacuum sealed. For each temperature and sampling time, three bags of the inoculated wafers of beef were submerged in a thermostatically controlled water bath and heated to an internal temperature of 54.4°C (130°F) for up to 90 min, to 60°C (140°F) for up to 4 min, or to 65.6°C (150°F) for up to 0.26 min. In lower fat wafers, D-values ranged from 13.5 to 23.6 min, 0.6 to 1.2 min, and 0.05 to 0.08 min at 54.4, 60.0, and 65.6°C, respectively. Heating higher fat wafers to 54.4, 60.0, and 65.6°C generated D-values of 18.7 to 32.6, 0.7 to 1.1, and 0.05 to 0.2 min, respectively. In addition, we observed reductions of ca. 0.7 to 6.7 log CFU/g at 54.4°C after 90 min, ca. 1.1 to 6.1 log CFU/g at 60.0°C after 4 min, and 0.8 to 5.8 log CFU/g at 65.6°C after 0.26 min. Thus, cooking times and temperatures effective for inactivating a serotype O157:H7 strain of E. coli in ground beef were equally effective against the seven non-O157:H7 Shiga toxin-producing strains investigated herein.

Systematic review, meta-analysis and economic modelling of diagnostic strategies for suspected acute coronary syndrome

BACKGROUND

Current practice for suspected acute coronary syndrome (ACS) involves troponin testing 10-12 hours after symptom onset to diagnose myocardial infarction (MI). Patients with a negative troponin can be investigated further with computed tomographic coronary angiography (CTCA) or exercise electrocardiography (ECG).

OBJECTIVES

We aimed to estimate the diagnostic accuracy of early biomarkers for MI, the prognostic accuracy of biomarkers for major adverse cardiac adverse events (MACEs) in troponin-negative patients, the diagnostic accuracy of CTCA and exercise ECG for coronary artery disease (CAD) and the prognostic accuracy of CTCA and exercise ECG for MACEs in patients with suspected ACS. We then aimed to estimate the cost-effectiveness of using alternative biomarker strategies to diagnose MI, and using biomarkers, CTCA and exercise ECG to risk-stratify troponin-negative patients.

DATA SOURCES

We searched MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations; Cumulative Index of Nursing and Allied Health Literature (CINAHL), EMBASE, Web of Science, Cochrane Central Database of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR), NHS Database of Abstracts of Reviews of Effects (DARE) and the Health Technology Assessment database from 1985 (CTCA review) or 1995 (biomarkers review) to November 2010, reviewed citation lists and contacted experts to identify relevant studies.

REVIEW METHODS

Diagnostic studies were assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool and prognostic studies using a framework adapted for the project. Meta-analysis was conducted using bayesian Markov chain Monte Carlo simulation. We developed a decision-analysis model to evaluate the cost-effectiveness of alternative biomarker strategies to diagnose MI, and the cost-effectiveness of biomarkers, CTCA or exercise ECG to risk-stratify patients with a negative troponin. Strategies were applied to a theoretical cohort of patients with suspected ACS. Cost-effectiveness was estimated as the incremental cost per quality-adjusted life-year (QALY) of each strategy compared with the next most effective, taking a health-service perspective and a lifetime horizon.

RESULTS

Sensitivity and specificity (95% predictive interval) were 77% (29-96%) and 93% (46-100%) for troponin I, 80% (33-97%) and 91% (53-99%) for troponin T (99th percentile threshold), 81% (50-95%) and 80% (26-98%) for quantitative heart-type fatty acid-binding protein (H-FABP), 68% (11-97%) and 92% (20-100%) for qualitative H-FABP, 77% (19-98%) and 39% (2-95%) for ischaemia-modified albumin and 62% (35-83%) and 83% (35-98%) for myoglobin. CTCA had 94% (61-99%) sensitivity and 87% (16-100%) specificity for CAD. Positive CTCA and positive-exercise ECG had relative risks of 5.8 (0.6-24.5) and 8.0 (2.3-22.7) for MACEs. In most scenarios in the economic analysis presentation, high-sensitivity troponin measurement was the most effective strategy with an incremental cost-effectiveness ratio (ICER) of less than the £20,000-30,000/QALY threshold (ICER £7487-17,191/QALY). CTCA appeared to be the most cost-effective strategy for patients with a negative troponin, with an ICER of £11,041/QALY. However, when a lower MACE rate was assumed, CTCA had a high ICER (£262,061/QALY) and the no-testing strategy was optimal.

LIMITATIONS

There was substantial variation between the primary studies and heterogeneity in their results. Findings of the economic model were dependent on assumptions regarding the value of detecting and treating positive cases.

CONCLUSIONS

Although presentation troponin has suboptimal sensitivity, measurement of a 10-hour troponin level is unlikely to be cost-effective in most scenarios compared with a high-sensitivity presentation troponin. CTCA may be a cost-effective strategy for troponin-negative patients, but further research is required to estimate the effect of CTCA on event rates and health-care costs.

FUNDING

The National Institute for Health Research Health Technology Assessment programme.

Adjusting for background mutation frequency biases improves the identification of cancer driver genes

A common goal of tumor sequencing projects is finding genes whose mutations are selected for during tumor development. This is accomplished by choosing genes that have more non-synonymous mutations than expected from an estimated background mutation frequency. While this background frequency is unknown, it can be estimated using both the observed synonymous mutation frequency and the non-synonymous to synonymous mutation ratio. The synonymous mutation frequency can be determined across all genes or in a gene-specific manner. This choice introduces an interesting trade-off. A gene-specific frequency adjusts for an underlying mutation bias, but is difficult to estimate given missing synonymous mutation counts. Using a genome-wide synonymous frequency is more robust, but is less suited for adjusting biases. Studying four evaluation criteria for identifying genes with high non-synonymous mutation burden (reflecting preferential selection of expressed genes, genes with mutations in conserved bases, genes with many protein interactions, and genes that show loss of heterozygosity), we find that the gene-specific synonymous frequency is superior in the gene expression and protein interaction tests. In conclusion, the use of the gene-specific synonymous mutation frequency is well suited for assessing a gene's non-synonymous mutation burden.

Bivalirudin for the treatment of ST-segment elevation myocardial infarction: a NICE single technology appraisal

The National Institute for Health and Clinical Excellence (NICE) invited the manufacturer (The Medicines Company) of bivalirudin to submit evidence for its clinical and cost effectiveness within its licensed indication for the treatment of adults with ST-segment elevation myocardial infarction (STEMI) intended for primary percutaneous coronary intervention (PPCI), as part of NICE's single technology appraisal (STA) process. The School of Health and Related Research (ScHARR) at the University of Sheffield was commissioned to act as the Evidence Review Group (ERG), which produced a review of the evidence within the manufacturer's submission to NICE. This article describes the manufacturer's submission, the ERG review and NICE's subsequent decisions. The main evidence was derived from one randomized controlled trial (RCT) of STEMI patients intended for PPCI, comparing bivalirudin with unfractionated heparin plus glycoprotein IIb/IIIa inhibitors (GPIs). Bivalirudin was associated with a significant reduction in cardiac mortality at 30 days (p = 0.03) and at 1-year follow-up (p = 0.005), and a significant reduction in major bleeding at 30 days (p < 0.001) and 1 year (p < 0.0001), compared with heparin plus GPI. Stent thrombosis up to 24 hours following PPCI was significantly (p < 0.001) more common with bivalirudin. However, there was no significant treatment effect for stent thrombosis from 1 to 30 days (p = 0.28), or at 1-year follow-up (p = 0.53). There were no significant treatment group differences at 30 days and at 1 year in stroke (p = 0.68 and p = 0.99, respectively), in myocardial infarction [MI] (p = 0.90 and p = 0.22, respectively), or in the need for the revascularization of the target vessel for ischaemia (p = 0.18 and p = 0.12, respectively). There were two decision-analytic models: the base-case scenario used 1-year follow-up data from the RCT; and a sensitivity analysis used 3-year follow-up data. Resource use was primarily drawn from this RCT. Health-related quality-of-life (HR-QOL) estimates were drawn from a UK cohort study. Both models evaluated the incremental costs and outcomes of bivalirudin compared with heparin plus GPI for patients with STEMI intended for PPCI. The analysis adopted a UK NHS perspective over a lifetime horizon. Unit costs were based on year 2009-2010 prices. The model adopted a decision-tree structure to reflect initial events for the initial period (stroke, repeat MI, minor/major bleeding events, repeat revascularization and death) and a two-state Markov component to simulate longer-term survival. The economic analysis suggested that bivalirudin is expected to dominate the heparin plus GPI strategy. This finding was consistent across the probabilistic sensitivity analysis and the vast majority of deterministic sensitivity analyses undertaken. Three exceptions to this finding were observed for the following sensitivity analyses: (1) the exclusive use of eptifibatide as the GPI (incremental cost-effectiveness ratio [ICER] = £1,764); (2) the combination of 100 % eptifibatide use, 100 % radial arterial access and no differential length between strategies for initial hospital stay (ICER = £4,106); and (3) a longer length of ward stay (increase of 0.33 days) for the initial hospitalization (ICER = £415). The Appraisal Committee (AC) gave a positive recommendation for bivalirudin for the treatment of adults with STEMI undergoing PPCI.

Correlated evolution of positions within mammalian cis elements

Transcriptional regulation critically depends on proper interactions between transcription factors (TF) and their cognate DNA binding sites. The widely used model of TF-DNA binding – the Positional Weight Matrix (PWM) – presumes independence between positions within the binding site. However, there is evidence to show that the independence assumption may not always hold, and the extent of interposition dependence is not completely known. We hypothesize that the interposition dependence should partly be manifested as correlated evolution at the positions. We report a Maximum-Likelihood (ML) approach to infer correlated evolution at any two positions within a PWM, based on a multiple alignment of 5 mammalian genomes. Application to a genome-wide set of putative cis elements in human promoters reveals a prevalence of correlated evolution within cis elements. We found that the interdependence between two positions decreases with increasing distance between the positions. The interdependent positions tend to be evolutionarily more constrained and moreover, the dependence patterns are relatively similar across structurally related transcription factors. Although some of the detected mutational dependencies may be due to context-dependent genomic hyper-mutation, notably CG to TG, the majority is likely due to context-dependent preferences for specific nucleotide combinations within the cis elements. Patterns of evolution at individual nucleotide positions within mammalian TF binding sites are often significantly correlated, suggesting interposition dependence. The proposed methodology is also applicable to other classes of non-coding functional elements. A detailed investigation of mutational dependencies within specific motifs could reveal preferred nucleotide combinations that may help refine the DNA binding models.