A 64-year-old Hispanic man was referred to the urology clinic by his primary care physician for evaluation of right upper pole kidney stones associated with recurrent urinary tract infections (UTIs; 

). The patient’s medical history was significant for well-controlled hypertension, type 2 diabetes mellitus, hyperlipidemia, and epilepsy, as well as a remote history of nephrolithiasis. In the past, he underwent several surgeries, including laparoscopic cholecystectomy, bilateral inguinal hernia repairs with mesh, and a percutaneous nephrolithotomy for a complete right staghorn renal calculus over 10 years ago, the latter of which was complicated by a pneumothorax requiring chest tube placement. The patient denied any workup or prophylaxis for stone formation. He denied any history of tobacco, alcohol, or substance abuse.

On evaluation, the patient had normal baseline renal function, with a serum creatinine level of 1.0 mg/dL. Follow-up urine cultures revealed recurrent, pan-sensitive 

 UTIs. The patient denied any urinary complaints, and a physical examination was largely unremarkable. A follow-up CT urogram of the abdomen and pelvis with intravenous contrast was obtained, revealing a 2–3 cm cluster of renal stones within a calyceal diverticulum at the upper pole of the right kidney. Cross-sectional imaging showed no other stones and/or masses, with a normal contralateral left kidney.

After discussing potential treatment options with the patient, including the risks and benefits, a right robotic-assisted laparoscopic upper pole partial nephrectomy was selected. During the procedure, the calyceal diverticulum was extracted, and residual kidney stones were removed en bloc with renorrhaphy; the collecting system was also closed. A 1-cm margin of normal renal parenchyma was also taken deep to the calyceal diverticulum during the resection. The surgery was successful, with no postoperative complications, and the patient was discharged 2 days later.

The kidney stones were calcium oxalate monohydrate in nature. However, follow-up histopathological evaluation of the resected tissue revealed poorly differentiated carcinoma arising from the right upper pole calyceal diverticulum with invasion through the urothelial wall into the underlying renal parenchyma, with a 2–3 mm focal positive surgical margin at the base of the partial nephrectomy resection site. Microscopic examination of the tissue by our pathologist revealed keratin pearl formation, intercellular bridges, and keratotic debris, and immunohistochemical staining demonstrated strong positivity for p63 (

). Based on these findings and the patient’s clinical history, a diagnosis of primary squamous cell carcinoma (SCC) of the right kidney was established.

Given the incidental primary squamous cell carcinoma within the upper pole calyceal diverticulum of the right kidney with a positive focal surgical margin at the base of the partial nephrectomy resection site, what is the most appropriate next step in management?

. Repeat right partial nephrectomy with negative histological margins

CORRECT ANSWER: C. Salvage right radical nephrectomy

Primary SCC of the renal pelvis and kidney is a very rare neoplasm, accounting for less than 1% of all malignant renal neoplasms.[1] SCC of the kidney occurs in men and women equally, particularly those between the ages of 50 and 70 years.[2] Most often, the etiology of SCC is long-standing nephrolithiasis, especially the formation of staghorn renal calculi.[3] The mechanism of carcinogenesis is assumed to be secondary to chronic irritation, inflammation, and/or infection, all of which can lead to squamous metaplasia. Other risk factors, such as hydronephrosis, vitamin A deficiency, chemical use, and hormonal imbalances, have also been implicated in the development of SCC of the kidney.[4,5] It is also worth noting that the incidence of renal SCC is higher in individuals with horseshoe kidneys.[6]

SCC of the kidney tends to be large, necrotic, sessile, ulcerated, and infiltrative of the renal parenchyma, as well as the perinephric soft tissue.[3] Diagnosis of this malignancy is restricted to tumors that show squamous differentiation and histological hallmarks of SCC, such as keratin pearl formation, intercellular bridges, and keratotic debris.[7] However, because most tumors are moderately or poorly differentiated, these characteristics may not be obvious.[3] Lee et al devised the classification of renal SCC as one of two groups: central or peripheral.[8] Central renal SCC is more intraluminal and is usually associated with metastases to the lymph nodes, while peripheral renal SCC is prominent in the renal parenchyma and may invade the perirenal fat before metastasizing to the lymph nodes or a distant site. Since primary SCC of the kidney is rare, it should be distinguished from metastatic SCC based on the patient’s clinical history, imaging, and histopathology.[9]

Establishing a diagnosis via imaging is difficult because SCC of the kidney exhibits no specific radiological features; in addition, given the rarity of this disease, other diagnoses, such as transitional cell/urothelial carcinoma of the renal collecting system or a complex kidney cyst, are much more likely.[10] The use of 18-fluorine fluorodeoxyglucose (18F-FDG) PET/CT imaging has been reported to be most helpful in confirming the diagnosis of these renal tumors, with a sensitivity of 62% and a specificity of 88%, although imaging cannot differentiate disease stage or grade.[11] For extra-renal SCC lesions, the sensitivity and specificity of 18F-FDG PET/CT increase to 84% and 91%, respectively, making this imaging much more useful for non-renal primary sites.[11] Overall, however, imaging is not helpful in determining the histology of primary SCC of the kidney, and diagnosis is typically made after nephrectomy.

Primary renal SCC tends to manifest at a more advanced disease stage compared with transitional cell/urothelial carcinomas and/or renal cell carcinomas, most likely due to the underlying aggressive nature of SCC, the reduced prevalence of symptoms such as hematuria or flank pain, and the lack of a measurable tumor marker in the urine or serum. Therefore, the prognosis of primary SCC of the kidney is, on average, incredibly poor, with the majority of patients presenting with locally advanced or metastatic disease and limited long-term survival.[12] Although the incidence of warning signs is low, patients may experience a constellation of symptoms, including flank or abdominal pain, hematuria, fever, weight loss, or an abdominal mass, and patients are occasionally diagnosed with localized disease incidentally.[13] Renal SCC may also be associated with paraneoplastic syndromes, such as hypercalcemia, due to the ectopic secretion of parathyroid hormone–like substances.[14]

Very little long-term data on renal SCC exist in the literature, the majority of which is contradictory. In addition, no gold standard of management exists, with many treatments being anecdotal or based on other SCC tumor sites.[12] For clinically nonmetastatic, localized primary SCC of the kidney, the best known treatment is surgical resection with radical nephrectomy (

). This is because partial nephrectomy, especially in the setting of positive histological margins, is unlikely to be curative and carries a high risk of recurrence within the kidney due to the diffuse, infiltrative nature of SCC into the renal parenchyma and perinephric soft tissue (

). Active surveillance with serial imaging would also not be warranted, even for smaller foci of SCC within the kidney, due to the aggressive nature of this disease and a lack of reliable imaging to accurately predict the extent of disease (

Although surgical resection is rarely curative, adjuvant chemotherapy or radiation is usually ineffective for renal SCC (

).[3] The most common chemotherapy agents used to treat this condition are platinum-based, with radiation most often reserved for palliative purposes in patients with metastatic disease and severe local symptoms. Currently, the combination of cisplatin, methotrexate, and bleomycin is employed for these individuals, but no benefit in survival has been observed.[4] One study reported a median survival of 7 months postoperatively, with a 5-year survival rate of 7.7%, but another demonstrated a median survival as low as 3.5 months.[3,7] In a large meta-analysis by Berz et al analyzing survival among patients with squamous cell malignancies of the upper urinary tract, the authors reported that the median overall survival time was 10 months compared with 63 months for transitional cell/urothelial histology.[12] Although the prognosis of renal SCC can be equivalent to transitional cell/urothelial carcinoma with appropriate treatment throughout each stage of disease, the average survival for primary SCC of the kidney is comparatively much worse due to more advanced disease at the time of presentation.[3]

The patient underwent salvage right robotic-assisted laparoscopic radical nephrectomy 2 months after his initial right partial nephrectomy with calyceal diverticulectomy surgery. His postoperative course was uncomplicated, and he was discharged home on postoperative day 2. Final histopathological analysis of the remaining right kidney revealed residual pT3a primary SCC with microscopic perinephric fat invasion with negative surgical margins circumferentially and no evidence of vascular invasion. No adjuvant treatment was given.

Currently, the patient is doing well 6 months after salvage right radical nephrectomy. He has remained recurrence-free based on cross-sectional surveillance imaging with contrasted CT scan of the chest, abdomen, and pelvis at 3 and 6 months postoperatively. Periodic imaging with follow-up will be continued every 3 months for the first year.

• Primary renal squamous cell carcinoma (SCC) is an extremely rare malignancy of the kidney, most often associated with long-standing nephrolithiasis, especially the formation of staghorn kidney stones, due to chronic irritation, inflammation, and/or infection, all which can lead to squamous metaplasia.

• Manifestation of primary renal SCC tends to occur at an advanced disease stage, with locally infiltrative and/or metastatic tumors that are difficult to accurately identify on imaging.

• The prognosis of renal SCC is poor, with an average median survival of less than 1 year. No proven survival benefit has been seen with the use of adjuvant chemotherapy or radiation treatment after surgical resection.

The authors have no significant financial interest in or other relationship with the manufacturer of any product or provider of any service mentioned in this article.

Li MK, Cheung WL. Squamous cell carcinoma of the renal pelvis. J Urol. 1987;138:269-71.

HolmÃ¤ng S, Lele SM, Johansson SL. Squamous cell carcinoma of the renal pelvis and ureter: incidence, symptoms, treatment and outcome. J Urol. 2007;178:51-6.

Bhaijee F. Squamous cell carcinoma of the renal pelvis. Ann Diagn Pathol. 2012;16:124-7.

Attalla K, Haines K, Labow D, et al. Squamous cell carcinoma of the renal pelvis: atypical presentation of a rare malignancy. Urol Case Rep. 2017;13:137-9.

Tyagi N, Sharma S, Tyagi SP, et al. A histomorphologic and ultrastructural study of the malignant tumours of the renal pelvis. J Postgrad Med. 1993;39:197-201.

Mizusawa H, Komiyama I, Ueno Y, et al. Squamous cell carcinoma in the renal pelvis of a horseshoe kidney. Int J Urol. 2004;11:782-4.

Bandyopadhyay R, Biswas S, Nag D, et al. Squamous cell carcinoma of the renal pelvis presenting as hydronephrosis. J Cancer Res Ther. 2010;6:537-9.

Lee TY, Ko SF, Wan YL, et al. Renal squamous cell carcinoma: CT findings and clinical significance. Abdom Imaging. 1998;23:203-8.

Ghosh P, Saha K. Primary intraparenchymal squamous cell carcinoma of the kidney: a rare and unique entity. Case Rep Pathol. 2014;2014:256813.

Kumar S, Tomar V, Yadav SS, et al. Primary squamous cell carcinoma of kidney associated with large calculus in non-functioning kidney: a case report. Urol Case Rep. 2016;8:4-6.

Wang Z, Yan B, Wei YB, et al. Primary kidney parenchyma squamous cell carcinoma mimicking xanthogranulomatous pyelonephritis: a case report. Oncol Lett. 2016;11:2179-81.

Berz D, Rizack T, Weitzen S, et al. Survival of patients with squamous cell malignancies of the upper urinary tract. Clin Med Insights Oncol. 2012;6:11-8.

Jiang P, Wang C, Chen S, et al. Primary renal squamous cell carcinoma mimicking the renal cyst: a case report and review of the recent literature. BMC Urol. 2015;15:69.

Diaz GonzÃ¡lez R, Barrientos A, Larrodera L, et al. Squamous cell carcinoma of the renal pelvis associated with hypercalcemia and the presence of parathyroid hormone-like substances in the tumor. J Urol. 1985;133:1029-30.

A 64-year-old man is diagnosed with primary squamous cell carcinoma within the upper pole calyceal diverticulum of the kidney. What are the best steps of management?

Incidental Primary Squamous Cell Carcinoma of the Kidney Within a Calyceal Diverticulum Associated With Nephrolithiasis

Cancer patients with depression experience more physical symptoms, pain, and fatigue; have a poorer quality of life; and are more likely to encounter negative thoughts compared with cancer patients who are not depressed. Accurate assessment and treatment of depression can have a positive impact on improving a patient’s quality of life. Pharmacotherapy for depression in patients with advanced cancer should be guided by a focus on symptom reduction, irrespective of whether the patient meets the diagnostic criteria for major depression. For effective treatment of a depressive illness, antidepressant medication and cognitive behavioral therapy need to be initiated sooner rather than later to reduce symptom burden and improve quality of life.

 to read an expert perspective from Mirat Shah, MD, Kerry O’Connor, MD, and Thomas J. Smith, MD, FACP.

Depression is the most common psychological condition in patients with cancer. To improve the rate of recovery among such individuals, it is important that clinicians learn to recognize and treat the symptoms associated with this condition. Several practice guidelines recommend screening every patient with cancer for depression. In terms of treatment, both psychotherapy and pharmacotherapy have been shown to be effective in treating depression in this population. When pharmacotherapy is warranted, selection of treatment agent(s) should be made based on the patient’s medical history, cancer type, and disease severity. Increased awareness about the risk of depression along with knowledge of how to identify symptoms and initiate treatment are essential to improve the overall quality of life for patients with cancer.

The reported prevalence of depression in patients with cancer varies from 3% to 38%.[1] Depressive symptoms in this population range from feelings of sadness to a clinical diagnosis of major depressive disorder (MDD). The rate of depression is highest in those closest to the end of life and in patients with certain types of cancer, including pancreatic cancer, lung cancer, gastric cancer, and oropharyngeal cancer.[2] MDD is present in 16% of individuals with cancer, while minor depression and dysthymia occur in about 22% of patients. This wide variability can be explained by the lack of consensus regarding the appropriate diagnostic criteria to utilize in the setting of advanced illness, as well as by the differences in various assessment methods.[3]

Psychological distress is a major cause of concern among patients with advanced cancer.[4] Depressive syndromes are highly correlated with a reduced quality of life, increased difficulty managing the course of disease, decreased adherence to treatment, and earlier admission to inpatient or hospice care.[5] A meta-analysis revealed that minor or major depression increases the rate of mortality by up to 39%; in addition, patients displaying even a few depressive symptoms may be at a 25% increased risk of mortality.[6] Depression weakens the patient’s ability to experience pleasure, sense, and connection; it also intensifies pain and other symptoms and causes suffering and worry in friends and family members.

Depression is underdiagnosed in patients with cancer due to the overlap in symptoms caused by these conditions, such as fatigue, insomnia, and anorexia.[1] The National Comprehensive Cancer Network (NCCN) provides the Guidelines for Distress Management, which are intended for any psychosocial problems found in patients with cancer. These guidelines recommend that every cancer patient be screened for depression during his or her initial visit to the oncologist. A screening tool called the “Distress Thermometer,” also available from the NCCN, can aid in this process. The tool asks patients to rank their level of stress during the past week on a scale of 0–10. A score of 4 or higher indicates that the patient should be referred to mental health services.[7]

Clinicians often rely more on the psychological or cognitive symptoms of depression (ie, worthlessness, hopelessness, excessive guilt, and suicidal ideation) than the physical/somatic signs (ie, weight loss, sleep disturbance) when making a diagnosis of MDD in patients with advanced disease. To effectively treat patients’ depression, it is also important that clinicians be able to recognize the risk factors that can lead to depression (see 

In addition to recognizing risk factors for depression, it is important that clinicians be able to monitor patients for an appropriate reaction to advanced disease and death. If any depressive syndrome is suspected, physicians should utilize appropriate screening tools to aid in the diagnosis. The Patient Health Questionnaire-2 (PHQ-2), a two-item depression screener, can be utilized among patients who screened positive for depression in order to diagnose the condition.[11] Self-report tools, such as the 21-item Beck Depression Inventory (BDI) or the visual analog scale, may be helpful if a direct face-to-face evaluation is not feasible.[12] However, it should be noted that, compared with a brief interview addressing the commonly accepted disease criteria, the efficacy of these tools as a diagnostic aid is inferior.

Guidelines and self-report tools for depression

The Diagnostic and Statistical Manual of Mental Disorders, 5th Edition

 (DSM-5) is one of the most common tools used to diagnose depression, particularly among non–terminally ill patients.[14] The DSM-5 outlines somatic complaints associated with the condition, such as changes in weight or appetite, sleep disturbance, fatigue or loss of energy, and difficulty thinking or concentrating. MDD is often not recognized in patients with cancer because the somatic symptoms of depression-including changes in appetite, energy, and/or sleep-may be attributed to normal cancer-related changes or to cancer treatment side effects. In addition, these somatic symptoms may overlap with the changes seen in patients with cancer who are not depressed.[14] This is why, in addition to the self-reported tools discussed previously, it is very important that physicians assess patients for signs of depression. This can be done by conducting an interview with the patient. Self-report tools may not provide accurate enough evidence on their own to diagnose depression. Clinicians must also ascertain that the patient’s symptoms are causing functional impairment.

In cancer patients, the treatment of depression should be individualized to address the patient’s depressive symptoms, as well as the disease-related and psychosocial factors contributing to the development of depression. Typically, antidepressant medication is most effective for severe depression, whereas psychotherapy is recommended for both mild and severe cases of depression.[15,16] Cognitive behavioral therapy (CBT) is the most commonly recommended psychotherapy; compared with other types of therapy, the most evidence from clinical trials involving patients with major depression is available for CBT.[17] The NCCN guidelines also recommend supportive psychotherapy.[7] Among patients recently diagnosed with cancer, CBT, relaxation strategies, and problem-solving approaches are recommended.

One specific type of supportive therapy, dignity therapy, has been shown to improve depressive symptoms in clinical trials.[18] Dignity therapy is an individualized type of psychotherapy that provides patients the opportunity to discuss their preferences, such as reflecting on what matters most to them or on how they would like to be remembered by their loved ones. Patients receive an edited transcript of the session to share with friends and family.[19] Lastly, the NCCN guidelines recommend that cancer patients complete family and couples therapy to lessen distress. Studies have shown that both of these types of therapy are associated with reduced distress and grief among families.[20]

 Several classes of antidepressants are available to treat depression, including:

• Selective serotonin reuptake inhibitors (SSRIs)

• Tricyclic antidepressants (TCAs) and tetracyclic antidepressants

• Serotonin-norepinephrine reuptake inhibitors (SNRIs)

• Atypical antidepressants (eg, mirtazapine)

When selecting an antidepressant, clinicians should consider the patient’s symptoms and concurrent medications. Characteristics of the available agents to treat depression are described in 

When it comes to selecting pharmacotherapy to treat depression in patients with cancer, no particular medication class has been proven to be superior in terms of efficacy. Thus, when choosing a medication to initiate treatment, physicians must consider a number of factors. First and foremost, an assessment of the patient’s symptoms should be completed, since certain medications may be more beneficial in treating specific ailments. For example, a patient with symptoms of both depression and insomnia may benefit from an antidepressant that has a more sedative effect. Clinicians should also inquire whether the patient has ever used antidepressant medications in the past and, if so, whether they were effective.

Other considerations when selecting pharmacotherapy for depression include potential drug-drug interactions and adverse side effects. Cost issues, as well as the patient’s comorbid symptoms-such as insomnia, neuropathic pain, or poor appetite-should be considered. Furthermore, due to inadequate studies among the cancer patient population, dosing of antidepressants must be highly individualized. Failure to adhere to antidepressant therapy leads to several high-risk outcomes,[21] particularly among those who are nonadherent during the first 6 weeks of therapy.[22] Individuals who adhere to antidepressant therapy early and continue to take their medication as prescribed are more likely to recover from depression and to avoid future relapse.

The preferred first-line therapy for depression is an SSRI, the most commonly prescribed class of antidepressants. The favorable efficacy and side effect profiles associated with SSRIs make these agents the most attractive for some patients with cancer who are diagnosed with depression. SSRIs are generally as effective as TCAs for the treatment of depression, and they have a wider margin of safety than TCAs in the event of an overdose.[23] Compared with TCAs, SSRIs are associated with fewer cardiac arrhythmias, hypotension, and anticholinergic effects.[24] In addition, SSRIs generally have reduced sedative and autonomic properties. SSRIs should be started at a low dose and then titrated to the minimum effective dose to reduce the potential for side effects, such as jitteriness, restlessness, anxiety, agitation, headache, sexual dysfunction, gastrointestinal symptoms (ie, diarrhea and nausea), and insomnia.[23]

Among the SSRIs, citalopram or escitalopram are considered to be first-line therapy because they are well tolerated and pose few drug-drug interactions. Furthermore, since these agents do not significantly inhibit cytochrome P450 (CYP450) 2D6 (CYP2D6), they can be used in patients taking tamoxifen.

In a study evaluating the use of escitalopram among cancer patients in palliative care, 18 patients who met the DSM-IV criteria for depressive disorder were treated with 10 mg for 2 weeks. Patients were evaluated using the Hospital Anxiety and Depression Scale (HADS) and the Mini-Mental Adjustment to Cancer Scale (Mini-MAC). At the endpoint, there was a significant reduction in anxiety, as measured by HADS, and in hopelessness-helplessness, as measured by Mini-MAC.[25]

Fluoxetine has limited clinical use in advanced illnesses due to its long half-life, since it takes 5 to 6 weeks to reach steady-state drug concentrations; in addition, its potential for significant drug-drug interactions is high because it inhibits hepatic drug–metabolizing enzymes, such as CYP450.[26,27] Paroxetine can be sedating and may lead to withdrawal phenomena with missed doses.[28] Sertraline has been shown to reduce fatigue, appetite, anhedonia, and suicidal thoughts in all types of depression in patients with cancer undergoing chemotherapy.[29] In older adults with advanced cancer, the initial SSRI dose should be reduced to approximately one-half that used in healthy patients.

An older class of antidepressants, TCAs have demonstrated efficacy in depressed, medically ill patients.[30] TCAs are also proven adjuvant analgesics for neuropathic and chronic low back pain.[31] While TCAs have been used frequently in the cancer setting, they are not as well tolerated as other antidepressants. This is due to their anticholinergic properties, which may cause symptoms such as dry mouth, blurred vision, urinary retention, and constipation. Older adults are particularly susceptible to confusion and hallucinations. TCAs may also induce delirium and prolong the QTc interval, and they can be dangerous in the event of overdose.[32] Therefore, their use is limited, since they may be problematic for patients with cardiac conditions and those taking other medications with anticholinergic properties. Of note, nortriptyline and desipramine have reduced anticholinergic properties compared with other tricyclics.

SNRIs are another class of antidepressants that may be helpful for patients with cancer. Research has shown that SNRIs are beneficial in managing neuropathic pain, vasomotor instability, or anxiety-predominant depression.[33] Venlafaxine is one example of an SNRI that is well tolerated in cancer patients. It is not significantly active in the CYP450 system, and, compared with other antidepressants, it is somewhat less protein-bound.[34] Venlafaxine is considered to be the first-line therapy in the SNRI class for those receiving tamoxifen due to its lack of CYP2D6 inhibition, as well as its ability to reduce hot flashes in those receiving chemotherapy or experiencing menopause as a result of taking tamoxifen. However, it should be used with caution in those with hypertension, as high doses of this agent may elevate blood pressure.

Duloxetine is another SNRI that is approved to treat anxiety, diabetic neuropathy, and chronic musculoskeletal pain.[35] A prospective, 12-week, case-control study compared the use of duloxetine in patients with depression and cancer vs depression alone. Patients received an initial dose of duloxetine 30 mg, which was then titrated to 60 mg after 1 week and up to 120 mg after 1 month, based on response. Patients were assessed using HADS, the Clinical Global Impression Scale-Severity (CGI-S), and the Montgomery–Ãsberg Depression Rating Scale (MADRS). In individuals with both cancer and depression, scores significantly improved on each of the depression scales at both 4 and 12 weeks.[36] Duloxetine is contraindicated in patients with heavy alcohol use or chronic liver disease. Milnacipran has been shown to be significantly effective in the treatment of depression and fibromyalgia.[37,38]

The antidepressant bupropion has been shown to decrease fatigue and improve depression symptoms in cancer patients. An open-label study examined
	sustained-release bupropion for depression and fatigue in the cancer population. Patients took bupropion for 4 weeks after titrating to 300 mg or the maximum tolerable dose. Nine of the patients were identified as depressed at baseline, based on a Hamilton Depression Rating Scale (HAM-D) score of > 17. Both depressed and non-depressed patients had significantly lower HAM-D scores at the endpoint, with depressed patients showing a greater decrease in HAM-D scores compared with non-depressed patients.[39] Bupropion is contraindicated for cancer patients with central nervous system disorders due to its association with increased prevalence of seizures.[40]

While trazodone is an effective antidepressant, its use is limited because it causes significant sedation.[41] However, it may be helpful among patients with sleep disturbances, as well as for those who require an adjunct analgesic effect in addition to an antidepressant effect. Nefazodone, another antidepressant, is associated with liver failure and has been withdrawn in many countries; therefore, it should not be used in medically ill patients.[42]

Although mirtazapine has a sedative effect, it has been shown to be effective for improving pain symptoms, as well as depression and quality of life in patients with advanced cancer.[43] When used at low doses, mirtazapine has also been shown to improve appetite and insomnia, particularly among patients on the 15 mg/day dose, and to provide sedation when warranted.[44]

Monoamine oxidase inhibitors (MAOIs) are generally considered to be less desirable for treating depression in patients with advanced medical illnesses due to the substantial number of adverse interactions associated with these agents. MAOIs are generally reserved for patients who have shown a past preferential response to them for depression.[45]

Antidepressants may not have sufficient time to reach their full effect in patients with a prognosis of days or weeks. In these cases, psychostimulants may be beneficial. Compared with antidepressants, psychostimulants have a more rapid onset of action of within 24 to 48 hours. They have been shown to improve attention, concentration, and overall performance on neurologic testing in the medically ill.[46] In addition, they can improve mood, appetite, and sense of well-being. Although generally well-tolerated, psychostimulants may cause side effects such as agitation, nausea, anorexia, insomnia (particularly if administered within 6 to 8 hours of bedtime), anxiety, and tremor. Modafinil has fewer sympathomimetic effects than amphetamines and is often a good choice for older patients.[47] Psychostimulants should be avoided in patients with delirium and used with caution in those with heart disease. Although the data on psychostimulants are somewhat mixed, controlled trials have shown their benefit as a monotherapy or to augment the effects of another antidepressant.[47]

A number of case reports and one small placebo-controlled trial of intravenous ketamine have shown that ketamine can rapidly improve depressive symptoms in patients with refractory depression in the setting of a terminal illness, particularly in the setting of chronic pain.[48,49]

Depression is underdiagnosed and undertreated in chronically ill cancer patients. It is imperative that healthcare providers be proactive in screening for depression, as well as in educating patients about the available treatment options for depression. Each individual case of depression is different and must be treated according to the patient’s medical history, cancer prognosis, and severity of symptoms. When selecting therapy, physicians must evaluate the side effect profile of each agent to ensure it is appropriate for the patient. A specific patient interview should also be conducted in order to carefully confirm the diagnosis and to avoid confusing the symptoms of progressing cancer with those of major depression. Once pharmacotherapy is initiated, close monitoring for adverse effects and the potential need for dose titration is warranted.

• For patients with neuropathic pain and depression, consider duloxetine or venlafaxine.

• If the patient has insomnia or anorexia, consider the use of mirtazapine.

• When polypharmacy is present, consider citalopram, escitalopram, or mirtazapine.

• Closely monitor patients initiated on an antidepressant for adverse effects and the potential need for dose titration.

• Refer to social work and/or spiritual support services if the depression appears to be escalating in relation to social or spiritual factors.

Ms. Ahmed has no financial interest in or other relationship with the manufacturer of any product or provider of any service mentioned in this article

Asking Matters Most: Depression in Advanced Cancer

Mirat Shah, MD, Kerry O’Connor, MD, and Thomas J. Smith, MD, FACP

Depression is common in patients with advanced cancer and leads to poor symptom control and decreased life expectancy, as highlighted in this article. Ms. Ahmed notes that depression is underdiagnosed in this group because some symptoms-loss of appetite, fatigue, and trouble sleeping-mimic the symptoms of advanced cancer. In our experience, there are additional factors that prevent diagnosis.

Oncology providers tend to think that depression is a normal reaction to advanced cancer, rather than something that can be treated. They may feel ill-equipped to manage depression once it is diagnosed, or that a different member of the care team is better-suited to the task. In addition, clinicians are pressed for time during clinic visits. Without a prompt, patients often don’t bring up these symptoms and instead try to cope on their own.

Asking the simple question “Are you depressed?” helps patients feel heard and valued, and has acceptable sensitivity and specificity for cancer patients.[1] After we asked this question of a 35-year-old woman with advanced uterine cancer at our clinic, she remarked, “Thank you for listening. No one ever listens. Listening helps.” Another man with metastatic colorectal cancer who was asked this same question remarked, “Thank you for being human.” These responses highlight that being present and actively listening to patients are therapeutic acts in and of themselves.

After asking the question, additional members of the care team can assist as needed, and, as noted in this article, therapy and medication may also help. But asking is the most important step. Although we sometimes may not be able to cure, by asking and listening, we can still heal.

 Chochinov HM, Wilson KG, Enns M, Lander S. “Are you depressed?” Screening for depression in the terminally ill. Am J Psychiatry. 1997;154:674-6.

Dr. Shah is an Oncology Fellow at Johns Hopkins Medicine, Baltimore, Maryland.

Dr. O'Connor is a Hospital and Palliative Medicine Fellow at Johns Hopkins Medicine, Baltimore, Maryland.

Dr. Smith is the Harry J. Duffey Family Professor of Palliative Medicine and a Professor of Oncology at Johns Hopkins Medicine, Baltimore, Maryland.

Massie MJ. Prevalence of depression in patients with cancer. J Natl Cancer Inst Monogr. 2004;57-71.

Boyd AD, Riba M. Depression and pancreatic cancer. J Natl Compr Canc Netw. 2007;5:113-6.

Akechi T, Okuyama T, Sugawara Y, et al. Major depression, adjustment disorders, and post-traumatic stress disorder in terminally ill cancer patients: associated and predictive factors. J Clin Oncol. 2004;22:1957-65.

Wilson KG, Chochinov HM, McPherson CJ, et al. Suffering with advanced cancer. J Clin Oncol. 2007;25:1691-7.

Wilson KG, Chochinov HM, Skirko MG, et al. Depression and anxiety disorders in palliative cancer care. J Pain Symptom Manage. 2007;33:118-29.

Satin JR, Linden W, Phillips MJ. Depression as a predictor of disease progression and mortality in cancer patients: a meta-analysis. Cancer. 2009;115:5349-61.

National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology: Distress Management (Version 2.2018). 

https://www.nccn.org/professionals/physician_gls/pdf/distress.pdf

Walsh D, Donnelly S, Rybicki L. The symptoms of advanced cancer: relationship to age, gender, and performance status in 1,000 patients. Support Care Cancer. 2000;8:175-9.

In this article, we discuss the importance of screening for depression in cancer patients, as well as the available treatment options and their pros and cons.

Antidepressants in Patients With Advanced Cancer: When They’re Warranted and How to Choose Therapy

Poly (ADP-ribose) polymerase (PARP) inhibitors are a class of oral anticancer medications that have the most evidence for use in patients with inherited germline mutations in 

 tumor suppressor genes. By inhibiting PARP, these agents form PARP-DNA complexes, resulting in DNA damage, apoptosis, and cancer cell death (

The first US Food and Drug Administration (FDA) approval of a PARP inhibitor, olaparib, occurred in 2014; there are now four FDA-approved PARP inhibitors: olaparib, rucaparib, niraparib, and talazoparib (

). Olaparib has four indications: three in ovarian, fallopian tube, and primary peritoneal cancer, and one in metastatic human epidermal growth factor receptor 2 (HER2)-negative, 

-mutated breast cancer after treatment with chemotherapy.[2] Rucaparib is indicated in ovarian, fallopian tube, and primary peritoneal cancer as a third-line treatment following chemotherapy or as maintenance treatment.[3] Niraparib is indicated only for maintenance treatment in platinum-sensitive ovarian, fallopian tube, and primary peritoneal cancer.[4] Talazoparib has one indication for treatment in locally advanced or metastatic HER2-negative, 

As this class of medication becomes more commonly prescribed and continues to gain new indications, it is important that providers address the most common adverse effects of these medications, which include anemia, neutropenia, gastrointestinal (GI) toxicity, and fatigue. When prescribing oral agents with known risk of GI toxicity, providers must proactively prescribe medications to prevent these toxicities, such as nausea and vomiting, in order for the patient to remain adherent to the prescribed dose without significant decline in quality of life.[6] We will outline here how to prevent and treat nausea and vomiting among patients being treated with PARP inhibitors.

In the 2014 Kaufman trial, as well as the SOLO-1 and SOLO-2 trials, nausea was the most commonly reported adverse event seen with olaparib treatment (all grades, 62%–77%; grade ≥ 3, 1%–6%), with a lower incidence of vomiting (all grades, 35%–40%; grade ≥ 3, 1%–3%).[7-9] In the ARIEL2 and ARIEL3 trials, nausea was also the most common adverse event in the rucaparib group (all grades, 75%; grade ≥ 3, 4%); vomiting was also less common (all grades, 37%–42%; grade ≥ 3, 2%–4%).[10,11] Similarly, the NOVA trial showed that the most common adverse effect with niraparib was nausea (all grades, 74%; grade ≥ 3, 3%), with vomiting of all grades occurring in 34% of patients (grade ≥ 3, 2%).[12] However, nausea and vomiting were less prevalent with talazoparib in the EMBRACA trial (for nausea: all grades, 49%, grade ≥ 3, < 1%; for vomiting: all grades, 24%, grade ≥ 3, 2%).[13]

As opposed to other toxicities wherein grade 3 and 4 are seen as more clinically significant, grade 1 nausea and vomiting can cause a significant impact on quality of life. The National Comprehensive Cancer Network (NCCN) guidelines for antiemesis categorize the emetogenic potential of oral anticancer agents based on the frequency of vomiting/emesis reported (

). Treatments are considered to have a moderate to high frequency if emesis occurs in ≥ 30% of patients, which places olaparib, niraparib, and rucaparib in this category. However, talazoparib would be considered to have minimal to low emetic risk.[14]

When assessing a patient’s risk for chemotherapy-induced nausea and vomiting (CINV), it is important to consider not only the emetic risk of the therapeutic agent, but also to take into account the individual patient’s risk factors. Women, younger patients, patients with lower alcohol intake history, and those with a history of motion sickness are at increased risk of experiencing CINV during treatment.[14,15] For oral cancer agents that fall in the moderate to high emetic risk category (including olaparib, rucaparib, and niraparib), the NCCN recommends that the patient take a serotonin (5-HT3) receptor antagonist, such as ondansetron 8 to 16 mg daily, ideally approximately 30 minutes before a PARP inhibitor dose.[14] All four of the PARP inhibitors currently approved can be taken with or without food; therefore, the patient should be counseled to take the medication after a meal if they experienced nausea taking it on an empty stomach.[2-5] Of note, olaparib and rucaparib are both dosed twice daily and may require the anti-nausea medication to also be taken twice per day.[2,3] Niraparib is taken only once daily and is recommended to be taken at night to help prevent CINV.[4] Compliance to antiemetic therapy should be encouraged when initiating new PARP inhibitor therapy to avoid drug intolerance and potential for unnecessary interruption in therapy. At each clinic visit, nausea and vomiting should be assessed to determine if any adjustments need to be made to the antiemetic regimen.

For patients who experience CINV during treatment, it should first be determined if they are compliant with their antiemetic regimen, refractory to the antiemetic therapy (experiencing CINV with the majority of doses), or experiencing breakthrough nausea with only occasional episodes. If first-line prevention with a 5-HT3 receptor antagonist is not sufficient, the patient should try alternative treatments for CINV. The general recommendation for patients who are refractory to initial therapy is to stop the first-line agent and switch to a drug with a different mechanism. For patients who experience only breakthrough CINV, it is recommended to add one agent from a different drug class to the current regimen.

The NCCN also recommends olanzapine, lorazepam, dronabinol, haloperidol, metoclopramide, scopolamine transdermal patch, prochlorperazine or promethazine, an alternative 5-HT3 receptor antagonist such as dolasetron or granisetron, or a steroid such as dexamethasone for the prevention of CINV. When deciding between these options, one should consider patient- and drug-specific factors. For example, olanzapine should be started with a low dose and it is recommended that it should be taken at night, since it can cause significant drowsiness. Age should be considered when using benzodiazepines such as lorazepam, since they can cause more complications in elderly patients, though they may be helpful if the patient is experiencing anticipatory nausea. Dronabinol may be useful if the patient is also experiencing decreased appetite. The scopolamine patch may help if the patient finds that they are more sensitive to motion sickness while on PARP inhibitor therapy. Granisetron is also available as a patch if the patient is vomiting and unable to keep down oral medications. Dexamethasone should be recommended with caution, since there are notable complications with long-term use of steroids and it may need to be used daily with PARP inhibitor–related nausea. If a patient vomits after a medication dose, it is important to counsel them to not take another dose until the next scheduled dose. For patients who continue to experience CINV after attempted antiemetic optimization and have compromised quality of life, dose reduction may be required (

Additional Considerations With PARP Inhibitors

Drug interactions should be taken into account when choosing which PARP inhibitor to prescribe. For example, a dose reduction for olaparib would be required if given with cytochrome P450 3A4 (CYP3A4) inhibitors.[2] Aprepitant or fosaprepitant, netupitant, and fosnetupitant should all be avoided in patients on olaparib treatment, since they have interactions with medications metabolized through CYP3A4. Patients should also be counseled to avoid grapefruit juice and Seville oranges when taking olaparib.[2] For patients on talazoparib, a dose reduction is required when given with P-glycoprotein inhibitors.[5]

Additionally, mild increases (grade 1 or 2) in serum creatinine are possible due to transporter inhibition with olaparib and rucaparib. Continuation of therapy during evaluation is recommended to rule out true renal dysfunction.[6] Mild increases in transaminases (aspartate and alanine transaminases) may be seen within the first 28 days of rucaparib treatment. Providers should note that they do not need to hold the medication unless transaminases exceed 5 times the upper limit of normal (ULN) or more than 3 times the ULN for bilirubin and more than 2 times the ULN for alkaline phosphatase.[6] Lastly, additional GI adverse effects that should be monitored with PARP inhibitor use include constipation, diarrhea, abdominal pain, dyspepsia, dysgeusia, and decreased appetite.[2-5]

Providers should be aware of the potential risk for GI toxicity with PARP inhibitors. Patients initiating olaparib, rucaparib, or niraparib should receive prophylactic antiemetic medication to help prevent nausea and vomiting throughout the treatment course. Patients should be monitored for compliance and efficacy of antiemetic therapy to help preserve quality of life and prevent inappropriate dose reduction. There are several different classes of medication available for use as antiemetic therapy if initial therapy is not sufficient or is contraindicated. Each patient is different and will require continued monitoring and potential alteration in antiemetic therapy to help them remain compliant and maintain a good quality of life.

The authors have no significant financial interest in or other relationship with the manufacturer of any product or provider of any service mentioned in this article

 Dziadkowiec K, GÄsiorowska E, Nowak-Markwitz E, Jankowska A. PARP inhibitors: review of mechanisms of action and BRCA1/2 mutation targeting. Prz Menopauzalny. 2016;15:215-9.

 Olaparib (Lynparza) prescribing information. 

https://www.azpicentral.com/lynparza_tb/pi_lynparza_tb.pdf#page=1

 Rucaparib (Rubraca) prescribing information. https://

clovisoncology.com/media/1094/rubraca-prescribing-info.pdf

 Niraparib (Zejula) prescribing information. 

https://www.zejula.com/prescribing-informati

 Talazoparib (Talzenna) prescribing information. 

http://labeling.pfizer.com/ShowLabeling.aspx?id=11046

 Gunderson CC, Matulonis U, Moore KN. Management of the toxicities of common targeted therapeutics for gynecologic cancers. Gynecol Oncol. 2018;148:591-600.

 Kaufman B, Shapira-Frommer R, Schmutzler RK, et al. Olaparib monotherapy in patients with advanced cancer and a germline BRCA1/2 mutation. J Clin Oncol. 2015;33:244-50.

 Moore K, Colombo N, Scambia G, et al. Maintenance olaparib in patients with newly diagnosed advanced ovarian cancer. N Engl J Med. 2018;379:2495-505.

 Pujade-Lauraine E, Ledermann JA, Selle F, et al. Olaparib tablets as maintenance therapy in patients with platinum-sensitive, relapsed ovarian cancer and a BRCA1/2 mutation (SOLO2/ENGOT-Ov21): a double-blind, randomised, placebo-controlled, phase 3 trial. Lancet Oncol. 2017;18:1274-84.

 Swisher EM, Lin KK, Oza AM, et al. Rucaparib in relapsed, platinum-sensitive high-grade ovarian carcinoma (ARIEL2 Part 1): an international, multicentre, open-label, phase 2 trial. Lancet Oncol. 2017;18:75-87.

 Coleman RL, Oza AM, Lorusso D, et al. Rucaparib maintenance treatment for recurrent ovarian carcinoma after response to platinum therapy (ARIEL3): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet. 2017;390:1949-61.

 Mirza MR, Monk BJ, Herrstedt J, et al. Niraparib maintenance therapy in platinum-sensitive, recurrent ovarian cancer. N Engl J Med. 2016;375:2154-64.

 Litton JK, Rugo HS, Ettl J, et al. Talazoparib in patients with advanced breast cancer and a germline BRCA mutation. N Engl J Med. 2018;379:753-63.

 National Comprehensive Cancer Network. Antiemesis. Version 3.2018. 

https://www.nccn.org/professionals/physician_gls/pdf/antiemesis.pdf

 Mustian KM, Devine K, Ryan JL, et al. Treatment of nausea and vomiting during chemotherapy. US Oncol Hematol. 2011;7:91-7.

Here, two oncology pharmacists discuss the side effects of PARP inhibitor treatment, and how to prevent and treat nausea and vomiting in this patient population.

Nausea and Vomiting: Managing Side Effects From PARP Inhibitors

 to read an expert perspective from Tufia C. Haddad, MD.

Artificial intelligence (AI), for years, has captured society’s imagination and generated enthusiasm for its potential to improve our lives. Presently, AI already plays an integral role in our daily routines and our interactions with media, transportation, and communications. There is an increasing interest in the applications of AI in healthcare to improve disease diagnosis, management, and the development of effective therapies. Given the large number of patients diagnosed with cancer and significant amount of data generated during cancer treatment, there is a specific interest in the application of AI to improve oncologic care. In this review, we introduce the fundamentals of AI and provide an overview of its current applications, pitfalls, and future directions in oncology.

Originally formalized in the 1950s, AI refers to the ability of a machine to perform tasks commonly associated with intelligent human behavior.[1] Including disciplines from both computer science and mathematics, AI can be considered a group of iterative, "self-learning" techniques, which discover relationships within data that can evolve and often be performed faster over time.

Machine Learning (ML): The Engine That Drives AI

A majority of the AI applications within healthcare involve the utilization of ML algorithms. As ML algorithms are exposed to more training data, they are able to appreciate hidden patterns within the data which can then be used to perform a task without explicit programming.[2] There are dozens of ML algorithms that have been proposed over the past several decades, and the most traditional forms of ML, such as logistic regression, have proven themselves as valuable tools for general clinical oncology research.[3]

ML tasks are often broadly dichotomized into supervised or unsupervised learning. In supervised learning, a labeled dataset of inputs and outputs is used to train the ML algorithm. The algorithm attempts to learn a general rule that maps input to output. Supervised ML algorithms can learn patterns in data for categorical outputs (classification) and continuous data (regression). Conversely, unsupervised ML algorithms use unlabeled data, with the goal of discovering structure in the input data. Unsupervised ML algorithms are often used to simplify (dimensionality reduction) or organize (clustering) data.

In traditional ML tasks, there is often some pre-engineered organization of raw input data into features that are inferred to have an impact on output.

Artificial Neural Networks and Deep Learning (DL): The Wave Comes to Healthcare

One drawback of traditional ML algorithms has been the need for pre-engineered organization of raw input data into structured datasets. The inability of certain ML algorithms to use unstructured data from the point of generation has limited their utility in clinical practice. One ML algorithm well suited to analyze unstructured data is known as DL (

DL algorithms are often synonymous with AI. DL is a form of ML that uses layered "artificial neural networks" to develop sophisticated models with the ability to understand data at different levels of abstraction.[4] Artificial neural networks (ANNs) were inspired by human neurobiology and the ability of the brain to use cascading, varying, and layered combinations of neurons to learn complicated patterns with a hierarchy of progressively more complex features (

). Modeling the human neuron in computers yielded the basic design of early ANNs.[5]

While traditional ML algorithms used pre-engineered features to develop predictions, DL algorithms are able to learn the optimal features that best fit the data through the training process, avoiding the need to use pre-engineering, unstructured data. This ability has allowed DL algorithms to outperform traditional ML algorithms in many common AI problems, including image classification, natural language processing, and sequence prediction.[4,6,7] Accordingly, the number of research articles published involving deep learning in the medical field has skyrocketed over the past few years (

Due to this DL renaissance, AI has now generated significant attention in healthcare.[8-11] In oncology, DL has become a natural partner in the pursuit of precision medicine,[12-15] leveraging vast, heterogenous datasets to better diagnosis disease burden, predict patient outcomes, and tailor management. AI can also be coupled to a multitude of emerging mobile health interfaces, such as smartphone apps and wearable devices, to develop “digital biomarkers” that can explain, influence, and predict clinical outcomes. In the following sections, we will delve into current AI applications in oncology, their limitations, and future implications.

Seeing better with convolutional neural networks

Image analysis has proven to be among the most effective methods in which AI has impacted society. AI powered by DL algorithms has provided us self-driving vehicles, mobile check deposit, and multiple other useful technologies. Given the large amount of digital imaging data present within medicine, there is increasing excitement about the application of similar techniques to imaging within oncology (

This revolution in image analysis was catalyzed by the development of a particular DL architecture, the Convolutional Neural Network (CNN). CNNs analyze pixel-level information from images. The added benefit of CNNs compared to other DL configurations is the ability to account for orientation of the pixels in relation to one another. This effectively allows the CNN to appreciate lines, curves, and eventually objects within images. CNN-based models have recently been shown to be equivalent to humans in picture classification and object detection.[6,16]

One of the initial papers highlighting the promise of DL in cancer imaging was in identification of skin cancer based on skin photographs.[17] The CNN trained on 130,000 skin images was able to classify malignant lesions with higher sensitivity and specificity than a panel of 21 board-certified dermatologists. Practical applications of detecting skin pathology utilizing patients as the generator of imaging input data has evolved.[18] Further use of CNNs to classify digital photography has been in the automatic detection of polyps during colonoscopy. One particular study highlights that CNNs can be used for not only image classification, but also to detect regions of clinical importance. Researchers found that CNN, which was trained on colonoscopic images from 1,290 patients, had a 94% sensitivity in polyp detection.[19]

Given one of the most effective applications of AI techniques has been in the field of computer vision, there is naturally excitement in the field of radiology, where there exist a number of digitized images. The goals of these AI algorithms have ranged from assisted diagnosis to outcome prediction.

AI algorithms have been found to be effective in streamlining cancer screening and detection. Automated lung nodule detection and classification has attracted significant attention and formed the basis of the 2017 Kaggle Data Science Bowl, an international competition for ML scientists.[20] Several successful CNN-based models resulted from this competition and other research groups, demonstrating accuracies in the 80% to 95% range and showing promise for lung cancer screening.[21–26] Additionally, CNNs have been shown to be successful in segmenting tumor volumes, which may have implications for radiotherapy treatment planning.[27] Improvement of breast cancer screening with AI has also been an active area of investigation, with its own data science competitions,[28] resulting in a CNN algorithm able to detect breast malignancy with a sensitivity of 90%.[29,30]

AI has also shown promise in detecting radiographic anatomic features of malignancies that goes beyond what can be reliably achieved by clinicians. While extranodal extension (ENE) of tumors in the head and neck cancer lymph nodes has been notoriously difficult to diagnosis radiographically by clinicians, a CNN-based model showed >85% accuracy in identifying this feature on diagnostic, contrast-enhanced CT scans.[31] Identification of ENE has high importance in prognostication and management for head and neck cancer patients, and thus this model shows promise as a clinical decision-making tool.

Going further than anatomic characterization, AI has shown promise in the burgeoning field of radiogenomics, where radiographic image analysis is used to predict underlying genotypic traits. This has recently been demonstrated using CNNs on brain MRIs of patients with low-grade glioma. CNNs have been able to predict both 

 methylation status with 85% to 95% and 83% accuracy, respectively, based on raw imaging data alone.[32,33]

DL also has the potential to predict response to treatment based on imaging findings. Recently, a CNN model showed success in predicting complete response to neoadjuvant chemoradiation with 80% accuracy.[34] Additionally, a radiomics signature using extracted features from CT data and a ML algorithm were able to predict underlying CD8 cell tumor infiltration and, remarkably, response to immunotherapy for a variety of advanced cancers.[35]

The increasing digitization of histopathologic tumor specimen slides provides a robust 2D image suitable for DL analysis. DL CNN algorithms have now been shown to diagnose breast cancer metastasis in lymph nodes with at least equivalent performance compared to a panel of pathologists, and in a more time-efficient manner.[36] DL has also been shown to be useful in automated Gleason grading of prostate adenocarcinoma Hematoxylin and Eosin–stained specimens, with a 75% rate of agreement between the algorithm and pathologists.[37]

DL algorithms have also gone a step further than pathologic diagnosis automation and have been used to characterize the underlying genotype-phenotype correlation within a tumor specimen. Using raw input data consisting of digitized formalin-fixed, paraffin-embedded tissue from lung cancer biopsies, a CNN was trained to predict six different genetic mutations (

), establishing that genotypic information may be gleaned from histopathologic architectural patterns.[38] These methods may assist pathologists in the detection of cancer gene mutations and may be cost-effective compared to direct mutational analysis.

Within clinical oncology, AI has increasingly been applied to harness the power of the electronic health record (EHR) (

). Specifically, AI-based natural language processing techniques have shown promise in predicting the development of diseases across large healthcare systems. A notable example from a group at Mount Sinai, a DL-based AI algorithm modeling EHR, was able to predict the development of a variety of diseases with 93% accuracy overall, including cancers of the prostate, rectum, and liver.[39]

There has additionally been interest in using DL to predict cancer treatment toxicity. Recently, a CNN approach was used to predict side effects of polypharmacy combinations based on databases of protein-protein and drug-protein interactions.[40] This study led to the discovery of at least five novel drug-drug interaction predictions, which were subsequently found to have supporting literature evidence. Use of AI to predict radiotherapy toxicity has generated significant interest over the past few years.[41] Basic neural networks, CNNs, and other ML methods have been explored, using clinical and dosimetric data to predict urinary and rectal toxicity resulting from prostate radiotherapy results,[42–44] hepatobiliary toxicity after liver radiotherapy,[45] and rectal toxicity for patients receiving radiotherapy for cervical cancer.[46]

Translational oncology is an area where AI is beginning to emerge. Over the past decade, there has been an expansion of biological quantitative or “-omic” data. Given the complexities and heterogeneity within this data, the use of DL in analysis has been appealing. DL neural networks have been utilized to predict protein structure,[47] classify cells into a distinct stage of mitosis,[48] and even predict the future lineage of progenitor cells based on microscopy images.[49]

Drug development and repurposing has become an attractive target for DL. One group used DL ANNs trained on transcriptomic response signatures to drugs to predict with high accuracy the likelihood of failure of a clinical trial of over 200 example drugs.[50] Another used an ANN to predict cancer cell sensitivity to therapeutics using a combination of genomic and chemical properties.[51] CNNs have also been employed to predict peptide-major histocompatibility complex binding,[52] which may have implications for oncologic immunotherapy development. See 

 for a summary of applications of AI in translational oncology that have been researched.

Given the increasing amount and pace of published research, clinical trial enrollment, drug development, and biomarker discovery in oncology in recent years, there is more opportunity than ever for AI to assist in synthesizing this data and to guide decision-making. There are several commercial applications in development that utilize DL and natural language processing to this aim.[53] These applications are being designed to link patient data to clinical trial databases and to match patients to appropriate clinical trials nationwide. Another algorithm utilizes ML to select the appropriate investigational drugs in development for a given patient. There has also been interest in utilizing AI coupled with patient data and national treatment guidelines to guide cancer management, with the most prominent example being IBM’s Watson for Oncology (WFO). WFO has demonstrated high concordance with tumor board recommendations for breast cancer patients, though has fallen short in other areas of oncologic decision-making.[54,55] While this area of AI application is in its nascent stages, performance continues to improve, and there is great potential to improve clinical practice.

Proving generalizability and real-world applications

While AI is rapidly being incorporated into oncologic research, work remains to be done to translate these studies into real-world, clinically meaningful applications. One of the biggest barriers is in external validation and proving generalizability of DL applications. Given the complexity of neural networks and the extremely large number of parameters (often in the millions), there is a high tendency for neural networks to create overfitted models that do not generalize across different populations. Additionally, because there is a significant amount of heterogeneity of medical data across institutions, multiple external validation sets may be required to prove the performance of an application.[56]

Directly contributing to this problem of overfitting are limitations with data access and quality. DL neural networks, more than any other ML algorithm, require large amounts of data. This can pose a problem in healthcare when attempting to apply AI to disease processes with less prevalence. Furthermore, data is often siloed within individual institutions. Contributing to this relative data drought are concerns with transmission of protected patient health information, along with lack of data-sharing infrastructure to link institutions, heterogeneity and incompleteness in the collection of data, and competition between institutions. These obstacles are beginning to be addressed, with more and more emphasis on streamlined data capture,[57] and a number of multi-institutional data-sharing agreements[58-61] Guidelines have been proposed to support FAIR (findable, accessible, interoperable, reusable) data use,[62] and there are now opportunities for research groups to publish their data itself, which may incentivize openness.[63]

Interpretability and the black box problem

One of the central limitations to adoption of AI in healthcare is the concern that these models, despite regularly achieving high performance, are somewhat opaque. For instance, a DL model may correctly predict that a patient will develop pancreatic cancer based on his past 2 years of EHR data, but why did it make that prediction? At the present, we are limited in our ability to determine the precise logic behind DL-based predictions. This is often referred to as the “black box” problem.[64] In medical practice, it has traditionally been essential in clinical decision-making to know the rationale for each decision. Traditional ML algorithms, like linear regression, have limited ability to model complex relationships, but offer this easy interpretability-in these algorithms, we have a set of pre-defined features and the resulting feature weights that characterize their effect sizes. In contrast, DL utilizes unstructured input data, and the bulk of knowledge generation occurs within the hidden layers. It thus becomes difficult to determine which specific characteristic(s) of the input data contributed to the outcome. This interpretability challenge has large implications for adoption of AI-based algorithms in healthcare, both from practitioner and regulatory perspectives.[65-68]

Tackling the black box problem has now become a major focus of research.[69] In AI image analysis algorithms, several methods have been developed, including feature visualization, saliency maps, class activation mapping, and sensitivity analyses, where certain parts of the image are hidden to the effect on prediction.[70] While these methods have advanced over the past few years, further work is needed to better elucidate the decision-making logic with deep neural networks.

To successfully merge AI with clinical oncology and maximize its impact, there are knowledge gaps that need to be addressed. Currently, physicians receive little training in data science and ML, limiting their ability to understand DL mechanisms, adopt algorithms appropriately, and conduct research.[71] Similarly, most data scientists have little experience with oncologic workup and management, limiting the ability to identify important and suitable clinical use cases. Further collaboration should be pursued between clinical oncologic departments and bioinformatics and data science divisions, and strategic partnerships with technology firms should be formed where appropriate.

Promoting AI in Oncology: Professional Societies and National Initiatives

In response to these challenges, several national professional societies have launched initiatives to bridge these knowledge gaps and promote the dissemination of AI in oncology.

The ACR has founded a Data Science Institute (ACR-DSI) with the mission of collaborating with radiologists, industry, and government agencies to facilitate the development of AI in imaging.[72] Within the ACR-DSI are several core goals: 1) providing standards for measuring performance of AI algorithms (“Touch-AI”), 2) independent, external validation of algorithms and navigating the regulatory landscape (“Certify-AI”), and longitudinal, prospective evaluation of deployed algorithm performance “(Assess-AI”). The ACR-DSI has additionally set up a series of use cases for recommended AI imaging applications with unmet clinical need.

American Society of Clinical Oncology (ASCO) and American Society for Radiation Oncology (ASTRO)

ASCO has launched a big data initiative named CancerLinQ, in partnership with oncologists, industry, and academia, with the goals of real-time quality of care tracking and treatment evaluation, as well as knowledge dissemination to oncologists in user-friendly ways.[73] The initiative’s backbone is a constantly growing database of de-identified patient information that can be mined and analyzed. In 2017, ASTRO partnered with CancerLinQ to provide radiation oncology expertise and uses for the database. In addition, Big Data Analytics and Bioinformatics is one of the core initiatives of the ASTRO Research Agenda for 2018.[74]

As part of the NIH Common Fund, the Big Data to Knowledge (BD2K) initiative was launched to support the research and development of tools to integrate big data and data science into biomedical research.[75] One of the central components of the initiative involves leveraging existing national datasets, such as the Library of Integrated Network-based Cellular Signatures (LINCS) and The Cancer Genome Atlas (TCGA), and applying ML techniques to discover patterns in the data that may result in heretofore unknown compounds for cancer therapeutics.[76]

Over the past decade, AI has undergone a reawakening. Due to an explosion of electronic data, advances in technological infrastructure, and groundbreaking research in DL neural networks, AI is poised to make practice-changing impacts on the medical field and oncologic care. At present, AI has shown promise in improving cancer imaging diagnostics and treatment response evaluation, predicting clinical outcomes, and catalyzing drug development and translational oncology. Obstacles remain-such as validation and proving generalizability, concerns over interpretation, and the widening knowledge gap between clinical and data science experts. If we can address these challenges, AI has the potential to transform oncology, harnessing the power of big data to drive cancer care into the 21st century and beyond.

Artificial Intelligence as a Solution to Burnout in Oncology

Developing applications of artificial intelligence (AI) and cognitive systems in oncology requires a collaborative, multidisciplinary effort that extends far beyond medicine and computer science. More than a few significant challenges, however, limit the translation of AI-related cancer research into meaningful clinical applications. Among these are the abilities to form partnerships across multiple industries, to gain equitable access to large volumes of annotated data, and to conduct unbiased training of machine-learning algorithms. The article by Aneja and colleagues provides a detailed overview of the evolving field of AI in oncology and serves as a primer for curious and reluctant oncologists alike, the vast majority of whom have not received formal training in data science.

While there has been hype that AI has the potential to replace doctors, early experience has suggested that AI can instead augment human intelligence, enabling doctors to perform with greater efficiency, engagement, and effectiveness. That said, a more credible, contemporary threat faces the oncology profession: unprecedented levels of burnout.[1] Although multiple factors have been associated with this state of stress, the ubiquitous involvement of computers in all aspects of medicine is recognized as a major driver. For example, while electronic health records (EHR), computerized prescribing, and order entry were established to improve quality and coordination of patient care, implementation of these systems has resulted in unintended consequences for providers: reduced efficiency and increased clerical and cognitive burden.[2] During office hours, physicians spend almost 2 hours engaged in EHR and desk work for every 1 hour of direct patient face time,[3] and it is estimated that an additional 1.4 hours of EHR interactions occur daily outside of business hours.[4]

The EHR contains an enormous volume of data, and new data sources have the potential to generate a tsunami of additional patient data points. These include tumor genome sequencing reports, electronic patient-reported outcomes responses, patient online messaging, and patient-generated health data from apps and wearables. Integration of these data within the EHR and clinical workflows is currently suboptimal or lacking for most practices, furthering the rift between providers and their computers. Concurrent with this explosion of data is the frenzied pace of knowledge gains in oncology, including a record of 51 new agents or new indications for existing agents that were approved by the US Food and Drug Administration for cancer treatment in 2018.[5]

It has been recognized that much of the existing digital workload can be delegated to other care team members or to data abstraction and documentation specialists. Alternatively, technology may be exactly what is needed to solve our technology problem. Specifically, natural language processing (NLP), a branch of AI, can interpret, augment, and transform free text so that it can be represented for computation.[6] In medicine, it leverages the EHR, including its most valuable asset: the clinical note. NLP serves as the engine for cognitive systems, but adoption of this technology requires expert training and supervision, as well as thoughtful implementation, with minimal disruption to clinical workflows.

Under these conditions, AI offers the potential for detailed EHR summarization in a single click, as well as clinical decision support at the point of care aligned with evidence-based care pathways. AI may also generate order sets, cancer registries, or personalized care plans for treatment or survivorship. Finally, it may provide optimized billing codes and quality/outcomes reporting for programmatic assessments and regulatory mandates. While extraordinary investments have aimed to create AI applications for cancer risk prediction, diagnosis, and treatment, perhaps an equally noble goal would be to develop the NLP capabilities of cognitive systems to eradicate the growing mindless hours spent navigating the EHR, thereby allowing oncologists to do what they do best: provide the highest level of cutting-edge, patient-centered care to those facing cancer.

Dr. Haddad has no significant financial interest in or other relationship with the manufacturer of any product or provider of any service mentioned in this article

 Shanafelt TD, Gradishar WJ, Kosty M, et al. Burnout and career satisfaction among US oncologists. J Clin Oncol. 2014;32:678-86.

 Shanafelt TD, Dyrbye LN, Sinsky C, et al. Relationship between clerical burden and characteristics of the electronic environment with physician burnout and professional satisfaction. Mayo Clin Proc. 2016;91:836-48.

 Sinsky C, Colligan L, Li L, et al. Allocation of physician time in ambulatory practice: a time and motion study in 4 specialities. Ann Intern Med. 2016;165:753-60.

 Arndt BG, Beasley JW, Watkinson MD, et al. Tethered to the EHR: primary care physician workload assessment using EHR event log data and time-motion observations. Ann Fam Med. 2017;15:419-26.

In this review, we introduce the fundamentals of artificial intelligence and provide an overview of its current applications, pitfalls, and future directions in oncology.

Oncology Vol 33 No 2