Predictors of depressive symptomology in head and
neck cancer patients
Master Thesis
9-8-2016 B.J.F. de Haan10268413
Supervision: Prof. dr. I.M. Verdonck-de Leeuw Dr. D. Sent Drs. F. Jansen Dr. L. Minne Drs. M. Vuurboom
P
REDICTORS OF DEPRESSIVE SYMPTOMOLOGY IN HEAD AND NECK CANCER PATIENTS
S
TUDENT
B.J.F. de Haan (Ben), 10268413
Faculty of Medicine, Academic Medical Centre - University of Amsterdam Meibergdreef 9, 1105 AZ Amsterdam
Period: September 2015 – June 2016
M
ENTORS
,
VU
U
NIVERSITY MEDICAL CENTRE
:
Prof. dr. I.M. Verdonck-de LeeuwOtolaryngology / Head & Neck Surgery
im.verdonck@vumc.nl
Drs. F. Jansen
Otolaryngology / Head & Neck Surgery f.jansen1@vumc.nl
(+31) (0)20 444 09 31
PK2Y114, De Boelelaan 1118, Amsterdam
M
ENTORS
,
F
URORE
:
Drs. M. Vuurboomm.vuurboom@furore.com
From February 2016 to June 2016 Dr. L. Minne
From September 2015 to February 2016 (+31) (0)20 346 71 71
Bos en Lommerplein 280, Amsterdam
T
UTOR
,
AMC-U
V
A:
Dr. D. Sent
Department of Medical Informatics d.sent@amc.uva.nl
T
ABLE OF
C
ONTENTS
Preface and acknowledgements ... 4
Summary ... 5
Samenvatting (in Dutch) ... 6
Chapter 1: Introduction ... 7
Objective and research questions ... 7
Outline ... 8
Chapter 2: Background... 9
Head and neck cancer ... 9
The head and neck cancer department ... 9
Depression and depressive symptoms ... 11
OncoQuest ... 11
Chapter 3: Predictors of depressive symptomology and the course of depressive symptomology in head & neck cancer patients: a systematic review ... 14
Abstract ... 14
Introduction ... 15
Methods ... 15
Data collection and analysis ... 16
Results ... 16 Description of studies ... 18 Quality Assessment ... 18 Identified predictors... 18 Discussion ... 28 Conclusion ... 29
Chapter 4: Identifying predictors of depression: data analysis ... 32
Abstract ... 32 Introduction ... 33 Methods ... 33 Data sources ... 33 Patient selection ... 34 Pre-processing ... 35
Testing for selection bias ... 35
Statistical methods ... 35
Results ... 41
Exploratory data analysis ... 41
Train-test split ... 43
10-fold cross-validation ... 45
Zero-inflated Poisson 10-fold cross-validation ... 47
Discussion ... 49
Conclusion ... 50
Chapter 5: Discussion ... 52
Main Findings ... 52
Reflection on using data from the Research Data Platform ... 52
Strengths and weaknesses ... 53
Future research ... 53
Conclusion ... 53
Appendix ... 54
Appendix A: Search strategy ... 54
Pubmed search ... 54
CINAHL + PsycInfo search ... 54
Appendix B: Quality assessment scoring list ... 55
P
REFACE AND ACKNOWLEDGEMENTS
At the time of writing this thesis, I am a Medical Informatics student with a specific interest in data warehousing, data transformation and data analysis. Furore has been working on the development of data infrastructures in hospitals and the collaboration between hospitals on data infrastructures. This is what prompted me to work with Furore, who granted me the opportunity to actively participate in building a knowledge network within the Medical Intelligence department.Via Furore I got in contact with the Department of Otolaryngology / Head and Neck Surgery at VU University Medical Center (VUmc), where I was granted the opportunity to participate in a research group studying quality of life for patients suffering from head and neck cancer. The research group observed that a disproportionate amount of patients with head and neck cancer patients suffers from depressive symptoms. The question is thus whether these depressive symptoms could be predicted in order to possibly initiate preventive or early treatment of depressive symptoms.
There was a specific challenge in combining knowledge from previous studies, databases from previous research, and new databases in order to shine a new light on the subject. This, of course, was highly appealing to me and prompted me to research the subject. If you are academically schooled and interested in depression in (head and neck) cancer patients, or quality of life or depression in general, this thesis is reading material written for you.
I want to personally thank my supervisors: prof. dr. Verdonck-de Leeuw, dr. Sent, drs. Jansen, drs. Vuurboom, dr. Minne, and prof. dr. Abu-Hanna (SRP coordinator). Without the feedback, suggestions, and opportunities you provided, it would not have been possible to participate in research at this level. I also want to thank R. Wertheim and R. Mulders for giving me the opportunity to gain experience in the private sector. Finally, I want to thank my girlfriend, parents, sister, brother, and friends and family, for the continuous encouragement to put the capstone upon these years of study.
S
UMMARY
Many patients with head and neck cancer suffer from depression. It is important to detect any signs of depression as early as possible to effectively prevent an episode of clinical depression. The goal of this thesis is to aid the identification of head and neck cancer patients with a high risk of depression. The research question is therefore: what are predictors of depressive symptomology and the course of depressive symptomology? Additionally, we will reflect on using the Research Data Platform, which can be used to request cleaned, validated datasets to utilise in research.Head and neck cancer is cancer that is found in the paranasal sinuses, nasal cavity, oral cavity, pharynx and larynx. 22-57% of patients with head and neck cancer suffer from depressive symptoms, which is associated with a poor quality of life. The data on psychological distress, depressive symptoms, and quality of life is retrieved through questionnaires, i.e. patient-reported outcome measures (PROMs), like the Hospital Anxiety and Depression Scale (HADS), the EORTC Quality of life questionnaire C30, and the EORTC Head & Neck 35. Firstly, we conducted a systematic review for discovering predictors for a high risk of depression (or depressive symptoms) and the course of depressive symptomology. We searched three databases: Pubmed, PsycINFO and CINAHL. 22 articles were included, and 48 predictors were extracted. There is strong evidence that the presence of depressive symptomology at an earlier point in time positively predicts depressive symptomology later on. Karnofsky performance status, available emotional support, and the extent of the patient’s social network negatively predict depressive symptomology. There is inconclusive evidence for any predictors on the course of depressive symptomology.
Secondly, in a data analysis we investigated which baseline factors could be used to predict the course of depressive symptomology (as measured with the HADS-D scale) in head and neck cancer patients. To this end, data from 209 patients was analysed. The data was gathered prior to our research in the time period from January 2008 to June 2014. This resulted in a longitudinal dataset with multiple measurements of depressive symptoms per person. A model that can predict depression using a combination of variables was created. To achieve this, a generalised linear mixed model (GLMM) was made by selecting variables based on the Akaike Information criterion in a forward selection procedure and selecting the final model in 10-fold cross-validation. A zero-inflated Poisson model yielded the best model based on root mean squared error. Sticky saliva, worse quality of life, worse physical functioning, weight gain, longer time since treatment, higher levels of anxiety, and speech problems can be used to predict the course of depressive symptomology.
Given the Research Data Platform, pre-processing data took little time. Since GLMMs only support complete case analyses, we were unfortunately not able to train the final model on the desired population size. This might have biased the parameter estimates, and thus the generalizability of the model. If the final model can be validated, it can be directly implemented to predict the course of depression. In future research, other methods, such as methods from the field of machine learning, can be used to improve predictions. The results of this thesis aid medical decision making for patients suffering from head and neck cancer by a better understanding of predictors for depression and the course of depression.
S
AMENVATTING
(
IN
D
UTCH
)
Veel patiënten met hoofd-halskanker hebben last van symptomen van depressie. Het is belangrijk om signalen van depressie zo snel als mogelijk te kunnen herkennen, om met een vroege interventie te voorkomen dat de patiënt een episode van klinische depressie krijgt. Het doel van deze thesis is om bij te dragen aan het identificeren van patiënten met een hoog risico op depressie. De onderzoeksvraag luidt dus: wat zijn voorspellende factoren voor symptomen van depressie en het beloop van symptomen van depressie in hoofd-hals kankerpatiënten? Ook zullen we reflecteren op het gebruik van het Research Dataplatform.Hoofd-halskanker is een groep van kankersoorten in de bijholtes, de neus, de mond, de stembanden en het strottenhoofd. 22-57% van deze patiënten met hoofd-halskanker heeft last van symptomen van depressie. Meer specifiek, de aanwezigheid van veel symptomen van depressie. Onze data, en data in de systematische review, komt van door de patiënt gerapporteerde uitkomsten (PROMs). Veelgebruikte uitkomstmaten worden verzameld door middel van vragenlijsten als de Hospital Anxiety and Depression Scale (HADS), de EORTC QLQ-C30 (kwaliteit van leven) en de EORTC QLQ-H&N35 (hoofd-halsklachten).
Eerst hebben we een systematische literatuurreview uitgevoerd om voorspellers van symptomen van depressie en het beloop van symptomen van depressie te ontdekken. Voor de systematische review doorzochten we drie databases: Pubmed, PsycINFO en CINAHL. Uiteindelijk zijn er 22 artikelen geïncludeerd, waaruit 48 voorspellers naar voren kwamen. Er is sterk bewijs dat eerdere depressieve symptomen een voorspeller zijn voor latere depressieve symptomen, en dat Karnofsky performance status, de beschikbare emotionele steun en het sociale netwerk van de patiënt minder symptomen van depressie voorspellen. Er is onvoldoende bewijs voor voorspellers van het beloop van symptomen van depressie.
Om uit te zoeken welke variabelen gebruikt kunnen worden om het beloop van depressieve symptomen te voorspellen doen we een data-analyse met behulp van data die beschikbaar is vanaf het eerste consult. Van 209 patiënten is informatie verzameld over behandelingen en kwaliteit van leven in de periode 2008 tot 2014, met data van Radiotherapie, Keel- Neus- en Oorheelkunde en het Research Dataplatform. Met een methode voor gegeneraliseerde gemengde lineaire regressie (GLMM) kunnen we een model bouwen door te kijken naar het Akaike-informatiecriterium en kruisvalidatie. Het uiteindelijke model bevat de voorspellers plakkerig speeksel, kwaliteit van leven, fysiek functioneren, aankomen, tijd, angststoornissen en spraakproblemen. Deze voorspellers kunnen gebruikt worden om het beloop van symptomen van depressie te schatten.
Dankzij data van het Research Dataplatform kostte het schoonmaken en koppelen van data relatief weinig tijd. Het uiteindelijke model is niet gemaakt met de hoeveelheid patiënten die gewenst was omdat GLMM regressie alleen complete metingen ondersteunt. Dit kan een afwijking opleveren in de schattingen van coëfficiënten. Als het gemaakte model gevalideerd wordt, kan het direct ingezet worden om de mate van depressie bij specifieke patiënten te voorspellen. In de toekomst kunnen andere methoden, bijvoorbeeld uit machine learning, ingezet worden om dit nog beter te kunnen voorspellen. Deze thesis draagt bij aan de besluitvorming door patiënten met een hoog risico op depressie beter te kunnen identificeren.
C
HAPTER
1:
I
NTRODUCTION
It is estimated that 22-57% of head and neck cancer patients suffer from depressive symptoms (1;2). Cancer patients with depression or depressive symptoms have a higher mortality during the first 19 months after the start of treatment. It is unclear whether depression (mostly) stem from psychological, physical, or social factors (3). it is important to detect any depressive symptoms early to effectively prevent an episode of clinical depression (4).
Though prevalent and a predictor of mortality, there is a lack of screening for depression in oncological patients (5). To assist in such screening, OncoQuest has been developed at the department of Otolaryngology-Head and neck surgery of VU University Medical Center (VUmc) (5). OncoQuest is used to monitor quality of life and psychological distress in head and neck cancer patients before start of treatment and during follow-up visits. This monitoring is done through so-called patient-reported outcome measures (PROMs), which are gathered by patient self-assessment.
The PROMs that are measured using OncoQuest include the Hospital Anxiety and Depression scale (HADS)(6) measuring psychological distress and the European Organization for Research and Treatment of Cancer (EORTC) general (EORTC QLQ-C30)(7) and HNC-specific (EORTC QLQ-H&N35)(8) module measuring health-related quality of life. These PROMs (except the HADS) are measured at the department of Radiation Oncology before, during and after treatment with (chemo)radiation. Combining all these PROMs of patients provides valuable information into the course of health-related quality of life and depressive symptomology in head and neck cancer patients.
O
BJECTIVE AND RESEARCH QUESTIONS
The main goal of this study is to aid early detection of depression in head and neck cancer patients primarily treated with (chemo)radiation in order to be able to effectively prevent an episode of clinical depression. One way to achieve this is by finding variables that are a good indicator for future depressive symptoms, which we will refer to as predictors. We will thus try to achieve the goal by creating a predictive model and identifying predictors of depressive symptoms, and predictors of the course of depressive symptoms. We will investigate whether the course of depressive symptoms is predicted by socio-demographic (for example: gender and age) and clinical characteristics (for example: tumour location, tumour stage, and type of treatment), and quality of life (QoL). The research question is therefore: what are predictors of depressive symptomology and the course of depressive symptomology in head and neck cancer patients?
The first part of this research is to systematically review the literature in order to analyse what is known about the predictors of depression. The second phase includes analysing the patient data from the
Furore, a company that provides specialised Healthcare IT knowledge and manpower, is currently working on a Research Data Platform (RDP) at the VUmc, which researchers and other stakeholders can use to request cleaned, validated datasets to utilise in their research. The purpose of the RDP is to provide a trusted, future proof data warehouse where different sources are integrated. Firstly, data from multiple sources is added. The second phase integrates sources including a timestamp. This creates a traceable, auditable data warehouse.
One of the first projects of the RDP at the VUmc was to make data of head and neck cancer patients available. There is already data available on diagnoses (ICD-10), chemotherapy, radiotherapy, surgery, body mass index, and supportive care provided at the VUmc.
When a request for data arrives, data is cleaned and validated according to detailed clinical models. Detailed clinical models aim to describe a healthcare entity in a model that can be used for data storage, data exchange and data distribution. If similar data has already been generated, the previous script can be re-used and elaborated upon. Data is pseudonymised depending on the role of the customer. These detailed clinical models can then be extracted from the data warehouse in order to provide researchers, business intelligence employees and other stakeholders with traceable, auditable, cleaned, pseudonymised, and validated data.
To summarise, we aim to aid early detection of depression by:
- Gathering and merging PROMs data (e.g. OncoQuest data) with socio-demographic and clinical data obtained from sources of the department of Otolaryngology, Radiation Oncology, and the Research Data Platform.
- Conducting a systematic review to gain insight into depressive symptomology and socio-demographic and clinical factors, and QOL associated with depressive symptoms in head and neck cancer patients. - Finding and providing insight into the appropriate statistical technique to answer the research
question (e.g. mixed-model analyses) - Creating a statistical model
O
UTLINE
This thesis is organised as follows: the introduction contains the objective and background of the thesis (Chapter 2). In Chapter 3, the predictors of depressive symptoms and predictors of the course of depressive symptoms are described in a systematic review. The data analysis contains an analysis of head and neck cancer patients treated at the VUmc and are presented in Chapter 4. We end this thesis with the overall discussion and main findings, discuss use of the RDP in this research, and address whether the data analysis conforms to results found in the systematic review.
C
HAPTER
2:
B
ACKGROUND
H
EAD AND NECK CANCER
Head and neck cancer is cancer that is found in the paranasal sinuses, nasal cavity, oral cavity, pharynx and larynx (9). Head and neck cancer accounts for 6% of all cancer cases worldwide (9). In the Netherlands, 2869 patients were diagnosed with this form of cancer in 2015, about 3% of all cancer cases (10). Most patients are male and diagnosed in their early 60s (9). Risk factors of head and neck cancer are smoking, alcohol consumption, and Human Papillomavirus (HPV). The severity of the cancer is specified by tumour staging. A combination of T-value (tumour characteristics such as the growth into nearby tissues), N- value (spread to nearby lymph nodes) and M-value (haematogenous metastases) leads to an overall tumour stage (also: TNM-classification) (11). A higher stage means that a patient has a more advanced cancer, and so, is more likely to die. The combined five-year survival across all tumour stages is approximately 60% (9). We will address whether a higher stage is also an appropriate predictor for depression in both the systematic review and the data analysis.
Curative treatment mostly consists of surgery and radiation treatment (9). For small malignancies, laser surgery is often used. Some patients may also receive cytostatic drugs, which can be part of an induction chemotherapy. In this case, the patient will receive cytostatics before starting a radiation therapy or undergoing surgery. Induction chemotherapy reduces the probability of distant metastases (9). The combination of tumour sites and treatment options makes medical decision making about head and neck cancer complex and requires collaboration among different disciplines (Otolaryngology-Head & Neck surgery, Medical Oncology, and Radiation oncology). A traditional functional structure, where healthcare is organised around skills and facilities, no longer suffices in such an environment (12). Thus, an integrated practice unit (IPU), which is organised around a medical condition, is a more appropriate way of delivering care in this situation (12).
T
HE HEAD AND NECK CANCER DEPARTMENT
The Head and Neck department is an integrated practice unit within the VUmc Cancer Center Amsterdam. The VUmc Cancer Center Amsterdam is one of eight Head and Neck cancer centres within the Netherlands and one of two within Amsterdam (13). To run this IPU, weekly multidisciplinary meetings of Otolaryngology-Head and Neck surgery, Medical Oncology and Radiation Oncology are held on Thursdays. During these meetings they discuss patients with head and neck cancer. The department of Otolaryngology-Head and Neck surgery is responsible for diagnostics, surgery, and follow-up. Radiation oncology is responsible for creating and executing treatment plans, and follow-up. Medical oncology creates and executes chemotherapy treatment plans.
intake. Here, the patient meets their radiation oncologist and they discuss the patient’s medical history. The radiation oncologist explains the upcoming procedure. During the second appointment, a personal mask is created for the patient. Here, the patient is carefully positioned on a table before a medical laboratory technician creates a mould. The patient will wear this mould during radiation treatment, where it will aid in positioning the patient for upcoming radiation treatment. The last meeting on the first day is a CT-scan. This CT-scan is used to create the radiation treatment plan. At the next meeting, radiation treatment is started. Every follow-up meeting, patients are asked to fill out a questionnaire provided by OncoQuest.
D
EPRESSION AND DEPRESSIVE SYMPTOMS
Depression, or a depressive disorder is a disorder where a patient suffers from a “sad, empty, or irritable mood, accompanied by somatic and cognitive changes that significantly affect the individual’s capacity to function” (14). Depression is part of the ‘depressive disorders’ in the DSM-V, a manual for diagnosing mental disorders. The exact sub-category of depression depends on duration, timing and cause (14;15). The diagnosis involves assessing the presence of depressive symptoms. We distinguish between clinical depression and depressive symptoms: depressive symptoms are often patient reported, e.g. by results of a questionnaire like the Hospital Anxiety and Depression scale. Clinical depression, on the other hand, is diagnosed by a physician. It is important to state that when depression is mentioned in this thesis, this should be interpreted as a high risk of depression because of a high number of depressive symptoms.
The classical form of depression is an episode of ‘major depressive disorder’ (14). To be diagnosed with major depressive disorder, the patient has to experience five or more symptoms of the following during a two week period: depressed mood most of the day, diminished interest or pleasure in activities most of the day, significant weight loss when not dieting or weight gain, insomnia or hypersomnia nearly every day, psychomotor agitation or retardation nearly every day, fatigue or loss of energy nearly every day, feelings of worthlessness or excessive or inappropriate guilt, diminished ability to think or concentrate, or indecisiveness nearly every day, recurrent thoughts of death, recurrent suicidal ideation, or a suicide attempt (14).
O
NCO
Q
UEST
OncoQuest measures three aspects (5;16): anxiety and depression through the Hospital Anxiety and Depression Scale (HADS), generic quality of life through the EORTC Quality of Life Questionnaire-C30 (QLQ-C30)(7), and head and neck cancer specific symptoms through the EORTC Head & Neck-35 (H&N-35). Measurement points include baseline (before treatment), and approximately three months, six months, nine months, twelve months, eighteen months, three years and five years after treatment.
H
OSPITALA
NXIETY ANDD
EPRESSIONS
CALEThe Hospital Anxiety and Depression scale (HADS), measures psychological distress in hospital patients. It has shown to be a reliable and valid tool for screening depressive symptomology (6;17;18). The Hospital Anxiety and Depression Scale (HADS) has two subscales: HADS-A for anxiety (7 items) and HADS-D for depression (7 items) (6). Together the subscales form a measure for psychological distress. The depression subscale has a cut-off point of 7, which indicates that the patient shows depressive symptoms and has a higher chance of being depressed. The HADS-D is useful as a screening tool, but cannot be used to diagnose clinical depression (19), partly because diagnosis depends on an assessment of cause, it involves assessment over time and an assessment of duration (14). Additionally, the Case-Finding utility (UI+), or positive utility, is very poor (0.27) (UI+ = sensitivity * PPV) (19;20). The Screening utility (UI-) , or negative utility, is good (0.79).
EORTC
QLQ-H&N-35
The EORTC QLQ-H&N35 module covers specific head and neck cancer issues and comprises 7 subscales: pain (4 items), swallowing (5 items), senses (2 items), speech (3 items), social eating (4 items), social contact (5 items) and sexuality (2 items). There are 10 single items covering problems with teeth, dry mouth, sticky saliva, cough, opening the mouth wide, weight loss, weight gain, use of nutritional supplements, feeding tubes, and painkillers. The scores of the QLQ-H&N35 are linearly transformed to a scale of 0-100, with a higher score indicating a higher (i.e., more positive) level of functioning or global HRQOL, or a higher (i.e., more negative) level of symptoms or problems (7;8).
Reference List
(1) Massie MJ. Prevalence of depression in patients with cancer. 2002 p. 29.
(2) Krebber AMH, Buffart LM, Kleijn G, Riepma IC, Bree R, Leemans CR, et al. Prevalence of depression in cancer patients: a metaanalysis of diagnostic interviews and selfreport instruments. Psycho-Oncology 2014;23(2):121-30.
(3) Satin JR, Linden W, Phillips MJ. Depression as a predictor of disease progression and mortality in cancer patients. Cancer 2009;115(22):5349-61.
(4) Hollon SD, Muñoz RF, Barlow DH, Beardslee WR, Bell CC, Bernal G, et al. Psychosocial intervention development for the prevention and treatment of depression: promoting innovation and increasing access. Biological psychiatry 2002;52(6):610-30.
(5) Verdonck-de Leeuw IM, de Bree R, Keizer AL, Houffelaar T, Cuijpers P, van der Linden MH, et al. Computerized prospective screening for high levels of emotional distress in head and neck cancer patients and referral rate to psychosocial care. Oral Oncol 2009 Oct;45(10):e129-e133.
(6) Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta psychiatrica scandinavica 1983;67(6):361-70.
(7) Aaronson NK, Ahmedzai S, Bergman B, Bullinger M, Cull A, Duez NJ, et al. The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology. Journal of the national cancer institute 1993;85(5):365-76.
(8) Bjordal K, Kaasa S. Psychometric validation of the EORTC Core Quality of Life Questionnaire, 30-item version and a diagnosis-specific module for head and neck cancer patients. Acta Oncol 1992;31(3):311-21.
(9) Argiris A, Karamouzis MV, Raben D, Ferris RL. Head and neck cancer. The Lancet 2008;371(9625):1695-709.
(10) IKNL. Incidence of Head and Neck Cancer in the Netherlands. IKNL, editor. 4-2-2016. Ref Type: Online Source
(11) Sobin LH, Gospodarowicz MK, Wittekind C. TNM classification of malignant tumours. John Wiley & Sons; 2011.
(12) Porter ME, Teisberg EO. Redefining health care: creating value-based competition on results. Harvard Business Press; 2006.
(13) Nederlandse Werkgroep Hoofd-Hals Tumoren. Hoofd-Hals Oncologische centra. 6-8-2010. 27-6-2016. Ref Type: Online Source
(14) American Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-5-«). American Psychiatric Pub; 2013.
(15) Klein DF. Endogenomorphic depression: a conceptual and terminological revision. Archives of General Psychiatry 1974;31(4):447.
(16) de Bree R, Verdonck-de Leeuw IM, Keizer AL, Houffelaar A, Leemans CR. Touch screen computer assisted healthrelated quality of life and distress data collection in head and neck cancer patients. Clinical Otolaryngology 2008;33(2):138-42.
(17) Bjelland I, Dahl AA, Haug TT, Neckelmann D. The validity of the Hospital Anxiety and Depression Scale: an updated literature review. Journal of Psychosomatic Research 2002;52(2):69-77.
(18) Moorey S, Greer S, Watson M, Gorman C, Rowden L, Tunmore R, et al. The factor structure and factor stability of the hospital anxiety and depression scale in patients with cancer. The British Journal of Psychiatry 1991;158(2):255-9.
(19) Mitchell AJ, Meader N, Symonds P. Diagnostic validity of the Hospital Anxiety and Depression Scale (HADS) in cancer and palliative settings: a meta-analysis. Journal of Affective Disorders 2010;126(3):335-48.
(20) Mitchell AJ. HOW DO WE KNOW WHEN A SCREENING TEST IS CLINICALLY USEFUL? Screening for Depression in Clinical Practice: An Evidence-Based Guide 2009;99.
(21) Fayers P, Bottomley AEOR, EORTC Quality of Life Group. Quality of life research within the EORTCthe EORTC QLQ-C30. European Journal of Cancer 2002;38:125-33.
(22) Archer JA, Hutchison IL, Dorudi S, Stansfeld SA, Korszun A. Interrelationship of depression, stress and inflammation in cancer patients: A preliminary study. Journal of Affective Disorders 2012 Dec 20;143(1-3):39-46.
(23) Arlot S, Celisse A. A survey of cross-validation procedures for model selection. Statistics surveys 2010;4:40-79.
(24) Hern+índez MA, Stolfo SJ. Real-world data is dirty: Data cleansing and the merge/purge problem. Data mining and knowledge discovery 1998;2(1):9-37.
C
HAPTER
3:
P
REDICTORS OF DEPRESSIVE SYMPTOMOLOGY AND THE
COURSE OF DEPRESSIVE SYMPTOMOLOGY IN HEAD
&
NECK CANCER PATIENTS
:
A SYSTEMATIC REVIEW
A
BSTRACT
Background: The incidence of depressive symptoms among head and neck cancer (HNC) patients ranges from 22% to 57% and is higher than in most other cancer patients. It is important to gain insight into the predictors for depressive symptoms and the course of depressive symptomology.
Objectives: To systematically review literature on predictors for depressive symptoms and the course of depressive symptoms in head and neck cancer patients.
Search methods: A literature search was conducted in Pubmed, PsycINFO, and CINAHL. The search was executed to include articles available in the database up to and including 2015-11-10.
Selection criteria: Studies were included if they included a group of adult head and neck cancer patients, had depression or depressive symptomology as outcome, listed predictors of depression or depressive symptomology, were of a prospective, longitudinal nature and when full-text was available in English.
Quality assessment and data collection: Two reviewers screened the articles; first on title and abstract, then full text. Included studies were subjected to a quality assessment. Predictors for either depressive symptoms or the course of depressive symptomology were then extracted and level of evidence was determined.
Main results: Twenty-two studies were included, of which 9 on the course of depression. In total, 48 predictors were extracted: eleven socio-demographic, fourteen clinical, and 23 patient-reported outcome measures. There is strong evidence that a high number of depressive symptoms at an earlier point in time predict a higher number of depressive symptoms in follow-up. There is also strong evidence that gender, education, and age do not predict depressive symptoms. Level of evidence was inconclusive for 32 other potential predictors of depressive symptoms. There is moderate evidence that age and concurrent chemotherapy are no predictors for the course of depressive symptomology. For all other predictors on the course of depressive symptomology, the level of evidence was inconclusive.
Author’s conclusions: Depressive symptoms at an earlier point in time positively predicts depressive symptoms later on. A more extensive social network, a better performance status and more available support were found to predict fewer depressive symptoms. Gender, age, education, treatment, tumour location, anxiety, physical functioning, and poor sleep quality were not found to predict fewer or more depressive symptoms. For all other variables, the evidence is inconclusive. These results partially contradict the findings of an earlier review (1). The patient’s history of depressive symptomology, their performance status, the extent of their social network, and their available support should be taken into account in decision making for psychological care. Further research on the found predictors is needed to determine their importance.
I
NTRODUCTION
Cancer patients who report depressive symptoms have a 29% higher probability to die (Relative risk (RR): 1.29, 95% Confidence Interval (CI): 1.12-1.40)(2), and patients who are diagnosed with a depressive disorder have a 39% higher probability to die (RR: 1.39, 95% CI: 1.10-1.89) (2). Depressive symptoms are reported by the patient (via a questionnaire), while a depressive disorder is diagnosed by a physician. The prevalence of depressive symptoms among head and neck cancer patients (22-57%) is higher than in most other cancer patients. For instance: lymphoma (8-19%), colon cancer (13-25%), and gynaecological cancer (12-23%) (3;4). The higher prevalence might be due to the cancer itself as well as its treatment. Both may cause head and neck cancer-specific symptoms, such as facial disfigurement, or difficulty speaking, eating, and swallowing. However, the association between these different factors and depressive symptoms is unclear (3-5).
Haisfield-Wolfe et al. (1) conducted a literature review in 2009, concerning factors associated with depression in head and neck cancer patients. They found that male sex, younger age, a lower education, decreased social support, smoking, unemployment, and being unmarried or living alone were found to be positively associated with depression. Furthermore, depression at the start of or before treatment, a high number of comorbidities, higher disease stage, a bigger tumour, and lower physical functioning were also reported to be associated with depression. The course of depression was out of scope for the literature review.
Since the review of Haisfield-Wolfe et al. in 2009 several new studies have been performed on predictors of depression in head and neck cancer patients, warranting an update of current the literature. In addition, more insight is needed regarding the predictive factors so that high-risk patients can be identified. In this study, we therefore aim to systematically review the predictors of depression. This leads us to two research questions: 1) what are predictors for depression in head and neck cancer patients? 2) what are predictors of the course of depression in head and neck cancer patients?
M
ETHODS
We conducted a literature search in PubMed, PsycINFO, and CINAHL using keywords, MeSH terms and subject headings. The most important keywords are: ‘Head and Neck Neoplasms’, ‘Depression’, ‘Anxiety’, ‘Distress’, ‘Quality of Life’, ‘emotional functioning’, ‘psychological functioning’, ‘associat*’, and ‘correlat*’. The complete queries are available in Appendix A. Duplicates from the search were identified through Reference Manager and by hand search. Identified duplicates were removed.
Studies were included if they met all of the following criteria: 1) included a group of adult head and neck cancer patients, 2) had depression as outcome, 3) reported on predictors of depression or predictors of the course of depression, 4) were of a prospective, longitudinal nature 5) full text was available in English and 6) were not randomised controlled trials. The latter is due to the fact that the effects of an intervention on depression were outside the scope of this study.
Article selection took place in two phases. In the first phase, article title and abstract were screened on eligibility and marked for further evaluation or were excluded. In the second phase, the full text of the articles that were marked for further evaluation were assessed for eligibility based on the inclusion and exclusion
analysis. All questions were scored positive (score ‘1’), or negative or unclear (score ‘0’). A total score per study was calculated by summing the scores resulting in a score ranging from 0 to 15. Validity and reliability of predictors of depressive symptoms could be scored anything between 0 and 1, rounded to two decimal places. If, for instance, a study had three determinants and one of them would be unreliable, the score would be 0.67. For some studies, articles on the validity and reliability were not retrievable. Therefore, they have been given a rating of 0 for any variables.
Any study scoring ≥70% of points (i.e. ≥10.5 points) would be categorised in the ‘high methodological quality’ group. Studies scoring <70% would be categorised in the ‘low methodological quality’ group (6;10). No sub-score per aspect was used (6).
D
ATA COLLECTION AND ANALYSIS
From all included articles, the following data were extracted: year of publication, study population, depression measuring instrument, measurement points (number and timeframe), used questionnaires (if any), and analysed potential predictors.
Variables that influence the probability of getting depressed, called predictors, were gathered from the selected articles. These potential predictors were categorized in two different classes: one for predicting depression and another for predicting the course of depression. Studies that had used a repeated measures analysis technique for identifying predictors were categorized in the latter category. Variables from studies that used (multivariate/multiple) regression without repeated measures techniques were categorized as predicting depression. Studies without repeated measures analysis generally use a baseline measurement to predict the follow-up depression score (and possibly dichotomize the depression outcome).
Since a multivariate technique corrects for confounders or estimates depression using multiple variables, while univariate techniques estimate depression by using one variable. Multivariate results were regarded as superior to univariate results, since multivariate results are more likely to contain possible confounding factors. Therefore, if both multivariate and univariate analyses were used, the multivariate results were collected. If both non-repeated measures techniques and repeated measures analysis were used, both types of variables were collected and categorized.
Level of evidence
Evidence for a predictor can be classified in a certain level of evidence, depending on the quality of studies supporting it (10). We used a best-evidence synthesis as defined in Hayden et al., 2006 (6). The levels are: 1) strong evidence, if a predictor was consistently supported by at least two high-quality studies, 2) moderate evidence, if a predictor was consistently supported by at least one high-quality study and at least one low-quality study, or if a predictor was consistently supported by at least two low-low-quality studies, and 3) inconclusive evidence, if a predictor was supported by only one study or results were inconsistent in multiple studies. A result is defined as consistent if at least 75% of supporting studies reported results in the same direction (11).
R
ESULTS
The literature search of PubMed yielded 621 studies, and the PsycINFO and CINAHL search yielded 72 studies (see Figure 1). Screening for duplicates revealed 39 duplicate studies. Duplicates were excluded. A first screening based on title and abstract with two reviewers resulted in the exclusion of 533 studies. 417 studies did not list predictors for depression, 409 studies were not longitudinal or prospective, 301 studies did not have depression as outcome and 292 studies did not include a group of adult head and neck cancer patients. After assessing the remaining studies on eligibility based on full text, 99 studies were excluded. 36 studies were not
longitudinal, 27 studies did not list predictors of depression and in 24 studies depression was not used as outcome variable (see Figure 1).
D
ESCRIPTION OF STUDIES
In Table 1, the characteristics of the studies are described. The studies used different measurement instruments for depressive symptoms: thirteen studies used Hospital Anxiety and Depression scale (HADS-D) (59%) (14;15;19-21;22-29) , five studies used Center for Epidemiologic Studies Depression scale (CES-D) (23%) (12;13;16;17;33), two studies used the Beck Depression Inventory (BDI) (10%) (18;20), two studies used the Symptom Checklist (SCL-90) (10%) (30;32), and one study used the Self-Rating Depression scale (SDS) (5%) (31). 15 studies reported on predictors of depression (12;15-20;23-27;29;31;32) and nine studies used a repeated measurements analysis and were categorised as predicting course of depression (13;14;19;21;22;27;28;30;33). Publication dates of included studies ranged from 1987 to 2015. Eighteen studies had a western study country. Six studies were performed in (partly) the same population. The studies of De Graeff et al. (12;13), and De Leeuw et al. (16;17) all used data from the same population. Neilson et al. (14;15) similarly studied the same population in their two included studies.
Q
UALITY
A
SSESSMENT
In Table 2, the quality assessment results are depicted. A column reflects the results of a study, while rows describe the different items. The ‘total’ column on the outer right shows the quality per item across studies (max = 22). The ‘score’ row at the bottom shows the score per study (max = 15). There are seven high-quality studies (13;22-25;31;32). Predictors from high-quality studies weighed more in assessing level of evidence. All studies scored positive on the provision of exact information on follow-up duration. Most studies (20/22) had also described their in- and exclusion criteria (12;13;16-25;28-33). The majority of studies did not have a baseline participation rate of at least 80% (12;14;15;18-20;23-27;29;33). In four studies non-response was not selective (13;16;17;22). Five studies (23%) reported point estimates with measures of variability (14;17;20;31;32). Five studies scored negative on the sampling frame because they did not mention the time period of recruitment (17;18;20;22;26). Four studies scored negative on the response rate at first follow-up because they completely excluded patients who were lost to follow-up (12;13;17;32).
I
DENTIFIED PREDICTORS
From the included studies, 48 predictors were identified. Eleven predictors were socio-demographic, fourteen were clinical, and 23 were patient-reported outcome measures on psychological functioning or quality of life (see Table 4). A predictor’s total N equals the number of studies in which it was used in analysis. N+ refers to the number of studies in which a predictor positively predicted depression. N- refers to the number of studies in which a predictor negatively predicted depression, and N0 refers to the number of studies in which a predictor was found to have no predictive value. The ‘positive’, ‘negative’ and ‘none’ columns list the studies that support the N+, N- and N0, respectively. If within one study contradictory results were found, the result was added in both columns (e.g. Female gender - De Graeff 6m, vs 12m) (12). If this was the case, the sum of N+, N-, and N0 exceeds the total N. Gender (n=11), age (n=11), disease stage (n=9) , education (n=4), and concurrent chemotherapy (n=4) were the most often researched predictors of depression. Concurrent chemotherapy (n=3), gender (n=2), and age (n=2) were the most often researched predictors of the course of depression.
Level of evidence
There was strong evidence that depression at an earlier point in time predicts depression in the follow-up period (see Table 3). There was also strong evidence that gender, age, and education do not predict depression. For treatment, tumour location, anxiety, physical functioning, and poor sleep quality there was moderate evidence that they do not predict depression. Better performance status, more available support and the extent of the social network all had a moderate level of evidence for negatively predicting depression.
For 9 predictors of depression, results were inconsistent. For the remaining 23 predictors, there was insufficient evidence.
There was no strong evidence for any predictor on the course of depression (see Table 4). All other variables had an inconclusive level of evidence. Female gender has inconsistent evidence (13;14). The other fifteen variables have insufficient evidence.
TABLE 1: CHARACTERISTICS OF INCLUDED STUDIES
First Author Year Population (n) Depression measuring instrument
Course of depression
Measurement points 1 Predictors Study locale
Aarstad, HJ (18) 2005 61 control, 79 HNC. 27 HNC in analysis BDI No 2: Pre-treatment, 6y follow-up
Anxiety, depression, pre-treatment depression, sense of humour Norway Archer, JA (19) 2012 HNC (56),
colectoral (34)2
HADS-D Yes 4: Pre-treatment, 6w, 12w, 24w post-treatment
Childhood trauma, Inflammatory markers (IL6, TNF-α, CRP, IFN-γ), Number of recent life events
UK
Chen, AM(20) 2009 40 Non-Metastized HNC HADS-D, BDI-II
No 3: Pre-treatment, last day of treatment, 3w post-treatment
Age, concurrent chemotherapy, education level, employment status, gender, income, living alone, marital status, treatment depression (BDI), pre-treatment depression (HADS-D), previous surgery, smoking history, toxicity, tumour stage
USA
Chen, SC (21) 2010 76: Oral cavity cancer treated with postoperative radiotherapy or
chemoradiotherapy
HADS-D Yes 4: Pre-treatment, 1m, 2m, 3m follow-up
Concurrent chemotherapy Taiwan
De Graeff, A (12) 2000 153: HNC treated with curative intent CES-D No 3: Pre-treatment, 6m, 12m follow-up
Age, gender, depression in follow-up, performance status (Karnofsky), previous surgery The Netherlands De Graeff, A (13) 2000 107: HNC treated with surgery and/or radiotherapy with curative intent
CES-D Yes 5: Pre-treatment, 6m, 1y, 2y, 3y follow-up
Age, concurrent chemotherapy, gender, previous surgery, tumour stage The Netherlands De Leeuw, JR (17) 2000 155: HNC treated with surgery or radiotherapy CES-D No 3: Pre-treatment, 6m, 1y follow-up
Age, available support, extent of social network, pre-treatment depression, general health symptoms, higher tumour stage, gender, head and neck related symptoms, physical functioning, gender, locus of control, positive coping behaviour, negative coping behaviour, treatment, received support
The Netherlands
De Leeuw, JR(16)
2001 197: HNC treated with surgery and/or radiotherapy with curative intent
CES-D No 5: Pre-treatment, 6m, 1y, 2y, 3y follow-up
Age, gender, tumour stage, received support, available support, extent of social network, pre-treatment depression, avoidance coping, general health symptoms, physical functioning, tumour related symptoms, openness to discuss cancer, treatment, recurrence, religious control
The Netherlands
Finizia, C(22) 2002 26: Laryngeal cancer HADS-D Yes 6: Pre-treatment, 1m, 2m, 3m, 6m, 1y post-treatment
Communication dysfunction (S-SECEL) Sweden
Hammerlid, E(23)
1999 357: HNC HADS-D No 6: Pre-treatment, 1m, 2m, 3m, 6m, 1y follow-up
Age, gender, performance status (Karnofsky), tumour stage Sweden, Norway Hammerlid,
E(24)
2001 232: HNC HADS-D No 6: 6 times in year 1, 3y Age, gender, tumour location, tumour stage Sweden, Norway
1
y = years, m = months, w = weeks, d = days 2
Hassel, AJ(25) 2012 35: Oral squamous cell HADS-D No 2: 39d-165d (T1), T1 + 1y
Oral health impairment (HNC related symptoms) Germany Humphris,
GM(26)
2003 87 HNC HADS-D No 2: 3m, 7m post-treatment
Age, gender, radiotherapy, tumour stage, UK Humphris,
GM(27)
2004 87: Oral and oropharyngeal HADS-D Yes 4: 3m, 7m, 11m, 15m post-treatment
Current smoker UK
Kobayashi, M(28)
2008 58: HNC treated with surgery HADS-D Yes 4: Pre-treatment, 7d-10d, 6w, 6m
Self esteem Japan
Llewellyn, CD(29)
2007 49: HNC HADS-D No 3: Pre-treatment, 1m, 6m-8m
Age, ethnicity, gender, Timeline (IPQ), self-blame (IPQ), identity (IPQ), cyclical timeline (IPQ), consequences (IPQ), personal control (IPQ), coherence (IPQ), emotional representation (IPQ), beliefs about medicine, satisfaction with information, optimism, positive coping behaviour, negative coping behaviour, tumour stage, marital status, education level, tumour location, treatment
UK Manuel, GM(30) 1987 35: HNC SCL-90 Yes 3: Pre-treatment, 4w-6w, 2m-3m Coping USA
Mo, YL(31) 2014 51: Nasopharyngeal cancer, treated with primary IMRT
SDS No 2: 1w pre-treatment, 1w post-treatment
Age, gender, tumour stage, education level, concurrent chemotherapy, pre-treatment depression, pre-pre-treatment anxiety, poor sleep quality (Pittsburgh Sleep Quality Index)
China
Neilson, KA(14) 2013 101: HNC treated with radiotherapy
HADS-D Yes 3: Pre-treatment, 3w, 18m
Age, gender, living alone, pain, concurrent chemotherapy, severity of physical symptoms
Australia Neilson, KA(15) 2010 102: HNC treated with
radiotherapy (75 in analysis)
HADS-D No 2: Pre-treatment, post-treatment (average 12.8w)
Age, gender, living alone, alcohol use, previous surgery, concurrent chemotherapy, pre-treatment depression
Australia
Qin, L(32) 2015 60: Local-advanced nasopharyngeal, completed radiotherapy and concurrent chemotherapy
SCL-90 No 2: Pre-treatment, post-treatment
Age, gender, education level, tumour stage, concurrent chemotherapy, poor sleep quality (Pittsburgh Sleep Quality Index)
China
Rhoten, BA(33) 2014 43: HNC CES-D Yes 4: Pre-treatment, post-treatment, 6w, 12w follow-up
Body image USA
TABLE 2: QUALITY ASSESSMENT (1 = POSITIVE, 0 = NEGATIVE) Study A ar st ad , 2005 (18) A rc he r, 2012 (5;19) C he n , 2009 (2 0) C he n SC , 2 010 (21) de G rae ff , 2000 (H aN )( 12) de G rae ff , 2000 ( TL) (13) de L ee uw , 2000 (17) de L ee uw , 2001 (16) Fi ni zi a, 2002 (22) H am m er lid , 1999 (23) H am m er lid , 2001 (24) H as se l, 2012 (25) H um ph ri s, 200 3 (26) H um ph ri s, 200 4 (27) Kob ay as hi M , 2008 (28) Ll ewe lly n, 2007 (29) M an ue l, 1987 (30) Mo , 2014 (31) N ei lso n , 2010 (15) N ei lso n , 2013 (14) Qi n , 2015 (32) R ho te n , 2014 (33) To tal
Study population and participation The sampling frame and recruitment are adequately
described (setting and geographical location)
0 1 1 0 1 1 0 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 15
Description of inclusion and
exclusion criteria 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 20
Positive if the participation rate at baseline was at least 80%, or if the non-response was not selective (clarify)
0 0 0 1 0 12 12 12 12 0 0 0 0 0 1 0 1 1 0 0 1 0 9
Adequate description of baseline study sample for
general characteristics (age, gender, cancer type, stage and treatment)
0 0 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 1 0 0 1 1 14
Study attrition
Provision of the exact number of participants at each follow-up measurement
1 1 1 1 0 0 0 1 1 1 1 1 1 0 0 1 0 1 1 1 1 1 16
Provision of exact information on
follow-up duration 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 22
Number of patients included in the
analysis > 100 0 0 0 0 1 1 1 1 0 1 1 0 0 0 0 0 0 0 0 1 0 0 7
Positive if the response at first
follow-up was at least 80%, or if 0 1
3 1 1 0 0 1 1 1 1 0 1 1 1 0 1 1 1 0 0 0 1 14
the non-response at first follow-up was not selective (clarify) Data collection
Positive if determinants of depression were measured with a reliable tool4
0.67 1 0.7 1 1 1 0.75 0.38 1 1 1 1 1 1 0 0.46 0 1 1 0 1 1 17 Positive if determinants of
depression were measured with a valid tool
0.67 1 0.7 1 1 1 0.63 0.13 0.5 1 1 1 1 1 0 0.46 0 1 1 0 1 1 16.1 Depression was measured by a
reliable tool 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 20
Depression was measured by a
valid tool 1 1 1 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 0 1 14
Data analysis
Multivariate analysis techniques
were used 0 1 0 0 1 1 1 1 0 0 0 0 0 0 0 1 0 1 1 1 1 0 10
Results were presented as point estimates
(mean
differences/Betas/correlation coefficients)
and measures of variability (SD, standard error or CI)
0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 5
Positive if number of samples is at least
10 times the number of independent variables
0 0 0 1 1 1 1 0 1 1 1 1 1 1 1 0 1 0 1 1 1 1 16
TABLE 3: PREDICTORS OF DEPRESSION AND EVIDENCE
(ASSOCIATION: + (POSITIVE), - (NEGATIVE), ? (INCONCLUSIVE), NONE (NO ASSOCIATION)
Predictors of depression
Socio-demographic N N+ Positive N- Negative N0 None Level of evidence Association
Female gender 11 3 (12 (6m); 16 (2y, 3y); 17
(6m))
0 11 (12 (12m); 15; 16 (6m, 1y); 17 (12m); 20; 23; 24; 29; 31; 32)
Strong None
Higher age 11 0 2 (20 (HADS-D);31) 10 ( 12; 15; 16; 17;20 (BDI-II); 23; 24;
26; 29; 32)
Strong None
Higher education 4 0 0 4 (20; 29; 31; 32) Strong None
Being married 2 0 1 (20) 1 (29) Inconclusive ?
Living alone 2 1 (20) 0 1 (15) Inconclusive ?
Being employed 1 1 (20) 0 Inconclusive ?
Higher income 1 0 0 1 (20) Inconclusive ?
Ethnicity 1 0 0 1 (29) Inconclusive ?
Alcohol use 1 0 0 1 (15) Inconclusive ?
Current smoker 1 1 (27 (7m, 11m, 15m)) 0 1 (27 (3m)) Inconclusive ?
Smoking history 1 0 0 1 (20) Inconclusive ?
Clinical N N+ Positive N- Negative N0 None Level of evidence
Association
Higher disease stage 9 3 (16 (1y, 3y); 17; 24) 0 7 (16 (6m, 2y); 20; 23; 26; 29; 31; 32) Inconclusive ?
Concurrent chemotherapy 4 2 (15;32) 0 2 (20;31) Inconclusive ?
Previous surgery 2 0 0 2 (15;20) Inconclusive ?
Group 1 1 (12) Inconclusive ?
Treatment 3 0 0 3 (16;17;29) Moderate None
Performance status (Karnofsky) 2 0 2 (12;23) Moderate Negative Radiotherapy 1 0 0 1 (26) Inconclusive ? Recurrence 0 0 1 (16) Inconclusive ? Toxicity 1 0 0 1 (20) Inconclusive ?
TNF-α 1 0 0 1 (19) Inconclusive ? IL-6 1 0 0 1 (19) Inconclusive ? CRP 1 0 0 1 (19) Inconclusive ? IFN-γ 1 0 0 1 (19) Inconclusive ? Patient-reported outcomes
N N+ Positive N- Negative N0 None Level of evidence
Association
Depression at earlier point in time
7 6 (12; 15; 16; 17; 20; 31) 0 1 (18) Strong Positive
Negative coping behaviour 3 2 (16 (avoidance); 29
(self-blame))
0 1 (17) Inconclusive ?
Positive coping behaviour (acceptance, seeking support)
3 1 (29) 0 2 (17) Inconclusive ?
HNC-related symptoms (OHIP, EORTC QLQ-H&N 35, FACT H&N)
3 2 (16 (6m, 1y); 25) 0 2 (16 (2y, 3y); 17) Inconclusive ?
General health symptoms (EORTC QLQ-30)
2 2 (16 (3y); 17) 0 1 (16 (6m, 1y, 2y)) Inconclusive ?
Anxiety 2 0 0 2 (18;31) Moderate None
Number of recent life events
2 1 (19 (6w, 24w)) 0 1 (19(12w)) Inconclusive ?
Received support 2 0 1 (16) 1 (17) Inconclusive ?
Available support 2 0 2 (16;17) 0 Moderate Negative
Social network 2 0 2 (16;17) 0 Moderate Negative
information content)) Optimism 1 0 0 1 (29) ? Communication dysfunction 0 0 0 0 Inconclusive ? Pain 0 0 0 0 Inconclusive ?
TABLE 4: PREDICTORS OF THE COURSE OF DEPRESSION AND EVIDENCE
(ASSOCIATION: + (POSITIVE), - (NEGATIVE), ? (INCONCLUSIVE), NONE (NO ASSOCIATION)
Predictors of the course of depression
Socio-demographic N N+ Positive N- Negative N0 None Level of evidence Association
Female gender 2 1 (13) 0 1 (14) Inconclusive ?
Higher age 2 0 0 2 (13;14) Moderate None
Living alone 1 0 0 1 (14) Inconclusive ?
Current smoker 1 1 (27) 0 0 Inconclusive ?
Clinical N N+ Positive N- Negative N0 None Level of evidence Association
Higher disease stage 1 0 0 1 (13) Inconclusive ?
Concurrent chemotherapy 3 0 0 3 (13;14;21) Moderate None
Previous surgery 1 0 0 1 (13) Inconclusive ?
TNF-α 1 0 0 1 (19) Inconclusive ?
IL-6 1 0 0 1 (19) Inconclusive ?
CRP 1 0 0 1 (19) Inconclusive ?
IFN-γ 1 0 0 1 (19) Inconclusive ?
Patient-reported outcomes N N+ Positive N- Negative N0 None Level of evidence Association
Positive coping behaviour (acceptance, seeking support)
1 0 1 (30) 0 Inconclusive ?
HNC-related symptoms (OHIP, EORTC QLQ-H&N 35, FACT H&N)
1 1 (14) 0 0 Inconclusive ?
Number of recent life events 1 1 (19) 0 0 Inconclusive ?
Childhood trauma 1 0 0 1 (19) Inconclusive ?
Communication dysfunction 1 1 (22) 0 0 Inconclusive ?
D
ISCUSSION
In this systematic review we aimed to identify predictors of depressive symptoms and predictors of the course of depressive symptomology in head and neck cancer patients. Our results included 22 studies discussing 48 different potential predictors for depressive symptoms in patients with head and neck cancer. 26 predictors with a moderate or strong level of evidence for depressive symptoms were identified. There is strong evidence that depressive symptoms at an earlier point in time predict depressive symptoms at a later stage. For education, age, and gender, strong evidence was found that they do not predict depressive symptoms. No strong evidence was found for any predictor on the course of depressive symptomology.
Moderate evidence was found that good performance status, more available emotional support and the extent of the patient’s social network predict fewer depressive symptoms. For treatment, tumour location, previous surgery, anxiety, physical functioning and sleep quality, there is moderate evidence that they are no predictors for depression. There is moderate evidence that age and concurrent chemotherapy are no predictors for the course of depression.
Inconclusive evidence exists for 22 predictors of depressive symptoms. These include: being married, living alone, higher disease stage, concurrent chemotherapy, negative coping behaviour, positive coping behaviour, HNC-related symptoms, general health symptoms, being employed, higher income, ethnicity, alcohol use, current smoker, smoking history, radiotherapy, recurrence, toxicity, TNF-α, IL-6, CRP, IFN-γ, number of recent life events, received support, locus of control, openness to discuss cancer, sense of humour, childhood trauma, illness perception, beliefs about medicine, satisfaction with cancer information, and optimism. Apart from age and concurrent chemotherapy, there is inconclusive evidence for all other variables on predicting the course of depressive symptomology investigated in this systematic review.
Our results were largely comparable to previous findings reported by Haisfield-Wolfe et al (1). These include depression at an earlier point in time (positively associated), the extent of the social network (negatively associated), and the available support (negatively associated). However, some of our findings contradicted those earlier findings. For instance: Haisfield-Wolfe (1) reported that a low education level, physical functioning, higher age, and being of male gender were found to be associated with depression. Our review found moderate to strong evidence that these factors do not predict depression. The difference in findings might be explained by the fact that Haisfield-Wolfe et al also included cross-sectional studies, which we excluded in our review.
The results of this review imply that depression at an earlier point in time, the (Karnofsky) performance status of a patient, the patient’s available support, and the extent of their social network could be taken into account in medical decision making for (preventive) psychological care to head and neck cancer patients.
The quality assessment reveals that there were 7 high quality studies (32%). most of the included studies (i.e. 15 of 22) have a small sample size. The majority of studies did not have a high baseline participation rate. Furthermore, some studies made use of the same study population, albeit for a different research question. If one population was used for research questions with overlapping factors, the conclusions derived from these studies could bias the results in this study. Another factor is selection bias (35): the main study population of included studies was curatively treated patients. This means that these results might not be generalizable to a palliative care population. Another limitation is that this review does not make a distinction between determinants and predictors, so no causal relationships can be identified from the results.
C
ONCLUSION
Depressive symptoms at an earlier point in time positively predicted depressive symptoms at a later stage in patients with head and neck cancer. A more extensive social network, a better performance status and more available support were found to predict fewer depressive symptoms. Gender, age, education, treatment, tumour location, anxiety, physical functioning, and poor sleep quality did not predict depression. For all other variables, the evidence is inconclusive: different studies found different results or within studies different results were reported per measurement point.
Since the evidence for most predictors is insufficient, further research is needed. A future study should ensure that sample size is sufficient for the number of variables under examination. Also, a meta-analysis should be conducted to determine the importance and possible risk for predictors that were found.
Reference List
(1) Haisfield-Wolfe ME, McGuire DB, Soeken K, Geiger-Brown J, De Forge BR. Prevalence and correlates of depression among patients with head and neck cancer: a systematic review of implications for research. ONCOL NURS FORUM 2009 May;36(3):E107-E125.
(2) Satin JR, Linden W, Phillips MJ. Depression as a predictor of disease progression and mortality in cancer patients. Cancer 2009;115(22):5349-61.
(3) Massie MJ. Prevalence of depression in patients with cancer. 2002 p. 29.
(4) Krebber AMH, Buffart LM, Kleijn G, Riepma IC, Bree R, Leemans CR, et al. Prevalence of depression in cancer patients: a metaanalysis of diagnostic interviews and selfreport instruments. PsychoOncology 2014;23(2):121-30.
(5) Archer J, Hutchison I, Korszun A. Mood and malignancy: head and neck cancer and depression. J Oral Pathol Med 2008 May;37(5):255-70.
(6) Hayden JA, Bombardier C. Evaluation of the quality of prognosis studies in systematic reviews. Annals of internal medicine 2006;144(6):427-37.
(7) Chinapaw MJM, Proper KI, Brug J, Van Mechelen W, Singh AS. Relationship between young peoples' sedentary behaviour and biomedical health indicators: a systematic review of prospective studies. Obesity reviews 2011;12(7):e621-e632.
(8) Proper KI, Singh AS, Van Mechelen W, Chinapaw MJ. Sedentary behaviors and health outcomes among adults: a systematic review of prospective studies. American journal of preventive medicine 2011;40(2):174-82.
(9) Jack K, McLean SM, Moffett JK, Gardiner E. Barriers to treatment adherence in physiotherapy outpatient clinics: a systematic review. Manual therapy 2010;15(3):220-8.
(13) de Graeff A, de Leeuw JR, Ros WJ, Hordijk GJ, Blijham GH, Winnubst JA. Long-term quality of life of patients with head and neck cancer. Laryngoscope 2000 Jan;110(1):98-106.
(14) Neilson K, Pollard A, Boonzaier A, Corry J, Castle D, Smith D, et al. A longitudinal study of distress (depression and anxiety) up to 18 months after radiotherapy for head and neck cancer. Psychooncology 2013 Aug;22(8):1843-8.
(15) Neilson KA, Pollard AC, Boonzaier AM, Corry J, Castle DJ, Mead KR, et al. Psychological distress (depression and anxiety) in people with head and neck cancers. Medical Journal of Australia 2010;193(5):S48.
(16) de Leeuw JR, de Graeff A, Ros WJ, Blijham GH, Hordijk GJ, Winnubst JA. Prediction of depression 6 months to 3 years after treatment of head and neck cancer. Head Neck 2001 Oct;23(10):892-8. (17) de Leeuw JR, de Graeff A, Ros WJ, Blijham GH, Hordijk GJ, Winnubst JA. Prediction of depressive
symptomatology after treatment of head and neck cancer: the influence of pre-treatment physical and depressive symptoms, coping, and social support. Head Neck 2000 Dec;22(8):799-807.
(18) Aarstad HJ, Aarstad AK, Heimdal JH, Olofsson J. Mood, anxiety and sense of humor in head and neck cancer patients in relation to disease stage, prognosis and quality of life. Acta Otolaryngol 2005 May;125(5):557-65.
(19) Archer JA, Hutchison IL, Dorudi S, Stansfeld SA, Korszun A. Interrelationship of depression, stress and inflammation in cancer patients: A preliminary study. Journal of Affective Disorders 2012 Dec 20;143(1-3):39-46.
(20) Chen AM, Jennelle RL, Grady V, Tovar A, Bowen K, Simonin P, et al. Prospective study of psychosocial distress among patients undergoing radiotherapy for head and neck cancer. International Journal of Radiation Oncology* Biology* Physics 2009;73(1):187-93.
(21) Chen SC, Lai YH, Liao CT, Lin CC, Chang JT-C. Changes of symptoms and depression in oral cavity cancer patients receiving radiation therapy. Oral Oncol 2010 Jul;46(7):509-13.
(22) Finizia C, Palme C, Bergman B. A longitudinal study of the Swedish Self-Evaluation of Communication Experiences after Laryngeal Cancer questionnaire in patients treated for laryngeal cancer. Acta Oncol 2002;41(3):262-8.
(23) Hammerlid E, Ahlner-Elmqvist M, Bjordal K, Biorklund A, Evensen J, Boysen M, et al. A prospective multicentre study in Sweden and Norway of mental distress and psychiatric morbidity in head and neck cancer patients. Br J Cancer 1999 May;80(5-6):766-74.
(24) Hammerlid E, Silander E, Hornestam L, Sullivan M. Health-related quality of life three years after diagnosis of head and neck cancer--a longitudinal study. Head Neck 2001 Feb;23(2):113-25.
(25) Hassel AJ, Danner D, Freier K, Hofele C, Becker-Bikowski K, Engel M. Oral health-related quality of life and depression/anxiety in long-term recurrence-free patients after treatment for advanced oral squamous cell cancer. J Craniomaxillofac Surg 2012 Jun;40(4):e99-102.
(26) Humphris GM, Rogers S, McNally D, Lee-Jones C, Brown J, Vaughan D. Fear of recurrence and possible cases of anxiety and depression in orofacial cancer patients. Int J Oral Maxillofac Surg 2003 Oct;32(5):486-91.
(27) Humphris GM, Rogers SN. The association of cigarette smoking and anxiety, depression and fears of recurrence in patients following treatment of oral and oropharyngeal malignancy. Eur J Cancer Care (Engl ) 2004 Sep;13(4):328-35.
